*Article* **IoT Open-Source Architecture for the Maintenance of Building Facilities**

**Valentina Villa 1, \* , Berardo Naticchia 2 , Giulia Bruno 3 , Khurshid Aliev 1 , Paolo Piantanida <sup>1</sup> and Dario Antonelli 3**


**Abstract:** The introduction of the Internet of Things (IoT) in the construction industry is evolving facility maintenance (FM) towards predictive maintenance development. Predictive maintenance of building facilities requires continuously updated data on construction components to be acquired through integrated sensors. The main challenges in developing predictive maintenance tools for building facilities is IoT integration, IoT data visualization on the building 3D model and implementation of maintenance management system on the IoT and building information modeling (BIM). The current 3D building models do not fully interact with IoT building facilities data. Data integration in BIM is challenging. The research aims to integrate IoT alert systems with BIM models to monitor building facilities during the operational phase and to visualize building facilities' conditions virtually. To provide efficient maintenance services for building facilities this research proposes an integration of a digital framework based on IoT and BIM platforms. Sensors applied in the building systems and IoT technology on a cloud platform with opensource tools and standards enable monitoring of real-time operation and detecting of different kinds of faults in case of malfunction or failure, therefore sending alerts to facility managers and operators. Proposed preventive maintenance methodology applied on a proof-of-concept heating, ventilation and air conditioning (HVAC) plant adopts open source IoT sensor networks. The results show that the integrated IoT and BIM dashboard framework and implemented building structures preventive maintenance methodology are applicable and promising. The automated system architecture of building facilities is intended to provide a reliable and practical tool for real-time data acquisition. Analysis and 3D visualization to support intelligent monitoring of the indoor condition in buildings will enable the facility managers to make faster and better decisions and to improve building facilities' real time monitoring with fallouts on the maintenance timeliness.

**Keywords:** digital twin; facility management; building information modeling; HVAC; fan coil; Internet of Things; predictive maintenance; fault detection; smart building

#### **1. Introduction**

Correctly managing the maintenance phase is the pivot to lower costs and energy waste [1] and to preserve the asset value over time and, above all, to maintain the level of performance required by users. Real Estate property managers are now realizing that investing in this activity brings tangible benefits in terms of investment preservation and a strong perception by customers/users. One of the early frameworks used to visualize the collected data was the building management system (BMS), developed in the 1980s [2]. The BMS is a system for controlling and monitoring a building's facilities, such as heating, lighting, electrical and mechanical services, safety and security [3]. Since the introduction

**Citation:** Villa, V.; Naticchia, B.; Bruno, G.; Aliev, K.; Piantanida, P.; Antonelli, D. IoT Open-Source Architecture for the Maintenance of Building Facilities. *Appl. Sci.* **2021**, *11*, 5374. https://doi.org/10.3390/ app11125374

Academic Editor: Lavinia Chiara Tagliabue

Received: 30 April 2021 Accepted: 30 May 2021 Published: 9 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

of BIM [4], the industry's awareness of the importance of adopting digital models for FM is increasing [5]. Thus, there is a need for faster, more efficient, and possibly error-free real-time visualization and analysis of collected data. In fact, building control together with maintenance operators need to analyze both stored and current data measured by sensors to track and monitor the overall performance of the building [6].

Maintenance is generally classified as corrective (run-to-failure), scheduled, preventive, predictive and proactive [7]. Corrective maintenance operates only in case of failures and in the event of interruption of a service; scheduled maintenance uses the estimated life span of each component to provide for its replacement according to predefined schedules [8]. Preventive maintenance involves system inspection and control at fixed intervals to lessen the likelihood of it failing unexpectedly [9]. Predictive maintenance, through sensors and machine learning algorithms, allows detection of abnormal behaviors well before accidents happen [10]. Presently, this is mainly applied in manufacturing industries [11]: out-of-range data [12] are identified, analyzed and intervention procedures are implemented to restore the system's correct behavior. Proactive maintenance focuses on the root causes of failure, not fault symptoms, in order to improve the system operation.

The breakthrough for improving the efficiency and effectiveness of FM is the integration with IoT technologies, i.e., building environmental data [13] and the use of wireless sensors to collect data from building systems [14].

Several technologies are introduced: for example, data visualization using a BIM platform that helps operators to efficiently navigate through buildings [15], or the use of BIM combined with wireless sensors [16] to monitor temperatures in the subway stations [17], or CO<sup>2</sup> levels [18] or occupancy levels [19] in the rooms.

Therefore, the technologies now developed should allow facility managers real-time analysis, optimization, and visualization of large data sets to better manage energy consumption, operation costs, and user comfort. In general, there is still limited research on the application of these new technologies during the building's service life [20], particularly concerning the FM sector. In most cases, even if the use of automated sensors/devices and databases is present, the information collected is not fully exploited [21,22].

Similar research has already been introduced [23–25], where some authors [23] represent automated IoT and BIM-based alert systems for comfort monitoring in buildings. They used humidity and temperature sensors to detect discomfort in rooms, whereas other authors [26] integrated IoT sensor data (temperature, illuminance and power consumption) to monitor indoor conditions of the buildings. The weak point of all the above works is that they do not provide building facility management services. This unavoidably results in laborious and inefficient processes [20].

In addition, FM operators still rely on paper maintenance and control sheets, and this again greatly increases both the time required for compilation and processing of the information [21]. Therefore, the proposal of a stream-lined and technology-integrated workflow methodology is essential for the improvement of FM processes and the effective improvement of the efficiency of plant systems. In the present case study, the proposed methodology will be tested on a HVAC device, specifically on a FC.

The study goal is to create an IoT network for FM in order to monitor components of the HVAC system and detect their failures. This work presents part of a larger research focused on developing a proof-of-concept implementation of a digital framework based on the connection between existing technologies, to support the building data digital transaction. The specific purpose of this paper is to define a data management methodology integrated with fault detection, tailored for the FM. A cloud-based user interface (UI), an integrated IoT and BIM model provides the building 3D model with information on room temperature, humidity and luminosity, and fan coil service condition. On the edge of the system, this integrated fault detection methodology notifies and sends alerts to the facility managers when any anomaly is detected. Moreover, maintenance operators can easily find the location of the faulty component on the building 3D model by room ID or room name, both of them integrated into the application. Using a cloud-based service, building supervisors and facility

managers can remotely monitor the thermal condition of fan coils and take necessary actions, e.g., if the operating temperature exceeds the pre-defined thresholds. As a result, this paper is organized as follows: a review of the literature related to the current research topic is reported in Section 2, the design architecture of the framework is presented in Section 3 together with the FM fault detection method. IoT and BIM application with data visualization within the case study is presented in Section 4, and finally the results of the IoT and BIM application for the considered case study are discussed.

#### **2. Literature Review**

#### *2.1. Operation and Maintenance in Facility Management*

Considering the whole life cycle, operation and maintenance (O&M) is the longest and most costly phase compared to others [27]. Therefore, the efficiency of control and monitoring systems is becoming increasingly important. An overall improvement in operational efficiency will be achieved due to improved technologies for data collection and inter-device communication, besides the decreasing cost of sensors with good data collection capabilities [28]. As the complexity and number of devices in HVAC systems increases, errors due to incorrect system configuration may increase; it is estimated that 40% of buildings have improperly configured devices [29] and there is a potential for 40% energy savings from correcting building errors [30]. This not only results in energy waste, but also in system malfunctions leading to discomfort indoors and also reduces air quality levels [31]. For this reason, fault detection and diagnosis (FDD) is becoming more and more important.

As FM is worth up to 85% of the building's entire life cycle cost [32], efficiency gains become crucial for cost containment and building quality preservation. Furthermore, facility management is a multidisciplinary topic involving many stakeholders and requires the collaboration and coordination of several different teams [33]. ISO 41011:2017 defines FM as an "organizational function which integrates people, place and process within the built environment with the purpose of improving the quality of life of people and the productivity of the core business" [34]. Most buildings currently do not operate efficiently due to standard procedures, fittings not customized for specific use, that generate a lack of information for operators [35]. In addition, there is habitually no historical data available to compare situations and to properly search for the cause for malfunctions or failures; the tools that come along with digital management are often still manually updated spreadsheets, making performance tracking challenging. Moreover, this mixed process is prone to error and confusion. Sometimes some important information is included in the digital models but they cannot describe the building systems' operation accurately. On the other hand, understanding and using real-time data is crucial nowadays, especially in high-performance buildings equipped with complex and multiple systems that require dynamic management to optimize their energy performance and provide adequate indoor environmental conditions for a large number of occupants [6].

HVAC systems are essential for efficient building operation and indoor air quality (IAQ) is critical to the habitability and comfort of spaces; many research efforts are now focusing on air quality monitoring for post-COVID-19 recovery [36]. HVAC systems also consume about 30% of primary energy in Europe [37], 14% of primary energy in the United States [38], and about 32% of the total amount of electricity generated in the United States [39]. Calculated energy waste as a result of defects in building HVAC systems is due to inappropriate operating procedures, equipment malfunctions, or design issues [40], while controlled and efficient management of HVAC systems is expected to save an average of 5–15% of total energy consumption if fully adopted in existing buildings [41,42].

#### *2.2. BIM and IoT as Digital Twins*

With the introduction of sensing and IoT in the construction industry, a new vision of building management tools has been emerging. In particular, a significant solution taken from the manufacturing sector has arisen: cyber-physical systems (CPS) [33]. CPS,

better known as digital twins (DT), are a revolutionary digital vision, where the most innovative technologies come together to create a virtual model that performs exactly like its physical "twin". This continuous information exchange between the data collected in the real building and those processed by the virtual model through self-learning algorithms, makes it possible to mirror the life of the real twin and the corresponding building, to predict the building components' health, their useful life, failures [43] and, in general, the building performance [44]. In the Industry 4.0 era, physical and virtual worlds are really growing together [45].

Asghari, Rahmani, and Javadi [46] define IoT as "an ecosystem that contains smart objects equipped with sensors, networking and processing technologies integrating and working together to provide an environment in which smart services are taken to the end-users".

IoT is an unprecedented disruptive technology that has led to interconnection between people and objects on a scale and pace unimaginable even a decade ago [47]. It enables new strategies to improve the quality of life [48], allows autonomous decisions to connected devices through algorithms and machine learning, and can properly inform users to make the best decisions [49], for example, in case of emergencies or major failures. IoT and data networks have great potential in optimizing FM activities, including document management, historical data cataloging, logistics and material tracking, building component lifecycle monitoring, and building energy controls [20]. Several studies have been conducted on the use of data from IoT devices (e.g., [50]), although many of them do not include any BIM integration, and all smart buildings and homes [51] are examples of IoT technology integration.

Actually, the DT can be simply configured as a dashboard with critical performance indicators (KPIs), e.g., data on temperature, pressure, tilt, power, voltage, etc., representing inputs from sensors located on systems' components or in specific environments. This allows the DT to actively participate from the design to the functional phase of any product or process. According to [52], the first appearance of the DT was in NASA's Apollo program. In that case, DT was defined as "an integrated multi-physics, multi-scale, probabilistic simulation of a vehicle or system that uses the best available physical models, sensor updates, fleet history, etc., to mirror the life of its flying twin" [53].

In this situation, NASA needed to reproduce the useful life of its assets in a precise manner through the statistical analysis of the data collected by the sensors [54]. Thus, a DT consists of an entity composed of a physical space, a digital space, and a connection layer.

The digitization permeates across all sectors, including construction, and aims to build systems as well as approaches for helping operators not only in the conceptualization, prototyping, testing, and design optimization phase, but also during the operational phase. The importance of numerical simulation tools in the first phase is unquestionable, however, the potential of real-time data availability in the operational phase is opening new possibilities to monitor and improve operations throughout the life cycle of a product. Grieves [55], in his white-paper, called the presence of this virtual representation a "Digital Twin".

The digital twin arose from the integration of sensor networks and the digitization of machinery and manufacturing systems in the manufacturing industry [56]. The main difference between a design-phase simulation and a digital twin is that the latter requires a physical asset and sensor network, whereas simulation lives in a completely virtual environment [43]. Accordingly, the study in [56] presented an extended definition: "digital twins will facilitate the means to monitor, understand, and optimize the functions of all physical entities, living and non-living, by enabling seamless transmission of data between the physical and virtual worlds." Research in [57] described the simulation aspect of digital twins as the collection of relevant digital artifacts involving engineering and operational data, as well as the description of behavior using various simulation models. Digital twins use these specific simulation models based on their ability to solve problems, derive solutions relevant to real-life systems, and describe behavior. In general, the study

in [57] defined the digital twin view as "a complete physical and functional description of a component, product, or system along with all available operational data."

Hicks [58] differentiates a digital twin from a virtual prototype and redefines its concept as an appropriately synchronized object of useful information (structure, function, and behavior) of a physical entity in virtual space, with information flows enabling convergence between the physical and virtual states. According to the authors in [44], digital twins represent real objects or subjects with their data, functions, and communication capabilities in the digital world.

Researchers in [59] defined the digital twin of a building as "the interaction between the real-world indoor environment of the building and a digital but realistic virtual representation model of the building environment, which provides the opportunity for real-time monitoring and data acquisition." In their definition, an indoor environment indicates information about air temperature, airflow, relative humidity, and lighting conditions, while a digital virtual environment indicates computational fluid dynamics and luminance levels. Furthermore, based on the study presented in [59], some of the notable advantages of creating a building digital twin are as follows: (1) collection, generation, and visualization of the building environment; (2) analysis of data irregularities; and (3) optimization of building services. The digital twin is continuously updated with sensor data in near real-time, and the data can be reprocessed with algorithms that make the data concise and more usable, even for customers or users. In addition, the large amount of collected data makes decision-making more informed, giving the ability to make predictions about how the object will behave in the future [60].

The DT can be applied to an asset for a technical simulation, with the purpose of integrating different model components to simulate almost every aspect of a complex system [45]. In a manufacturing plant in the industrial sector, for example, a DT offers the ability to simulate and improve the production system, considering the logistical aspects and optimization of the production process.

According to an Oracle report [61], the digital twin has many application areas: real-time remote monitoring and control; increased efficiency and safety at work; risk assessment; synergy and collaboration among team members; informed decision support; customization of products and services; better documentation, data collection and communication; and, most importantly for us, predictive maintenance and scheduling. The IoT paradigm is paving the way for smart cities and smart buildings. The definition of a common model of web-based protocols for data exchange allows transferring this data between objects, making the various network parts interactive. This is where the digital twin concept was born, which is a candidate to revolutionize the building management model through predictive analysis and dynamic simulations based on real-time data [62].

A full digital twin will ensure sensors are collecting data and the onset of failures can be detected well in advance through intelligent analysis of that data. This will allow for better maintenance scheduling.

Digital twin is a concept that can be exported from manufacturing to many fields and technologies [63]. Moreover, digital twin is one of the top ten strategic technology trends, and, according to future research predictions, the digital twin market will reach USD 15 billion by 2023 [57,64].

Once DT has been introduced, a look at how it can be linked to BIM methodology is provided. The U.S. National Institute of Building Sciences (NIBS) defines BIM as "The digital representation of the physical and functional characteristics of a structure. As such, it serves as a shared knowledge resource for information about a structure, forming a re-liable basis for decisions throughout its life cycle from inception onward" [65].

BIM, up from DT is widely used in the construction industry. BIM models allow for integrated information management throughout the building life cycle, including the FM phase [66]. BIM provides a collaborative platform to manage not only the project and its components, but also the relationships between stakeholders [67]. The BIM model acts as a collector of all building information, thus for facility managers it is of great support, working on information that is always up-to-date and shared by all team members, overcoming the uncertainty of paper data and information fragmentation.

However, BIM is much more widely used in the design phase, while in the FM phase it is still underutilized.

The most significant causes hindering this integration are: (1) the use of BIM only as a three-dimensional model, which has no added value in the maintenance management phase [68]; (2) FM workers are not involved in the creation of the model so the information contained within the BIM system is not useful [69]; (3) the need for interoperability between BIM and FM technologies and the lack of open systems [70]; (4) the lack of clear roles, responsibilities, contract and accountability framework [70]; and (5) the information contained in BIM models is static and not dynamic, as FM requires. Data is provided during the design phase but is not updated during the building life cycle [71].

Currently, as reported above, the BIM process is mainly used for visualization, construction, coordination, quantity calculation, planning, and project cost evaluation [72]. BIM is basically a repository for project information, with the major limitation of not transferring all this information to the facility management phase [73] and not being able to handle real-time data.

However, the great potential of a three-dimensional BIM model is the integration with all the virtual reality (VR), augmented reality (AR) and mixed reality (MR) systems. Research can be found discussing the development of collaborative BIM approaches based on AR/MR/VR [74–76]. One example is the creation of a framework based on building information modeling, mixed reality, and a cloud platform to support information flow in facility management [77].

#### **3. Materials and Methods**

#### *3.1. IoT and BIM System Architecture*

Automated data acquisition (DAQ) technology has seen significant advancements in hardware and software in recent years, even if the majority of available technologies are still expensive and not open-source, namely, a sort of "black box" where users have no access to alter and modify the implemented algorithms. Furthermore, a free access to the data previously stored is frequently impossible without buying a specific software. To address the aforementioned issues and overcome the limits of commercial technologies, studies were recently carried out to investigate and develop customized design of automated DAQ systems. This chapter proposes an automated data acquisition system within the IoT and BIM integration methodology based on open-source technologies.

All the sensors for building facilities and rooms are hosted by Arduino based microcontrollers and a Raspberry Pi single board computer with integrated Wi-Fi module to acquire the planned data in real time and store them on the server. The collected data is then processed to monitor and detect anomalies in building facilities operation.

A digital twin is next implemented by linking the stored data with a dashboard for trend visualization; a BIM model is used to visualize the component position and to have an overall view of the building.

The general architecture of the IoT integrated into the BIM is shown in Figure 1. The architecture exhibits IoT building facilities' sensors continuously sending data to the sensor nodes. In the hardware and networking section, raw data coming from sensors are processed by the sensor nodes' microcontroller and raw data becomes human-readable data. The following sensor nodes send processed data to the server that is connected to the gateway. Users can gain access to the local and cloud sensor node dashboards through this gateway providing data visualization and fault detection features. In the last section of the architecture, IoT data is integrated into BIM and combined IoT-BIM data is visualized on the BIM dashboard.

**Figure 1.** IoT and BIM integration system architecture.

#### *3.2. IoT System Components*

The proposed IoT system is based on open-source tools and market components. This system is flexible to set up, to add extra sensors and sensor nodes in real case-building facilities, as well as to add IoT sensors data into the 3D digital model of the building, according to any requirement of end-users.

Coming to the specific application, physical parameters must be acquired to monitor the room's internal conditions and to control the FC operation. For that reason, sensors shown in Table 1 are placed in the room and connected to their sensor boards.


**Table 1.** Applied sensors with technical specifications.

To oversee the temperature, relative humidity, and illuminance of the room, the DHT22 sensor and a light-dependent resistor (LDR), are connected to the ESP8266 module. FC monitoring sensors are connected to the RPIZCT4V3T2 board. Furthermore, both sensor boards are connected to the server board Raspberry Pi 3B (Rpi3B) to store and display incoming data locally and remotely using wireless sensor nodes (WSN).

The networking and hardware connectivity block diagram of the system is displayed in Figure 2.

The ESP8266 module is powered by a simple universal serial bus (USB) cable and automatically connects to the Rpi3B by means of installed credentials. In case of wireless network failure or unavailability, the board tries to connect again every five seconds until a successful connection is established. After that, the ESP8266 starts to collect data from the deployed sensors through the defined pins and the acquired data are sent through a serial monitor to the server with a sampling frequency of 1 Hz.

Ω

Ω

An RPIZCT4V3T2 sensor board is used to monitor the FC of the room; Figure 3 shows a simplified block diagram of the sensor board.

**Figure 3.** Simplified block diagram of the RPIZCT4V3T2 sensor board.

The RPIZCT4V3T2 board hosts an Arduino microcontroller (MCU) that is connected to two types of temperature sensors and current/voltage sensors that are connected to the MCU through an amplifier and analog to digital converter (ADC). Additionally, the RPIZCT4V3T2 board is connected to the Raspberry Pi (Rpi) Zero W. The MCU collects all the raw data, computes necessary values, sending the final computation to the Rpi Zero W using the universal asynchronous receiver-transmitter (UART) serial port; Rpi Zero W supports Wi-Fi and the board connects to the server Rpi3B through IP address. The sampling frequency of each sensor is specific and explained later in the next section.

The main component of the IoT and BIM system is the Rpi3B, which is a small singleboard computer; it operates as server, networking router, middle communicator, hosts a dashboard, and a database. The Rpi3B system connects with other sensor nodes through the Wi-Fi network and logs the received data from the sensors to a database. On the Rpi3B system, Node-Red [6] is installed providing the ability to access all the sensor variables through serial protocols and displaying them on its own local dashboard. For each sensor node, a topic is assigned that is responsible for sending (publishing) a message to the main server (Rpi3B) which will act as a receiver (subscriber) using Message Queue Telemetry Transport (MQTT) protocol. MQTT protocol is an OASIS standard messaging protocol

for IoT; it is designed as an extremely lightweight publish/subscribe messaging transport that presents as being ideal for connecting remote devices with a small code footprint and minimal network bandwidth [27].

Moreover, DNSmasq free software is installed to use the Rpi3B as a router and to provide a communication bridge between internal (sensor nodes) and external (internet network) components using Internet Protocol (IP) addresses. By setting the SSID, password, and an IP address on the Rpi3B using DNSmasq, the system becomes visible on the network to the publisher and subscriber.

To store sensor data locally on the Rpi3B, MySQL database, PHP interpreter, and Apache web server are utilized. By powering the Rpi3B, MySQL gets the IP address with a configured port number and waits for Node-Red to send the data to be collected; then, MySQL allocates the received data to the linked tables.

Later MQTT, MySQL, and Node-Red utilize the same credentials to background run and connect to the network, getting the IP address from DSNmasq. Communication between sensor nodes being established, the MQTT protocol on the configured port of the server receives all the subscribed topic's data from publishers through TCP protocol and provides publishing devices access to the port. Simultaneously, Node-Red starts on port 1880 using the same IP address to manage and monitor the data flow to the server and also sends data to the BIM dashboard by using forge nodes in JSON format or using a data driven approach, in CSV format.

Thus, acquired data from sensors will be displayed on the dashboard of the Rpi3B and using MQTT on the internet, and MySQL can be monitored and viewed through any device connected to the same network by opening the IP address followed by the port number.

#### *3.3. FC Fault Detection Methodology*

In HVAC systems, FCs are used as heating/cooling elements of rooms, which are a very common systems, especially in office buildings, hospitals, and schools. In order to have precise control of the FC operation, sensors must be inserted in the various components of the FC. The choice of which elements to monitor was made in relation to the identification of anomalies through collected data. Figure 4 shows the methodology for detecting faults in the FC and subsequent maintenance planning according to the condition of the FC. The image also shows the possible failures of the FC linked to the collection of anomaly data of the FC parts. The most common anomalies of the FC system are blocked motor, insufficient air flow, dirty filters, capacitor failure, insufficient water flow, etc. Each of these anomalies require an appropriate action. Particularly, the maintenance actions to be executed by operators are to clean the filter and battery, to change bearings, to replace capacitors, and to check valve adjustment and presence of pipe sediments. Thus, based on the anomaly types and data collected by sensors, a FC condition monitoring system is implemented. In addition, centralized collected data can be used to inform FM about the condition of any FC and provide preventive maintenance services, if needed.

Figure 5 demonstrates the anomaly detection algorithm and alarm system implemented on the RPIZCT4V3T2 sensor board. The maintenance alarming system is composed of three main sections: installed sensors on the FC, management and monitoring system of the FC components, and alarming FMs or end-users when an anomaly has occurred. In the sensors section, three voltage sensors, type EU: 77DE-06-09 are responsible for acquiring data from the three motor speeds (v1, v2, v3); three current sensors SCT-013-000 acquire data (i1, i2, i3) from the three speeds; and temperature sensors DS18B20 are responsible for monitoring T1 (delivery water temperature), T2 (return water temperature), and T4 (outlet air temperature) in the range 0–90 ◦C and T3 (inlet air temperature) in the range 0–50 ◦C. T5 (motor case temperature) is monitored with an PT100 temperature sensor in the range of 0–200 ◦C. On the controller of the RPIZCT4V3T2 sensor board, the management and monitoring algorithm of each component of the FC is implemented. The algorithm makes decisions depending on the sensors' signal values compared to the installation conditions.


**Figure 4.** Kinds of faults and data collection for preventive maintenance of the FC.

**Figure 5.** Integrated anomaly detection flow chart.

Likewise, Table 2 shows the sampling frequency for each FC component depending on the power condition. For example, on the server side, delivery pipe, return pipe, air inlet, and air outlet components temperatures are monitored every 30 min by sending average values when the voltage is at least 200 V, while the motor temperature must be monitored every 10 s and send average values of 30 min to the server. The motor voltage

″

″

″

″

″

″

″

″

″

″

″

″

must be monitored more frequently and the sensor board must send average values to the server every 3 min if the voltage is in the range between 0 and 200 V. On the other side, electric power (current) shall be monitored in the first 10 s, when the voltage is above 200 V. More detailed sampling frequencies on the server-side of the system integrated on the microcontroller of the sensor board and sensor locations in the FC is given in Table 2.

The given measure ranges are implemented on the RPIZCT4V3T2 board using Node-Red flows and displayed on the dashboard too. The system also informs the end-user if any anomaly is detected on the FC.


#### **Table 2.** Sampling frequency and FC sensor allocation.

#### *3.4. Data Transmission and Visualization 3D Model and IoT Data*

The 2D and 3D models of the fan coil unit and the building model used for this research are both built in Autodesk Revit authoring tool. To connect IoT sensor data with BIM, the Autodesk Forge Platform is utilized: it is a cloud-based platform and provides application programming interface (API) services. A free trial of Forge APIs lasts 90 days and gives access to 100 cloud credits, 5 GB storage, and services such as authentication (two-legged authentication), data management, Model Derivative, Model Viewer, all of them necessary and sufficient to create the customized application. By registering on the Forge platform, the end-user receives a client ID, client password, and all the services necessary for application development. BIM–IoT integration and the visualization process using the Forge platform is described in Figure 6: the right column records the services used for the custom application on the Forge platform, while the left column shows the applications layer.

**Figure 6.** BIM–IoT data integration and visualization approach diagram.

The 3D building model can be uploaded onto the Forge platform by inserting credentials provided by the authentication API and using the upload.html file. Afterwards, the Model Derivative API translates the uploaded 3D model using the viewer.html file, so the uploaded model can be accessed by typing a static address (e.g., localhost:9000) on any browser.

The middle column of Figure 6 displays the browser visualization on the Forge Viewer: it can be customized by inserting different extensions, buttons, or plots using the JavaScript programming language that supports the Open-source VS Code editor. Furthermore, IoT sensor data in JSON or in CSV formats can be added through plugins to the uploaded 3D building model. The model properties will appear in the viewer as soon the model itself is uploaded on the Forge Viewer.

A custom web application of Forge Viewer can be created by accessing the object's properties in the model and relying on the database identifier (DbId) for the model objects.

#### **4. Case Study**

This section explains the case study setup, the allocation of sensors, the algorithms to monitor and detect anomalies in building facilities, and summarize the achieved results.

The case study has been operated at the Politecnico di Torino in the DISEG Laboratory. The laboratory room is located below ground in the department. Figure 7 shows the case study building: (a) geolocation and (b) BIM model Revit software representation. In all the rooms, fan coils are located under the windows.

**Figure 7.** Case study building: (**a**) geolocation on the maps.google.it and (**b**) 2D view of the case study room with the location of the FC.

The case study fan-coil motor is made by EMI (EuroMotors Italia), type FC83M-2014/1 with a 4 possible speeds. The FC motor rotates in an anti-clockwise direction at 1100 revolutions per minute (RPM) at maximum speed. Moreover, FC has a cooling or heating battery and a filter that must often be monitored; to provide an automatic preventive maintenance system, the FC is equipped with sensors to acquire real-time data.

More technical details of the FC are provided in Table 3.


**Table 3.** Technical specifications of FC Motor FC 83M-2014/1.

The ESP8266 sensor node is placed on the FC together with sensors. All the collected data from the room facilities are sent to the BIM using the Forge API; Autodesk Forge API provides the client ID and client password to access the uploaded customized BIM on the cloud; using the callback URL everyone can access the IoT-integrated BIM dashboard. A general schematic diagram of sensors' locations of the FCs, sensor boards, and data acquisition system of the BIM model is showed in Figure 8.

The system is powered with simple USB cables and starts collecting data from the installed sensor nodes in the room. The Rpi3B mainboard then connects to the ESP8266 and RPIZCT4V3T2 sensor nodes and starts acquiring data from sensors and at the same time registers these data to the MySQL database. Using the developed extensions on the Visual Studio Code for Autodesk Forge Viewer, collected IoT data and the 3D building model of Revit can be visualized together on the Forge Viewer through a static URL and port provided by Forge API. The functional flowchart of the whole system is shown in Figure 9.

**Figure 8.** Sensors' location in the FC and building facilities.

**Figure 9.** Functional flowchart of the system.

FC with wireless sensor nodes (b), FC battery with installed temperature sensors (a), the motor with a temperature sensor (c), and room sensors with the sensor node (d) are shown in Figure 10.

**Figure 10.** Case study sensor nodes placement to the room and FC: (**a**) FC battery and temperature sensor, (**b**) FC with RPIZCT4V3T2 board, (**c**) motor with temperature sensor, and (**d**) room equipped with ESP8266 sensor node.

Data are measured by fan coil and room sensors and then displayed on the dashboard of the Rpi3B server using Node-Red flows in Figure 11.

**Figure 11.** Node-Red flows of the FC and room sensors data acquisition system.

Measured and integrated variables with their descriptions on the Node-Red flow besides the dashboard are provided in Table 4.

To verify the IoT and BIM integrated application dashboard, one voltage (V1) and one current (I1) sensors are connected to the first port of the sensor's board and all temperature sensors including RTD are allocated to the FC as shown in Figure 8.


**Table 4.** FC Measured physical parameters.

#### **5. Results**

The proposed framework was tested on the fan coil in the room with connected sensors. The FC sensors were linked to the set-up speed and each time the speed was adjusted. The final result of the dashboard with the acquired data at speed 1 is shown in Figure 12a—the motor power consumption values are as follows: V1 = 244.1, I1 = 0.1 A, RP1 = 32.3 and PF = 0.899. Fan coil data at speed 2 on the dashboard are shown in Figure 12b—the motor power consumption values are V1 = 241.5, I1 = 0.2 A, RP1 = 38.1 and PF1 = 0.924. Finally speed 3 on the dashboard is shown in Figure 12c—the motor power was measured as V1 = 244, I1 = 0.2 A, RP1 = 47.8 and PF1 = 0.944. The average temperature of the FC motor was RTD = 23.43 ◦C and average outlet temperature was indicating as T2 = 29.45 ◦C because the fan coil was heating the room. Rooms' dashboards with physical parameters acquired from sensors are visualized in Figure 12d.

The Rpi3B server board memory was not sufficient for the big data storage To avoid overloading the local server memory, only daily maximum and minimum values were sent from the sensor board to the cloud and BIM server. The fault detection methodology described in the previous section was applied to the system. If the values supplied by sensors exceed the predictable ranges, the system sends alarm signals or real-time notifications to the facility managers and operators using Telegram or SMS.

On the Forge platform, IoT data are integrated into the BIM using the Forge reference application (https://github.com/Autodesk-Forge/forge-dataviz-iot-reference-app, accessed on 6 April 2021) and two NPM modules (React UI components and Client-Server Data-Module-Components): NPM is the packet manager for Node.js and it is an opensource project helping to support JavaScript developers.

Installation and running of the Forge reference application started by cloning the application from GitHub repository (git clone https://github.com/Autodesk-Forge/forgedataviz-iot-reference-app.git, accessed on 6 April 2021) to the VScode using lines in the console of the editor.

The project folder is composed of client-side codes, a guide on how to upload the Revit model, router configuration, and client-server configuration files that speed up work for developers on their custom applications. The structure of the folders and files of the reference application is shown in Figure 13.

In the reference application an .env file is created and added to Forge CLIENT\_ID, CLIENT\_SECRET, and Forge\_BUCKET; in this way the custom 3D building model is added to the project (Figure 14). By running the project, the custom 3D model is then uploaded in a static browser (http://localhost:9000/upload, accessed on 6 April 2021).


(**d**)

**Figure 12.** Node-Red dashboard results: (**a**) FC at speed 1; (**b**) FC at speed 2; (**c**) FC at speed 3; and (**d**) room dashboard values coming from the ESP8266 sensor node.


**Figure 13.** Reference application directory structure.

**Figure 14.** The .env file with authentication credentials and the added sensor values in the model DbId objects in JavaScript.

To interact with uploaded model objects on the Forge Viewer, DbIds are required.

This is because most API methods to manipulate entities require the argument DbId (or array), such as isolate, hide, highlight, etc.; knowing the DbId array, a map with the model hierarchy node, a unique ID can be built and then custom functions can be written to connect IoT sensor variables to the BIM model.

The objects; dbIds can be acquired using functions getSelection() that returns a list of DbIds, and getProperties() expecting DbId as input, etc. The extraction process of DbIds from model objects is shown Figure 15.

**Figure 15.** Uploaded custom model DbId's extraction.

To identify each fan coil DbId's was added as a list in customized JavaScript code. It included sensor parameters such as position, type, etc. The added custom IoT sensor data were implemented through additional functions as shown in Figure 15.

Finally, Figure 16 shows IoT and BIM applications developed on the Forge platform. In the application, the fan coil color changes according to data coming from the sensor nodes. The color is "green" if the fan coil is in a good condition, "red" if overheated, and "blue" if the FC is cooling.

**Figure 16.** Final result of implemented heatmap function to the fan coil data and BIM.

#### **6. Discussion and Conclusions**

Cloud services for data analytics, including platforms for data visualization, are now available from providers like Microsoft Azure, Amazon Web Services and Google Cloud. These services can gather data and generate alerts, notifications, and graphs, but they are not flexible and cannot interact with the BIM model or sensor data.

This paper introduced a fully automated integrated framework for the maintenance of building facilities using open source IoT technologies. The IoT part of the proposed framework is composed of sensors to be installed in building facilities and wireless sensor network nodes that continuously send received data through gateways to the local and cloud servers according to configured sampling frequencies. The proposed fault detection methodology was integrated into the server and sends alarms to the end user or managers when any anomaly occurs for it to be fixed effectively; BIM was utilized to view the monitored HVAC FC component's condition and the room's physical parameters such as temperature, humidity, and luminosity using wireless-connected remote devices. A case study was used to test the new framework implementation. The integration of an FC monitoring system into the BIM model would improve the building's maintenance plan by helping the facility managers to inspect the monitored building environments inside the 3D model. Thus, facility managers can profit from the proposed framework to resolve maintenance problems in the following aspects: (a) using building facility anomaly or failure signals, facility managers can plan and schedule maintenance work in advance; (b) conditional and real-time data coming from sensors allow for more accurate maintenance; (c) data visualization and real-time data on the dashboard create the possibility of avoiding the risk of disastrous breakdowns and reduce unplanned forced outages of building components; (d) IoT sensor data for building components within the BIM model makes maintenance work more convenient, e.g., it would be easy to find the location of the failure component in the real-time BIM model; and finally, (e) collected indoor and facility sensor data is the initial tool to perform predictive maintenance actions.

The study conducted in this paper aims to fill the gaps of the following researches: limited number of sensors [23], an absence of automation [24], and lack of data acquisition system [25]. The final results of the fully automated framework composed of IoT sensors, dashboard, IoT and BIM integrated application, and implemented preventive

maintenance methodology of the building facilities on the server prove the proposed framework's applicability.

Nevertheless, this work has several limitations, which are as follows:


Future research work will attempt to use collected data from the proposed system for predictive maintenance management of building facilities. Decision support systems for facility predictive maintenance management systems, artificial intelligence (AI), specifically machine learning (ML) tools, and algorithms should be used.

Time series sensor data, either historical failures, anomalies, or both, data can be used as ML inputs. For example, building facility anomalies or failures can be predicted using classification analyses. The use of regression analyses can utilize time series sensor data to forecast physical parameters of the building facilities' components. Thus, either predicted values, failures, anomalies, or combination, of building facilities offer the FM managers the ability to provide even more effective maintenance services before failures.

**Author Contributions:** Conceptualization, V.V., B.N., and G.B.; methodology, V.V. and P.P..; software, K.A.; validation, P.P. and D.A.; formal analysis, P.P.; investigation, G.B., B.N. and V.V.; resources, P.P. and K.A.; data curation, D.A. and P.P.; writing—original draft preparation, V.V. and K.A.; writing review and editing, B.N., D.A. and P.P.; visualization, V.V.; supervision, V.V.; project administration, B.N. and V.V.; funding acquisition, B.N. and V.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Italian government, through the PRIN2017 project of the Ministero dell'Istruzione, dell'Università e della Ricerca (MIUR). The project entitled "Distributed Digital Collaboration Framework for Small and Medium-Sized Engineering and Construction Enterprises" is coordinated at the national level by Berardo Naticchia.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Anahita Davoodi \* , Peter Johansson and Myriam Aries**

Department of Construction Engineering and Lighting Science, Jönköping University, SE-551 11 Jönköping, Sweden; Peter.Johansson@ju.se (P.J.); myriam.aries@ju.se (M.A.) **\*** Correspondence: anahita.davoodi@ju.se; Tel.: +46-73-910-1581

**Abstract:** Validation of the EBD-SIM (evidence-based design-simulation) framework, a conceptual framework developed to integrate the use of lighting simulation in the EBD process, suggested that EBD's post-occupancy evaluation (POE) should be conducted more frequently. A follow-up field study was designed for subjective–objective results implementation in the EBD process using lighting simulation tools. In this real-time case study, the visual comfort of the occupants was evaluated. The visual comfort analysis data were collected via simulations and questionnaires for subjective visual comfort perceptions. The follow-up study, conducted in June, confirmed the results of the original study, conducted in October, but additionally found correlations with annual performance metrics. This study shows that, at least for the variables related to daylight, a POE needs to be conducted at different times of the year to obtain a more comprehensive insight into the users' perception of the lit environment.

**Keywords:** building performance simulation; lighting simulation; lighting quality; visual comfort; office field study; evidence-based design

#### **1. Introduction**

In the field of architecture, evidence-based design (EBD) is defined as the process of basing design decisions about the built environment on credible research and learning from previous evidence to achieve the best possible outcomes [1,2]. One of the easiest, quickest, and inexpensive methods to provide evidence for predicting or evaluating values in a built environment is using computational modeling. In a framework for evaluating evidence in EBD developed by Pati [3], computer simulation is categorized under 'experiment level', which confirms the application of simulation for providing evidence. For example, in the study by Jakubiec and Reinhart [4], simulation tools were used for the prediction of occupants' visual comfort within daylit environments. The results illustrated that it is possible to use current simulation-based visual comfort predictions to predict occupants' long-term visual comfort assessments in a complex daylit space.

Investigation regarding the application of simulation tools in the EBD process is ongoing [5–9]. A conceptual framework was developed to integrate the use of lighting simulation within the EBD process in a systematic way: the EBD-SIM framework [6]. The study concluded that the translation between the user evaluation and the simulationbased evaluation was a critical step in the integration of lighting simulation with EBD. Therefore, an initial validation study [7] was dedicated to a first application of the EBD-SIM framework in a post-occupancy evaluation (POE) step.

POE is defined as 'the process of evaluating buildings in a systematic and rigorous manner after they have been built and occupied for some time' [10] (p.3). POE's purpose and methodology are varied, often depending on the type of building. For example, POEs of office buildings are, in most cases, interested in occupants' comfort and productivity, utilizing both subjective and objective evaluations of indoor environmental quality (IEQ) [11]. The recent literature review conducted by Dam-Krogh et al. [12] investigated

**Citation:** Davoodi, A.; Johansson, P.; Aries, M. The Implementation of Visual Comfort Evaluation in the Evidence-Based Design Process Using Lighting Simulation. *Appl. Sci.* **2021**, *11*, 4982. https://doi.org/10.3390/ app11114982

Academic Editors: Lavinia Chiara Tagliabue and Ibrahim Yitmen

Received: 29 April 2021 Accepted: 25 May 2021 Published: 28 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

methods applied in previously performed POEs in office buildings with special attention to IEQ and productivity to compare and evaluate successful practices of POEs in office buildings. In more specific and recent POE studies, lighting quality of office buildings was investigated by [13–17]. These studies were concerned with obtaining better insight on occupant satisfaction and acceptance about daylighting by conducting POEs while using photosensors, shading, and/or control systems. This is a key factor for proper lighting design by utilizing daylighting without neglecting human comfort and achieving responsive buildings. Having more information available about occupants' behaviors, needs, and preferences enables architects to design better responsive elements of a building for optimal rates, scales, and types of changes.

While POEs are popular tools for the evaluation of different aspects of building from the occupant's point of view, most POEs are one-off studies [11]. Additionally, using POEs in a framework of EBD is not studied well (yet). One attempt to strengthen the EBD knowledge base by developing standardized POE tools was conducted by [18]. In this study, based on a review of over 100 research publications, a conceptual framework and a set of standard tools were established to comprehensively evaluate building performance in terms of eight key design areas, including air quality, visual environment, thermal comfort, acoustic environment, hazardous materials, conservation of resources, overall climate response, and building envelope (façade).

A real-time case study in a fully operational office building was conducted to analyze visual comfort from a subjective ('the user') and an objective ('the simulation') point of view to provide a systematic performance evaluation using the EBD-SIM framework [7]. It covered both long-term and short-term evaluations of the light environment. The results showed that, although illuminance preference varies significantly among individuals, there was a positive correlation between the overall lighting quality perceived by the occupants and the amount of light on the task area. Regarding performance metrics, the results showed the highest correlation with point-in-time horizontal illuminance (Eh), especially on the task area (Eh-task) and human perception. The implementation of the study results in the EBD-SIM process model is schematically illustrated in Figure 1.

**Figure 1.** Study results [7] implementation in the EBD-SIM model addressing the workflow during the post-occupancy evaluation steps.

The initial validation study [7] suggested that a POE should be conducted more frequently to obtain a better insight into user perception of daylight and, subsequently, the use of the new evidence to improve the design of the EBD-SIM model further. Therefore, the field study was repeated to perform a POE for a second time, including the same

participants but at a different time of year to investigate the possible improvement in the reliability of a POE study (within-subjects study [19]). In addition, a larger sample with a similar procedure was selected in which users experienced only one condition each (between-subject-study [19]) to investigate if previous findings are confirmed with a larger sample or that results are just a coincidence and more detailed simulation of the actual/intended use of a space is required to forecast the visual comfort.

Even though a long-term goal is to show how the EBD-SIM framework can be used to incrementally develop evidence through several projects, the research questions for this project were as follows:


Note that the first two questions are repeated from the preceding study [7], though with a larger sample. The most frequently used visual comfort metrics were identified based on the state-of-the-art literature review conducted by [7].

The third question was added specifically for the follow-up POE study.

#### **2. Materials and Methods**

#### *2.1. Research Design*

A study in a fully operational office building focusing on visual comfort analysis was conducted to explain the application of the EBD-SIM framework in the POE step. Objective data were collected via computational modeling and subjective data by using an online questionnaire. Calculated performance metrics and questionnaire results were collected similarly to the previous study [7]. Previously collected data were included in the study as part of the repeated measures method.

#### *2.2. Procedure*

At first, the physical environment was modeled, and performance metrics were calculated. The simulation results were compared against a limited and randomized set of illuminance measurements in the real building to check if they were in the same range.

The results showed a difference between simulated and real values of 20 ± 11%, which is within normal variability levels for simulated objects [20]. Secondly, using an online questionnaire, occupant characteristics were recorded, and their feedback regarding visual comfort was gathered. Finally, user feedback and simulation output were compared. This comparison was performed to investigate how metrics measured by simulation tools (to assess visual comfort) correlate with actual visual comfort perceived by the users, and to answer the research questions. The first sample of the population (*N* = 15) filled out the questionnaire in October, and the second sample of the population (*N* = 46) filled it out in June. For *N* = 10 of these people, all working in one wing of the survey location, the questionnaire was a repetition. In total, *N* = 51 unique people participated in the study, which means, in total, *N* = 61 complete survey results were gathered.

An overview of the three research questions and the results of the comparison between POE and the performance metrics (PM) for different groups of participants at different data collection moments is illustrated in Figure 2.

**Figure 2.** Three main research questions and the results of POE vs. PM comparison for different groups of participants during the first and second times (October and June).

#### *2.3. Site*

The study was carried out in a four-story academic building (with a basement) located in Jönköping, Sweden (lat. 57.778168◦ N, long. 14.163526◦ E). To the left and right of the building are buildings of comparable height. Physical data were collected from 51 private office rooms on the second, third, and fourth floors. Most of the rooms are approximately 15 ± 5 m<sup>2</sup> and are furnished mainly by a large desk, a chair, and one or more bookshelves. The rooms have windows in the north-east (21), south-west (19), north (9), and south-east orientations (7). Three rooms on the north-east and south-west side have a second window in the same direction. Five rooms on the south-west side have a second window on the south-east side. The number of rooms with two windows is highlighted in the parenthesis in Table 1. The size of the windows on the second floor on the north-east side is one third larger compared to that on the third and fourth floors. There is permanent solar shading at the building's south and south-west sides (see Figure 3). All rooms are equipped with conventional suspended luminaires and were used according to the users' needs during the survey, but they were not modeled nor further included in the study.

**Table 1.** Number of rooms based on the location of the windows; rooms with two windows are shown within parenthesis.


**Figure 3.** Photo of the real building (**top**) and a 3D render of the building model (**bottom**) showing the south and south-west façades including the permanent overhang. Note that on this side of the building, the basement level is above the ground.

#### *2.4. Survey Participants*

Occupants of four wings of the building were invited to participate in the survey. They were all academic employees whose work involves research and teaching/education, mainly using computers. The invitation email was sent to 150 people, of which 61 people (39 male/22 female, average age 45 ± 20 years) responded (response rate 40.7%). For *N* = 36 participants, this was the first time they filled out the survey; *N* = 15 people answered during the first study in early October [7], and *N* = 10 of this group participated for the second time in this study, see Table 2. (in parenthesis: the respondents who were inquired twice). Most participants in this study worked for more than one year in their current office room, and more than half of the participants (*N* = 36) reported wearing vision aids (near vision = 13, distance vision = 10, both = 11, trifocal = 1, other = 2).

**Table 2.** Number of respondents per floor and wing—within parenthesis: the respondents who were inquired twice.


#### *2.5. Performance Metrics for Visual Comfort*

Seven currently used performance metrics for visual comfort were obtained, including five metrics related to illuminance values (Ev-eye, Eh-room, Eh-task, Daylight Autonomy DA, Spatial Daylight Autonomy sDA) and two glare indicators (Daylight Glare Probability DGP, Annual Sunlight Exposure ASE), see Table 3. Further details are explained in [7].


**Table 3.** Collected visual comfort metrics (daylighting).

#### *2.6. Simulation Model*

The building was modeled in Autodesk Revit. For detailed daylight analysis, the model was exported to Rhinoceros 3D to perform lighting analysis with the DIVA plugin. In addition to the simulated building wing, direct surroundings were also modeled to consider potential shading. The material properties and simulation parameters used for the calculations are described in Table 4.



For each workplace area on the second, third, and fourth floors, a grid of 1.45 × 1.45 m at workplace height (*h* = 0.75 m) was used to calculate the mean, minimum, and maximum horizontal illuminance under CIE Standard Clear Sky, CIE Intermediate Sky, and CIE Standard Overcast Sky conditions [21]; for the rest of the paper, these are referred to as 'clear', 'overcast', and 'intermediate' sky conditions. The same setting was applied for the room area of each room for horizontal illuminance (Eh-room), ASE, and DA calculations. One point in the middle of each room with a viewing direction towards windows was selected for the vertical eye illuminance (Ev-eye) and DGP calculations under clear, intermediate, and overcast sky conditions. As the questionnaires were filled out in early October and early June, a day in early October (11) and early June (4) was chosen for the simulation to have a comparable day length and sun path. In total, 67.2% of the questionnaires were filled out in the morning; hence, for the simulation, a time in the morning was selected (10 a.m.).

#### *2.7. Questionnaire*

The questionnaire was a web-based form in English to collect feedback regarding the visual comfort perception of office occupants. It contained questions regarding office characteristics, satisfaction with the lit environment in general ('annual'), satisfaction with the lit environment at the time of response ('momentary'), user preferences, and personal information. For 15 satisfaction variables of the lit environment 'in general' (annual) and 'at the time of response' (momentary), a 7-point Likert satisfaction scale [22] was used (1 = very satisfactory to 7 = very unsatisfactory). All variables were included in the analysis, see Table 5.

**Table 5.** Description of the variables included in the analysis with their variable name. Satisfaction variables are all on a 7-point Likert scale (1 = very satisfactory to 7 = very unsatisfactory).


#### *2.8. Analysis*

The correlation of all 15 satisfaction variables as well as the correlation among these variables with seven groups of performance metrics was explored by calculating Pearson correlation coefficients using IBM SPSS Statistics 24. To determine whether there is statistical evidence that the mean difference between paired observations on a particular outcome is significantly different from zero, a paired sample t-test was run for ten participants who filled out the questionnaire two times during early October and June. Only significant correlations are reported.

#### **3. Results**

In this section, the objective, subjective, and correlation analysis of the follow-up POE study for ten participants who filled out the questionnaire two times during early October and June as well as the larger sample (*N* = 46) who filled out the questionnaire only once in June is reported.

#### *3.1. Within-Subject Analysis (Follow-Up POE Investigation)*

#### 3.1.1. Objective Comparison

1. Horizontal illuminance on a room area (Eh-room)—The overview of the results for the mean value of horizontal illuminance Eh-room of the calculated sensor points of the ten rooms' areas at the work plane height is shown in Figure 4a,b. Where the mean Eh-room values in October were below 200 lux for all sky conditions, the amount of daylight increased, as expected, during the second POE in June. On average, the rooms received 200 lux or more horizontal illuminance during June. Since the participants mostly worked under the clear sky condition, more details of the results for the clear sky condition are illustrated in the right image.


**Figure 4.** *Cont*.

**Figure 4.** Overview of the results for the mean value of (**a**) horizontal illuminance of the calculated sensor points of the ten rooms' areas at the work plane height (Eh-room), (**c**) horizontal illuminance on the task area of the calculated sensor points of the ten rooms' areas (Eh-task), and (**e**) mean value of vertical illuminance in the center point of each room with viewing direction towards the daylight opening of the ten rooms (Ev) under three different sky conditions for 11 October and 4 June both for 10 a.m. (**a,c,e**). The details of the mean values under clear sky condition (**b**,**d**,**f**).

**Figure 5.** (**a**) DGP values under clear sky (blue), overcast sky (red), and intermediate sky (green) conditions: imperceptible level (DGP < 0.3), perceptible level (0.3 < DGP < 0.35), disturbing glare (0.35 < DGP < 0.4), intolerable glare (DGP > 0.45); (**b**) The details of the mean values under clear sky condition.

3.1.2. Subjective Comparison

Descriptive statistics—Descriptive statistics analysis for all 15 satisfaction variables shows that the mean values of all variables except 'Glare sun' (satisfaction with glare from the sun at the moment) and 'A-artificial' (annual satisfaction with artificial lighting) were a bit lower for the second POE in June compared to October. On average, for the satisfaction values, the change was 0.4 points. This means people were more satisfied with the built environment in general in June (1 = very satisfactory, 7 = very unsatisfactory). Similarly, in October, the highest satisfaction belonged to 'satisfaction with the job', which increased slightly in June. Then, 'satisfaction with daylight' (both 'annually' and 'at the moment') had the highest satisfaction rate in June (*M* = 2.3, *SD* = 1.5). Satisfaction with daylight 'at the moment' with a 1.8 point improvement from October shows the highest increase in satisfaction. The least satisfaction in June belonged to artificial light (*M* = 3.5, *SD* = 1.58). See Table 6 for more details on other variables.

Paired samples *t*-test—To determine whether there is statistical evidence that the mean difference between paired observations on a particular outcome is significantly

different from zero, a paired sample *t*-test was run for ten participants who filled out the questionnaire two times during early October and June. The results show that, although the mean satisfaction values are increased from October to June for almost all variables, a significant difference was only found in scores for natural light from October to June (*sig* = 0.016, *M* = 1.80, *SD* = 1.93). In addition, a marginally significant difference was found for 'A-Glare sun' (*sig* = 0.052, *M* = 0.50 *SD* = 0.71) and 'Light desk' (*sig* = 0.053, *M* = 0.80 *SD* = 1.13).

**Table 6.** Descriptive statistics analysis of all 15 satisfaction variables (1 = very satisfactory, 7 = very unsatisfactory) of the first POE in October (*N* = 10) and the second POE in June (*N* = 10) of within-subjects study.


Correlation analysis—The results show that there were 'very strong' (0.8 < *r* < 1) to 'strong' (0.6 < *r* < 0.79) degrees of correlations for all the variables between annual and at the moment evaluations except for 'Natural light' in October. In June (*N* = 10), the highest correlation was found for 'Glare sunlight' (0.95, *p* < 0.01) and 'Glare artificial light' (0.92, *p* < 0.01). This means that participants either perceived glare as similar all year round or found it hard to recall the difference between glare at the moment and annually.

#### *3.2. Between-Subjects Analysis*

All participants who responded to the survey in June (*N* = 46) worked in individual office rooms. These rooms are located on the second, third, and fourth floors of the building's four wings. Based on the location of the window, these rooms are categorized into four groups which are presented in Table 7.

**Table 7.** Number of rooms based on the location of the windows—within parenthesis: the rooms with two windows.


There are three rooms at the north-east side and one room at the south-west side with two windows on the same side. Five rooms at the south-east side have a second window at the south-west side. These rooms are highlighted in parenthesis in Table 7.


east side are in the range of 'intolerable glare' (DGP > 0.45). For more details, see Figure 9.

**Figure 6.** Overview of the results for the mean value of horizontal illuminance of the calculated sensor points of the room areas at work plane height.

*Appl. Sci.* **2021**, *11*, 4982

**Figure 7.** Comparison of the amount of daylight incident at the task area against the mean value of Eh-room.

**Figure 8.** The overview of the results for the mean value of vertical illuminance in the center point of each room with viewing direction towards the daylight opening of the ten rooms under three different sky conditions on 4 June at 10 a.m.

**Figure 9.** DGP values under clear sky (**blue**), overcast sky (**red**), and intermediate sky (**green**) conditions: imperceptible level (DGP < 0.3), perceptible level (0.3 < DGP < 0.35), disturbing glare (0.35 < DGP < 0.4), intolerable glare (DGP > 0.45).

#### 3.2.2. Subjective Evaluation Using Questionnaires

Descriptive statistics—The sample of all participants in June (*N* = 46) shows a similar trend to the smaller sample. This means that the highest satisfaction in June belonged to satisfaction with 'A-Job' (*M* = 2.4, *SD* = 1.87). After that, daylight (both annually: 'A-Natural', and moment: 'Natural') with (*M* = 2.9, *SD* = 1.75 and *M* = 2.8, *SD* = 1.72) and 'Glare artificial' (*M* = 2.8, *SD* = 1.6) had the highest satisfaction rates. For more details, see Table 8.



Correlation analysis—The correlation of all fifteen satisfaction variables was explored by calculation of Pearson correlation coefficients, and only significant correlations are reported here. The results of correlation analysis for the larger samples (*N* = 46 and *N* = 61) show that there were 'very strong' (0.8 < *p* < 1) degrees of correlations for all the variables between annual and at the moment variables. Similarly, for the smaller samples (*N* = 10 for both studies in June and October), the highest correlation was found for 'Glare artificial light'. For daylight (A-Natural vs. Natural), a correlation was found only for June in both small and large samples. For more details, see Table 9.

**Table 9.** Correlation analysis for variables between 'annual' and 'at the moment'.


\* *p* ≤ 0.05; \*\* *p* ≤ 0.01.

All the samples consistently showed the strongest correlation between 'Light quality' and 'Light desk' for both annual and moment situations. For the large sample in June (*N* = 46), all correlations of the variables for the moment situation were stronger than the same annual situation. For example, the correlation of 'Light desk' with 'Light quality' was (*r* = 0.87, *p* < 0.01), and the correlation of 'A-Light desk' with 'A-Light quality' was (*r* = 0.85, *p* < 0.01).

Both daylight and artificial light seem to contribute to the user assessments of the light quality in the room, with the correlation between 'Artificial light' and 'Lighting quality' (*r* = 0.81, *p* < 0.01) being slightly higher than the correlation between 'Natural light' and 'Lighting quality' (*r* = 0.77, *p* < 0.01). Both sources of glare also contributed to the assessments of the lighting quality in June with the same weight (*r* = 0.71, *p* < 0.01).

Unlike in October, when the assessments of the amount of light on the desk area ('Light desk') seem to only be linked to the amount and quality of artificial lighting, in June, both light sources seem to have contributed to the assessments of the 'Amount of light on desk area'. For both light sources, the correlations were stronger for at the moment assessments compared to the annual assessments. 'Amount of artificial light (moment)' and 'Amount of daylight (moment)' showed strong correlations with the 'Amount of light on desk area' (*r* = 0.80, *p* < 0.01 and *r* = 0.71, *p* < 0.01, respectively). 'Glare from artificial light' and 'Glare from sunlight' showed a lower correlation with 'Light desk' (*r* = 0.76, *p* < 0.01 and *r* = 0.64, *p* < 0.01, respectively).

#### *3.3. Objective–Subjective Correlation Analysis*

The correlation between performance metrics and perceived visual comfort for the sample in June (*N* = 46) shows that, in total, eight subjective variables including satisfaction with 'Glare sun', daylight (both annual and at the moment: 'A-Natural' and 'Natural' variables, respectively), 'A-View', total light at the desk ('A-Light desk' and 'Light desk'), and lighting quality ('A-Light quality' and 'Light quality') had a moderate (0.3 < *r* < 0.5, *p* < 0.001) relationship with at least one of the simulated performance metrics. All variables except glare had positive correlations with user satisfaction. Note that it is shown as negative numbers since user satisfaction was defined in inverse order, with 1 being the highest satisfaction level and 7 being the lowest satisfaction level.

The variable 'Glare sun' had the highest correlation with point-in-time horizontal illuminance (momentary) at task area (Eh-task) for mean values under the clear sky condition (*r* = 0.38, *p* < 0.01) and the intermediate sky condition (*r* = 0.33, *p* < 0.01).

Satisfaction with daylight for both the annual and momentary situations ('A-Natural' and 'Natural') showed a significant correlation with Eh-room, ASE, and DGP. The highest correlations for Eh-room with 'A-Natural' were for the mean value of illuminance under clear sky and intermediate sky conditions (*r* = −0.30, *p* < 0.01), both classified as moderate correlations. Additionally, 'Natural' and 'A-Natural' showed a moderate correlation with DGP (*r* = −0.41, *p* < 0.01).

Since the view to the outside is often inextricably linked to daylight, it was included in the analysis as well. The 'A-View' variable correlated with vertical illuminance (Ev-eye), horizontal illuminance at task area (Eh-task), and horizontal illuminance for the room area (Eh-room) as well as with DA, sDA, and ASE. The highest correlation was found for the mean value of Eh-room under 'clear' sky conditions (*r* = −0.46, *p* < 0.01).

'A-Light desk' and 'Light desk' showed correlations with the horizontal illuminance at task area (Eh-task) and room area (Eh-room) as well as with DA and sDA. Additionally, 'Light desk' correlated with vertical illuminance (Ev-eye) and 'A-Light desk' with ASE. The highest correlation for 'A-Light desk' was found for the minimum value of Eh-room under intermediate sky conditions (*r* = −0.47, *p* < 0.01). The highest correlation for 'Light desk' was found for the minimum value of Eh-room under 'intermediate' sky conditions (*r* = −0.47, *p* < 0.01).

Finally, 'Light quality' and 'A-Light quality' had significant correlations with Eh-room, sDA, and ASE. The highest correlation for 'Light quality' was found for the minimum value of Eh-room under 'clear' sky conditions (*r* = −0.36, *p* < 0.01).

#### **4. Discussion**

This follow-up study, conducted in June, is the second validation test of the EBD-SIM framework to provide new evidence to obtain better insights about the lit environment, specifically concerning visual comfort in office rooms. The effects of a larger sample size and having two POE studies in two different seasons (October, June) were investigated to find out if previous findings are confirmed (research questions 1 and 2), and to elucidate

the usefulness of having two continuous POE studies for better analysis of visual comfort in office environments (research question 3). The results are categorized into two groups: within-subject study and between-subject study.

The main purpose of conducting a between-subject study in June with a larger sample (*N* = 46) was to verify previous findings in October (*N* = 15) and provide new or updated evidence for the EBD-SIM framework. Comparison between the two measurement moments showed similarities as well as differences.

The correlation analysis of subjective variables for both studies shows a 'strong' to 'very strong' degree of correlations for all annual and momentary variables except daylight (A-Natural vs. Natural), for which a correlation was not found in October. These strong to very strong correlations can be interpreted in that it is difficult for people to remember a lighting situation throughout the year, and the current situation dominates their feeling regarding the lit environment. Additionally, for daylight (A-Natural vs. Natural), it seems that occupants can better distinguish the difference between annual and momentary situations during dark seasons. From a subjective point of view, for all variables, it was consistently observed that the overall lighting quality perceived by the occupants had the highest correlation with the amount of light on the task area. While in October, the assessment of 'light quality at the task area' seemed to only be linked to the 'amount and quality' of artificial lighting, in June, in addition to artificial lighting, a strong correlation was found with 'Natural light' and 'Glare sun'.

Regarding the first question related to the correlation between visual comfort metrics and perceived occupant visual comfort, similar to the results obtained in October, the results confirm the previous finding that point-in-time-horizontal illuminance had the highest correlation with perceived visual comfort by occupants. Moreover, in the larger study (June), annual performance metrics showed some degrees of correlation. This means that it is worth calculating sDA, DA, and ASE to provide a moderate forecast on occupant perception of lit environments, especially for 'Light desk', 'Light quality', and 'Natural'. This analysis is in agreement with other studies, e.g., [15,24]. Since the results of vertical eye illuminance (Ev) and DGP are sensitive to the point of view of the occupants [25], the calculation of data specifically for the occupants' position and viewpoint improves the accuracy of the data.

Regarding the second question related to instantaneous and annual visual comfort perception and simulated comfort assessment metrics, it was found that for the large sample in June, all correlations of the subjective variables for the moment situation were stronger than the equivalent annual situation. Additionally, Eh, which measures the instantaneous situation, showed the highest number of correlations with perceived variables compared to the annual performance metrics. This could be interpreted in that human visual comfort perception for instantaneous situations and modeled comfort performance metrics have a higher agreement than human perception for annual estimations and modeled performance. In the future, POEs can be integrated into, e.g., (artificially) intelligent building control systems, providing direct feedback to the control agent so that each occupant is provided with their preferred lighting and, in a broader sense, with other desired IEQ conditions. Additional input from occupants such as user characteristics (e.g., age, gender, light sensitivity), user behavior, and user preferences can be collected to analyze the effects of the built environment on occupants in greater depth. Logging this feedback data and storing it in a database can provide a set of valuable evidence from the instantaneous feedback of occupants, which in turn can help with better prediction of human comfort and improvement of lighting design. For example, in an innovative study conducted by Newsham et al. [26], along with lighting simulation tools, a humanoid robot was used to attract the attention of the occupant about their real environmental situation and provide them with personalized suggestions to improve their well-being. If a building is responsive to the requirements and behavior of occupants and organizations, either via a POE and/or via continuous monitoring, it can become a truly intelligent building. Conducting metadata analysis on all data collected from evidence and designing building

interfaces or (self-)learning systems based on this evidence-based knowledge would make the interaction of occupants with buildings more convenient.

Regarding the third research question related to the usefulness of a POE with an increased frequency, the results show that in June, people were generally more satisfied with the lit environment compared to answers given in the original study performed in October. As daylight levels are higher in June, it seems that having more daylight has a positive impact on user satisfaction in general. The importance of daylight on user satisfaction was also shown by other researchers, e.g., [27–30]. For example, a study was conducted by Day, Futrell, Cox, and Ruiz [28] that measured physical data and surveys to assess occupants' subjective visual comfort. The results of their survey study (*N* = 1068) showed that occupants who were more pleased with (their access to) daylight were also more likely to have a higher level of satisfaction. This finding also indicates that, depending on the time of year, a single POE study can overestimate or underestimate different lighting quality metrics. The paired sample *t*-test analysis shows a significant difference for the daylight 'Natural' variable and a marginally significant difference for the daylight 'A-Glare sun' and 'Light desk' variables, which means at least for these variables, it is worth running a POE at least twice at different times of the year.

#### **5. Conclusions**

This study addressed implementing subjective–objective results in the evidence-based design process using lighting simulation tools in a POE step of the EBD-SIM framework. The POE focused on assessing occupant visual comfort in an individual office space to provide a systematic approach to repeatedly gather evidence in this field and build a knowledge database that can help improve how the results are analyzed, presented, and interpreted. As this study shows, running the EBD-SIM framework each time can provide new evidence. In the meantime, it helps to tailor the next run based on lessons learned in the previous run.

The results confirm the previous finding that the overall lighting quality perceived by the occupants had the highest correlation with the amount of light on the task area. In parallel, E<sup>h</sup> (point-in-time horizontal illuminance) showed a consistently positive correlation with the highest number of subjective variables. Moderate correlations between annual performance metrics and some of the subjective variables were also observed during June, which were non-existent in the October study.

This study suggests that, at least for daylight-related variables (e.g., 'Natural', 'A-Glare sun', and 'Light desk'), it is worth running the POE more than once at different times of the year to obtain a better insight into user perception of the lit environment.

As described in the EBD-SIM framework [6], it is preferable to use the simulation output in the POE step so that the situation present at the time of conducting a POE study can be approximated using lighting metrics, and so that the result of the survey can be analyzed in greater depth.

The scope of this research was limited to the visual aspects of lighting quality, and the simulation results included only daylight aspects. In the future, the study can be extended to include both electric lighting and daylight as well as visual effects and light effects beyond vision through a long-term evaluation.

**Author Contributions:** Conceptualization, A.D., P.J., and M.A.; methodology, A.D., P.J., and M.A.; software, A.D.; validation, A.D., P.J., and M.A.; formal analysis, A.D.; investigation, A.D., P.J., and M.A.; resources, A.D. and M.A.; data curation, A.D.; writing—original draft preparation, A.D.; writing—review and editing, A.D., P.J., and M.A.; visualization, A.D.; supervision, P.J. and M.A.; project administration, M.A.; funding acquisition, P.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Region Jönköpings Län's FoU-fond Fastigheter and the Bertil and Britt Svenssons Stiftelse för Belysningsteknik.

**Institutional Review Board Statement:** Ethical review and approval were not requested for this study as collection and analysis of data could not be used to identify participants, did not collect any sensitive personal data, did not include physical contact with participants, would not provide any risk of discomfort, inconvenience, or psychological distress to participants or their families, did not recruit from vulnerable groups, and did not include data collection undertaken overseas. Participants were asked to give informed consent.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Data not available. The authors do not have permission to share data.

**Acknowledgments:** The authors would like to acknowledge the survey participants' cooperation in this study as well as the valuable comments by the reviewers and editors of the journal of *Applied Sciences*.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


**Ibrahim Yitmen 1, \* , Sepehr Alizadehsalehi 2 , ˙ Ilknur Akıner <sup>3</sup> and Muhammed Ernur Akıner 4**


**Abstract:** In the digital transformation era in the Architecture, Engineering, and Construction (AEC) industry, Cognitive Digital Twins (CDT) are introduced as part of the next level of process automation and control towards Construction 4.0. CDT incorporates cognitive abilities to detect complex and unpredictable actions and reason about dynamic process optimization strategies to support decisionmaking in building lifecycle management (BLM). Nevertheless, there is a lack of understanding of the real impact of CDT integration, Machine Learning (ML), Cyber-Physical Systems (CPS), Big Data, Artificial Intelligence (AI), and Internet of Things (IoT), all connected to self-learning hybrid models with proactive cognitive capabilities for different phases of the building asset lifecycle. This study investigates the applicability, interoperability, and integrability of an adapted model of CDT for BLM to identify and close this gap. Surveys of industry experts were performed focusing on life cyclecentric applicability, interoperability, and the CDT model's integration in practice besides decision support capabilities and AEC industry insights. The evaluation of the adapted model of CDT model support approaching the development of CDT for process optimization and decision-making purposes, as well as integrability enablers confirms progression towards Construction 4.0.

**Keywords:** cognitive; digital twins; building lifecycle management; artificial intelligence; IoT; decision support; self-learning; optimization

#### **1. Introduction**

Computerization and digitization are beginning to significantly affect how physical/engineering properties are handled during their life cycles [1,2]. The capture, exchange, use, and control of data and information during an asset's entire life (design, construction, Operation and Maintenance (O&M), and disposal/renewal) are among the most challenging aspects of implementing Building Information Modeling (BIM), so-called BIM in asset management [3]. Intelligent, innovative asset life cycle management has arisen during the last years in the Architecture, Engineering, and Construction (AEC) industry [2]. Digital twins (DT), the blockchains, and the Internet of Things (IoT) draw interest because of their synergistic and information management functionality [4].

Cognitive computing is machines' ability to mimic the human capacity to sense, think, and make optimal decisions in a given situation [5]. While the path reaching fully cognitive systems is still in its early stage, there are several application areas where the technology has already been implemented in many applications such as chatbots by the service sector to provide optimal responses to customer feedback [6].

The DT is already in the early stages, mainly used for prototyping, and includes modeling, simulation, verification, evaluation, and confirmation of the physical artifact

**Citation:** Yitmen, I.; Alizadehsalehi, S.; Akıner, ˙ I.; Akıner, M.E. An Adapted Model of Cognitive Digital Twins for Building Lifecycle Management. *Appl. Sci.* **2021**, *11*, 4276. https://doi.org/10.3390/ app11094276

Academic Editor: Andrea Carpinteri

Received: 19 April 2021 Accepted: 7 May 2021 Published: 9 May 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

using a simulated replica [7]. Analysis emphasis has been heavily placed on simulations and what-if analyses to advise implementation and eventual physical product refinement by continuous monitoring and data assimilation. According to Zhang et al. [8], encompassing intelligence and cognition in a DT is a requirement to realize disruptive technology's potential accurately and to produce integration, calibration, and symbiotic connectivity in the environments, the physical and virtual replica. According to intelligence and cognition, mental abilities and mechanisms that utilize complex information management and synergy across physical and digital settings will manipulate and strengthen the "twining" structure. Dimensions can be listed as stimuli, interaction, aims, time, and situation switching. The main goal is to foster self-adaptive assessment and smart, proactive decisionmaking through the two realms in an info-symbiotic way and work on the more wealthy and finer-grained information base. Cyber-Physical Systems (CPS) and socio-technical environments, for example, may benefit from this view because their activities are marked by ambiguity, dynamism, and confusion. Cyber Foraging will represent intelligent analysis and planning simulation difficult and costly computational to achieve this in reality with limited computing resources.

In the digital transformation era in the AEC industry, Cognitive Digital Twins (CDT) are introduced as part of the next level of process automation and control towards Construction 4.0 [9]. CDT incorporates cognitive abilities to detect complex and unpredictable actions and reason about dynamic process optimization strategies to support decisionmaking in building lifecycle management (BLM). Nevertheless, there is a lack of awareness of the real impact of CDT integration, Machine Learning (ML), CPS, Big Data, Artificial Intelligence (AI), and Internet of Things (IoT), all connected to self-learning hybrid models with proactive cognitive capabilities for different phases of the building asset lifecycle. This study investigates the applicability, interoperability, and integrability of an adapted model of CDT for BLM to identify and close this gap. Four research questions are raised in line with the study's goals:


As the immense contribution, the knowledge domain understands how a CDT model operates and how it connects to most BLM fields. Besides, the study of integrability enablers and professionals' perceptions of the technical ecosystem's accessibility is facilitated by synthesizing industry professionals' questionnaire perspectives. Understanding DT's interoperability value, IoT, AI, big data, and sophisticated building management systems could also help build life cycle management.

As shown in Figure 1, this article is structured as follows: in section two (theoretical background), BLM's essential concepts and CDT are depicted. Section three describes the adapted model of CDT for BLM. In section four, an evaluation of the CDTsBLM model is presented. Section five offers the discussions on decision support capabilities, integrability enablement, and practical implications. In section six, the conclusions, recommendations, and future road map are presented.

**Figure 1.** Research framework of the adapted model of Cognitive Digital Twins for Building Lifecycle Management.

#### **2. Theoretical Background**

This section outlines key related works in the three main areas relevant to this article: (1) Building Lifecycle Management (BLM)'s essential concepts; (2) Digital Twins (DTs) in build environment and; (3) Cognitive Digital Twins (CDT) are presented.

#### *2.1. Building Lifecycle Management (BLM)*

The building lifecycle mainly includes the design, construction, operation, maintenance, and end-of-life stages. Each step can be separated into superimposed information layers that entail efficient data/information exchange strategies for interoperability throughout all lifecycle phases [10]. BLM refers to a method of integrating and handling the different stages of a construction project's lifecycle [11]. BLM is a strategic planning process that supports the development, operation, and maintenance of buildings and their associated infrastructure, including building planning, design, construction, operation, and maintenance. It aims to reduce costs and improve efficiency by ensuring that buildings are built, operated, maintained, and replaced in the most cost-effective and timely manner. The BLM is an integrated approach to building management that considers all the building activities, the building's surroundings, and the impact of these activities on the environment. The BLM, directly and indirectly, affects many aspects, such as buildings or infrastructures' operation and efficiency, operational risks, the environmental impact of buildings, people's quality of life, safety, and businesses. Such a complex and complicated process needs to be real-time, accurate, intelligent, and automated to monitor, detect, learn, analyze, simulate, validate, and operate. There are disintegrated data and information in every phase of the construction project, which contains a significant amount of design and cost information in the design process and decision-making steps; in the implementation step, a considerable amount of material consumption beyond the data generated during the design process and decision-making steps; and a vast data and information in operation and maintenance step.

As a result, to enforce BLM, a management mechanism must be developed that communicates each participant's expertise and phase of the development project to avoid a lack of sufficient and timely information connection and dissemination at different life cycle stages [12]. BLM sometimes starts with a physical analysis of the structure to generate a numerical representation (e.g., CAD documents). In this sense, at the beginning of its lifecycle, developing an effective BLM system is more than enough to avoid data loss during the building's construction, use/maintenance, and disposal [13]. By offering an interactive IT environment to handle the whole construction lifecycle, BLM seeks to migrate and enhance knowledge exchange in all phases of the building process [14]. Energy management, facilities management, maintenance management, and product/information traceability management are part of a scalable BLM scheme that allows users to incorporate

and reuse building knowledge, domain expertise across a building's life cycle [13–16]. A scalable networking infrastructure that offers uniform interfaces for sharing all forms of data consumed or generated by the participants, corporations, or information systems, in general, participating in the building lifecycle is a vital component of a functional BLM framework. BLM must interact with any intelligent items/systems that are part of the building lifecycle (sensors, actuators, RFID, databases) [10]. According to BIM, a technique that seeks to manage a building's entire life cycle in a particular data environment, proper data digitalization may optimize knowledge management and share within the multidisciplinary team [17].

#### *2.2. Digital Twins (DTs) in the Built Environment*

The CPS is realized through the DT for visualization, modeling, simulation, analysis, prediction, and optimization. DT contains three main components to create a practical loop: a physical entity, a virtual entity, and a data link [18]. Usually, there are two approaches to dynamic mapping in the DT. Inspection data are gathered in the physical world and subsequently transmitted to the virtual world for further analysis. Simulation, prediction, and optimization are achieved in the virtual model by learning data from multiple sources, offering prompt solutions to guide the realistic process and adapt to the changing context.

Based on Alizadehsalehi and Yitmen [19], DTs have various features in the AEC industry such as Real-time (gather and present real-time data of physical assets), Analytics (store data, run continuous analytics from historical data, and provide helpful insight), Simulations (utilize to run various data-driven simulations), visualization (overlay real-life and live 3D BIM models, images, and videos of the physical asset and also the foundation for immersive visualizations), Automation (a bi-directional system that can manage the behavior of physical assets), and Predictions (provide predictions of assets' future behaviors using historical data and analytics of various scenarios assets). As a comprehensive summary, Table 1 presents DT applications in the AEC industry that appeared in the recent literature (2019–2021).


**Table 1.** Digital Twins applications in the AEC industry that appeared in the recent literature.


**Table 1.** *Cont.*

#### *2.3. Cognitive Digital Twins (CDT)*

The DT concept allows the physical equivalent to be mirrored in virtual space, including exchanging data between them [7]. CDT expresses an evolution of the DT concept. It has been crafted to fit the requirement of monitoring complex industrial processes and apply the same trade model, shadow, and thread of DTs [6]. The balance between rapidity, resolution, and exception handling is crucial from any industry's economic perspective [44]. Virtualization in a dynamic, run-time process allows the digital counterpart's behavioral model to be constantly modified to mimic the physical element's actions, resulting in the CDT [45]. Virtualization is a dynamic design-time process involving computational approaches to model the physical feature, evolving into the complex, run-time process that allows the digital counterpart's behavioral model to be constantly adjusted to mirror the physical element's actions, resulting in the CDT. The CDT is a DT with cognitive abilities, including detecting anomaly and behavioral learning, and the power to determine physical twin actions to improve measures defining its state or function [46]. Therefore, a CDT uses

optimization approaches to aid decision-making and data from the physical twin analyzed using analytics or ML.

To put it another way, the CDT is envisioned as a robust monitoring and control mechanism and an essential part of the decision-making action that leads to system optimization. Using optimization techniques inside the heart of the cognitive twin and its impact is the primary crucial differentiating point instead of currently available DT solutions [46]. To make the transition from physical assets in the form of digital replicas to cognitive advancement, Abburu et al. [6] used a three-layer structure to describe the types of twins needed: digital, hybrid, and cognitive. The need to build isolated models of systems for anomaly detection, connect the models for predicting unusual behavior, and problem-solving skills to deal with uncertain situations constitutes the three-layer separation. CDT is characterized in DT through advanced semantic abilities to detect the mechanisms of virtual model evolution, enhance DT-based decision-making, and foster the interpretation of virtual model interrelationships [47]. The CDT ensures that assets are adequately managed and that problems outside technical stakeholders are resolved by implementing Internet of Things (IoT) systems [48].

CDTs may have a high degree of intelligence, allowing them to mimic human cognitive processes and perform conscious acts with little or no human intervention [8]. The Knowledge Core of CDT has semantic-driven recognition, learning, inference, estimation, and decision qualifications consisting of a series of prediction and ML models developed using the data from multiple sources such as physical equivalents and sensors from all aspects of operational conditions of the industrial systems. Besides, it incorporates temporal supply chain data and processes as well as experts' domain knowledge. As a result, the CDT can train and improve to represent and depict the physical asset's current state and operating conditions in real-time. Furthermore, in both the digital and physical worlds, the CDT can identify, analyze, deduce, forecast the twinned physical system's present and potential actions, and produce decisions by interrelating machines and humans.

Lu et al. [49] suggested a new cognitive twins so-called CT definition and a knowledge graph-centric framework for the CT process. Du et al. [50] explored how to build individualized information systems for future smart cities using a human-centered DT simulation approach of cognitive behaviors. Eirinakis et al. [46] suggested an Enhanced Cognitive Twin so-called ECT introducing advanced cognitive skills to the DT asset that allow assisting choices to allow DTs to respond to internal or external stimulation in the context of process industries. The ECT can be used at varying levels of the supply chain hierarchy, including sensor, device, process, workforce, and manufacturing stages, and can be integrated to allow lateral and vertical interaction.

The concepts of the Hybrid and Cognitive Digital Twin (COGNITWIN) toolbox were developed by Abburu et al. [6] to cover cognitive skills for efficient management and operation of processing equipment, for lowering production costs, and efficiency improvements in the process industry. A sensor network can constantly track and capture data from different plant processes and properties stored in a standard setup database. The COGNITWIN project mainly aims at adding the cognitive component to process control systems, thus enabling them to self-organize and provide solutions in case of unexpected behaviors. Figure 2 shows the different stages of DT to CDT. A DT is a formal digital representation of an asset, process, or system that captures any systems' attributes and behaviors through IoT-based various reality capturing sensors suitable for communication, storage, interpolation, and processing to measure, simulate, and experiment with the digital replica to understand its physical counterpart. A DT for monitoring, diagnostics, and prognostics to optimize asset performance and utilization uses sensory data combined with historical data, human expertise, and fleet and simulation learning to improve prognostic outcomes. A DT gets data from physical entities and applies them to the model.

**Figure 2.** Different stages of Digital Twins to Cognitive Digital Twins.

A Hybrid Digital Twin is usually defined as the DT comprised of combined multiple models. A Hybrid Twin (HT) extends the DT by intertwining different models to take advantage of both physics space and data-driven modeling. HT gets the data from the physical entities and uses them in several models jointly. The way to increase the degree of influence and scope of DT is to have cognitive features, such as reasoning, planning, and learning. Digital twins based on data analytics require immense amounts of data for accuracy, and while physics-based simulation models are highly accurate, they take an incredible amount of time to run. New hybrid systems are combining the best of both worlds for a digital twin that is both quick and exact.

Although HT has a lot of different models, there are so many parameters that influence the processes that, in some situations, are not covered by existing models. CDT represents the next step in evolving the DT concept in the AI era, incorporating cognition aspects to deal with unforeseen situations effectively. Revolutionary DTs will arise as a result of intertwining distinct models to accomplish advanced predictive capabilities and finding solutions to problems to be encountered by integrating expert knowledge. CDT gets data from physical entities and compares them with models, including models of expert knowledge.

Table 2, as a comprehensive summary, exhibites diverse CDT applications in various fields of industry based on the latest research (2019–2021).


**Table 2.** Diverse Cognitive Digital Twins applications in various fields of industry.


**Table 2.** *Cont.*

#### **3. Methodology**

#### *3.1. Adapted Model of aCognitive Digital Twin for Building Lifecycle Management (CDTsBLM)*

This paper reviews previous work on BLM, DT in the built environment, and CDT and presents an adapted framework developed by Lu et al. [48] and Abburu et al. [6] to improve BLM with CDT in the AEC industry. The adapted framework in this research is referred to as the CDTsBLM Model of the framework. This framework's processes, as shown in Figure 3, are discussed in detail in this section. The CDT is a capabilities-driven digital representation of its physical twin. It should be a capability augmentation and an intelligent digital companion cycle and evolution phases. CDT facilitates cognition towards improving the behavior of the complex process systems inherent in planning, design, construction, and operations. An ML pipeline automates the ML workflow by facilitating data to be converted and associated into a model that can be processed to automate the ML model's outputs and input data completely. As shown in Figure 3, the conceptual framework facilitates the implementation and evaluation of consistent CDT in BLM by integrating various pipelines of ML and analytical tools at various stages from planning to the whole operations through the processing phases during operations.

*Appl. Sci.* **2021**, *11*, 4276

#### 3.1.1. CDTsBLM Framework

The first section of CDTsBLM is CDT and adaptive dynamic process modeling. In this section, all IoT-based systems, consisting of reality capturing sensors, other constructionrelated technologies, networks, and computational composition, are considered hybrid systems, including continuous systems and discrete systems. DT is an incorporated structure of mathematical models and data ensuring that real physical systems and their virtual entities are synchronized in real-time. Such a method can be characterized as whole workflows where the computing composition and other plant nodes are connected. A process modeling and simulation approach is applied to enact these workflows and simulate the hybrid system behaviors in this arrangement.

Knowledge Graph (KG) helps to represent the data that can achieve cognitive learning by machines. Knowledge is awareness or familiarity, someone or something gained by the experience of a fact or situation. On the other side, a Graph represents how any data are stored in the form of associations. KG is a term of how the engine builds relationships between people, technology, and facts. The KG models are focused on topological relationships between physical and cognitive entities. Ontologies for KG models are created before designing KG models to describe semantics and syntax. KGs will serve as the core mechanism for ML flows, extending data manipulation to enable practical consumption through CDTsBLM. KGs and ML techniques provide the required abstraction layer to clarify better (a) the context of each method and (b) the complex interactions that represent machine-understandable data and ML algorithms to make it easier for data and information extraction tools to communicate.

Artificial Intelligence (AI) APIs, historical data, process models with dynamics, and KG models are integrated to produce CDT models. CDT models aim to support decisionmaking for dynamic processes of physical entities. The use of dynamic process simulation has been developed as a reliable and effective tool to examine the transient behavior of process systems.

In the CDT and analytics for the process simulation stage, optimization tools will support process optimization through real-time data and CDT models. The result of this optimization is implemented to make decisions for physical entities manipulating.

A service-oriented interface for the data interoperability approach is offered to develop interfaces for heterogeneous data, and for that reason, all the assets and business domain data should be converted into integrated formats through the established interfaces. It means that all generated and captured data at any stage of projects need to be converted to a common data environment.

#### 3.1.2. Layers in CDT

The architecture of CDTsBLM has essentially four layers with each of them providing a set of services as follows.

Model Management Layer is in charge of three different kinds of models: (1) firstprinciple models for processes based on underlying physics; (2) analytical models based on various AI methods and ML; and (3) information-driven models focused mainly on their detailed work experience based on tacit knowledge of the domain and human operating experts. This *Layer's* primary role is to ensure that various services, including modeling, data-driven, and human experts, provide efficient storage and access for multiple models.

The Service Management Layer makes effective use of all available services to solve the fundamental domain issues. It is focused on a complicated organization of services, combining data-driven model-based services to create value-added pipelines. It contains a registry service, enabling the rapid discovery of the orchestral services required. Service results should be made public, and practical and scalable communication of service can be ensured.

The User Interaction Layer is a digital definition of a physical device simulating its actions. It's critical to assist a user in discovering a CDT's data and models, as well as its characteristics. To put it another way, intuitive yet exploratory user interaction should be possible.

A twin represents a dynamic framework that should be handled effectively, and the Cognitive Twin Management Layer models a physical system's behavior. A Twin can define the system's actions as a digital description of a physical system by offering a standard behavior model. Contrarily, a Twin is a digital entity whose life cycle is affected by a physical system; in other words, physical design behavior changes should be replicated in double structure models as soon as the physical environment's corresponding data become apparent.

#### 3.1.3. CDT realization within Cognitive Building Lifecycle Environment (CBLE)

Figure 4 depicts the CDTsBLM conceptual architecture built on service pipelines from which accessible data flow. The use of data streams implements the cognitive center of the CDT through one pipeline, which provides learning, event identification, and prediction and reasoning skills. A second pipeline for each CDT allows analysis and justification of vast quantities of raw data from different sources. A meta-structure improves the CDT by allowing multi-source data flow interoperability, higher reasoning, cognition pipelines since they interconnect through KGs.

**Figure 4.** Outline of the process flow of Cognitive Digital Twin for Building Lifecycle Management (CDTsBLM) (adapted from the developed framework by Lu et al. [48]).

KGs enrich and direct the relationship between these two data-driven modeling approaches. ML algorithms, data analytics, and KGs form the foundation of a robust cognitive computing framework that allows for fine-tuned outcomes and increased process and reasoning abilities. The semantic models augment a set of data with features that enable cognitive processes to be far more agile. Combining quantitative-driven ML, qualitative KGs, and data analytics combines machines' computing power with the human intuition and experiences needed to solve various Construction 4.0 use cases. Optimization can be exploited, related to the scope of the activity, the time horizon, providing CDTs with the capability to resolve optimized production issues, such as short-term real-time reorganization and reconfiguration of entire systems, mid-term timing, and lots for individual activities or whole construction and long-term capacity planning.

In the transition from CDTsBLM Conceptual Architecture to Technical Architecture, there is a data communication framework for collecting, processing, integrating, and managing multi-source, multi-scale, and multivariate data from production assets. CDT module interaction and API-based communication with the business domain can also be supported by a messaging and operation bus.

Actuators can be implemented to execute real-time decisions dynamically and communicate them directly from CDT to physical twin. Several enablers must characterize the CDT definition: (1) A profile describes the twinned asset and descriptive details, as well as models; (2) the connections between CDT and other CDT construction, as a network of linked CDTs, a factory or process or different architecture; (3) facets of cognition such as thinking, modeling, estimation, and optimization; (4) aspects of confidence to ensure correct knowledge transmission; (5) status and future notification visualization for end-users; (6) computation requirements as well as implementation aspects; (7) Identifying the CDT lifecycle.

These enablers tackle various stages of a cognitive factory model. The first is to simulate the construction or even other development contexts as a network of interconnected CDTs, for example, the workstation, process, and machine. Data from different sources, including ERP, Physical Twins, Human Operators, were initially added. Detection services (CEP), which are combined with Simulation and Optimization Services in the Cognitive Core functionality, use data-driven process models to allow CDTs to (1) detect a natural anomaly, such as an impending system malfunction, (2) forecast possible response steps with ideal outcomes, (3) simulate the optimized outputs and future consequences gradually, (4) return a well-thought-out proposal for the future course of action, which will be submitted to the appropriate stakeholders or actuaries for approval or denial.

#### 3.1.4. CDT and Cognition

The definition, which is data-oriented, resides at the heart of the CDT. Construction 4.0 needs a greater cognitive increase in assets to allow continuous improvement of the data-driven process. CDTsBLM uses a modern architecture for generation construction data analytics to integrate cognition into DTs and as a meta-platform to help create and implement a range of building applications, such as quality management and predictive maintenance. The CDTsBLM approach depends on a new DT data analytics, in which an innovative CDT-driven metaphor represents a system model, improving the DT base and the integrated CDT structures to understand and solve situations that cannot be modeled, for example, by design allowed in numerical models or experience in the context of numeric models. Based on the current process data in real-time, this integrated approach will explain the issue, including tool deterioration for each particular machine and product type. This cognitive function is assisted in rare cases by the just-in-time process status simulation to measure an anomaly's assurance that needs to be resolved in an extraordinary circumstance. This data-driven simulation would also indicate whether an anomaly is triggered by a particular scenario, meaning how long current process settings will remain unchanged. This novel approach describes data-driven model simulations from twins using novel predictive clustering methods and advanced inductive database mining rules.

#### 3.1.5. CDT and Data

For the analytical models implemented in CDTsBLM, extensive data sources are required. The well-known critical issue in developing analytical models is processing and modeling various types of data in real-time. A complex architecture must adhere to the other methods to construct a universal analytical structure in Construction 4.0 for real-time data stream processing. It is also possible to synchronize and optimize data until they are fed into the analysis models. CDTsBLM seeks to provide an exhaustive and modern, multi-level model of uncertainties and causal relationships that include the following submodels at multiple levels: model content fluxes, statistical capability models, technological process models, deviation models between optimal technical simulations, observations, and logistics demand projects. Material flow pattern models consider the lead time of input

materials and generate uncertainties for understanding product transport and logistics and the various technical procedures to be used. Machinery availability, breakdown model, and models for employee insecurity will be considered for statistical capability models. Models of technical processes can incorporate domain awareness. Due to its state-of-the-art data-guided online processing algorithms for broad re-in-time data streams, the CDTsBLM uses the framework as its analytical tool to handle data requirements for CDTs. It aims to address multimodal data fusion, data preparation, optimization, and the analytical design of a manufacturing process as an entity that generates a typical analytical structural model for intrinsic, interrelated process variables. The architecture allows CDTs, when introduced, to seamlessly use many multimodal data types.

#### 3.1.6. CDT and Optimization

When converting a DT to a CDT, the implemented optimization that allows the CDT to generate optimization functions is a critical enabler and differentiator. The vast majority of batch processing, construction planning optimization issues are NP-hard, which applies to Construction 4.0. Consequently, using conventional algorithmic and mathematical programming approaches to generate a proper solution to a real-life issue is computationally intractable. It is not always possible to have precise values of optimization criteria such as inventory supply, production times, costs, human resource efficiency, equipment durability, and construction industry specifications, or to be mindful of future diversities in material order preference, equipment failures due to a lack of information or the changing existence of actual construction sites. As a result, an optimum solution for approximate parameters could be inadequate until the parameters are realized. This complexity is present, especially in the process industries, where the quality of a given material inside a construction cannot be calculated with certainty before the component is processed. As a result, dealing with complexity is almost as critical as making the model itself, as it can be used to verify mathematical models and maintain production viability during operations.

The proposed solution dealing with decision circumstances involving complexity treats all potential realizations of parameters as part of the feedback. This collection is referred to as a scenario set, and each parameter completion is referred to as a scenario. As a result, a scheme reflects a possible condition of the universe. Since the cost of a solution is determined by a situation, its value is therefore unknown. Ex-post research compares a solution to an optimal solution that might have been obtained if the parameters had been realized in their original form. Decision-makers who do not want to take risks are more interested in avoiding the worst situations in the real world. A robust optimization methodology, under a discrete or intermittent uncertainty and the regret criteria of max– min or min–max, is a critical modeling approach for meeting the above requirements.

Another important aspect is the manufacturing process's performance. Its capability primarily determines the control system's capability to adjust schedules to changing conditions, especially at the short-term decision support level for real-time adaptive optimizing. These dynamic problems with re-optimization can be handled via reactive and proactive frames, in which the optimization process is progressively conducted at some intervals and dynamically evolves into integrated new or old knowledge. These methods can address complex optimization problems effectively with input parameters and variables that have not been completed, uncertain or unknown, that are modified simultaneously by the development of the real-time solution process. A suggested architecture for this path utilizes vigorous optimizations of different models (1) builds on well-established concepts such as negligence in which unique input parameters are not known to provide a universal solution which is efficient to optimize the worst-case solution and optimize overall actuality in all realizations of hidden parameters, (2) implements the adaptive, efficient, multi-stage optimization technique for planning where optimized decisions on unknown criteria and action on recourse depend on the realization of insecurity.

CDTsBLM is designed to deliver a complete CDT optimization toolkit based on a local hybrid search, evolutionary calculation, and data-driven techniques to scalable resourceaware planning and optimization algorithms that can be utilized to solve complicated planning issues with a variety of constraints, including utilities, renewable resources, and machinery service restraints. They also possess the potential to hierarchically address various schedule targets, including time and energy-conscious combinations. They often have a high degree of precision by strengthening their forecasting capabilities by utilizing (1) multiple design processes as decision variables to help monitor construction site factors such as efficiency and length while controlling several scheduling parameters such as processing time, energy usage, and operational expense, (2) a variety of execution types, including, for example, alternate routings and resource demand variations for each construction operation. A modified algorithm is designed to complement the prior algorithm set in the Optimization toolkit to endorse the CDT for rigorous online preparation problems, easily extended to resource-conscious purpose multifunctional optimization alternatives.

#### **4. Evaluation of the Proposed CDTsBLM Model**

Testing the proposed CDTsBLM model aims to recognize its effectiveness in practice and thus validate it. In this assessment, a digital survey was established with the literature review as a basis and distributed to industry professionals across countries. The survey's core theme was to provide practitioners with an insight into the life cycle-centric applicability and integrality of CDTs with existing BLM practices.

#### *4.1. Sampling*

AEC increasingly involves multiple stakeholders ranging from Design Manager to Design Coordinator, Designer, BIM Manager, BIM Coordinator, Digitalization specialist, Project Manager, Construction Manager, Asset Manager, Asset Administrator, and Asset Controller. The longevity of assets may mean that the stakeholders or even the type of usage may change over time; this poses challenges in how these assets are managed over their life and specific challenges to the way data and information about them are handled. Therefore, the study focused on private organizations dealing with building projects operating in the United States (USA), United Kingdom (UK), and Sweden. The sample includes only large firms.

#### *4.2. Data Collection*

The data collection was confined to actors that have vital roles in capturing, delivering, and using the information in the building life cycle and technology domain innovation projects. Design, project management, contracting, and facilities management firms were compiled by searching for geographic position cataloging enterprises. The survey included owners and consultants for asset management. The national inclusions improve the validity of questions as they represent the different cultures, experiences, and ways of working of corporate and national groups. The proportions of company positions, sizes, and regions are shown in Table 3.


**Table 3.** Company region, size, and role in percentage.

LinkedIn contacted 271 businesses, and a single representative from each was requested to participate in the questionnaire. Contributors were apprised about the search's aims, and their answers were kept private and anonymous. A total of 45 percent of completed queries were collected. Experts were asked to talk regarding their work, observations, and organizations. Participant experts used a five-point Likert scale to rate their agreement with BLM digitalization-central statements, with one being the most disagreeable and five being the most agreeable.

#### *4.3. Descriptive Statistics*

Descriptive statistics reporting the mean values and standard deviations of questionnaire responses are presented in Table 4. The summarized statistics speficies interesting comprehensions as an overview of the AEC industry's perception of the concepts. According to the results, the mean scores for 16 of the 20 questions were higher than 3.65 out of 5.00. The proposed model's overall mean rating was 4.06, which means that industry professionals support approaching CDT development for process optimization and decision-making purposes and that integrability enablers confirm progression towards Construction 4.0.

The argument that sought respondents' opinions on real-time analytics for data-driven models enhanced with cognitive resources was conducted to support decision-making and aid learning, optimization, and reasoning had the highest mean of 4.45 in the relative importance of the variables. Through reason, learning, and optimization, CDTs can monitor, project, modify, and make better choices in real-time. CDT is a robust monitoring and tracking method, and the overall system is optimized with a mean value of 4.38. CDT covers existing process control systems with cognitive elements that allow them to organize themselves and provide a so-called indication of unanticipated actions at an average of 4.34. Overall, respondents agree that CDT should provide cognitive features that enable it to sense complex and unpredictable movements and reason about dynamic process optimization strategies to aid decision-making in BLM.


#### **Table 4.** Descriptive statistics, factor analysis and reliability test.

#### *4.4. Factor Analysis*

Functionalities of the BLM, CDT, IoT, and Process optimization and achievable levels of interoperability ad integrability of the proposed model as rated by the industry professionals are presented in Table 4. Confirmatory factor analysis boosts trust in the assessment's precision and quality. Table 4 lists the items that were used to calculate each element. A five-point Likert scale was applied to measure all objects, and they were

found perceptual. All factor loadings between 0.703 and 0.892, as well as all Cronbach's coefficients less than 0.70, were considered to be adequate.

According to Table 4, CDT was ranked as the highest, Process Optimization as the second, IoT as the third, Self-learning as the fourth, and BLM as the fifth factor for life cycle-centric applicability and integrability of CDTs with current BLM practice, exploring decision support capabilities and AEC industry insights.

#### *4.5. Correlation Analysis*

Spearman's rank-order correlation was used to validate the relationships, and the evaluation of the matrix shows a correlation. A positive linear relationship exists within BLM for improved productivity and sustainability, CDT for enhanced decision-making, IoT for real-time connectivity, and self-learning by applying new knowledge on the existing data, models, methods, and optimization simulation decision support. The highest correlation occurs between CDT for improved decision making and IoT for real-time connectivity in ρ < 0.01 (r = 0.812). The second significant positive correlation exists between CDT for enhanced decision making and optimization and simulation for decision support in ρ < 0.01 (r = 0.799). The correlation calculations of respondents' perception of CDT decision support abilities are depicted in Table 5.

**Table 5.** Correlational analysis of Cognitive Digital Twins' perception of decision support capabilities.


Notes: *N* = 85. Correlations have a (2-tailed) level of significance "Sig. < 0.000". Correlation is significant at the 0.01 level.

#### **5. Discussion**

The motivation for this research came from the novelty of the DT concept and its future applications, which will establish the adaption of CDTs in the AEC industry. Besides, the lack of attention paid in the literature to CDT in AEC project management led the authors to investigate this research. The adapted model in this study provides a viable solution to the identified problem. Process modeling has been used to explain the steps and significant aspects of the CDTsBLM framework. This study presents a novel adapted model that integrates CDT and BLM concepts and allows all project stakeholders to identify and collect the right data sets and implement them properly to optimize the system. The

proposed model attempts to improve the BLM performance compared to the traditional and current methods.

In BLM, the CDT can be used to represent any physical unit. Buildings, process phases, total procedures, and ultimately an entire construction operation can be virtualized using CDT. CDTs can be elicited at various hierarchical levels, with CDTs combining horizontally and vertically to form an aggregated structure. The Cognitive Building Lifecycle Environment (CBLE), built by combining CDTs, shares significant knowledge horizontally. Only important decision-making material, on the other hand, is transmitted vertically to upper levels. A mission-critical building's CDT (monitoring and managing its condition and actions) supplying input to a particular process phase that feeds the building design process's CDT is an example. These CDTs will act and respond when sharing data with the various exchanged data sets and their semantics. Hence, the respective CDTs must be coordinated by a supervisory check, resulting in the CBLE, with market requirements, time horizons, and the essence of various activities that must be processed at any given time determined. This study examined the implementation, integrability, and interoperability of CDT in existing BLM practices in the life cycle, exploring decisionmaking skills and AEC industry insights. It is anticipated that the CDTsBLM model will promote the qualifications mentioned to allow better knowledge, analysis, optimization, and decision-making, which will concentrate on evaluation. The CDTsBLM model will enable re-evaluation, projection, and re-evaluation in a dynamic and complex world, with the possible planning, design, structuring, and operating. The operational processes' environmental effect must be reduced in the AEC industry by optimizing the building lifecycle processes by the CDTsBLM model.

One path forward to achieving new operational efficiencies depends on the reality that much of the human-dependent work activity can be significantly reduced by automating repetitive activities such as data acquisition, base data analysis, and the need for physical presence at physical locations, yielding a faster and safer approach to gathering data as well as reducing the time it takes to correlate and analyze that information. This rich collection of information is accessed, maintained, and controlled by humans for three primary activity streams. Analytics involves various analytical models, technologies, and approaches providing historical, current, and predictive insights from the data gathered. The workflow requires information about and procedures to inspect, maintain, modify and repair the physical asset. Visualization involves information, including the spatial geometry used for primarily planning and engineering.

The cognition, interpretation, and optimization of decision-making skills are fundamental to CDTsBLM. The connections between the ecosystem capacity perceived as a collective framework and all of the capabilities contribute to establishing a crucial cyclecentric application with inclusive aspects that contribute to explain the value of technology integrity from a professional inducible usability perspective. The CDTsBLM uses its models to evaluate data from the current framework to provide feedback and support decisionmaking. Depending on the study, the data and intelligence displayed are performed by the CDTsBLM.

The quantitative analysis of the data collected from Design Managers, Design Coordinators, Designers, BIM Managers, BIM Coordinators, Digitalization specialists, Project Managers, Construction Managers, Asset Managers, Asset Administrators, and Asset Controllers indicated that there is a willingness to use this type of CDT technology and related models. This analysis justified that the CDTsBLM model framework helps to provide a real-time analysis for data-driven models augmented by cognitive resources, which was conducted to facilitate decision-making and improve understanding, optimization, and thinking. Further, it shows that CDTsBLM is a valuable monitoring and control mechanism that helps in overall device optimization. It helps managerial levels of projects self-organize and respond to unpredictable activities and aid in decision-making for complex systems, including physical actors. The AEC industry can revolutionize how to design, build truly, and operate in a complex project environment. The AEC industry will inevitably adapt

cognition, analytics, self-learning, and optimization techniques due to the emergence of DT, Cognitive computing, AI, ML, and cloud-based systems. Table 6 shows such a system's process, opportunities, and challenges. This research indicates that the CDTsBLM is an intelligent system that seamlessly connects engineering operational data, information, and models utilized over the whole building asset lifecycle with self-learning and predictive capabilities. It then makes the results readily available in real-time and the proper context for all related stakeholders to prevent or solve potential issues proactively. The findings from this research could serve as a base to pave the way for promoting progression towards Construction 4.0.

**Table 6.** Sample of Cognitive Digital Twins cognition, analytics, and optimization processes in Building Lifecycle Management.


#### **6. Conclusions**

Construction projects and their data from the first stage to the last day of AEC projects are becoming huge and more complex to gather and manage. It is becoming exceedingly difficult, if not impossible, to identify and collect the right data sets and put them in the proper context to enable the optimization of the system. However, with help from a DT that can sense, reason, and act, such intelligent systems will help projects' stakeholders make the right decisions or autonomously trigger the right actions in the digital or physical world. Increasing complexity in terms of the DT becomes apparent when looking at the various application streams and their need for precise and real-time data. The DTs' highest level is the CDT connected with the top-level cognitive engineering maturity, including AI and ML. This article introduced the newly adapted CDT paradigm, including BLM's capabilities in cognition, analytics, and optimization for construction 4.0. The most significant advantage of cognition is the ability to solve problems preventively unknown. The CDT provides a toolkit for optimizing based on its cognitive components that enables CDT to carry out optimization tasks and delivers valuable results that other CDTs or process actors consume. Industry practitioners have examined the technical framework of CDTsBLM

and taken full advantage of additional CDT capabilities, including construction schedules, preventive maintenance, and other goals, in the traditional DT sense. The benefits of implementing the CDT concept in the construction industry are intended by opening the optimization tool kit inside the CDT and enhancing real-time or almost real-time choices by interplaying optimization and simulation. Finally, the CDT description and conceptualization formalities will be further evolved alongside this implementation and assessment by tailoring every application scenario specific to the technology, ability, and KPIs that show this CDT effect.

The findings demonstrate the applicability of the CDTsBLM integration for a variety of AEC analysis scenarios. Future research directions could focus on investigating the processes and sub-processes of CDTsBLM applications in various AEC projects. Utilizing this system's legal and financial aspects will also lead to future research opportunities of CDTsBLM. Researchers might want to explore the processes and integrability of various construction technologies with CDT for various purposes.

**Author Contributions:** Conceptualization, I.Y., S.A.; methodology, I.Y., S.A., ˙I.A., M.E.A.; software, I.Y., S.A.; validation, I.Y., S.A., ˙I.A., M.E.A.; formal analysis, I.Y., S.A., ˙I.A., M.E.A.; investigation, I.Y., S.A., ˙I.A., M.E.A.; resources, I.Y., S.A., ˙I.A., M.E.A.; data curation, I.Y., S.A., ˙I.A., M.E.A.; writing—original draft preparation, I.Y., S.A., ˙I.A., M.E.A.; writing—review and editing, I.Y., S.A., ˙I.A., M.E.A.; visualization, I.Y., S.A., ˙I.A., M.E.A.; supervision, I.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** not applicable.

**Informed Consent Statement:** not applicable.

**Data Availability Statement:** not applicable.

**Acknowledgments:** The authors would like to thank all the survey respondents in the AEC industry.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Towards an Occupancy-Oriented Digital Twin for Facility Management: Test Campaign and Sensors Assessment**

**Elena Seghezzi 1, \*, Mirko Locatelli 1 , Laura Pellegrini 1 , Giulia Pattini 1 , Giuseppe Martino Di Giuda 2 , Lavinia Chiara Tagliabue <sup>3</sup> and Guido Boella 3**


**Abstract:** This study focuses on calibration and test campaigns of an IoT camera-based sensor system to monitor occupancy, as part of an ongoing research project aiming at defining a Building Management System (BMS) for facility management based on an occupancy-oriented Digital Twin (DT). The research project aims to facilitate the optimization of building operational stage through advanced monitoring techniques and data analytics. The quality of collected data, which are the input for analyses and simulations on the DT virtual entity, is critical to ensure the quality of the results. Therefore, calibration and test campaigns are essential to ensure data quality and efficiency of the IoT sensor system. The paper describes the general methodology for the BMS definition, and method and results of first stages of the research. The preliminary analyses included Indicative Post-Occupancy Evaluations (POEs) supported by Building Information Modelling (BIM) to optimize sensor system planning. Test campaign are then performed to evaluate collected data quality and system efficiency. The method was applied on a Department of Politecnico di Milano. The period of the year in which tests are performed was critical for lighting conditions. In addition, spaces' geometric features and user behavior caused major issues and faults in the system.Incorrect boundary definition: areas that are not covered by boundaries; thus, they are not monitored

**Keywords:** Building Management System; Digital Twin; Post-Occupancy Evaluations; facility management; asset management

#### **1. Introduction**

The operation and maintenance (O&M) phase of buildings and civil infrastructures ranges between 20–30 years for buildings, but it can cover more than 50 years of the whole lifecycle [1]. It is essential to ensure an actual and efficient management of buildings during the O&M phase. Occupancy and actual use of spaces strongly affect the organizational effectiveness and functioning during the operational phase [2,3]. Typically, standardized and fixed values of occupancy are considered during design phases, e.g., maximum occupancy values from fire regulations or scheduled occupancy for energy models [4]. Consequently, actual occupancy and space use levels may significantly vary from and rarely correspond to the values considered during the design phase. Occupancy strongly influences use and cleanness of spaces, which in turn are related to well-being, satisfaction, and productivity of users [5,6]. In recent years, a consistent number of studies investigated the segment of the performance gap between expected energy consumptions, defined during the design phase, and actual consumptions, due to human-building interaction and variable occupancy [6–17]. However, other promising fields in building management include security, safety, cleanness, and space management. These aspects can have a crucial role,

**Citation:** Seghezzi, E.; Locatelli, M.; Pellegrini, L.; Pattini, G.; Di Giuda, G.M.; Tagliabue, L.C.; Boella, G. Towards an Occupancy-Oriented Digital Twin for Facility Management: Test Campaign and Sensors Assessment. *Appl. Sci.* **2021**, *11*, 3108. https://doi.org/10.3390/app11073108

Academic Editor: Jorge de Brito

Received: 28 February 2021 Accepted: 23 March 2021 Published: 31 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

especially in light of current sanitary emergencies related to the spread of the COVID-19 pandemic: space monitoring is a key aspect to guarantee safety in existing buildings [18].

In this context, the aim of the ongoing research project here presented is to define a Building Management Systems (BMS) based on an occupancy-oriented Digital Twin (DT), evolving from and enriching the Building Information Model (BIM) and integrating occupancy levels and additional relevant data from Post-Occupancy Evaluations (POEs). Analyses and simulations of the occupancy-oriented DT would support the decisionmaking processes during the O&M phase.

The case study for the application of the methodology is an existing office building hosting the Department of Architecture, built environment, and construction engineering (DABC) at Politecnico di Milano, Italy, used by people working at the university and performing their research and administrative activities in the indoor spaces of the building. The maintenance and cleanness of the distribution spaces and offices is a very important aspect in the facility management of the building and department business plan; strong variations in occupants' flows are experienced by the users and particularly during the pandemic.

The IoT network of sensors that represents the source of data for the occupancyoriented DT and that was tested and calibrated as described in this article was provided and installed by an external consulting company (Laser Navigation srl). They provided the hardware part of the system that is the camera-based sensors with an embedded deep learning algorithm, the installation, and the technical settings of the sensors. They also provided an online platform named SophyAI and integrated with the IoT system, that allows to visualize, store, and download collected data.

This paper focuses on the preparatory phases for the definition of the DT, i.e., sensor system calibration and collected data quality validation. In fact, a fundamental characteristic of a DT is the connection, alignment, and reciprocity between the physical and virtual part [19]. Therefore, a key aspect is the data collection process, ensuring data quality on which the correct digital representation of the physical phenomenon depends [20,21], since, in order to obtain satisfactory results, is essential to ensure the quality of input data [22]. In this perspective, fundamental steps are the selection of sensor types that are most suitable for the specific application [23], the spatial distribution of sensors in the indoor spaces [24], and the setting and calibration of the IoT sensor system [25,26], to allow a correct detection and collection of data.

Given the importance of data quality for the proper digital representation of the building occupancy phenomenon, the objectives of the research are: optimization of spatial distribution and orientation of sensors for system planning and installation, identification of issues and faults of the detection system, and resolution of issues and faults by performing an assessment of the detection system through test campaigns. This study proposes method and evaluation criteria for system calibration and data quality validation, also defining parameters for occupancy analysis. Two test campaigns were performed until all major faults have been checked and solved, allowing for the verification and validation of collected data quality to monitor building occupancy. The study also describes and tests the use of the platform SophyAI for real-time visualization of data during the test campaigns.

#### **2. Literature Review**

#### *2.1. Evolution, Main Applications, and Features of Post-Occupancy Evaluations*

Post-Occupancy Evaluations (POEs) aim at assessing building performances, users' behavior, and feedback regarding existing buildings during the operational phase and once the building has been occupied for some time [27–31]. POEs were first introduced in the UK and US in the 1960s in order to assess building performances from user perspective, by means of interviews, questionnaires, photographic surveys, and walk-through surveys [27,28]. The major developments of POEs were during the 1980s, aiming at analyzing and optimizing the facility management and design [29]. POEs had been performed in the US, mainly in the public sector, UK, New Zealand, and Canada [32], and, since a

correlation between workplace features and worker productivity was proposed in 1985, they have been also applied in the private sector to improve costumer and worker satisfaction and to optimize the workplaces [30]. In the mid-90s, the interest moved from analyses during the operational phase alone to an entire building life cycle process, i.e., Building Performance Evaluation (BPE) [33,34]. Insights and findings from POEs could be applied in the subsequent design and building life cycle process [34–36].

In the last two decades, POEs have mainly been applied to assess and optimize building energy performances, and to reduce the building environmental impacts [37]. A less investigated but promising research field is the optimization of occupancy patterns, and cleaning activities and contracts. Space features and workplace cleanness have been classified as basic factors affecting user satisfaction [5], and, consequently, user productivity. The variable "interior use of space" can account for around 43% of the variance in employees' enjoyment at work, well-being, and perceived productivity [6].

As above mentioned, POEs are analyses of the built environment, aiming at defining the effectiveness and functionality of spaces for users, building performances, and user satisfaction and perception regarding facilities in general and workplaces in particular [27,38]. There are three levels of POEs depending on accuracy, time needed to be performed, tools, and levels of invasiveness of user privacy [7,30]:


Despite being expensive and invasive for user privacy, especially when performing diagnostic analyses, POEs can have several benefits, ranging from the optimization of the operational phase in terms of performances and user satisfaction, to an increased facility adaptation to organizational change and growth over time [27,30], and to the definition of design criteria and requirements based on actual user and space needs for similar buildings [39].

#### *2.2. From Building Information Models to Digital Twins for Asset Management*

In recent years, a major evolution of Building Information Modelling (BIM) occurred in the construction sector. BIM models are parametric models, centralized sources of information mostly for the design and construction phases, and instruments to improve collaboration among specialists and document management [40]. The application of BIM for facility management can result in several benefits: customer services improvement, time and cost reduction resulting from better planning capabilities, and higher consistency of data [41,42]. The integration of POEs in a BIM approach enables the connection between POE data and the digital model [8,12,43], with the advantages of defining a single source and storage of POE and building data, integrating structured data into the BIM, and identifying POE data and related issues in a visual representation of the building space [8,44,45]. Despite the advantages of adopting BIM during the operational phase, a BIM approach for asset management lacks of information richness, analysis, and simulation capability, which are usually manually implemented and time-consuming when using a BIM model [20]. In addition, an effective and efficient management of buildings during the operational phase strongly rely on continuous flows of real time data regarding the building, its performances, and conditions [20,46]. However, BIM models present limitations for the integration with different data sources and systems, e.g., sensor data, and lack of automatic updating and evolution over time [20]. Therefore, in order to overcome these limitations, the definition of a Digital Twin is investigated.

#### *2.3. Evolution of the Digital Twin Concept*

The Digital Twin (DT) concept dates back in 2002 when the idea of a virtual space containing the information of and linked to the real space emerged in the field of study of complex systems, in particular regarding the Product Lifecycle Management (PLM) [19]. When the concept emerged, it was not referred to as DT, but it was presented as the "Conceptual Ideal for PLM" [19], evolved then to Mirrored Spaces Model in 2005 [47] and to Information Mirroring Model in 2006 [48] and 2011, when also the term Digital Twin was first used to describe the model [49]. In recent years, the concept of a DT has been studied also in the aerospace sector: the DT represents an ultra-realistic digital replica of real flying vehicles, considering one or more interconnected systems allowing for probabilistic simulations that take into account physical characteristics and models, sensor data, and history of previous flights and vehicles [50–52]. Recent definitions of DTs can be found in various sectors, with a wide use and diffusion of the concept of a virtual replica of physical entities whose purpose is to manage, optimize, and control the physical asset itself. In the infrastructure sector, DT was defined as a realistic virtual representation of the corresponding infrastructure, adding the built or natural context in which the object is contained and to which it is connected [53]. In the manufacturing sector, the idea of the connection between physical components and virtual models is widened, adding the necessary mono- or bi-directional flow of data between the physical asset and its virtual counterpart in order to real-time monitoring the actual object, supporting simulations, analytics, and control capabilities of the dynamic virtual model [54]. The construction industry can be still considered in its beginning regarding the definition of a DT for buildings. Despite the various attempts to define a DT in construction industry [21,24,53,55,56], a comprehensive definition was proposed by Al-Sehrawy and Kumar [57]: "an approach for connecting a physical system to its virtual representation via bidirectional communication (with or without human in the loop) using temporally updated Big Data [ . . . ] to allow for exploitation of Artificial Intelligence and Big Data Analytics by harnessing this data to unlock value through optimization and prediction of future state". This definition includes all the fundamental parts of a DT, which are described in detail in the following paragraph.

#### *2.4. Elements and Characteristics of a Digital Twin*

As stated, a DT is composed by some elements. A list of components for DTs in construction industry is provided as follows:


In addition, some characteristics are fundamental for the correct definition of a DT:

• Synchronization between physical and virtual component [40], with data flowing at least in one direction allowing for analyses, control, and simulation on the virtual model [20,46,64]. Any change in the monitored characteristics or conditions of the asset is detected and, through data flow, is reflected in the virtual counterpart [20,21];


As previously specified, one of the main characteristics of a DT is the direct connection between physical and virtual entity, with the concept of twinning as alignment and reciprocity between the two components [19]. Therefore, a fundamental aspect for the definition of a DT is the data collection process, i.e., data quantity, quality, and granularity, on which depends the correct detection of changes of the real object over time, and thus the correspondence of the digital object to the real one and its continuous evolution through the building lifecycle [20–22]. A first fundamental step is the selection of sensor types that are most suitable for each specific application [23]. In addition, the spatial distribution of sensor network, i.e., the spatial distribution of sensors in the indoor spaces, is another theme that should be faced [24]. Furthermore, in order to allow a correct detection and collection of data, the IoT sensor system should be properly set and calibrated [25,26] to ensure the quality of data collected. In fact, the output of an analysis strongly depends on the data that are used as input for the system or algorithm; therefore, to obtain satisfactory results, data quality is essential [22]. Nonetheless, existing studies tend to focus on different phases and aspects of the DT definition and creation, while IoT sensor system definition is a less investigated aspect, almost taken for granted [23].

As stated above, a fundamental preliminary step is a detailed analysis of sensor types in order to identify the most suitable ones for the research objectives. Such analysis is described in the following paragraph, in which existing studies regarding types of sensors for occupancy detection are analyzed, highlighting features, pros, and cons. In addition, a brief review of the concept of occupancy detection is provided. The investigation supported the selection of the sensor type for the case study, as explained in the following methodology section.

#### *2.5. Occupancy Detection: Analysis of Occupancy Monitoring Systems*

Occupancy detection consists in the definition of occupancy levels and patterns of buildings during the operational phase. Occupancy patterns consist of occupancy values at room-level and user movements inside the building [65]. Monitoring occupancy patterns and optimizing the use of spaces and cleaning activities, based on occupancy data, can increase user satisfaction and productivity at work. In fact, occupancy levels of buildings have a strong influence on cleanness and use of spaces that, in turn, are strongly related to well-being, satisfaction, and productivity of users [5,6]. Table 1 focuses on IoT monitoring sensor systems studies, highlighting main features, pros, and cons.

As shown in Table 1, camera-based sensors and PIR (Passive Infra-Red) sensors present the best accuracy levels, followed by CO<sup>2</sup> sensors, but they are also affected by detecting and privacy issues [9], such as the Hawthorne effect for camera-based sensors. It mainly causes alterations of behavior when users are aware of being observed and, if ignored, can affect the reliability of collected data [26]. One strategy implies the combination of more types of sensors, some of which may already exist in the building, having been previously installed for other purposes [9]. Additionally, system implementation costs can be reduced by previously analyzing the building with Indicative POE analyses in order to identify the most critical areas to be further analyzed [27,32] by means of sensor systems and other techniques.


**Table 1.** Sensor systems and related features to monitor occupancy.

The highlighted advantages and disadvantages of existing sensor types supported the selection of the type of sensors for the methodology and case study, as described in the following section.

#### **3. Research Project Stages**

This paper presents some stages of an ongoing research project. The aim of the research project is to define a Building Management Systems (BMS) based on an occupancyoriented Digital Twin (DT), evolving from and enriching the Building Information Model (BIM) and integrating occupancy levels and additional relevant data from Post-Occupancy Evaluations (POEs). The expected results of the research project are: monitoring occupancy and defining building occupancy patterns, optimizing current O&M management, building space use and organization, cleaning activities, and, as possible future implementations, applying Smart Contract to cleaning and maintenance services and extending the IoT network with other kind of sensors for safety and quality control. The research project stages are presented in Figure 1.

The first two stages, "definition of BIM guidelines" and "BIM model creation", have been previously analyzed in a publication by Di Giuda et al. [75]. "Preliminary analyses", "system installation", and "test campaigns" are presented in this paper, as they are critical to provide the foundations upon which the occupancy-oriented DT should be based.

The "occupancy-oriented DT" set of activities is currently under development. In future steps of the research, collected data will be analyzed to identify occupancy patterns of the building spaces, and evaluate current management of spaces in terms of people permanence and cleaning frequency. In addition, benchmarks to evaluate optimization strategies will be defined together with the subjects in charge of O&M in the case study building, a fundamental step to evaluate advantages and results of the methodology [76]. The defined occupancy-oriented DT will be the base for the subsequent phase, i.e., "FM scenario definition and optimization", that will allow the optimization of cleaning activities and contracts that are currently based on the building floor areas, and to reach a better organization and planning of space usage.

**Figure 1.** Stages of the research project.

Figure 1 also provides "possible future implementations" of the research project. The integration of other kinds of sensors, such as "sensors for safety (man down) and quality control" will allow the monitoring and optimization of different aspects of the building management, resulting in a complete report of building conditions and indoor environmental quality. In addition, a possible future implementation of the system will be the definition of "smart contracts for facility management and cleaning activities" that would be based on the actual need of cleaning defined in previous stages. Smart Contract based on Blockchain technologies and on the occupancy-oriented DT data will provide relevant advantages, i.e., increased network security, reliable data storage, traceability [77], and the possible automation of payments for cleaning activities [78,79].

#### **4. Method**

This section provides the methodology applied for the "preliminary analyses" and "test campaigns" stages, analyzed in detail in this article. The "system installation" task was performed by an external consulting company that provided and installed the IoT sensor network, and the platform SophyAI for visualization, storage, and download of collected data.

#### *4.1. IoT Network of Camera-Based Sensors*

The "sensors analysis and selection" phase relies on the proposed literature review. As previously stated, most analyzed recent applications aimed at optimizing energy performances and consumptions rather than building operation and use [9,12,16]. Nonetheless, existing studies allowed objectively comparing several available sensor types, supporting the selection of the most suitable type for occupancy monitoring.

Camera-based sensors were selected considering their high accuracy and the possibility to perform other kind of analyses, such as security and safety monitoring, thus allowing for further implementations of new features in the system, increasing the scalability of the system itself.

The limitations of camera-based sensors that have been presented in the literature review section, and how they have been overcome, are described as follows:


The sensors can detect occupancy; in particular, they can visualize real time movements of users that are instantaneously transformed into anonymous virtual agents.

The detection of anonymous real-time movements of users is limited to common spaces, i.e., circulation areas and corridors, and they can be visualized in the online platform SophyAI, but are not stored in the database (DB), to protect the users' privacy. On the contrary, sensors count and store the number of agents that are entering or leaving rooms, which are the main objects of monitoring.

Two values are recorded by the sensors for each monitored room:


#### *4.2. Visualization and Analysis Platform*

A critical theme for real-time monitoring is the possibility of plotting sensors data for visualization, verification, and analyses [23]. Data visualization is a primary subject to support decision-making processes and to help people who are in charge of O&M in reaching management goals, since they may not possess the technical ability to effectively use the indexes and information directly extracted by the sensor system [58].

As shown in Figure 2, the described monitoring system is intertwined with the online platform SophyAI. The platform can:


Data stored in the online platform DB can then be downloaded as CSV files that contain the values of O and T for all days of a specified period of time, which could be a week, a month, or a year. In addition, data can be processed in graphs and diagrams and visualized through the online platform.

The online platform displays a 2D visualization of the spaces. Each monitored area is contained in a 2D boundary, which defines the contours of the area itself. The check between the area displayed in the 2D visualization and the 3D view of the same area detected by cameras is a key aspect to correct the optical distortion between the 3D view of the camera and the 2D view of the online platform. The check between 2D boundary and 3D view was performed as a part of the study during the system test and calibration, as described in following paragraphs.

**Figure 2.** Data and information flow.

#### *4.3. Preliminary Analyses Based on Post-Occupancy Evaluations and Building Information Modelling*

This phase applies Indicative POEs to preliminary analyze the building by means of general and low-invasive analyses, using the BIM model:


The use of the BIM model as a source of information and simulation tool to perform the Indicative POE ensures the minimization of user privacy invasiveness. In addition, the sensor system plan is optimized by comparing different configurations.

#### *4.4. Test Campaign Methodology for Data Quality Evaluation*

The preliminary analyses allow for an efficient planning and installation of sensors, selecting critical areas to be monitored, and optimizing spatial distribution, orientations, and fields-of-view of sensors. Nonetheless, after the first phases of data collection, some errors and faults, described in the following paragraph, may occur, and the system needs to be calibrated to ensure data quality. An incorrect calibration would lead to incorrect data collection and to an erroneous modelling of the occupancy patterns of the spaces, with repercussions on the whole BMS.

The iterative process to perform the "test campaigns" (Figure 1) is presented in detail in Figure 3 and described in the following paragraphs.

**Figure 3.** Data collection, test campaign, and adjustments application iterative process.

Once the system is installed as planned with the support of preliminary analyses, data are collected for a representative period of time that should be identified for each case study. Then collected data are downloaded in CSV format from the online platform DB. Collected data are analyzed in order to identify the possible data errors and related system faults, as described in Table 2. If no faults that could compromise the following analyses are detected, the system is properly functioning and calibrated. Otherwise, test campaigns are performed to verify the errors detected in the collected data.

The real time test campaign involves two operators. One operator (operator A) monitors through the online platform the position and movements of the other operator (operator B) inside the building. The two operators are constantly connected via earphones to communicate and coordinate with each other. In particular, operator A guides operator B towards the areas where errors were previously detected in the collected data. Moving inside the building and entering/exiting the rooms, operator B tests the detection of user movements and the room occupancy count (O) by the system. At the same time, operator A monitors the response of the system by checking the real-time displayed user movements and instantaneous values of O of the rooms through the online platform. Consequently, the operators search for detection errors and system faults in order to identify the causes, as described in Table 2. System faults can be classified as missing data, outliers, stuck values, and noise. Each fault can be identified in collected data or during test campaigns according to specific values of O. In addition, noise can be detected only during real-time test campaigns, by comparing the movements of operator B and his anonymous digital counterpart displayed on the online platform. Some examples of the causes of the errors and faults are camera malfunctioning in the case of missing data and extreme lighting contrast in the monitored area, which impedes a correct detection and causes noisy data.


**Table 2.** System faults [22] and related data errors. Data errors are divided into errors observed in collected data and errors detected during real-time test campaigns.

> Once the causes of data errors and faults are identified, some adjustments are proposed and applied to the system. Then the system must be verified again, in an iterative process, until no errors are detected and consequently the required data quality level is reached. This iterative process is also useful to check overtime the effectiveness of improvement solutions or to check the system after geometry changes in the building, e.g., in the case of refurbishments.

> Regarding the possible adjustments to solve the errors and the related causes, some general rules were identified to define a hierarchy of possible solutions.

> Generally, the most preferable solution would be not acting on the hardware of the system: in the case of a recently added physical obstacle that prevents the camera-based sensor from detecting, the most preferable solution would be moving the object before moving the sensor. In addition, before acting on the hardware part of the system (e.g., adding or replacing cameras), the camera settings could be checked, and the software system would be improved. An example of camera setting adjustment is the modification of contrast and luminance settings of the camera in the case of extreme lighting contrast in the monitored area. In addition, modifying the software is faster, less invasive, and cheaper than working on the hardware. Specifically, the deep learning algorithms of the embedded artificial intelligence system of the cameras for image recognition could be improved and optimized with the support of the external consulting company Laser Navigation srl. Consequently, the adjustments are hierarchized based on those general rules using the following symbols: from the most preferable solution, identified with (++), to the least preferable one, identified with (–).

#### **5. Case Study**

The building chosen as case study hosts the Department of Architecture, Built Environment and Construction Engineering (DABC) of Politecnico di Milano, and is located in Milan (Italy). It is a four-story building, hosting administrative offices, research spaces, and university staff offices, for a total of 4300 square meters of gross floor area. Rooms have variable dimensions depending on their use. The building has a symmetrical layout, with a common space in the center and two side corridors. The offices and workspaces are located on either sides of the corridors. Each floor houses at least one bathroom. Before the current study, the building has never been monitored. Therefore, neither data regarding the actual occupancy patterns, nor information about actual cleaning and maintenance activities are currently analyzed and optimized. Furthermore, no space optimization has been performed in relation to the use of available rooms and the actual occupancy indexes at room-level in the building. This case study building acts as prototype for a future application of the proposed method to other university's buildings.

The case study section is divided in two subsections: the first one describes the application and results of preliminary analyses on the building that supported the planning and installation of the IoT sensor system; the second subsection describes the two test campaigns with specific focus on the detected system faults and related proposed adjustments.

#### *5.1. Preliminary Analyses: Sensors Spatial Distribution and Orientation*

A preliminary study of the building ("Indicative POE supported by BIM model" phase as in Figure 1) was performed to identify critical areas to be monitored and to optimize number, position, and orientation of sensors, which in turn allowed the reduction of implementation costs and proper planning of the IoT sensor system.

As described in the methodology section, the preliminary analyses included the following activities that are analyzed in detail in the following paragraphs:


#### 5.1.1. Analysis of the Functions of Spaces

The analysis of the building through the BIM model allowed the identification of number and type of rooms of the building, as shown in Table 3. The BIM model had been previously defined and modeled, as described in Di Giuda et al. [75], who also performed a complete survey to update the as-built documents and to ensure the correspondence between the BIM model and the building. Equipment rooms, storage closets, and archives were excluded, since the only users are cleaning services employees or technicians in charge of maintenance activities. The analyses highlighted that sensors installed in common spaces, i.e., corridors, would be sufficient to monitor room occupancy, i.e., the count of users entering and leaving rooms. Anonymous real-time agent movements are detected only in corridors and can be visualized in the online platform, but are not stored in the DB, to ensure and protect the privacy of users. On the contrary, as regards rooms, the count of number of users (O) and time of occupancy (T) is recorded, as shown in Table 3. The occupancy of critical rooms is monitored to optimize their use, cleanness, and maintenance, while corridors are considered only as circulation areas. As shown in Table 3, 70 rooms out of 87 were selected as critical areas to be monitored.


**Table 3.** Type and quantity of rooms and necessity to be monitored.

5.1.2. Analysis of the Geometry of Spaces and Simulation of Sensors Location and Orientation

During the preliminary phases regarding the system planning, the BIM model of the building was used to optimize locations and orientations of the camera-based sensors. The geometry analysis highlighted that the building corridors are long, low ceiling, and narrow (length: 32 m; height: 2.40–2.70 m; width: 1.60 m). Three simulations of the interrelated position of the sensors in a corridor have been performed to define the best configuration. Virtual objects representing the sensors were added to the BIM model in different locations according to the three possible configurations. Each virtual sensor was then linked to a field-of-view that allowed to virtually check through the model the area covered by each sensor. The BIM-based simulation analyzed three possible configurations, as shown in Figure 4:


The chosen solution was the second one, since in many cases there are doors near the end of corridors, thus excluding the first solution. In addition, due to the reduced width of the corridors, one single camera could struggle in identifying two people walking lined up. Therefore, the third solution, which involved only one camera, was also less preferable than the second one.

The chosen sensors are High Quality Bullet Pro Camera PoE, with the following features: they provide HD quality images; the Power over Ethernet (PoE) allows to supply power and network connection to the camera with a single cable; a Wide Dynamic Range (WDR) allows to compensate problems due to exposure to light; the view angle of the camera reaches a maximum of 110 degrees. The system is installed in a dedicated Virtual Local Area Network (VLAN), and a static IP is provided for each element of the system. As stated in the Introduction, the sensor system was provided by a third-party organization, Laser Navigation Srl, who operated in full compliance with EU General Data Protection Regulation (GDPR). In fact, the deep learning algorithm does not record images, but only metadata regarding the anonymous movements and count of users are processed by the system, inhibiting the recognition of the observed subjects.

The 20 sensors were installed directly in the ceiling, i.e., at height 2.40/2.70 m depending on the level of the building, ensuring the maximum coverage area. Figure 5 shows the plan of the IoT sensors system in the case study building.

**Figure 4.** 2D visualization of the three simulations of sensors positioning in a typical corridor through the BIM model.

The BIM model also allowed for an optimization of the field-of-view of the sensors, as shown in Figure 6. The virtual camera field-of-view simulation supported the definition of the best orientation, i.e., the best tilt angle of each camera on x- and y-axis. This ensured that all the offices and bathrooms defined as critical, whose occupancy needed to be monitored, were correctly detected by the sensors.

**Figure 5.** Ultimate spatial distribution of camera-based sensors inside the case study building.

**Figure 6.** Comparison between the simulation of the virtual sensor field-of-view in the BIM model (**left**) and the actual field-of-view of the installed sensor (**right**).

#### 5.1.3. Analysis of Electrical and Data and Communication Systems

The last preliminary analysis performed with the BIM model was the check of the electrical and data system equipment and wiring distribution already available in the building. The analysis showed that since cameras would be installed in corridors, all the necessary wiring was already available. Therefore, no implementation was needed to install the system.

#### *5.2. Test Campaigns*

Once the preliminary analyses had been performed, the system had been installed. First data collection was performed, and collected data were analyzed to identify issues and faults. Data were collected during a three-month period, i.e., the representative period, as it is the minimum period of time to encounter all possible activities conducted by the users of the department. A qualitative analysis was conducted on the collected dataset to identify rough errors.

Figure 7 shows a graph of collected data about a bathroom during one day. Stuck values are identified since 4 p.m., because the occupancy raises but never decreases. It is a stuck value because the bathroom can host only one person at a time; therefore, a continuous occupancy of four people represents without any doubt a blunder in the detection. The solution to this specific problem is provided in Table 4.

After faults of the system were detected, a first test campaign was performed to understand the causes and propose improvements for the system. Detected data were tested in real time by two operators, as described in the method, to identify the causes of system faults. An example of the visualization of real time data in the online platform is shown in Figure 8.

**Figure 7.** Graph of collected data regarding occupancy overtime in a bathroom (O-T).

**Figure 8.** Visualization of real time data in the online platform during test campaign.

This first test campaign was followed by a second test campaign, to properly calibrate the system and ensure detected data quality, as shown in Figure 9. The two test campaigns were carried out during different periods of the year. This represented a key aspect for the recognition of lighting contrast issues. The first test campaign was performed in June 2020, with data collection for a three-month period from November 2019 to January 2020. The first test campaign was performed after the end of the first Italian shutdown period due to COVID-19 pandemic (early March–early June 2020). The second test campaign was performed in November 2020, with data collected for another three-month period from July to October 2020, excluding August, during which the building is usually under-occupied due to summer holidays. After the shutdown period March–June 2020, administrative and research activities have been resumed. Therefore, all the data collected for system test and calibration can be considered reliable.

**Table 4.** Identified issues and effects on the data collecting system during the test campaigns. Table legend: (a): data collection phase; (b): real-time test campaign phase; O: occupancy values at room-level, number of people occupying the room (p); T: period of time in which users occupy a room (minutes/hours); adjustments hierarchy scale ranging from (++) most preferable system adjustment to (–) least preferable system adjustment.



**Table 4.** *Cont*.

**Figure 9.** Preliminary analyses and two test campaigns process.

#### **6. Results and Discussion**

Table 4 provides a resume of errors, related evaluation criteria, fault identification and classification, causes of the faults, and proposed solutions, hierarchized and listed from the most preferable one (++) to the least preferable one (–) of the two test campaigns.

The first test campaign highlighted the following issues:


The identified issues are mainly due to the geometry of corridors, which are low ceiling, long, and narrow. Due to the limited height of the corridor, the cameras struggle in detecting people walking in groups or lined up (Figure 10). The issues led to an incorrect user detection affecting the displayed data in the online platform. However, while the possibility of having people walking lined up or in groups in corridors is relatively high, the probability of two or more people entering a room simultaneously is low, due to the standard dimensions of the doors, that allow the entrance of one person at a time. For this reason, it is possible to ignore these issues. To overcome the issue related to irregular detection of users due to the distance of areas from the camera, improvement in the deep learning algorithms for image recognition were implemented.

**Figure 10.** Location and orientation of camera-based sensors in corridor.

Considering the online platform, a major issue was related to the values indicating the presence of people in the rooms showing a negative value or a high positive value. This means that, according to the collected data, many people were entering or leaving the room in a very short time. Automatic routines to solve and mitigate incorrect data have been implemented to the software:


The improvement of the automatic routine solves the related problem of users' behavior that deceive the detection system, such as standing in front of the door when opening the office or talking right in front of the entry of a room.

After the modifications and improvements applied, data have been collected for a period of three months from July to October 2020, excluding August for lower building occupancy due to summer holidays, to verify the effectiveness of the strategies adopted.

The qualitative analysis of the second dataset highlighted a general improvement in detection capabilities of the system, since technical issues were not identified anymore, but some faults occurred anyway. Therefore, a second real-time test campaign was carried out in November 2020, resulting in the following sensor-related issues:


**Figure 11.** Unexpected blind zone generated by unusual occupants' behavior.

Figure 12 presents the percentages of error types detected during the two test campaigns. During the first test campaign, 30 out of 70 monitored rooms presented detection

faults, while during the second test campaign, 38 rooms presented detection issues. The reason of the higher number of faults during the second test campaign is mainly due to the lighting issues that emerged only during the second test campaign.

During the first test campaign, 70% of errors were related to technical issues: 17% of errors due to not-working cameras, and 53% of errors due to difficulty detecting users in areas too far from the camera-based sensors. As shown in Figure 12, these types of technical issues were completely fixed by adjusting the system settings. Once all cameras were properly working and correctly set, those issues did not occur in subsequent analyses.

Another type of issue detected in the first test campaign was related to unexpected user behavior, resulting in 30% of errors. These errors could be adjusted with some improvements in the system. However, a 3% of errors due to unexpected user behavior occurred also during the second test campaign. Despite the error percentage being significantly lower in the second test campaign, this type of error could not be completely avoided because of the unpredictable nature and high variability of user behavior.

Regarding the second test campaign, elevated lighting contrasts between different areas of corridors caused 68% of errors. This kind of error was never detected during the first test campaign because of the different period of the year when the test was performed, a key aspect for proper calibration of a camera-based sensors system.

The remaining 30% of errors of the second test campaign were related to difficulties detecting the cleaning employee (18%) and to obstructions and obstacles that impeded the detection (11%). Cleaning employee detection issues were classified as a technical error in the first campaign. After the resolution of technical issues, the error persisted, and the actual cause was identified, highlighting the importance of a multi-stage testing of the

system. Detection issues related to obstacles and obstructions appeared because some pieces of furniture were moved or spaces were reorganized. The frequent check of the correct functioning of the system overtime is fundamental to verify newly appeared issues and consequently adjust and improve the system.

A comparison of these results with other systems could be helpful to provide an assessment of the proposed system. As previously underlined, the setting and calibration of monitoring system has frequently been neglected in existing literature, and the accuracy of IoT systems applied to DTs is often taken for granted. Available data regard, as shown in Table 1, the accuracy of specific sensors' typologies, but do not address the accuracy of systems, which depends on several variables, e.g., building features and use, number of sensors, etc. The test campaigns here presented have been used to explore and improve the efficiency of the entire systems of DT.

For this reasons, these results obtained from the case-study building cannot be compared to other systems, based on different typologies of sensors. The provided case study application is useful to define guidelines to calibrate IoT camera-based sensors system.

#### **7. Conclusions and Further Developments**

This work presents the development of first steps of an ongoing research project to define a Building Management System (BMS) for facility management, especially regarding the occupancy and cleaning activities in office buildings that would be based on an occupancy-oriented Digital Twin (DT). The proposed BMS would ensure better space management, organization, and cleaning, since the system would detect actual occupancy levels and related needs for cleaning activities. The advantages result in optimizations of cost and space use, as well as customized cleaning activities and contracts.

In particular, this study presented the IoT system calibration phases, i.e., the preliminary analyses to optimize the planning of the IoT camera-based sensors system, and the test campaigns, in order to ensure the system efficiency and accuracy to monitor occupancy. A key aspect of the definition of a DT has been in fact identified in the data connection between physical asset and virtual counterpart, the main components of a DT. In addition, the data quality is a critical aspect to ensure the quality of the results of the analyses, simulations, and predictions performed on the virtual model.

The case study section highlighted that the preliminary analyses, i.e., Indicative Post-Occupancy Evaluations (POE) supported by the use of the BIM model, were important to plan the IoT system, in particular as regards number, locations, and orientation of the sensors. The analyses allowed the identification of offices and bathrooms as main spaces to be monitored. In addition, the observed configuration of building spaces allowed planning the sensors installation only in corridors, from which it is possible to detect entries and exits from the different rooms. The BIM model allowed for simulations of sensors location and fields-of-view.

As regards the two test campaigns results, some system faults and related causes were identified and solved. The issues generated by user behavior were the least predictable, trivial, and at the same time the most difficult and expensive to solve, requiring the installation of new cameras. The variability of human behavior inside a building is very high; the calibration of the system must cover a sufficient period of time to bring out all problems related to human behaviors. Considering the complexity of the monitoring system and the high dynamicity of the variables involved (e.g., fast-changing spatial conditions and user behavior), a multi-stage test and calibration campaign was fundamental for the correct setting of a camera-based sensor system.

Another interesting aspect resulting from the test campaigns was the influence that the period of the year had on the test itself, due to changing lighting conditions.

Other relevant aspects are the geometric features of the to-be-monitored spaces. For example, the limited width and height of the corridors led to some difficulties in detecting more users moving together. However, those issues did not have critical effects on the collected data. The boundary conditions of the system should be carefully checked, as they could have negative consequences on collected data and on data analyses.

The use of an online platform was useful to real-time check and evaluate data during the test campaigns, as well as to remote controlling the monitoring system.

Once the system is tested and assessed, further developments of the research will regard the proper monitoring of the building. As of now, qualitative analyses have been performed on collected dataset to identify rough errors, and by means of the two test campaigns, the causes of the faults have been identified and solved. During the next phases of the research project, quantitative analyses will be conducted on collected datasets, which will be the basis for the definition of the occupancy-oriented DT. DT analyses and simulations, and resulting optimization scenarios, will be proposed and analyzed to identify real advantages and limitations of the proposed methodology. The proposed method, once completely tested and refined on the case study building, could be extended to large building stock, supporting the decision-making process of building owners and building managers.

Potential applications of the system would entail the integration of other kind of sensors to monitor Indoor Air Quality (IAQ), carbon dioxide, temperature, humidity, and Volatile Organic Compounds (VOC) levels, resulting in a more complete evaluation of the building conditions and Indoor Environmental Quality (IEQ). Sensors could play an important role for safety management purposes. The combined use of the system with Smart Contract and Blockchain technology could ensure increased network security, reliable data storage, traceability of data, and the possible automation of payments for cleaning activities. Cleaning contracts could in fact be customized based on the actual use of spaces, detected by the proposed system.

**Author Contributions:** Conceptualization: E.S., M.L., and L.P.; methodology: E.S., M.L., and L.P.; software: L.P. and G.P.; validation: E.S, M.L., L.P., G.P., G.M.D.G., L.C.T., and G.B.; formal analysis: E.S., M.L., and L.P.; investigation: M.L., L.P., and G.P.; resources: G.M.D.G.; data curation: M.L. and L.P.; writing—original draft preparation: E.S., M.L., and L.P.; writing—review and editing: E.S, M.L., L.P., G.P., G.M.D.G., L.C.T., and G.B.; visualization: E.S., M.L., and L.P.; supervision: G.M.D.G., L.C.T., and G.B..; project administration, G.M.D.G.; funding acquisition, G.M.D.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Restrictions apply to the availability of these data. Data was obtained from Department of Architecture, built environment and construction engineering and are available from the authors with the permission of Department of Architecture, built environment and construction engineering party.

**Acknowledgments:** The authors thank Eng. Marco Schievano and Eng. Francesco Paleari, and the Department of Architecture, Built Environment, and Construction Engineering for the support in the ongoing research. This research is developed in collaboration with Laser Navigation srl.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

