Advanced Electronic and Optoelectronic Sensors, Applications, Modelling and Industry 5.0 Perspectives

Abstract

This review will focus on advances in electronic and optoelectronic technologies by through the analysis of a full research and industrial application scenario. Starting with the analysis of nanocomposite sensors, and electronic/optoelectronic/mechatronic systems, the review describes in detail the principles and the models for finding possible implementations of Industry 5.0 applications. The study then addresses production processes and advanced detection systems integrating Artificial Intelligence (AI) algorithms. Specifically, the review introduces new research topics in Industry 5.0 about AI self-adaptive systems and processes in electronics, robotics and production management. The paper proposes also new Business Process Modelling and Notation (BPMN) Process Mining (PM) workflows, and a simulation of a complex Industry 5.0 manufacturing framework. The performed simulation estimates the diffusion heat parameters of a hypothesized production-line layout, describing the information flux of the whole framework. The simulation enhances the technological key elements, enabling an industrial upscale in the next digital revolution. The discussed models are usable in management engineering and informatics engineering, as they merge the perspectives of advanced sensors with Industry 5.0 requirements. The goal of the paper is to provide concepts, research topics and elements to design advanced production network in manufacturing industry.

1. Introduction

Technological advances in industry are in different production sectors of the supply chain. Today, many optical, electronic, and mechatronic technologies can be applied to the industrial scenario, improving the Industry 4.0 framework based on the digital transformation process. Some recently advanced technologies are in metastructures oriented to optical computing [1], nanoparticles in metasurface applications [2], Epsilon–Near-zero (ENZ) metamaterials [3,4], graphene-based transistors for biomedical applications [5], plasmonic devices [6,7], and deep learning that supports decision-making, as seen in COVID-19 risk management processes [8,9]. These technologies could be integrated into industrial processes or considered as innovative products for manufacturing industries. An upscale in production was gained by Industry 4.0 and by innovative facilities of Industry 5.0. The Industry 5.0 era has been characterized by a full digitalization and transformation process which contributed to fully changing organizational and production processes. This change is mainly focused on the implementation of Artificial Intelligence (AI) algorithms, which tailor production and processes in a self-adaptive modality. Specifically, the new advances in optoelectronic and mechatronic technologies show a possible evolution in the management of industrial processes, thus optimizing resources according to the dynamic market. Today, a strategic market plan should be designed first in the short term, to follow better the market’s trends and trajectories. The goal of this proposed short review is then to define production models based on advanced technology discussed in the paper, and offer an approach to designing a possible future framework which includes Industry 5.0 facilities that can start an analysis of the scientific state of the art profiled here to serve actual industry technology needs, and to orient further research. Figure 1 illustrates a block diagram sketching a basic Industry 5.0 framework, which will be discussed in this paper. The diagram, deduced from the analysis of the state of the art and from emergent topics in research, consists of the following blocks or grouped modules:

-

A sensing module composed of innovative optoelectronic sensors (block 1), industrial mechatronic sensors (block 2), and detection algorithms (block 3);

-

Control and actuation facilities (block 4) which address production, quality, safety, process management, raw materials management, and Process Mining (PM);

-

An AI engine managing production line actions (blocks 1, 2, 3, 4, 5);

-

An Advanced robotic industrial platform (block 5).

The proposed auto-consistent framework of Figure 1 highlights the relationships between all mentioned blocks. Particular attention is given to new advanced sensors such as optoelectronic sensors suitable for high-velocity data transfer, industrial mechatronic sensors supporting the control and actuation mechanisms of advanced production machines and industrial robots, and engineered intelligent processes requiring AI decision-making procedures. The proposed work is structured as follows:

-

Discussion of the adopted methodology of the searching approach of the state of the art;

-

Discussion of the blocks of the framework of Figure 1, which will by provide a perspective of the implementation of each framework in a future Industry 5.0 scenario for each topic;

-

Description of a complex model of an Industry 5.0 framework defined by the topics found in the state of the art (framework constructed by supposing possible evolutions/implementations of the analyzed technologies and by hypothesizing a production line in the manufacturing sector);

-

Simulation of the information flux of the designed framework by means of the estimation of the diffusion heat parameters;

-

Design of Business Process Modeling and Notation (BPMN) workflows applied to this example of the Industry 5.0 framework, which will be useful for enhancing the different technological levels that will upgrade the production;

-

A BPMN model explaining, with more details, the processes of the robotic and machine control exploiting predictive maintenance and quality assessment processes designed with the Industry 5.0 framework.

Methodology

The method of searching for works in the literature is based on the scheme of Figure 2, and is defined by the following main steps:

-

The topics found in the literature are related to the requirements of the industry research projects (regional, national and European research topics) outlining the following macro-topics: Mechatronic Systems, Industry 4.0, Digital Transformation, Agriculture 4.0, Internet of Things (IoT), Human & Machine Interfaces (HMI), Quantum Computing, Energy, Edge Computing, Artificial Intelligence, Dynamic Business and Strategic Marketing, Additive Manufacturing, Key Enabling Technologies (KET), Photonics, Micro-Technologies, Nano-Technologies, Advanced Materials, Smart Materials, Technologies and Advanced Production Systems, Innovative Management Frameworks, Change Management, Product Quality Assessment, etc.;

-

Different keywords matching with specific industrial application fields (papers found in literature) are extracted, such as: Micro-Sensors, Nano-Sensors, Optoelectronic Sensors, Image Vision Techniques, Leakage Detection, Sensing and Actuation AI Systems, Nanocomposite Sensors, Defect Detection, Defect Prediction, Embedded Electronic Devices and Systems for the Automatic Control of Assembly Processes, Embedded Microelectronics, Integrated Systems for Applications of Remote Control, Multi-Sector Environment, Soft Robotics, Infrared Thermography, Multi-Spectral Analysis, AI Control and Actuation, etc.;

-

New research topics following the correlations between the different extracted keywords are defined, such as: rapid prototyping, reverse engineering, Industry 5.0, process mining applied to industrial production processes and on quality assessment, electronic assisted production management, etc.;

The proposed models integrate innovative Industry 5.0 topics found in litterature, are designed by the draw.io open source tool (BPMN standard), and are simulated by the Cytoscape tool (simulation of a complex Industry 5.0 framework which estimates the diffusion heat parameter).

The next sessions discuss the papers selected from literature about the topics and the keywords matching with the framework of Figure 1. For each topic found in these works, possible implementations in Industry 5.0 scenarios are proposed, and innovative aspects of perspectives in possible implementations are discussed.

2. Sensing Field

Many works found in literature address the sensing field in industrial environments (blocks 1, 2, and 3 of Figure 1). Innovative sensors such as optoelectronic sensors are a part of the sensing field and of innovative electronic/mechatronic systems that are interconnected to detection algorithms (see Figure 1). Specifically, Complementary Metal-Oxide Semiconductor (CMOS) and laser-based technologies [10,11,12] are typically applied to implement and to improve image vision techniques detecting colors, temperatures, and defects in manufacturing production processes. Looking forward-, advanced systems could include cameras integrated to other sensor systems, switching automatically in cases of defective cameras and detecting intelligently specific defect regions. For example, the scanning of small regions could be performed by means of intelligent algorithms controlling zooming. Nanocomposite optoelectronic sensors [13,14,15,16,17,18,19,20,21,22,23,24,25] are characterized by a fast detection response and by a high sensitivity, and are suitable for the detection of gases, energy-harvesting applications, pressure sensing, detection of notches and surface defects such as cracks, and three dimensional (3D) object morphologies and colors. High sensitivity and a fast response rate are very important in workpiece processing requiring a micrometric manipulation and very low processing tolerances (high-precision workstations), and for fully optical systems requiring a working frequency in the THz band. Advanced solutions could be implemented in worker security systems that detect small quantities of gases or liquids, thus enabling automatically alerting systems and security procedures. Concerning the robotic handling process, a high sensitivity pressure response allows the use of nanocomposite sensors for soft handling or for production systems characterized simultaneously by both hard and soft pressure forces; an intelligent system could calibrate the robotic handling by reading simultaneously strong and soft pressure forces to accurately control the product’s handling, and to detect possible millimeter-sized defects on the surface. The possibility of controlling the chemical composition of polymeric materials would allow tuning of the sensitivity of the optical response, which would allow the design of sensors which have different pressure working ranges that can adapt to the system of control. The control logic to be implemented could be improved by an auto-adaptive feedback control system which could intelligently drive robotic motors. An advanced nanocomposite sensor system could include both touch and no-touching sensors that detect 3D object morphology and other parameters such as colors and temperatures.

The research topics give particular attention to the leakage detection of oil, water and gases, including pipeline and electrical leakages [26,27,28,29,30,31,32,33,34,35,36,37,38]. Advanced solutions could include the adoption of AI tools for the pollution and leakage detections (as for production processes using liquids), and the integration of different technologies interconnected to renewable energy routing systems (sustainable production systems).

Finally, optoelectronic and fiberoptic systems are applied in manufacturing processes by controlling cutting machines, assessing the quality of surfaces of processed workpieces, and in additive manufacturing processes [39,40,41,42,43,44]. Advanced Industry 5.0 solutions could be addressed on reverse engineering approaches that automatically optimize laser cutting operations, that also integrate image vision techniques, or AI data driven systems that enable self-adaptive manufacturing processes. Specifically, additive manufacturing processes could be controlled intelligently with an AI self-tuning the machine parameters.

Table 1. State of the art: some advanced optoelectronic sensors adaptable for industrial applications in Industry 5.0 scenario.

Optoelectronic Technologies	Topic	Description	Reference	Possible Implementation in Industry 5.0 Proposed in This Work
CMOS RGB photodetectors	High-temperature detection	Acquisition of a digital images representing the temperature field	[10]	Quality check of product surface at high temperature by scanning automatically very small regions (intelligent control).
	Colorimetric gas detection	Accurate monitoring of the color change	[11]	Integration with other sensors detecting gas leakage in industrial environment.
	Inspection of camera defects	Automatic inspection system for color defect check in manufacturing process	[12]	Real time checking of defect line and color defect for array of cameras working for quality check (auto-changing of camera in cases of defective cameras)
Nanotechnology	Silicon nanowires	Silicon nanowires read the percentage of resistance variation in the presence of gases (laser interaction) [13], and are able to detect different gases such as CO, C6H6, NO2, etc. [14,15]	[13,14,15]	Security in air quality control in research laboratories: integration of different technologies to detect all gases characterizing a prototype or a chemical processes.
	Carbon-based nanocomposite flexible devices	Mechanical sensing and solar energy harvesting	[16,17]	High-sensitivity pressure sensors integrating simultaneously with solar energy harvesting generators.
	Polymeric nanocomposite liquid sensing	High sensitivity in detecting variation of the optical transmittivity response related to different liquids	[18]	Use of this typology of sensors to measure different percentages of contaminating liquids.
	Polymeric pressure sensors	High-sensitivity pressure sensors detecting soft pressure forces	[19,20,21]	Implementation of hard- and soft-pressure sensors obtained by changing the chemical composition of the nanocomposite material. Both soft- and hard-pressure force sensors can be implemented into a unique controlled system that addresses the calibration of forces.
	Optical nanocomposite sensors contacting surfaces	Sensors detecting small notches and shear forces	[22,23,24]	Implementation of intelligent AI algorithms classifying information about the morphology of defect discontinuity on products or in general on surfaces
	Optical nanocomposite sensors not contacting surfaces	3D object morphology and color detection	[25]	Implementation of intelligent algorithms scanning the object to detect by means a specific sampling pattern depending on the object morphology, and possible integration with image vision techniques to extract other object features.
Leakage and pollution sensors	Oil and gas leakage of from submerged structures	Estimation of the temperature of power lines.	[26,27,28]	Application of the proposed technology for production plants working with water or in liquids.
	Optical sensor detecting water pollution	Light-Emitting Diode (LED) source measuring turbidity [29] or an optical fiber transmitting- and receiving-system for oil spill applications [30]	[29,30]	Simultaneous use of turbidity and liquid substance detection to extract features of complex liquid pollution (use of AI algorithm for extracting information).
	Exhaust gas optical detection	Exhaust gas detection performed by luminescence method (automotive sector)	[31]	Integration of the technology to automate quality control in engine production
	High voltage leakage currents	Current leakage in 500 kV transmission lines	[32,33]	Optimization of electrical networks powering big industries linked to renewable energy sources (smart grid network balancing operations).
	Pipeline leakage	Fibre Bragg Grating Sensor monitoring pipeline leakage by the wavelength shift due to a variation in liquid or gas flows [34]; gas flow detection [35,36]; liquid flow detection network systems [37,38]	[34,35,36,37,38]	Integration of different technologies (capacitive based, electromagnetic, infrared thermography, acoustic or visual) optimizing flow detection and decreasing the leakage detection error.
Optoelectronic sensors in manufacturing	Trajectory distortion in laser cutting operations	Measurement of mirror’s inclinations with high resolution	[39]	Application of reverse engineering methods to optimize automatically the laser cutting operations.
	Optical Sensors improving Industry 4.0 facilities	Fiberoptic sensing in Industry 4.0	[40,41]	Integration of automatic actuation following fiber optic sensing; adopting self-adaptive manufacturing processes by means of AI data processing.
	Laser and vision techniques	Quality control of wood surfaces by using Laser TriangulationMethod (LTM)	[42]	Automation of waste of wood by simultaneously processing information of infected areas and of processing defects.
	Optical inspection in Electronics	Check for defects in electronic devices (wafer defects, solder defects, etc.)	[43]	Automatic opticalinspection systems based on an image vision AI algorithm that classifies defects in electronic manufacturing.
	Acusto-OptoFluidic (AOF) system	Laser Direct-Writing (LDW) approach in additive manufacturing	[44]	Additive manufacturing control and actuation managed in real time by a powerful algorithm adjusting step-by-step precision and laser machine parameters (power, pulse signal delay, fluency, etc.)

reports more details about the optoelectronic systems and their possible implementation in an Industry 5.0 framework.

Our selected state of the art discussion of electronic and mechatronic sensors is about optomechatronic systems [45,46,47,48], printed and flexible electronics [49,50,51,52], gas sensors [53,54], sensor network systems [55,56,57,58,59,60], and other sensors applied in different production processes such as food and agriculture; and manufacturing processes such as drilling, milling and cutting [61,62,63,64,65,66,67,68,69,70,71]. Perspectives in Industry 5.0 frameworks involve the possibility of integrating different technologies into a unique platform, and adopting spatially adaptable electronic components for specific production plants. Furthermore, the evolution of production processes requires advanced diagnostic to be implemented by AI predicting risks and by applying PM models, auto-calibration of the manufacturing machine parameters, and auto-adaptive control systems monitoring tool wear.

Table 2. State of the art: some electronic and mechatronic sensors adaptable for industrial applications in Industry 5.0 scenario.

Electronic and Mechatronic Technologies	Topic	Description	Reference	Possible Implementation in Industry 5.0 Proposed in This Work
Optomechatronic Systems	High-resolution lighting and optical models	Increase of image quality	[45,46,47,48]	Combination of more procedures and models addressing the optimization of the optomechatronic systems.
Printed Electronics (PE)	PE Sensors	Definition of smart active object; adding a data processing unit	[49]	PE-integrated circuit in actuation boards.
	Flexible electronics	Adoption of stretchable electronic circuits suitable for harsh environments	[50]	Combination and simultaneous processing of data acquired from different production environments to find correlations that optimize quality.
	Pressure sensors on flexible substrate	Piezoresistive pressure sensors printed on a polymeric substrate	[51,52]	Adoption of the same flexible substrate for the integration of other printable circuits
Electronic gas sensors	MOSFET-based gas sensors	Detection of CO and NOx gases	[53]	Reduction of gas emissions of by industrial plants by activating an automatic alerting system (sustainable production plants).
	Gas sensor	Fabrication of quantum cascade laser-based photoacoustic (QCL-PA) detectors, Gold Nanoparticle-based Field-Effect Transistor (Au-NP-FET), and III-V semiconductor-based circuits, detecting NOx gases	[54]	Integration of different technologies based on the specific sensitivity of gas detection.
Sensors systems and networks	Multiple pressure sensors monitoring fluid networks	Network monitoring system connected to SCADA units (pressure sensors monitoring water leakage);	[55]	Prediction of fluid leakage by means of AI algorithms (classification of the signals detected from different parts of the monitoring network).
	Sensors monitoring railway infrastructures	Multiple sensors detecting temperature or mechanical anomalies of the infrastructure	[56,57]	Use of AI diagnostic algorithms to predict risks.
	Wireless sensor network	Wireless sensor network for assembly process monitoring and for process management.	[58]	Process mining addressing assembling processes enabled by AI decision-making engine.
	Smart sensor system monitoring production lines	Smart factory integrating temperature, pressure, position, force, gas, color, light, flow, nuclear, micro- (MEMS), and nano- (NEMS) sensors	[59,60]	Data processing to find correlations between more signals to build production fault prediction.
Sensors in production processes	Roasting process	Temperature [61] and Near-Infrared Spectroscopy (NIRS) [62] to check food quality in roasting processes	[61,62]	Auto-calibration of temperatures ensuring a uniform roasting process (production lines made up of different ovens).
	Agriculture 4.0	Precision agriculture combining different technologies such as multispectral imaging, infrared thermography, and Unmanned Aerial Vehicles (UAV)	[63,64,65]	Implementation of Decision Support Systems (DSS) based on data fusion and local management of agriculture.
	Drilling process	Switch-activated and sensor-triggered proximity sensors detecting the arrival of the workpiece [66], and image processing [67]	[66,67]	Improvement of a reconfigurable manufacturing process by means of AI algorithms
	Food processing	X-rays for the detection of unknown elements in food products, thermal imaging (temperature and humidity check), volumetric sensors (check of raw materials in silos), NIR spectrometry (controlling moisture), image processing	[68]	Feedback control applied to an adaptive framework based on an AI engine optimizing the whole supply chain
	Hole machining process	Drilling, broaching, countersinking, and horning processes controlled by sensors detecting cutting forces, vibration, current/power, acoustic emission and temperature; and other tools such as microscopes, laser scanners, surface roughness testers, dynamometers, scanning galvanometers, profilometers, and internal micrometers	[69]	AI prediction of hole defects adjusting in real time machine parameters; predictive maintenance of machines
	Cutting Process	Strain gauge measuring cutting forces, cutting power	[70,71]	Auto-adaptive control predicting tool wear and tool breakage

lists more information about some electronic and mechatronic sensors and their possible implementation in an Industry 5.0 framework.

Concerning the industry sensing field, the analyzed state of the art finally addresses the important topics of algorithms supporting leakage detection [72,73,74,75,76,77,78,79,80,81], the digital transformation process of Industry 4.0 [82,83,84,85,86,87,88,89,90,91,92,93,94,95,96], and the data processing regarding energetic consumption in production processes [97,98,99,100,101,102,103]. Possible improvements to integrate into an Industry 5.0 framework relate to automated AI decision-making processes which enable interventions and procedures, production simulation setting machine parameters, and auto-calibration of the machines used in manufacturing processes. In the proposed scenario, an important role is assigned to big data systems interconnected to AI engines and to data-fusion approaches. The AI-based detection algorithms could provide an important contribution to managing complex production systems by estimating many variables such as energy required for the machined production, product defects, machine failure features, and more. Some efficient AI algorithms are Artificial Neural Network (ANN), and Long Short-Term Memory (LSTM). Other useful algorithms are Discrete Fourier Transform (DFT), and K-Means clustering algorithm (machine learning unsupervised algorithm). Two possible upgraded systems are in the management of energy used in production, and in energy digital twins using quantum computing. A possible future professional skill would be an energy manager that optimizes electrical consumption in industrial processes. Machine learning algorithms are also suitable for defect classification and defect prediction.

Table 3. State of the art: detection algorithms and possible implementation of processes in an Industry 5.0 scenario.

Detection Algorithms/Models	Topic	Description	Reference	Possible Implementation of Processes in Industry 5.0 Frameworks (Proposed in This Work)
Leakage detection	Pipeline leakage detection (liquids, gases or solids suspended in a liquid)	Laguerre fuzzy proportional-integral-derivative (PID)observation system	[72]	Automated processes enabling pipeline repair interventions.
	Natural gas leakage images	Thermal readings and images analyzed by machine learning algorithms	[73,74]	Alerting security processes enabled by AI algorithms.
	Gas leakages data	Spatial and temporal neural network model	[75]	Alerting security processes enabled by AI algorithms.
	Gas leakage images	Visual background extractor algorithm applied to mid-infrared.	[76]	Alerting security processes enabled by image vision algorithms.
	Water leakage	Data processing algorithms extracting information from infrared thermography and Ground Penetrating Radar (GPR) images	[77]	Data fusion processed could extract secure information about water leakage.
	Water/fluid leakage	Pressure data detected in whole water network	[78]	Data processing of correlations between all the measurements acquired along the whole water network (with application to a generic fluid network).
	Water/liquid leakage	Computational Fluid Dynamics (CFD) model	[79]	CFD simulations are included into reverse engineering processes to optimize production of pipeline network parts.
	Water/fluid leakage	Denoising method	[80]	Optimization process for algorithms detecting leakage.
	Water/fluid leakage	Genetic algorithm modelling leak nodes of the water network	[81]	Automatic data processing optimization that simultaneously considers all information extracted from all network nodes.
Industry 4.0	Data-driven manufacturing	Machine learning algorithms	[82,83,84]	Auto-calibration of machines and manufacturing processes.
	Integration of technologies	Digital twin approach: product, asset, system, process twins	[85,86,87]	Creation of front-end Human Machine Interfaces (HMI) supporting hardware and software integration and ensuring data collection in big data systems.
	Algorithms processing data at different steps of levels	Pre-processing, descriptive analytics, predictive analytics and prescriptive analytics	[88]	Implementation of multi-level data processing systems managed by AI decision-making processes.
	Quality	Machine learning quality prediction and anomaly detection	[89]	Implementation of automated quality procedures with the goal of avoiding production failures and performing automated preventive intervention.
	Smart Manufacturing	Deep reinforcement learning approaches	[90]	Integration of different algorithms learning approaches for optimizing machine learning data processing.
	Machine failure	CNN-LSTM forecasting models	[91]	Implementation of automated predictive maintenance procedures.
	Defect classification	Discrete Fourier Transform (DFT), K-Means clustering and Long Short-Term Memory (LSTM) classifying tire defects	[92]	Algorithm fusion for improving defect detection.
	Defect mapping and prediction	p-charts mapping defect trends and Artificial Neural Network (ANN) predicting defects	[93,94,95,96]	Integration of traditional processes for mapping and controlling defects with AI algorithms.
Energy and production	Digital Twin	Machine learning detecting anomalies [97], and predicting a system’s dynamic behavior with data-driven methods [98]	[97,98]	Improvement of flexible and auto-adaptive powerplant optimization, ensuring high product quality.
	Quantum computing	Renewable energy and energy source management by quantum computing approaches	[99,100,101]	Large-scale energy-balancing approach optimizing energy for production districts.
	Smart energy systems and Key Performance Indicators	Long Short-Term Memory (LSTM) [102] and Artificial Neural Network (ANN) energy forecasting [103]	[102,103]	Simulations of complex energy systems (theory of complex systems) based on AI energy-forecasting to optimize energy in production plants and buildings.

discusses some topics about detection algorithms and their possible implementation in an Industry 5.0 framework.

3. Supply Chain Processes and Advances

The research topics in industrial production mainly address an intelligent way to manage raw materials [104,105,106,107,108,109,110,111], new models of AI decision-making in processes controlling production [112,113,114,115,116,117,118,119,120,121,122,123], and new procedures/methods oriented to control and actuation actions [124,125,126,127,128,129,130,131,132]. The topics are indicated in block 4 of Figure 1. Possible implementations of the analyzed topics in Industry 5.0 perspectives are mainly in process automation and in processes controlled by AI decision model engines that improve production, security, and quality. Other important aspects of industry upgrades are in the integration of different technologies and in intelligent machine reconfiguration systems. The integration of AI in the decision-making processes allow to the implementation of PM models; AI data driven by production phases is a new advance for process switching and process management, which includes new organizational models. Image vision techniques based on infrared thermography technologies are important for improving inline monitoring processes controlling defects and automatically reconfiguring machines and robots.

Table 4. State of the art: raw materials management advanced approaches, process mining, control/actuation methods and possible implementations of processes in Industry 5.0 scenarios.

Raw Material Management Models, Process Mining and Control & Actuation Improvements	Topic	Description	Reference	Possible Improvements of Processes in Industry 5.0 Frameworks (Proposed in This Work)
Raw materials check and management	Quality check of raw wood pellet materials	Fourier Transform InfraRed (ATR-FTIR) spectroscopy applied to the analysis of chemical composition of wood pellets	[104]	Automated processes for managing raw materials and product quality assessment.
	Raw materials in pharmaceutics	Use of technology such as Near-Infrared (NIR) spectroscopy and fiberoptic probes for quality control	[105]	Data fusion techniques and AI algorithms to process optical signals.
	Raw material check in food industry	Temperature and humidity check	[106]	Continuous process monitoring of raw materials in all the phases of the supply chain by enabling automated waste.
	Optoelectronic sensor checking linear density of raw materials	Checking of linear density of cotton	[107]	Fast and accurate quality processes applied to raw materials.
	Estimation of the quality level of raw materials	Application of the Six Sigma methodology	[108]	Integration of traditional management processes with AI decision-making engines.
	Management and reduction of stone waste	Finding of technologies for reducing waste	[109]	Implementation of automatic sustainable processes based on waste control and reuse.
	Mathematical model controlling raw material releasing	Control of quality levels in manufacturing	[110]	Ideation of new algorithms as alternatives to AI ones to control raw materials in the whole supply chain.
	Raw materials management	Process management of raw materials	[111]	Process mining applied to raw materials.
Processes mining	Quality prediction	Machine learning algorithms applied to quality processes	[112,113]	Auto-calibration of quality processes by means of AI quality predictions.
	Decision Support Systems (DSSs) enabling process mining	Automated improvement of processes concerning product quality, worker security, and machine parameter setting	[114,115]	Production, security, quality and organizational processes enabled automatically by AI algorithms.
	Predictive maintenance process	Automated predictive maintenance by machine learning matched with Industry 4.0 technologies	[116,117]	Predictive maintenance procedures enable automatic interventions, which completely avoids the risk of machine failure.
	Data quality	Internet of Things (IoT), data processing and methods to correct errors	[118]	Implementation of data pre-processing algorithms automatically cleaning the production dataset by ensuring good performance of the AI training models (use of big data systems).
	Process mining manufacturing	Process mining focused on advanced business models	[119,120,121]	Dynamic adaptation of production to customer needs and markets in real time.
	Industry control	Model to extract information for detection of anomalies [122] and alarms [122,123]	[122,123]	Integrated alarm systems applied for different types of alarms and worked into a multi-level risk-alerting system.
Control and actuation improving production, quality and safety	Realtime control of fluid flow process	Tomography technique analyzing distribution of phases	[124]	Integration of tomography technique into a sensor network (finding correlations between different parameters).
	Inline quality control	Deep learning applied to a porosity check of polymers	[125]	Inline automated systems could check quality and synchronize successive machines (actuators are synchronized with all machines of the production line).
	Quality control in diary industry	Infrared spectroscopy improving quality control	[126]	Integration of the tomography technique into a sensor network. (Finding correlations between different parameters).
	Image vision in quality control	Imaging techniques in industry	[127]	Integration of image vision techniques into a sensor network (finding correlations between different parameters).
	Additive manufacturing	Control of laser parameters [128], three-staged control approach [129], and quality control in metal processing [130]	[128,129,130]	Control and actuation processes driven by AI.
	Fault-tolerant control in manufacturing	Recurrent neural network-driven control reconfiguration	[131]	Reconfiguration of the whole supply line according to the data processing of all working robots and machines.
	Mechatronics in reconfigurable production systems	Actuation in manipulation, assembly and packaging processes	[132]	Reconfiguration of the whole supply line according to the data processing of all working robots and machines.

has the works found regarding the state of the art around raw materials management, PM models, control and actuation processes, and provides possible implementations in an Industry 5.0 framework.

4. Advanced Robotic Industrial Platforms

Many developments in research fields are in robotic platforms [133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149,150,151] (see block 5 of Figure 1). Collaborative robot (Cobot), soft robot, robot agents, and industrial robots are today controlled by innovative sensors and techniques (as for writing 3D processes in additive manufacturing). Neural networks, reinforcement learning methods, and AI algorithms are possible new solutions for improving the handling and object manipulation of robots. Possible projections in Industry 5.0 scenarios are in synchronization of the whole supply chain’s machinery, in self-adaptive robot control systems, in the use of quantum computing to increase AI efficiency, and in the digital twin models assisting production and marketing strategies. AI data processing could be processed into an Industry 5.0 framework by edge computing systems. Advanced technological platforms are also platforms that accurately control motion tracking, information systems suitable for integration of heterogeneous data from different technologies such as the Enterprise Service Bus (ESB), and the Supervisory Control and Data Acquisition (SCADA) system. All the systems could be integrated into a unique one controlling and synchronizing a whole production. In

Table 5. State of the art: advanced robotic systems and technological platforms, and Industry 5.0 perspectives.

Advanced Robotic Industrial Platform Facilities	Topic	Description	Reference	Possible Improvements of Processes in Industry 5.0 Frameworks (Proposed in This Work)
Robotics	Soft robotics in automated food handling	Sensors improving gripper processing	[133]	Robot data read by sensors could be used to synchronized production-line velocity and machine parameters.
	Industrial robot agents	Reinforcement learning of robot	[134]	Improvement of reinforcement learning as an approach to optimizing the self-adaptive processes of robots
	Robot trajectory tracking	Neural networks improving control	[135]	Self-adaptive robot control systems could adjust real time trajectories.
	Collaborative robot (Cobot)	Artificial Intelligence applied in industrial Cobots	[136,137,138,139,140]	AI edge computing approaches optimized by quantum computers could optimize the Cobot’s processes.
	3D writing robotic systems	Two-photon polymerization (TPP) process	[141]	Integration of innovative laser techniques in precision manufacturing.
	Optical sensors implementable in robotics	3D measurement approach [142] and optical inspection [143]	[142,143]	Complex operation in the 3D space improved by multiple optical sensor 3D systems (high-precision manipulation).
Advanced platforms	Robotic manipulator	Motion tracking algorithms and dynamic models [144], and calibration systems [145]	[144,145]	Intelligent manipulations by AI control.
	From Industry 4.0 to Industry 5.0 scenario	Improvement of human-machine relationships	[146]	Reconfiguration of HMI according to specific processing programs.
	Personalized production	Manufacturing of a personalized product	[147]	The possibility of personalizing the product according to real-time market needs, customer segmentation, and sales predictions.
	Digital twin platform architecture	Digital twin platform connecting sensors and actuators with a Supervisory Control and Data Acquisition (SCADA) system	[148]	Digital twin model assisting and optimizing the production.
	Data quality	Adaptive algorithm removing uncertain data	[149]	Automatic selection of data processing optimizers (optimization of AI choice and of related hyper-parameters based on the cleaned dataset).
	Additive manufacturing platform	Integration of data driven approaches	[150]	Implementation of a precise feedback system automatically optimizing the additive manufacturing process by fast data processing engines.
	Information infrastructure	Enterprise Service Bus (ESB) system interfacing different information systems including big data	[151]	Integration of edge computing approaches to accelerate partial data processing.

, we discuss the works found in the state of the art about advanced robotic platforms and possible improvements in an Industry 5.0 framework.

5. Discussion: Industry 5.0 Framework and Complex Systems Simulations

This section designed and simulated an advanced Industry 5.0 framework integrating different advanced technologies. The discussion is completed by a description of the limitations and perspectives related the analyzed layout, by through a description of examples of applications matching with the Industry 5.0 scenario.

5.1. Industry 5.0 Workflow Design and Process Simulation

The works selected from the state of the art literature allow us to define a possible framework of how Industry 5.0 could involve a lot of previously mentioned technologies. The hypotheses of possible improvements of processes in Industry 5.0 frameworks allowed us to design the production layout, while including the main important variables found in the literature. A modern framework is characterized by a large number of variables, so we structured a complex system characterized by a “capillary” information flux between all the variables named nodes. An approach used to simulate complex models [103] is the calculation of the diffusion heat parameter [152,153,154]. The diffusion heat parameter is a probabilistic definition of the information flux distribution flowing between all nodes. Diffusion is typically a phenomenon associated to “particles” moving in an environment. In the specified case, the environment is modelled by nodes, and edges connecting all the nodes of the complex model. The probability of the distribution of the diffusion heat parameter is the position distribution of particles, which simulates the information flux, after a time t. In the analyzed model, a parameter of t = 0.038 s is enough for the transient calculation of the simulated complex Industry 5.0 manufacturing network. The hypothesized network is assumed to be a single production line consisting of three production machines and a robot, connected in a series configuration as illustrated in Figure 3a. The raw materials are processed by a first machine (Machine 1 node indicated by M1), then the workpiece is handled by a robot (Robot 1 node indicated by R1), sent to be processed by a second machine (Machine 2 node indicated by M2), which realizes a semi-product, which is then processed by a final, third machine (Machine 3 node indicated by M3), which produces the final product. As illustrated in Figure 3a, the machines define a first operating level named Production Line Level which is interconnected to a second one associated to sensing and actuation processes (Sensing and Actuator Level made up of sensors S and actuators A). A third level, named Industry 5.0 Level, represents the upgrade from Industry 4.0 to the Industry 5.0 scenario, and is consists of the following main elements (nodes): Quantum Computer, Big Data, External Big Data, AI Edge Computing, and an Actuator Engine (all the elements suggested for an upscale in production). The simulated network is better modelled by the graph of Figure 3b, which indicates nodes and edges. The simulated nodes are:

• Machine 1 (M1): the first machine, processing raw materials loaded at the input of the production line.
• Sensor M1_1: sensor monitoring Machine 1.
• Sensor M1_2: sensor monitoring Machine 1.
• Actuator M1: actuator that relays the sensing and actuation processing of from Machine 1.
• Power Meter_M1: power meter reading the electrical power of Machine 1.
• Robot 1 (R1): robot handling the workpiece processed by Machine 1.
• Sensor R1_1: first sensor monitoring Robot 1.
• Sensor R1_2: second sensor monitoring Robot 1.
• Actuator R1: actuator which relays sensing and actuation processing from Robot 1 (includes sensing of from both the sensors Sensor R1_1 and Sensor R1_2).
• Machine 2 (M2): second machine processing the workpiece after robotic manipulation.
• Sensor M2_1: sensor monitoring Machine 2.
• Sensor M2_2: sensor monitoring Machine 2.
• Actuator M2: actuator relaying sensing and actuation processing from Machine 2.
• Power Meter_M2: power meter reading the electrical power of Machine 2.
• Semi-Product (SP): semi-product output of Machine 2.
• Image Vision SP: camera implementing image vision algorithms to detect the defects of the Semi-Product.
• Machine 3 (M3): third machine processing the semi-product.
• Sensor M3_1: sensor monitoring Machine 3.
• Sensor M3_2: sensor monitoring Machine 3.
• Actuator M3: actuator relaying the sensing and actuation processing of Machine 3.
• Power Meter_M3: power meter reading electrical power of Machine 3.
• Product (P): final product of the whole production line (output of Machine 3).
• Image Vision Product: camera implementing image vision algorithms for detecting defects of in the final product;
• Big Data: internal big data collecting all production data (data from the whole production line).
• AI Edge Computing: edge computing nodes processing all data collected into the internal big data system by AI algorithms; the AI algorithms are optimized by quantum calculus (Quantum Computer).
• Quantum Computer: the quantum computer processing data collected in Big Data and External Big Data.
• External Big Data: dataset collected from the cloud by other third parties related to this specific production (external backend systems).
• Actuator Engine L1: engine synchronizing all the actuators: Actuator M1, Actuator R1, Actuator M2, Actuator M3 (with synchronization supported by the AI algorithms).

The complex network of Figure 3b is characterized by the parameters listed in

Table 6. Summary statistics concerning the parameters of the complex network of Figure 3b.

Parameter	Value
Number of nodes	28
Number of edges	51
Avg. number of neighbors	3.643
Network diameter	5
Network radius	3
Characteristic path length	2.370
Clustering coefficient	0.215
Network density	0.135
Network heterogeneity	0.645
Network centralization	0.413
Analysis time t (s)	0.038

.

The complexity of the operations of the network of Figure 3b can be simplified by the workflow of Figure 4, which sketches all the operations of the simulated Industry 5.0 framework and facilitates comprehension of the information flux trajectories. The workflow is sketched by a standard BPMN (standard ISO/IEC 19510:2013 [155]), typically adopted for process mapping in industrial applications such as predictive maintenance [156] and security [157]. The BPMN workflow of the analyzed model is structured into three pools (representing the three levels indicated in Figure 3a), and highlighted in red are the data processes of the Industry 5.0 level, which upgraded production processes. The BPMNs model integrating AI decision-making procedures are named BPMN Process Mining (PM) models. The adopted BPMN tool was illustrated with the open-source program, draw.io [158].

Figure 5 illustrates the result of the simulation of the complex Industry 5.0 network of Figure 3b: the calculation of the diffusion heat parameters of the whole information flux is summarized by the heat map in Figure 5. In this map, it is possible to observe that the elements that majorly upgrade the industrial production processes are: Big Data, Actuator Engine L1, Ai Edge Computing, and Quantum Computer, which are all the elements highlighted in red in the BPMN workflow of Figure 4, which confirms these nodes belong to the Industry 5.0 level (upgrading level). The simulation was performed by the open source Cytoscape tool [159].

The BPMN approach is then adopted to “explode” in detail the processes involving robot/machine control, predictive maintenance, and quality assessment. By hypothesizing a production line constituted by a single robot that handles a workpiece that will be processed successively by a machine (simplification of the production line layout of Figure 3a), the BPMN workflow of Figure 6 illustrates the interaction between the three processes that integrate the AI facilities (highlighted in red). The main process is represented by the first pool, which describes the robot and machine control flux performed by sensors, and actuation actions performed by the AI by adjusting by according to feedback systems from the robot and machine parameters during the continuous production. Digital production data are transferred to a data analysis engine able to activate predictive maintenance interventions by adopting AI based decisions (AI predicting product defects and machine failure) and processing simultaneously the electrical power data of machines. An anomalous electric load could represent a machine start failure. The data, stored in a big data system, are also processed for quality assessment processes. Also, in this case, the AI algorithms could improve the whole production process by using its quality prediction.

The proposed approach is based on the Industry 5.0 framework design and simulation. The framework conceptualization was extracted by analyzing the state of the art concerning the specific application field and technologies associated to an industry in the high-tech manufacturing sector. The limitations of the framework are mainly in the interoperability between all sub-processes.

Table 7. Framework limitations and perspectives to overcome these limitations.

Limitations	Perspectives
Use of a unique informatics infrastructure integrating all data protocols (different technologies use different protocols and infrastructures).	Design of ESB-based information networks integrating all data protocols and managing synchronization.
Synchronization of all robots and machines (the different sampling times of sensors could generate problems around the efficiency of synchronization of all robots and machines of a production line).	Development of codes managing the sampling time and time machines according to sensor technologies and production velocity.
Computational cost for real-time data processing (a large number of sensors provide a big quantity of data to be processed at the same time, thus incurring a high computational cost).	Cloud computing could be inappropriate for the real time data processing and big data analytics, but an edge computing approach could be a good alternative.
The gap between the velocities of production machines and high-velocity optical system detecting data.	Production machines that integrate all opto-electronic systems (fully optical sensing and actuation systems).
Efficiency of quantum computing (standalone qubit processor could be not appropriate for the data processing of industrial datasets).	Combined use of AI algorithms executed by edge computing methods and a quantum computer optimizing AI hyperparameters.

lists some important limitations and the associated perspectives as possible solutions.

5.2. Examples of Applications Matching with Industry 5.0 Framework

Table 8. Examples of application fields matching with the Industry 5.0 framework.

Application Fields	Device Typology	Industry 5.0 Framework
Telemedicine	Platform based on software enabling communication between doctors and patients	The telemedicine platform integrates different datasets performing data fusion operations, improving the homecare process.
Medical diagnostics	Wearable systems and medical images for diagnostics	An AI-based Point of Care (POC) platform computes different variables to provide a diagnosis with a low probabilistic error (high accuracy and precision).
Biomedicine	Molecular microscope	Systems of microscopes linked to provide in new pharmacological solutions in short times (precision medicine).
Nanomedicine	Biocompatible nanoparticles	Adoption of innovative automated tools focused on the execution of measurements in nanomedicine.
Smart Cities	Sensors detecting electrical power and environmental pollution	Integrated sensor complex systems automatically enable public management procedures.
Telecommunications	Wireless systems	A reconfigurable wireless network improves transmission security between different nodes (AI-based algorithms define network configuration).
Industrial Measurements Protocols	Industrial IoT devices	Adoption of innovative automated tools focused on the execution of measurements of production quality using AI automation.
Nanoelectronics	Smart circuits	Arrays of driven microcircuits providing punctual data on a micrometer scale (very accurate measurements).
Automotive	Sensors improving driving security	An AI system linked to the car motion control improves security.
Agriculture	Field sensors detecting hydric stress	Data fusion of spectral data and field data automate fertigation processes.
High Tech Manufacturing	Optical and laser-based devices	Production lines controlled by a unique platform which synchronizes production machines (see example in Section 5.1).
Renewable energy	Sensors reading electric power	An AI-based network that correctly balances the energy in grid connections.
Industrial Robotics	Sensors enabling control and actuation	AI-supervised and -unsupervised algorithms could improve robotic control and synchronization.
Logistics (services)	GPS sensors	AI based platform predicts fleet maintenance and security by optimizing logistics services (efficiency in transportation times, decrease of fuel consumption, flux optimization based on predictions, etc.).
Learning	E-learning platforms	AI based E-learning platform automatically optimizes the learning process.

indicates some examples of potential application fields and possible solutions for integrating the associated technologies into an Industry 5.0 scenario.

6. Conclusions

The goal of this paper was to provide an overview of possible technologies and approaches that could potentially be adopted to construct an advanced production framework which included the main important variables and topics found in the cited literature. Some technologies analyzed in the state of the art were selected to construct an Industry 5.0 framework, while defining possible variables used to model and simulate a high-tech manufacturing layout. Specifically, we defined a large number of variables that allowed us to design a complex Industry 5.0 production layout by creating a manufacturing production line composed of three machines and a robot. A simulation of the diffusion heat parameter has defined the information flux of the whole complex network, and we highlighted some innovative elements, such as edge computing, quantum computing, big data and AI data processing, which improved production processes and quality. Finally, the paper discussed an approach to model Industry 5.0 processes by means of graphs constructed using nodes and edges and BPMN workflows, and described in detail the interaction between these processes. The processes highlighted in the Industry 5.0 model are predictive maintenance, machine control and actuation, and quality assessment. The selected literature is useful for constructing industrial research projects. The discussed models could support scientists in finding possible solutions for the design of Industry 5.0 frameworks using some tools suitable for designing a complex network and defining information fluxes of the main processes, thereby improving production.