Research on Diagnosis and Assessment Processes and Methods for Existing Residential Buildings Based on Intelligent Assistance Models

Abstract

As renovating existing residential buildings shifts towards more detailed methodologies, conducting comprehensive diagnostic assessments before renovation is crucial for achieving successful outcomes. This research introduces an innovative large-scale diagnostic assessment method for existing residential buildings, addressing the inefficiencies, redundancies, and subjective biases present in traditional diagnostic processes through intelligent assistance modeling. The proposed method focuses on five key elements: construction year, exterior walls, windows, balconies, and shading devices, categorizing assessment levels into four grades (A, B, C, D) based on varying renovation intensities. Evaluation criteria are established for service life, thermal performance, degradation, and aesthetic quality. An intelligent assistance model, constructed using training datasets, enables rapid large-scale assessments, significantly reducing the evaluation time while maintaining an accuracy rate of over 95%. Empirical testing on residential buildings in Shenzhen confirmed the model’s effectiveness, demonstrating its superior accuracy and efficiency compared to traditional methods. A weighted analysis revealed that the impact of each factor on the building’s condition was as follows: exterior wall thickness (0.38), exterior finish deterioration (0.35), window deterioration (0.29), balcony deterioration (0.28), and construction year (0.23). Additionally, an interactive software application integrating the intelligent assessment model was developed. This study employed an interdisciplinary approach, combining machine learning and big data, offering a new perspective on building assessment and providing significant reference value for future residential renovation and urban planning.

1. Introduction

In response to the widespread issue of substandard existing residential buildings, China’s “14th Five-Year Plan” has delineated requirements for quality enhancement and low-carbon renovation at the national level. In recent years, China has witnessed a multifaceted approach to vigorously advancing the renovation of existing residential areas, achieving significant progress in standards, policies, scientific projects, and technological dissemination [1]. However, statistical data from the “China Building Energy Efficiency Annual Development Research Report 2023” reveal that while China’s building area ranges from 400 to 600 billion square meters, urban residential energy consumption ranks first due to the growth in both scale and average energy intensity [2]. In response to the widespread issue of poor quality in existing residential buildings, the government has taken multiple measures in recent years to vigorously promote the renovation of old neighborhoods, leading to ongoing research and practice in this area. Existing research experience indicates that the complete process of renovating residential systems includes preliminary investigation, diagnostic evaluation, decision-making, implementation, and post-renovation follow-up. Among these, diagnostic evaluation is the foundation and key to the success of the renovation. Conducting surveys and diagnostics on a large number of existing residences requires not only increased precision and objectivity but also enhanced efficiency [3]. This presents both new opportunities and challenges with the integration of information and intelligent technologies.

The quality of residential living is determined by multiple factors, including functionality, performance, adaptability to change, and maintenance strategies. International initiatives such as the United Nations’ “2030 Agenda for Sustainable Development”, the American Institute of Architects’ “Architecture 2030”, and the European Union’s “Renovation Wave” emphasize the enhancement of safety, functionality, infrastructure, and environmental quality in residential buildings. Equally important are user perceptions related to comfort, well-being, aesthetic value, and overall quality [4]. The lack of comprehensive data on the current state of large-scale residential areas makes predicting housing performance a challenging task, often relying heavily on subjective factors such as stakeholder experience and judgment [5,6]. Developed countries initiated research on diagnostic evaluation of existing residential buildings earlier, focusing on thorough preliminary investigations to establish comprehensive diagnostic and assessment databases and operational procedures for proposing rational renovation schemes [7,8]. Previous studies have compared diagnostic assessments, participation methods, existing issues, auxiliary tools, and sustainable technologies in residential renovations across various case studies, highlighting the critical role of status diagnosis and stakeholder involvement [9,10]. A two-stage diagnostic approach for existing residences has been proposed, consisting of an initial diagnosis (survey organization, questionnaires, visual inspection) and a secondary diagnosis (physical investigation and experimentation) [11]. The research further revealed that only 17% of maintenance interventions were based on the physical deterioration of buildings, while 44% were driven by subjective criteria [12]. Additionally, two key aspects affecting residential buildings—physical deterioration and aesthetic concerns—were explored, indicating that changes in aesthetic preferences often influence renovation decisions [13,14]. Generally, building components are replaced before their service life ends, highlighting a maintenance approach based on obsolescence rather than aging [15,16].

The construction industry is currently entering the 4.0 era, characterized by intelligent technologies at its core, where smart construction techniques are being integrated with research on improving the quality of existing residential buildings. Recent technological advancements, including advanced sensor technologies, big data analytics, and artificial intelligence methods, enable real-time acquisition and comprehensive analysis of building and environmental data [17]. The critical information elements of existing residential buildings are typically expressed in textual and visual formats. Different machine learning algorithms are employed to process these distinct forms of data. For image classification, convolutional neural networks (CNNs) [18,19,20], decision trees [21,22], and random forest algorithms [23,24] are commonly utilized, while for text classification, deep neural networks (DNNs) [25,26] and K-nearest neighbor (KNN) [27] algorithms are often preferred. However, despite the widespread application of these new technologies in the design, construction, and operation phases of existing residential building renovations, the preliminary planning and diagnostic evaluation stages remain relatively traditional and fragmented.

This research aimed to develop an intelligent assistant model to overcome the limitations of existing methods and enhance the efficiency, accuracy, and objectivity of evaluations. The goal was to create a machine-learning-based intelligent assistant model for the detailed diagnostic evaluation of existing residential buildings. We sought to achieve our objectives by (1) integrating data indicators across multiple dimensions, such as building functionality, performance, adaptability, and aesthetics, to formulate a combined qualitative and quantitative evaluation framework, providing detailed assessment results; (2) utilizing machine learning algorithms to automatically grade the aging level of residential building facades, significantly improving evaluation efficiency; (3) developing a user-friendly interactive software application, enabling non-experts to easily conduct evaluations; furthermore, the model’s accuracy was ensured through dataset training, localization of parameters, expert validation, and self-assessment; (4) conducting broad empirical testing within Shenzhen’s diverse architectural environment, which validated the model’s applicability and accuracy; (5) and making a significant improvement on existing methods. Compared to traditional approaches, the intelligent assistant model demonstrates substantial enhancements in evaluation efficiency, accuracy, and objectivity, which is particularly crucial in the renovation of large-scale residential areas. This innovative approach will provide data-driven decision support for the renovation of existing residential buildings, contributing to a more efficient and objective evaluation process.

The structure of this paper is as follows: Section 2 provides a comprehensive literature review related to the proposed methodology; Section 3 details the methodological framework; Section 4 presents the case study used for validation purposes; and, finally, Section 5 summarizes the key conclusions and outlines future research directions. This structured approach ensures a logical progression from theoretical foundation to practical application, facilitating a clear understanding of the development, implementation, and impact of the intelligent assistant model on the diagnostic evaluation of existing residential buildings.

2. Literature Review

This section provides a critical examination of current research in building renovation assessment and intelligent-assisted models in architecture. Through a thorough analysis of existing methodologies and their limitations, we aim to highlight the research gaps that our study addresses, with a particular focus on large-scale residential buildings in China.

2.1. Evaluation of Building Renovation Strategies and Pre-Assisted Decision-Making

Recent years have witnessed significant developments in the field of building renovation assessment; however, challenges persist, especially in large-scale applications. Jarosław Szulc’s [28] comprehensive study on over 400 buildings in Poland established the crucial groundwork for diagnostic methods in large-panel structures. While groundbreaking, this approach proves resource-intensive and challenging to scale for widespread use, particularly in rapidly developing urban areas like those in China.

Through an extensive literature review, Sanduni Peiris [29] identified building renovation performance assessment as a key research area. This finding underscores the critical need for effective assessment tools, especially in determining the success of renovation projects. Building on this, Pilar Mercader [30] developed a multi-dimensional decision support system for high-rise residential buildings. Their weighted model, incorporating 12 indicators across technical, social, economic, and environmental dimensions, represents a significant step towards holistic assessment. However, its complexity limits its applicability in rapid, large-scale evaluations necessary for China’s urban renewal projects. Anne N. [31] categorized existing decision support tools into six domains: setting sustainability goals, weighting criteria, building diagnostics, generating design alternatives, evaluating performance, and assessing design alternatives. This categorization reveals the multifaceted nature of renovation assessment while highlighting the lack of integrated, user-friendly tools for comprehensive evaluation. In the realm of heritage buildings, Blanca Tejedor [32] applied non-destructive techniques (NDTs) and advanced modeling technologies. While these technical tools aim to provide a deeper understanding of a building’s overall condition, their implementation methods are often complex, time-consuming, and demand high levels of professional expertise from users. Consequently, they are unsuitable for large-scale residential buildings and fail to provide users with a clear understanding of the current housing status or facilitate the development of satisfactory renovation proposals. Moreover, traditional methods struggle with efficiency and scalability when addressing the evaluation needs of large-scale residential areas.

Organizing the diagnosis methods related to the objectives of this study as shown in

Table 1. Diagnostic evaluation studies in developed countries.

Country	Diagnosis Method	Abbreviation	Year	Diagnostic Purpose	Diagnosis Content	Diagnostic Factors
U.K.	Housing Condition Report	HCR	2007	Identify key anomalies and health and safety risks to building elements	Lists the defects and potential risks of the house, with the house conditions ranging as green, orange, and red from best to worst	Electricity, gas and water supply, sewerage, etc.
Portugal	The state of conservation of residential buildings with controlled costs	Custos controlados	2008	Assessment of the degree of degradation of each functional element and the state of conservation of the building	Classify the building condition; determine the building anomaly index, conservation status, and restoration costs	Structure, facade, wall, floor, staircase, doors and windows, etc.
Netherlands	Assessing the state of conservation of buildings	NEN2767	2006	Standardized assessment of the state of conservation of building components and facilities	The building status	Insect pests, roof, doors and windows, basements, exterior facade, asbestos, ventilation, etc.
EU	Energy performance, indoor, environmental quality and renovation	EPIQR	1998	For diagnostic, evaluation, and decision support for a residential building rehabilitation intervention	Building protection status assessment, restoration engineering database, and restoration engineering cost database	Building protection status assessment, restoration engineering database, and restoration engineering cost database
Japan	Residential repair and judgment benchmark	/	2003	To judge the old decay degree of the residential building	Make the simple judgment and the expert judgment on the residence in five parts	Make the simple judgment and the expert judgment on the residence in five parts

helped to highlight recent advances and identify research gaps that need to be addressed. A critical analysis of the literature revealed several key challenges in current assessment methodologies: Complexity and Time Intensity: Most existing tools, while comprehensive, are often too complex and time-consuming for practical application in large-scale urban renewal projects. High Expertise Requirements: The implementation of current assessment methods typically demands a high level of professional expertise, limiting their widespread use. Limited Scalability: Many tools are designed for individual buildings or small-scale projects, rendering them unsuitable for the large-scale residential complexes common in Chinese cities. Contextual Limitations: Assessment systems developed in Western countries often focus on low-rise villas with well-established construction standards, making them difficult to apply directly to China’s large-scale collective housing projects with different social and technical aspects. Lack of Integrated Approaches: While quantitative algorithms have improved, they often fail to incorporate qualitative factors such as resident satisfaction and cultural considerations.

2.2. Status of Artificial Intelligence—Aided Modeling in Architecture

Recent advancements in BIM and AI offer promising avenues for addressing some of these challenges. Rebekka Volk’s [33] comprehensive review highlighted the underutilization of BIM in existing building renovations compared to new constructions, indicating significant potential for growth in this area. Stephen Fai [34] demonstrated the advantages of BIM in cultural heritage documentation, showcasing its superior geometric modeling and visualization capabilities. However, the application of these techniques to large-scale residential complexes remains largely unexplored. Hamidreza Alavi [35] proposed an integrated data model for building condition risk assessment within BIM, addressing interoperability issues. While innovative, this approach still requires adaptation for large-scale residential evaluations. Par Johansson [36] emphasized the critical importance of baseline information surveys and diagnostics in residential renovation projects through an extensive review of the literature and case studies. While these technical tools aim to provide a deeper understanding of building conditions, their implementation often proves complex, time-consuming, and demanding in terms of user expertise. Consequently, they fall short of providing users with clear insights into current housing conditions and satisfactory renovation proposals. Previous studies have explored automated solutions for facade stain diagnosis [37,38,39], but these approaches are predominantly applicable to single buildings and unsuitable for large-scale aging residential complexes in China.

Our study addresses several key research gaps identified in the literature: Scalability: We propose a method suitable for large-scale residential complexes, addressing the limitations of existing tools designed for individual buildings. Efficiency: By leveraging text information classification algorithms over image recognition techniques, our approach offers a more efficient and flexible solution for multi-label, multi-dimensional assessments. Contextual Relevance: Our methodology is specifically tailored to the unique characteristics of Chinese urban housing, addressing the contextual limitations of Western-developed systems. Integration of Qualitative and Quantitative Factors: Unlike purely quantitative approaches, our method incorporates both objective measurements and subjective factors like resident satisfaction. Accessibility: By utilizing deep neural networks (DNNs), our approach reduces the high expertise requirements of current methods, making it more accessible for widespread use. In conclusion, while significant progress has been made in the informatization and intelligentization of residential renewal, there remains a critical need for refined, efficient, and scalable diagnostic evaluation methods. Our study aims to fill this gap by proposing an innovative approach that leverages advanced technologies while addressing the unique challenges of large-scale residential assessment in the Chinese context.

3. Methodology

3.1. Diagnostic Evaluation Process Based on Intelligent Assistive Modeling

In the field of existing residential regeneration, diagnostic evaluation has evolved from primarily qualitative Phase 1.0 to the combined qualitative and quantitative Phase 2.0 (Figure 1). Despite the progress made during this evolution, limitations such as low efficiency, high subjectivity, and insufficient refinement of processes persist. In response, this paper proposes Phase 3.0 of diagnostic evaluation—a process based on an intelligent auxiliary model aimed at fine-tuning management processes to optimize quality, increase efficiency, and reduce costs. It is important to note that this study involved interdisciplinary collaboration with a team specializing in computer engineering. The partner team was responsible solely for the machine learning code development and interactive app creation, while the authors handled the construction of the intelligent evaluation framework, formulation of evaluation rules, and dataset creation.

The diagnostic evaluation process based on an intelligent auxiliary model can be divided into five key stages (Figure 2): fine data collection, multi-source data integration, current condition diagnosis, pre-renovation evaluation, and regeneration design decision-making. In this study, the fine diagnosis encompassed the first three stages, while the intelligent evaluation was central to the pre-renovation evaluation. The stages of data collection and integration leveraged advanced information technologies to supplement traditional survey methods, enhancing the comprehensiveness of information and survey efficiency. The current condition diagnosis stage combined qualitative and quantitative diagnoses to form more accurate diagnostic opinions. The pre-renovation evaluation stage incorporated machine learning technology to enhance the objectivity and efficiency of evaluations. Finally, regeneration design decisions were based on the diagnostic and evaluation results, with evaluation outcomes guiding renovation strategies and diagnostic results informing specific renovation measures. These five stages constituted a comprehensive fine-tuned diagnostic evaluation process, achieving precision through the integration of information flow and data continuity.

In the intelligent assistant model proposed in this paper, Figure 3 illustrates the process of assigning maintenance grades to each building renovation project. The intelligent evaluation focuses on individual residential buildings, encompassing the entire process from diagnostic input to the output of intelligent evaluation grades and design decision references. The framework is divided into three key components: the input phase, the machine phase, and the output phase (Figure 3). In the input phase, the intelligent evaluation system receives multi-level diagnostic information on existing residences, including construction age, structural systems, facade systems, equipment systems, interior systems, and public spaces. The critical task at this stage is to ensure the acquisition of accurate and comprehensive data on the basic characteristics of existing residences, providing reliable data support for subsequent machine-phase evaluations. The machine phase, as the core of intelligent evaluation, utilizes machine learning techniques to learn from expert experiences, such as rules for aging and deterioration grading, standards for renovation strategy classification, renovation methods, priority orders for renovations, and regulatory standards. The output phase primarily generates evaluation grades for various levels of existing residences. The system ranks the renovation priorities according to these grades, aiding in the selection of elements to prioritize under limited funding. More importantly, it outputs a renovation strategy and a toolkit of renovation measures corresponding to each evaluation grade, achieving a one-stop diagnostic evaluation. The facade serves as the external image of a residence, reflecting the occupants’ individuality and influencing the property’s asset value. Therefore, in existing renovation projects, the facade of existing residences often becomes a focal point. This research begins with the intelligent evaluation of residential facades as the first step in developing the intelligent evaluation model. The process aims to serve as an advanced demonstration and visionary model for future practices in the regeneration of existing residences.

3.2. Selection and Definition of Key Elements

Research indicates that urban renewal projects in China, particularly those focusing on facade systems, constitute over 60% of specialized renovation efforts. Given the high urgency for such renovations, this article begins by examining facade systems as a primary example of urban renewal initiatives. In the transformation of existing residential buildings in regions with hot summers and warm winters, the primary focus should be on passive design strategies to reduce air-conditioning energy consumption. Priority should be given to enhancing the thermal insulation and shading performance of exterior windows [40], improving the thermal insulation of west-facing walls, and simultaneously optimizing natural ventilation conditions. Controlling the solar heat gain coefficient of exterior windows can significantly affect the amount of solar radiation entering the interior, thereby influencing the building’s energy consumption. Based on the literature [41,42], five key factors were identified that significantly impact the energy consumption of existing residential buildings in Shenzhen: construction era, exterior wall structure, exterior window type, balcony design, and shading method (Figure 4). These factors were used to develop a hierarchical renovation scheme (

Table 2. Diagnostic indicators for key elements.

Part	Diagnostic Indicators for Key Elements
Building Information	Era of construction	1980–1986, 1987–1993, 1994–2000
	Structure type	Brick–concrete, steel–concrete, frame structure
	Building number of storys	Multi-story (4–6), Mid-rise (7–9), High-rise (10 stories or more)
Wall	Thickness of external walls	120, 180, 240, 300
	Availability of reflective insulation	With reflective heat insulation layer, without reflective heat insulation layer
	Degree of damage and appearance beauty of the exterior finish layer	Finish intact, no deterioration
		The finish is more complete, deteriorated and dirty, affecting appearance beauty
		The finish is partially damaged or cracked, deteriorated and dirty, affecting appearance beauty
		The finish is cracked or peeled off in a large area, which seriously affects the appearance beauty
Window	Form of external windows	Wooden frame window, aluminum alloy window, plastic steel window, broken bridge aluminum window
	Degree of damage and appearance beauty of external windows	The whole window is in good condition and can be opened and closed normally
		The deterioration of external windows affects the appearance beauty, but can be opened and closed normally
		The deterioration of external windows affects the appearance beauty, and a small portion of them cannot be opened and closed normally
		Most of the windows are damaged and cannot be opened or closed normally
Balcony	Form of balcony	Open balcony, closed balcony, semi-closed balcony
	Degree of damage and appearance beauty of balcony	Components more complete, does not affect the aesthetic
		Small part of the component is partially damaged or cracked, deterioration and dirt affecting appearance beauty
		Most of the components are partially damaged or cracked, deteriorated and dirty, affecting appearance beauty
		Most of the components are damaged, seriously affecting appearance beauty
Sunshade	Whether there is a sunshade	No sunshade, fixed sunshade, glass coating, movable sunshade
	Degree of damage and appearance beauty	Components more complete, does not affect the aesthetic
		Small part of the component is partially damaged or cracked, deterioration and dirt affecting appearance beauty
		Most of the components are partially damaged or cracked, deteriorated and dirty, affecting appearance beauty
		Most of the components are damaged, seriously affecting appearance beauty

). Additionally, thermal insulation performance indicators were extracted from the “Design Standard for Energy Efficiency of Residential Buildings in Hot Summer and Warm Winter Regions”.

3.3. Data Collection and Integration

The intelligent model, essential for data recording and visualization on diagnostic evaluation platforms, faces unique challenges with existing buildings due to data capture and lifecycle updates [43]. When original building drawings are available and clear, manual modeling using information modeling software is feasible. For missing or outdated drawings, 3D laser scanning and aerial surveys provide a point cloud model for reverse modeling. This approach integrates real-state data and a deterioration phenomena database, allowing for the overlay of missing information and renovation history, thereby enabling a comprehensive, iterative visualization of residential deterioration, diagnosis, and repair [44] (Figure 5).

The primary functions of BIM are integration and visualization [45], which can be used for multiple purposes and over an extended period. This model is utilized throughout all stages of residential regeneration. In this experiment, Revit was the software employed for BIM modeling. The specific steps are as follows: (a) Perform hierarchical base modeling, dividing the model into structural systems, envelope systems, interior systems, and equipment systems. (b) Input non-geometric attributes, integrate material information of a specific residence, and mark deteriorated areas with corresponding real-world images. (c) Conduct phased modeling, where the multiple maintenance records of a residence are presented in 3D, such as rooftop canopy installations, balcony expansions, elevator additions, and partial structural augmentations. These records are saved in different phases within the Revit model, allowing the display of various stages by selecting the corresponding phase. (d) Create design option models, such as multiple shading options for exterior windows. Selecting different design options will display various effects and automatically generate corresponding cost tables, providing valuable references for subsequent renovation design decisions. While the initial investment may be higher, in the long run, it can significantly improve efficiency, and reduce errors and rework, thereby saving time and costs. BIM can be utilized throughout the entire project life cycle, providing valuable data support for subsequent maintenance, renovation, and management [46]. BIM technology also has a high degree of scalability—even in resource-constrained regions, simplified BIM tools and processes can be adopted. Numerous open-source and low-cost BIM solutions are constantly emerging, making this technology more prevalent and accessible. BIM is not used in isolation but can be combined with other technologies and methods, such as drone surveying and 3D scanning. It can also serve as a central platform to integrate multiple data sources and analytical tools [47].

3.4. Classification of Assessment Levels and Rulemaking

Formulating intelligent evaluation rules for regenerating existing residential buildings encompasses three main tasks: grading criteria definition, indicator selection, and rule establishment. Different building components require specific indicator selections for intelligent evaluation rules. For instance, in evaluating the deterioration of exterior facades of existing residential buildings, both performance and appearance beauty are crucial dimensions impacting facade quality improvement. Based on these two dimensions, this study selected three objective indicators—transformation intensity, service life, and thermal properties—and two subjective indicators—degree of damage and appearance of beauty. From these, seven diagnostic points and their pathological symptoms were identified (Figure 6).

The aging degree of existing residential buildings determines their regeneration design strategies. This study categorized regeneration strategies based on renovation intensity into four types: maintenance, minor repair, medium repair, and major repair. Corresponding renovation methods include cleaning, repair, addition, and replacement. Cleaning can be performed using physical or chemical methods; repairs aim to restore the performance and functionality of deteriorated areas; additions refer to incorporating new elements into existing structures; and replacement involves the complete substitution of the enclosure structure.

For the classification of maintenance management levels of residential buildings, Professor Matsumura [48] from Japan divides them into four levels based on different renovation intensities [49]. Taking the classification of the deterioration degree of exterior facades as an example, since the regeneration target is the exterior facade, specific measures generally include cleaning (wall cleaning, component cleaning, etc.), repair (facade painting, component repair, surface patching, etc.), addition (adding shading components, thermal reflective layers, raising railings, adding air conditioning unit covers, etc.), and replacement (replacing exterior windows, railings, shading devices, and surface finishes). Based on different renovation intensities, the evaluation grades for the deterioration degree of existing residential facades are divided into four levels: A, B, C, and D (

Table 3. Key element diagnostic indicator parameters and levels.

Part	Thermal Properties	Service Life	Degree of Damage	Appearance Beauty	Level of Assessment	Retrofit Intensity	Retrofitting Measures
Wall	- Thickness of external walls - With or without reflective insulation	30–45	- Degree of finish deterioration	-Is it aesthetically pleasing	A1/B1/C1/D1	A/B/C/D	① Simple painting of walls ② Addition of heat-insulating and reflective layer ③ Addition of heat-insulating layer to walls and beautification of façade ④ Addition of high-performance heat-insulating layer to walls
Window	- Exterior window forms	30	- Degree of damage of external windows		A2/B2/C2/D2		① Dilapidated glass replacement ② Small partial replacement of external windows ③ Whole window replacement ④ High-performance windows
Balcony	- Balcony form	/	- Degree of balcony deterioration		A3/B3/C3/D3		① Remodeling of balcony partition wall ② Redesign of balcony elevation ③ Remodeling of interior space ④ Addition of new picket balcony
Sunshade	- Whether there are sunshade elements - Shading form	/	- Degree of external sunshade deterioration		A4/B4/C4/D4		① Glass coating ② Adding sun-shading components ③ Adding high-performance sun-shading components ④ High-performance motorized movable sunshade
Residential Overall	A. Maintenance: ① no repair, ② wall and accessory components’ cleaning, ③ simple wall painting. B. Minor repairs: ① partial damage to the finish or accessory components needs to be repaired, ② additional heat-insulating reflective layer or sun-shading components, ③ small-scale partial replacement or replacement of only one accessory component. C. Medium repair: simultaneous cleaning, repair, addition, and replacement of accessory components in several parts. D. Major repairs: including the complete replacement of the finish layer (except for the structural layer of the wall).

). The renovation intensity increases progressively from grade A to D, with typical cases illustrated in

Table 4. Detailed description of assessment levels.

Reference Indicators	State of Affairs	Level of Assessment
Service life	The reference service life of exterior finishes is 30–45 years, i.e., the 1980–1986 exterior finishes of existing dwellings have exceeded their service lives	Maximum B
	The reference service life of timber-framed windows is 30 years, i.e., timber-framed windows in existing dwellings from 1980 to 1993 have exceeded their life expectancy	Maximum B
Thermal properties	The heat transfer coefficient and thermal inertia of walls above 240 meet the code values even without reflective insulation; 180-thick brick walls with reflective insulation meet the code values	Maximum A
	The 180-thick brick wall has an insulating reflective layer, but if the exterior finish layer is partially or extensively damaged, its thermal properties will not meet the code	Maximum B
	The 120 walls, even with reflective insulation and external shading elements on the windows, still do not meet the energy requirements	Maximum B
	External sunshade components can significantly reduce the building’s energy consumption; therefore, there is no reflective insulation layer of the 180 wall if there is external window installation of external sunshade components	Maximum A
Degree of damage and appearance beauty	Remove finishes and redo	Maximum C
	Wall finishes 180 and above are intact and unobtrusive	Maximum C
	Localized cracking or extensive peeling of the finish, poor substrates, and the addition of reflective insulation require the finish to be removed and redone	Maximum C
	In the event that the thermal properties requirements are not met and the finish layer is partially or extensively damaged, and if the external windows and balconies are severely damaged, a complete replacement of the external wall including the removal and redoing of the finish layer is required (in addition to the structural layer of the wall)	Maximum D
Retrofit intensity	① No need for repair, ② walls and accessory components’ cleaning, ③ simple painting of walls	A
	① Localized damage to finishes or accessory elements requiring repair, ② additional heat-insulating and reflective layers or sun-shading elements, ③ small-scale partial replacement or replacement of only one element (except for finishes)	B
	Simultaneous cleaning, repairs, additions, and replacement of ancillary components in multiple sections	C
	Full replacement including removal and redo of finishes (except structural wall layers)	D

The remaining ambiguous and subjective rules need to be evaluated by experts based on their experience, and ultimately the degree and proportion of importance of the elements of each part will be determined based on the evaluation results.

.

In this paper, we demonstrate the proposed intelligent auxiliary model’s significant advancements in the diagnostic evaluation of residential buildings, particularly in assessing renovation intensity and classifying evaluation grades, compared to traditional methods. Conventional evaluation approaches often rely heavily on experts’ subjective judgments, utilizing standardized scorecards or checklists. While detailed, these methods can lead to inconsistencies in evaluation results due to differences in individual experience and interpretation. Additionally, some traditional methods may focus excessively on quantitative data, overlooking the qualitative aspects of a building’s functionality, aesthetics, and sociocultural value. In contrast, the intelligent auxiliary model adopts an innovative evaluation approach that integrates both qualitative and quantitative assessment techniques. Leveraging machine learning and big data analysis, this model not only processes and analyzes vast amounts of historical and real-time data but also considers the specific conditions and unique needs of each building. This provides a reference for future more precise renovations tailored to individual buildings. When evaluating renovation intensity, the model takes into account the building’s physical characteristics—such as the thickness of external walls, presence of reflective insulation, service life, extent of damage, and aesthetic appeal—and provides a range of renovation suggestions, from simple maintenance to major repairs, based on a comprehensive analysis of these factors. Regarding the classification of evaluation grades, the intelligent auxiliary model employs a refined set of indicators, offering more precise guidance for external wall renovations by subdividing evaluation grades into categories such as A1 to D1. This grading not only considers quantitative data, like thermal performance metrics of the walls, but also incorporates qualitative factors such as aesthetic appeal. This multifaceted approach enables the model to deliver more personalized and dynamic evaluation results, thereby enhancing the decision-making process. The model reduces user barriers by offering more visualized and dynamic evaluation results through a comprehensive assessment approach. This enhancement increases the scientific rigor and participatory nature of the decision-making process.

Overall, the intelligent auxiliary model presents a scientifically robust and comprehensive new method for assessing renovation intensity and classification. Combining the precision of quantitative analysis with the thoroughness of qualitative evaluation, it overcomes the limitations of traditional methods, offering a more effective and reliable tool for the diagnostic evaluation of residential buildings.

3.5. Construction of Intelligent Auxiliary Model and Interaction App Design

The development of the intelligent evaluation algorithm model for existing residential buildings involves seven main steps (Figure 7): data preparation, feature engineering, model construction, model training, model evaluation, tuning and optimization, and model deployment. Taking the development of a DNN algorithm model as an example, the process is divided into two parts: training and prediction.

To construct a comprehensive and representative dataset, this study identified 12 diagnostic points based on those above five key elements, further divided into 42 sub-labels (Table 2). These sub-labels were combined across the 12 evaluation tags, resulting in 362,880 unique combinations, each representing a different scenario of facade deterioration in existing residential buildings, effectively covering the spectrum of potential deterioration cases. Given the vast number of combinations, the integration of machine learning was crucial for model development. Initially, three experts were invited to evaluate each combination according to established criteria rigorously. The number of evaluated combinations for each element was as follows: 2025 for walls, 1017 for windows, 765 for balconies, and 1017 for shading devices. The hierarchical sampling of 1400 combinations representing the overall condition of the residential buildings was then continued for grade assessment and finally calibrated with random data sampling. This base dataset was the initial training material for constructing the model.

In the algorithm development process (Figure 8), an automated data-reading program was established, enabling the algorithm model to automatically access data tables stored in the corresponding computer folders. Upon reading the dataset, the algorithm divides it into a training set and a test set, with the split ratio adjustable as needed. For this experiment, the dataset was divided into an 8:2 ratio, which aids in validating the model during the training process to ensure it generalizes well rather than simply overfitting to the training data. After the dataset is imported into the algorithm model, the computer automatically encodes the text data and trains the model based on the provided dataset. The intelligent evaluation model learns the mathematical relationships between data points through multiple training iterations. Through these steps, the model is refined through manual calibration and extensive data learning and iteration. In this experiment, the K-nearest neighbors (KNN) algorithm and the deep neural network (DNN) algorithm were utilized, achieving evaluation accuracies of 88.8% and 94%, respectively. This experiment represents an effort to build on the foundations of qualitative and quantitative diagnostic assessments, with further refinement of the principles and model still needed.

The core step of the self-check for the training dataset involves using the initial training dataset to train the intelligent evaluation model. The trained model is then used to predict the evaluation results for the same dataset (Figure 9). Inconsistent evaluation results are filtered and compared using Excel, allowing for the correction or optimization of evaluation rules and enhancing the accuracy of the evaluation model.

To further verify the accuracy of the evaluation rules, this study employed SPSS software (version number 25.0) to conduct correlation analysis on the dataset, aiming to determine the correlation between various evaluation factors and the deterioration degree of exterior facades. We used the experiment set “evaluation grade” as the dependent variable and seven evaluation factors, including “construction year” and “wall thickness”, as independent variables, with a dataset of 1400 samples. The residual histogram from the analysis clearly shows a normal distribution (Figure 10), indicating strong sample representativeness and a good model fit.

The Pearson correlation heatmap analysis (Figure 11) revealed that the correlation coefficients of the evaluation factors with the grading level were ranked as follows: wall thickness (0.38) > exterior surface deterioration degree (0.35) > window deterioration degree (0.29) > balcony deterioration degree (0.28) > construction year (0.23) > presence of fixed external shading components (0.10) > presence of reflective insulation layer (0.05). This result shows that wall thickness, exterior surface deterioration degree, windows, balconies, and construction year have significant impacts on facade deterioration grading. This result supports the previous research hypothesis and statistically verifies the reasonableness of the evaluation rules. It is important to note that correlation coefficients only partially reflect the impact of evaluation factors on grading levels and do not directly indicate the importance of a factor for renovation quality improvement. For instance, while the presence of reflective insulation layers and fixed external shading components have a minor impact on grading levels, they significantly improve the thermal properties of facades.

To further enhance the usability of the intelligent evaluation model, this study designed a visualization application software with an interactive interface—the Intelligent Evaluation APP. The core design concept of this app is to simplify the technical operation process through a user-friendly interface, allowing users, including non-experts, to easily conduct assessments of existing residential buildings.

The core of the Intelligent Evaluation APP is driven by the DNN algorithm-based intelligent evaluation model. On this basis, an intuitive and user-friendly interactive interface was developed. The interface features a clear navigation menu and flexible drop-down options, with preset search options for various evaluation factors, enabling users to quickly set parameters for diagnosing the condition of existing residential buildings (Figure 12). For example, if users need to evaluate a building constructed in 1985, they can select the “1980–1992” period from the drop-down menu as the assessment condition. Additionally, the app provides a wealth of case studies and reference materials tailored to users’ needs. The display of actual cases not only deepens users’ understanding of the assessment process but also aids in making more accurate decisions when selecting evaluation factors. Each option is accompanied by a help icon linked to detailed explanations and relevant cases, ensuring that even first-time users can quickly get started and perform effective assessments.

In the visualization app, the Revit software API (version 2022) is utilized to access the pre-existing residential BIM model. By bypassing the relevant element IDs—such as exterior walls, windows, and balconies—through the API, an information window is triggered, displaying detailed assessment data for the selected element. These data include technical parameters extracted from the BIM model, such as material properties and dimensions, which are then displayed on the right side of the app interface. This integration method not only makes the assessment process more intuitive and efficient but also ensures that the assessment data remain synchronized with the BIM model, providing a comprehensive and accurate foundation for subsequent analysis and decision-making. Users can easily switch between different elements, quickly accessing and comparing the condition of various building components, thus offering strong support for the planning and implementing of building renovation projects.

4. Case Study—Residence A in Shenzhen as an Example

Taking Shenzhen, a representative city in hot summer and warm winter regions, as an example, data from the Shenzhen Statistical Yearbook 2021 reveal that, due to its limited urban area and rapid development [50], Shenzhen’s residential construction peaked as early as 2003. This led the city to enter the era of stock housing prematurely and forced it into high-density residential development. A large number of old multi-story brick–concrete residences built between 1980 and 2000 are concentrated in the “inside” area of Shenzhen, where surrounding facilities are well-developed, public services are comprehensive, and transportation is convenient, but energy consumption is high. Therefore, the energy renovation of old residences is a crucial component in supporting China’s low-carbon development strategy [51].

4.1. Selection of Research Subjects

Preliminary research on 93 existing residential buildings in Shenzhen led to the classification and statistical analysis of building forms (

Table 5. Statistics of existing residential survey.

Traffic Combination	Ratio of Lifts to Households	Building Type	Total		Structure Type
Staircase	One staircase, two households		33		Brick–concrete, frame, special-shaped column frame
	One staircase, three households		12		Brick–concrete, frame
	One staircase, four households		5		Brick–concrete
Corridor	Long corridor,		7	10	Brick–concrete
	long inner corridor		3		Brick–concrete
Detached	One staircase, four households		16		Brick–concrete

), with one-story, two-house brick–concrete structures being the most prevalent (35.48%). Data from the Shenzhen Municipal Bureau of Statistics show that before 2000, the average number of floors in Shenzhen’s residential buildings was 6–10 floors. Based on this, a typical southern residential unit prototype was selected as the “basic unit” for study. Basic unit data are shown in

Table 6. Information on the study population.

Residential A
- Project name: Residence A - Year of construction: 1988 - Floor plan: one-staircase, four-family - Structure type: Brick–concrete	- Number of storys: 6 - Residential floor height: 2800 mm - Roof type: flat roof - With or without elevator: None

, and enclosure structure construction and parameters are detailed in

Table 7. Enclosure construction.

Enclosure System	Form of Construction	Heat Transfer Coefficient W/(m2-K)	Standardized Indicators
Wall	10 mm cement mortar + 240 brick wall + 10 mm cement mortar	2.41	2.04
Window	Timber-framed, single-glazed windows (original)	4.70	2.50
Roof	10 mm cement mortar + 150 reinforced concrete + 10 mm cement mortar	1.14	0.95
Sunshade	None	-	0.7

.

4.2. Diagnosis of the Current Status of the Case

The residential building constructed in 1996, with less than 30 years of use, and previously maintained with exterior wall painting in 2014, shows relatively good maintenance (

Table 8. Diagnostic conclusions of the current situation.

Reference Indicators	Diagnostic Points	Diagnostic Conclusions
Service life	Age of construction	Built in the period 1992–2000, the external finishes are not past their useful life
Thermal properties	Wall thickness	The wall thickness is 180 mm
	Reflective insulation	There is no reflective heat insulation layer in the exterior finish layer
	Exterior sun-shading elements	The external windows on the north and south elevations are in the form of convex windows, which also have the effect of shading; the east and west elevations have fewer building openings, and all of them have external sun-shading components
Degree of breakage and appearance beauty	Exterior finishes	The exterior finish layer is intact, and deterioration affects the appearance beauty
	Exterior sun-shading elements	Most of the external sun-shading elements are slightly damaged
	External windows	The overall external windows are intact and can be opened and closed normally
	Balcony	The balcony’s railing handrails and other components are more complete and do not affect the appearance beauty; however, the form of spontaneous remodeling is not uniform, which affects the degree of tidiness

). The wall thickness is 180 mm, with standard plastering and no insulation layer, resulting in inadequate thermal properties. However, the north and south facades feature bay windows, while the east and west facades have gables with few openings, effectively reducing direct sunlight exposure. Overall, the building meets basic requirements. The exterior finishing layer shows small cracks and localized dirt but remains largely intact and aesthetically pleasing. The windows are in good condition and function properly, and some have been replaced by residents. Most balconies have been converted to enclosed types, and railings and other components are complete and safe. Overall, the building performs adequately in terms of service life, thermal properties, and appearance beauty, but there is room for improvement (Figure 13). Future renovations could consider using new insulation materials to enhance thermal properties while preserving the building’s appearance [52,53,54]. This comprehensive diagnosis provides a solid information foundation for subsequent intelligent evaluations and renovations, aiding in the development of targeted renovation plans to improve overall building performance and livability.

4.3. Intelligent Assessment Results and Analysis

For the intelligent evaluation of residential building A, the foundational information is derived from detailed diagnostic results. The fine diagnosis relies primarily on manual completion, while the evaluation phase utilizes the intelligent model to perform a comprehensive assessment based on these detailed diagnostics. During the facade fine diagnosis process, detailed diagnostic conclusions were obtained for fourteen key factors of the residential building: construction year, wall thickness, reflective insulation layer, exterior finishing layer, external shading components, windows, and balconies. These diagnostic conclusions are directly inputted into the interactive interface of the intelligent evaluation model. By selecting the corresponding options, this information is fed into the model. The intelligent evaluation model then performs efficient analysis and computation based on the input data, ultimately providing a rapid assessment of the facade deterioration level. In the current experiment, the facade deterioration assessment level for the residential building was output as B (Figure 14).

The assessment of the facade deterioration level for the residential building shows that both the manual and intelligent evaluations result in a B grade, confirming the feasibility and accuracy of the intelligent evaluation model. The model, based on machine learning, enhances prediction capability through multiple training iterations and achieves end-to-end data matching, significantly improving evaluation efficiency. However, the accuracy of the model is influenced by the quality of the evaluation rules and the accuracy of the training data. To continuously optimize the model, a feedback mechanism should be established to incorporate new data obtained from actual renovation processes. This intelligent evaluation method not only greatly reduces evaluation time but also provides a more efficient and accurate decision-support tool for facade regeneration of existing residential buildings, contributing to the advancement of intelligence in the building renovation field [55,56].

4.4. Exploration of Regeneration Design Decisions

The assessment grade provides a basis for selecting regeneration strategies, and the diagnostic results guide the choice of specific measures [57]. The residential building in question, with a facade assessment grade of B, indicates that the current condition of the building is relatively good and it remains functional. However, there is still room for improvement in appearance beauty and thermal comfort. Given that the building has already had an elevator installed, a large-scale overall renovation is not suitable. Instead, a localized quality enhancement approach, such as minor repairs, is more appropriate [58]. The building is located within a well-facilitated residential area with convenient transportation and is currently partly rented and partly owner-occupied, meaning the requirements for facade specification are not stringent. For facade regeneration measures, a comprehensive upgrade is recommended [59], which can enhance both indoor thermal comfort and appearance beauty (

Table 9. Improvement-type retrofit measures.

Grade	Specific Parts	Remodeling Measures	Specifics
Well-rounded	Exterior finishes	Tile finishes	Materials: through-body tiles, glass tiles, glazed tiles
		Energy-saving stucco	Material: reflective coating, heat-insulating coating
	Balcony	Installation of sun-shading elements	Style: horizontal, vertical, covered
		Enclosing balconies	Material: wood, aluminum alloy
	Exterior window	Replacement of energy-saving windows and doors	Style: fixed, adjustable
		Installation of sun-shading elements	Style: half-wall glass window, floor-to-ceiling glass window, glass window with guardrail

).

For the “minor repair” quality enhancement strategy, the focus is on improving the exterior finishing, external shading components, and the cleanliness of balconies [60]. Specific actions include washing and cleaning the exterior finishing, repairing cracks, and applying reflective insulation coatings; repairing and uniformly painting external shading components, and installing new components where they are missing; enhancing balcony cleanliness, and uniformly replacing security grilles or installing glass curtain walls on enclosed balconies. The renovation design takes into account the building’s duplex characteristics, with facade treatment standardized to two layers, requiring the removal of all security grilles and some spontaneously modified balcony components [61,62]. This renovation plan is more cost-effective and quicker than a large-scale overhaul, providing significant improvements in facade appearance and thermal comfort, while addressing both resident needs and facade cleanliness (

Table 10. Design decision exploration.

Retrofitting Strategies	Key Elements	Rehabilitation Measures	Pre-Remodeling	After Remodeling
Minor repairs for localized quality improvement	Wall	Scrubbing and cleaning to remove soiled parts, and repairing small localized cracks. Adding a reflective heat insulation layer on the east, west and south facades.
	Balcony	The balconies are unified to replace the steel wire security net with good invisibility effect, and at the same time, according to the current situation, some layers of balconies are transformed into closed balconies, and some layers of balconies are restored to their original form.
	Exterior windows	Installation of external sunshade components on the external windows that are not installed locally, and integration of external sunshade components with air-conditioning units, reduce the influence of air-conditioning units on the neatness of the façade.

).

5. Conclusions and Future Outlook

This research has achieved significant results on both theoretical and practical levels, successfully constructing the Diagnostic Evaluation 3.0 framework and validating the effectiveness of a refined diagnostic evaluation method through quantitative analysis.

Under the guidance of the intelligent evaluation framework, the development logic and process of the intelligent evaluation model were conceived using the intelligent evaluation of existing residential facades as a case study. The development logic of the intelligent evaluation system follows a systematic process, starting from the formulation of evaluation rules, collecting datasets through manual evaluation, developing the intelligent evaluation model, and finally, applying the model for predictions and validating and refining the evaluation rules. The intelligent auxiliary model demonstrated an accuracy of over 95% in evaluating the external walls of residential buildings, significantly outperforming traditional methods.

The evaluation rules were formulated by comprehensively selecting two key dimensions that impact the enhancement of facade quality: performance and aesthetics. Based on these two dimensions, three objective indicators—renovation intensity, service life, and thermal performance—along with two subjective indicators—damage level and aesthetics—were chosen. Key diagnostic indicator parameters and classification levels were then established accordingly. Through the application of machine learning algorithms, the model achieved automatic classification of the aging degree of residential building facades, enabling large-scale, rapid assessments of multiple buildings and reducing the evaluation time from several weeks to just a few hours.

The model performed a comprehensive analysis of multiple key indicators of buildings, calculating the weight of each parameter after self-checking and adjustment. The results showed the following weight distribution: external wall thickness (0.38) > deterioration of exterior finish (0.35) > window deterioration (0.29) > balcony deterioration (0.28) > construction age (0.23) > presence of external shading components (0.10) > presence of reflective insulation layer (0.05). This outcome indicates that factors such as wall thickness, deterioration of exterior finish, windows, balconies, and construction age have a significant impact on the facade deterioration rating. The results satisfy the initial research hypotheses, statistically validating the rationale behind the evaluation criteria and providing robust data support for the regeneration design of residential buildings.

In terms of innovation, this research has the following key features: It is the first to apply machine learning and big data analysis technologies to the diagnostic evaluation of residential buildings. This interdisciplinary methodological innovation introduces a new perspective to the field of building assessment. The intelligent auxiliary model we developed not only enhances the efficiency and accuracy of evaluations but also offers more comprehensive and objective results by integrating both qualitative and quantitative data. Our study provides theoretical foundations and technical support for the low-carbon renovation of residential buildings, aligning with the current trends in sustainable development.

In conclusion, this research has yielded significant outcomes on the quantitative level and demonstrated methodological innovation. We believe these achievements will provide critical references and guidance for future residential building renovation and urban renewal projects.

Future research will continue to optimize the intelligent auxiliary model, exploring the potential of multidimensional data analysis and further integrating user participation and feedback mechanisms to enhance the practicality and accuracy of the evaluation tool. Through these efforts, we anticipate that the intelligent auxiliary model will become an indispensable tool in residential building renovation and urban renewal projects, providing theoretical and technical support to improve the quality of urban residential buildings in China.