An Intelligent Redesign Method for Used Products Based on Digital Twin

Abstract

Remanufacturing used products is an important technological approach in sustainable development and circular economy. Meanwhile, redesign is the key component of remanufacturing, as it can innovate the function and structure of used products. However, due to the uncertain quality, variety, and small batches of the returned used products for remanufacturing, it is difficult to generate a sound redesign scheme to satisfy the customer demand quickly and dynamically. In addition, it is unpredictable whether the redesign scheme is suitable for the remanufacturing processes, which may lead to additional remanufacturing costs. In order to improve the efficiency of design and obtain the optimal design scheme, it is necessary to use intelligent technology to quickly generate and optimize the redesign scheme. To address this, an intelligent redesign method for used products based on the digital twin is proposed in this paper. Digital twin (DT) technology can connect the physical world with the virtual world and use the virtual model to simulate the redesign process, which is conducive to the dynamic adjustment and optimization of the redesign scheme. Firstly, the redesign process framework is constructed based on the axiomatic design (AD) method, and the redesign features of the used products are analyzed to determine the redesign problems. Then, based on the connotation of a digital twin, an intelligent redesign framework is constructed, which provides detailed guidance for building the digital-twin-driven redesign system. Henceforth, the application of the redesign process based on a digital twin is discussed, a technical approach of the digital-twin-driven redesign is proposed, and data processing methods, such as data cleaning, data integration, and data analysis, are used to realize the redesign scheme decision. Finally, the feasibility of this method is verified by the redesign of a used clutch.

1. Introduction

Remanufacturing, a process of returning used products (often called core) to an at least as-good-as-new condition, is increasingly recognized as an important part of the circular economy [1,2,3]. Moreover, redesign is the key component of remanufacturing which can innovate the function and structure of remanufactured products with the goal of maximizing the reuse of used products and parts and materials and added values. Hence, redesign can directly affect remanufacturing efficiency, quality, cost, resource efficiency, and environmental pollution [4]. Therefore, redesign is of great significance to the smooth implementation of product remanufacturing, which has attracted widespread attention in the field of sustainability.

Although redesign has huge advantages, the redesign process is complex and tedious, involving demand analysis, function planning, structural parameters solution, remanufacturing process scheme formulation, and redesign scheme evaluation. In addition, the quality of the cores is highly uncertain, and customer demand is also diverse, which means that the redesign scheme needs to be updated in real time to cater for customer demands and product quality. These make the redesign process very cumbersome.

Nowadays, many redesign methods have been proposed. Du [5] proposed a reuse-oriented redesign method of used products based on axiomatic design theory and Quality Function Development (QFD) to optimize and standardize the design process. Jiang [6] proposed an improved programming equation based on fuzzy nonlinear regression which can analyze the fuzzy relationship between customer requirements (CRs) and redesign parameters, and the fuzzy correlations among redesign parameters. Cao [7] presented a reuse-oriented redesign method of used machine tools based on the matter–element theory, which can organically combine the redesign schemes of each subfunction to construct the redesign scheme of the whole machine tool. Cong [8] proposed a quantitative method of value recovery, which can identify and eliminate disassembly bottlenecks by improving the existing design to reduce the time and cost required for parts and material recovery. The above-mentioned literature shows that many scholars conduct research on redesign methods from the perspectives of value reuse, product function, and customer demand. Undoubtedly, the proposed methods greatly contributed to the research of redesign; however, these approaches are still manual, are labor-intensive and low efficient, and cannot guarantee design quality.

Compared to traditional design, the quality of used products for remanufacturing has a high degree of uncertainty, and they are generally small batches of various types; thus, it is difficult to accurately extract redesign targets and rapidly generate corresponding redesign schemes, thereby affecting design quality and design efficiency. There is an urgent need to apply intelligent technology for improving the quality and efficiency of the redesign. Currently, many intelligent technologies have been applied to complex design; for instance, Hao [9] proposed a method of combining genetic algorithm and case-based reasoning for complex product design, which can accelerate the speed of product development and improve efficiency. Li [10] proposed an intelligent acquisition method based on k-means clustering and a feature selection algorithm based on mutual information, which can allow designers to quickly search for corresponding product genes for improving conceptual design efficiency when performing similar functional design tasks. Liu [11] proposed a product-level parameterized management and rapid design method based on the idea of holographic modeling to realize the product-level parameterized design. Hao [12] proposed a systematic method for the intelligent design of mechanical components using methods, such as knowledge acquisition, knowledge representation, and knowledge reuse, which can improve design efficiency and shorten design cycles. The above literature applied genetic algorithms, case-based reasoning, clustering algorithms, and other methods for intelligent design. There is no doubt that these methods can improve design accuracy and efficiency. However, these methods mainly use the original design knowledge and data to solve new design problems, which may not be able to solve the new design schemes to meet design requirements, or the solved design schemes may not be optimal. Meanwhile, the new redesign scheme may have errors and may be in conflict with the design of new products, and the above methods cannot accurately check the errors and adjust. Furthermore, the redesign of used products is a multiconstraint and multiobjective design process and needs to be adjusted in real time based on the individual situation of used products. In addition, it is necessary to consider whether the existing remanufacturing technology, equipment, and workshop environment can realize the redesign scheme. Unfortunately, none of the current methods have attempted to solve these problems.

Hence, it is necessary to apply intelligent technologies to generate the redesign scheme of used products, which can reduce the trial-and-error cost of the redesign and adjust the redesign scheme dynamically in real time under the comprehensive considerations of physical workshop environment, used product quality, and design constraints. Digital twin (DT), acting as a mirror of the real world, provides a means of simulating, predicting, and optimizing manufacturing system and process [13] and has been widely used in various fields, such as product assembly [14,15], 3D printing [16], fault diagnosis [17], and remanufacturing paradigm [18]. As a key to two-way mapping, dynamic interaction, and real-time connection between virtual and real environments, DT can map the attributes, structure, state, performance, function, and behavior of physical entities to the virtual world [19]. Many scholars have applied DT to the design field. For instance, Liu [20] proposed a digital dual-drive method for the rapid and personalized design of automated flow shop manufacturing systems, in which virtual models are utilized to represent real manufacturing systems. Zhang [21] proposed a rapid and personalized design method for insulating glass production lines based on DT, which is employed to provide engineering analysis capabilities, supporting system design, and solution evaluation decisions. Schleich [22] presented a comprehensive reference model based on the concept of skin model shape, which serves as a DT of physical products in the design and manufacturing processes. Tao [23] proposed a comprehensive design framework based on the DT to connect physical products and virtual products, which is expected to be most useful for iterative redesigns of existing products. It is clear from the above literature that DT can be utilized to establish the connection between the virtual space and the physical space, and the virtual model can truly map the physical model and simulate the whole design process to enable the visualization and dynamic adjustment of the design process. Meanwhile, the DT model can feed back and iteratively optimize the physical design process based on twin data, thereby obtaining the optimal design scheme and reducing the trial-and-error cost of the manufacturing processes. However, despite the great success of DT in design and manufacturing, to date, few efforts have been devoted to exploring the application of DT in the redesign of used products.

In order to accurately generate redesign schemes and improve design efficiency, this paper proposes an intelligent redesign method for used products based on the digital twin, which can provide a new technological model for used product redesign and promote the development of the circular economy. This method consists of “redesign process framework and features analysis”, “intelligent redesign framework based on digital twin”, and “application of redesign process based on digital twin”, based on which the redesign schemes could be rapidly generated and dynamically adjusted. The proposed research is composed of four steps: (1) The DT is integrated with the redesign process of used products based on AD, and a specific implementation framework for the redesign of used products based on the DT is proposed. (2) Under incomplete information, reverse engineering technology is used to construct a virtual model of used products, and the incomplete parts are fitted to restore the original shape of the used products. (3) Case-Based Reasoning (CBR) technology is used to intelligently generate the structural design parameters of used products; meanwhile, SolidWorks software is used to analyze the assembly and dimensional error of parts model, and ANSYS software is used to analyze the performance of remanufactured products. (4) Back Propagation Neural Network (BPNN) is used to establish the prediction model of remanufacturing process parameters, which takes the used product type, failure features, and customer demand information as input and remanufacturing process parameters as output. To verify the feasibility of the method, this method has been applied to the redesign of a used clutch.

2. Redesign Process Framework and Feature Analysis

2.1. Redesign Process Framework of Used Products

The redesign process of used products includes the design of remanufacturing repair scheme and innovative scheme design. The former is to use high technology to restore the original performance of the product, mainly for remanufacturing process design, while the latter is an innovative design based on the structure and function of the original product to obtain new functions and structures. For standardizing the redesign process, the axiomatic design (AD) method is used to redesign the used products, which is mainly composed of customer domain, function domain, physical domain, and process domain. Axiomatic design is a systematic design method used to execute the design process, support communication and coordination in the context of team design, and provide basic principles for analysis, comparison, and solution selection, so that design rationale can be generalized and made accessible to all the team members [24]; the redesign process framework is as shown in Figure 1.

2.2. Analysis of Features of the Redesign Process

The redesign of used products for remanufacturing is different from traditional design, because redesign requires the collection and analysis of used product information, and the process of redesign is constrained by factors such as the original product structure and materials. In addition, redesign not only just restores the appearance and performance of the original product but also maximizes resource reuse and carries out innovative redesign based on customer demand. Therefore, the redesign of the used product includes both the remanufacturing scheme design and the innovative design. The redesign process mainly has the following features.

2.2.1. The Uncertainty of Used Product Information

Since the used products have been in service for many years, the operating conditions of the products during the service are not available, and the damage and performance of each part are also unknown, which leads to uncertainty in the quality of the used products. The problem that used products lack much information may cause uncertainty in the redesign processes and results; meanwhile, this could lead to uncertainties in the remanufacturing process, which may be fed back to the design scheme. As a result, the structural parameters, function planning, and process design parameters of the design scheme cannot be determined in advance.

2.2.2. Customization of the Product

The remanufacturing of used products should not only restore the original performance but also meet the diverse and personalized demands of customers, including the appearance, structure, and function of the products. To cater to individual customer demands, customers are offered opportunities to participate in the redesign process. Through continuous communication between designers and customers, a customized redesign scheme can be obtained to meet the individual customer demand.

2.2.3. Integration of Redesign Process

Product remanufacturing processes are highly individualized. Due to the differences in product quality and customer demand, even for the same type of product, remanufacturing processes may be different. In addition, some redesign schemes can be realized through remanufacturing, while others cannot be implemented beyond the current technical level, equipment capacity, and processing cost. In view of the above problems, the redesign process of the used products needs to be integrated with remanufacturing processes, emphasizing the real-time interactions and feedback between redesign and remanufacturing to optimize the redesign scheme and improve the viability of the redesign scheme.

Due to the uncertainty, customization, and integration in the redesign process, the redesign process is very complicated, and designers often fail to quickly and adaptively generate appropriate redesign schemes for used products based on customer demand. Meanwhile, it is unknown whether the redesign parameters can be successfully implemented in the remanufacturing process. Therefore, it is necessary to use appropriate intelligent technology to generate a redesign scheme and verify the viability of the scheme in the virtual remanufacturing process.

3. Intelligent Redesign Framework Based on Digital Twin

For realizing the dynamic generation and real-time optimization of the redesign scheme rapidly, it is necessary to analyze and summarize the life cycle data of the product and use the data to predict and optimize the redesign parameters. Meanwhile, the virtual model should be built to simulate the redesign process and verify the feasibility of the redesign scheme.

DT technology aims to use advanced information technology to process physical space data and to describe, diagnose, predict, and make decisions on physical space through software definition, so as to realize the interactive mapping between physical space and digital space and realize intelligent decision-making. Therefore, DT technology can be applied to the redesign process of used products for remanufacturing. By analyzing and processing the original product information and historical design data, digital space is used to make design scheme decisions, product modeling, and performance prediction. Meanwhile, the virtual model can simulate the redesign process, including remanufacturing processes, usage process, and performance simulation. The data from the digital space can be fed back to the physical space (such as designers, customers, and remanufacturing workshop) to realize the interactive mapping between the physical layer and the digital layer of the redesign process. By continuous iterative optimization of design parameters between physical space and digital space, the optimal design scheme can be obtained.

3.1. Digital-Twin-Driven Redesign Process Framework

In order to successfully implement the digital-twin-driven redesign of used products, it is necessary to establish a framework for the DT-driven redesign process, which is mainly composed of five parts: physical world, virtual world, twin data, data management, and application. The physical world represents the data generated in the redesign process, including failure feature data, product structure design data, and remanufacturing process data. The virtual world is a virtual mapping of the physical world, including product virtual models, virtual function planning data, virtual remanufacturing process data, and product virtual operation data. The virtual world covers the entire life cycle data of the products. Twin data mainly refer to the fusion of virtual data and physical data through data processing technology, and the two-way connection, interaction, and real-time drive of virtual and real data through advanced technologies, such as the Internet of Things (IoT) or cloud computing. Data management is the process of virtual and real data processing, including the collection, analysis, integration, and fusion of product life cycle data. The service of DT is to realize the visualization of the redesign process, dynamic adjustment of design parameters, and prediction of product performance. The framework of the digital-twin-driven redesign process is shown in Figure 2.

3.2. The Construction Process of DT-Driven Redesign System

For a used electromechanical product, the DT-driven redesign system construction process mainly consists of six steps, as shown in Figure 3. Through these six steps, a fully functional DT system can be formed. The specific process is as follows:

Step 1: Build the virtual model of the physical product

The virtual model is mainly constructed by computer-aided design (CAD) and reverse engineering (RE). The virtual model generally includes three parts: elements, behavior, and rules. The element layer of the virtual model includes the geometric model, physical model, operational environment of the product, etc. For example, in the redesign process, the virtual model includes the used product model, remanufactured product, operation environment, and remanufacturing workshop environment. The behavior layer includes the usage mode, product operation mode, remanufacturing mode, etc. The designer must not only analyze the operation habits of the user but also analyze the impacts on operators after the product redesign scheme is optimized. The rule layer includes design rules, operation rules, remanufacturing rules, etc., based on which remanufactured product performance evaluation model, redesign scheme optimization, and evaluation model can be constructed.

Step 2: Collection of product-related data

The DT model requires product-related data for construction which is generally divided into product development data, manufacturing process data, customer demand data, remanufacturing process data, etc. Usage and service data include failure data, repair data, operation process data, etc. Product development data include geometric data, function data, performance data, etc. Customer demand data include disassembly demand data, function upgrade demand data, performance upgrade demand, etc. Recycling and reuse data include product quality, component information, remanufacturing data, etc. Intelligent sensors, IoT technology, and virtual simulation technology can be used to collect the relevant data of the product, and machine learning and intelligent algorithm can be used to predict and analyze the data of product performance, disassembly time, remanufacturing cost, etc. These data contribute to the establishment of a fully functional virtual model. The specific product data are shown in Figure 4.

Step 3: Data processing to achieve redesign decision

The collected data need to be analyzed, integrated, and visualized to realize the virtual redesign process. Firstly, the collected data need to be standardized and formatted, so that the data can be converted into information that can be directly identified by the designer. Secondly, since data from a single source can rarely identify the complete information of a product, it is necessary to integrate data from different sources, e.g., using ontology technology to encapsulate the data from various resources. Thirdly, the redesign process information formed after data analysis and integration is visualized through a virtual model to make the design process more intuitive. Finally, intelligent technologies could be used to enhance the DT’s decision-making in the redesign process.

Step 4: Simulate product behavior in a virtual environment

Virtual/hybrid reality enhancement technology (VR/AR), such as 3DMAX, SolidWorks, and Croe, is used to build the virtual scene of product operation. According to the product operation rules and customer operation mode, the product behavior is simulated to realize the direct interaction between users and products, for predicting customer satisfaction with the products in advance.

Step 5: Physical product executes virtual model instructions

The main role of DT is to perceive, analyze, and make decisions based on physical data in the virtual model and then fed back the decision-making data to the physical world and execute instructions. Through the data transmission module and communication protocol, the physical model receives the decision-making instructions of the virtual model and adjusts the function, behavior, and structure of the physical world through the executive mechanism. Intelligent sensors and actuators are important parts of DT; the former is used for intelligent perception and data acquisition of the physical world, and the latter is used to execute the virtual model, which is used to adjust the physical world to a reasonable state. In engineering applications, the commonly used actuators include hydraulic, pneumatic, electronic, and mechanical transmission components to drive the operation of the remanufacturing process.

Step 6: The connection between the virtual world and the physical world

The connection between the virtual and physical world is mainly divided into four parts: Firstly, the data communication channel of the virtual and real design process is constructed by using MQTT, RMST, and Modbus protocol, so that the virtual and real data can be transmitted freely. Secondly, it is necessary to establish the connection between virtual and real design process and twin data; the former connects with twin data through MySQL database, NoSQL database and OPC, MTConnect, and other protocols. Then, the virtual and physical worlds are connected with the redesign process service platform through cloud computing of virtual and real data, and services, such as failure feature analysis, product function planning, and design parameter solution, can be realized. Lastly, since the collected data are multisource and heterogeneous, dimensionality reduction, noise reduction, and visualization processing are required; the main enabling technologies include Storm, Hadoop, XML (Extensible markup language), K-means, etc. In addition, the redesign process data are a confidential document of remanufactured products, which is of direct interest to developers; therefore, it is necessary to ensure the security of data management and interaction.

4. Application of Redesign Process Based on DT

Axiomatic design (AD) can provide designers with a logical and rational thinking method and guide them to make correct decisions in the design process. However, due to a large number of used products and the presence of many types, the traditional axiomatic design method struggles to quickly generate the corresponding redesign scheme. In addition, redesign schemes are prone to product design errors, and it is difficult for designers to find errors quickly, resulting in increased remanufacturing trial-and-error costs. In order to solve these defects of traditional AD, a DT redesign process model of the used electromechanical product based on axiomatic design is proposed. Firstly, the original product model can be obtained by using reverse engineering technology, and the product model is adjusted according to the customer demand data and the adjusted product model is fed back to customers. Feedback iterations can constantly improve the quality of the product model. Then, the CBR technology is used to retrieve the product structure modules and the corresponding redesign parameters which can satisfy the functional requirements. Meanwhile, the three-dimensional virtual model is used to optimize the structural module combination and parameter adjustment, so that the new structure can be integrated with the original structure of used products. Finally, BPNN is used to establish the prediction model between structural characteristics and remanufacturing process parameters which can predict the remanufacturing process parameters of new product structure. Meanwhile, based on the remanufacturing process parameters, a DT model of the remanufacturing process is constructed to truly simulate the remanufacturing process and optimize the process parameters. In addition, the optimized remanufacturing process parameters are transferred to the remanufacturing shopfloor, where they can be used to verify the remanufacturing process scheme in accordance with the on-site remanufacturing environment. The remanufacturing data can be collected by intelligent sensors and can be fed back to the virtual model. Through the continuous iterative optimization between virtual and real models, the optimal remanufacturing process scheme can be obtained.

4.1. Virtual Model Construction for Used Products

Due to the varying degree of damage in used products during service, the amount of physical information lost (e.g., wear leads to loss of size or shape data) also differs. In order to construct the virtual model of used products and realize the visualization of redesign process, a method of used product model construction based on incomplete information reconstruction is proposed. Firstly, it is necessary to summarize the defect information of used products, which may include geometric information in the failure state and mechanical information, such as residual stress as shown in Figure 5. Then, reverse engineering technology may be used to obtain the point clouds of used products, using a scanner, e.g., handy scan 300 precision. Finally, based on reverse engineering, the data of defective parts (point cloud or triangular mesh surface) are collected, and the two-dimensional plane of the three-dimensional model of the component is segmented and matched to determine the damaged area. Otherwise, the damage depth is determined by three-dimensional reconstruction, and the original point cloud model is reconstructed. The specific process is shown in Figure 6.

4.2. Product Function Planning Based on DT

Customers have various requirements for products, covering all aspects of the product life cycle. In addition, in the Internet era, there are diverse customer demands which may be identified via online comments, telephone return visits, and online questionnaire surveys. This has led to explosive growth in customer demand data, and traditional customer demand analysis methods are difficult to accurately and quickly extract product function data. In order to quickly and accurately extract product functional demand, firstly, intelligent collection technology is used to extract customer demand data. For instance, the vector space model (VSM) could be used to model customer demand data, which is transformed into computer-recognizable language. K-means algorithm may be used to classify customer demand. Finally, QFD is often used to transform customer requirements into engineering features. Then, the virtual model is employed to visually express and configure functions, which can identify the conflicts between functions and products. These conflicts could be solved by the TRIZ approach. Based on the product’s functional requirements, the virtual model of the used product is modified which is fed back to designers and customers for evaluating the information related to the appearance, structure, and materials. Meanwhile, the evaluation results are fed back to the virtual model for product function adjustments, such as function upgrade, material replacement, and structural improvement. Through continuous feedback and iterative optimization of the functional planning information between the virtual space and the physical space, the functional planning scheme of the product is improved, which is conducive to the accurate solution of structural parameters. The product planning process is shown in Figure 7.

4.3. Solving Process of Structural Parameters Based on DT

The structured domain in axiomatic design describes the entire structural design process of the product, and the design parameters of the product mainly refer to the product structure parameters (such as the diameter of the shaft and the length of the guide rail) that can realize the function. In order to improve the efficiency of solving structural parameters in the redesign process, a DT-based method is proposed. Firstly, the historic data of remanufacturing process are collected by using intelligent sensors and RFID technology, and the knowledge base of remanufacturing process and redesign is established by using ontology technology. Then, the three-tuple model is used to establish the target case which meets the functional requirements and design constraints. Meanwhile, CBR technology is used to identify the structural design scheme which is most similar to the target case. Finally, according to the structure design scheme, the DT model of the product is constructed, including product material model, part model, assembly model, and product behavior model. Through structural analysis and performance analysis of the product model, the design scheme that meets the design requirements is fed back to the designers, and the product trial production and testing are carried out based on the design scheme. Meanwhile, the remanufactured product trial production and testing results are fed back to the virtual world for redesign scheme adjustment, and the virtual model simulation is carried out based on the adjusted design scheme. The physical solution and the virtual solution are continuously iteratively optimized until the optimal structural design solution is solved which meets the design constraints and customer demand. The structure design process based on DT is shown in Figure 8.

4.4. Design of Remanufacturing Process Scheme Based on DT

The remanufacturing process scheme directly affects the success rate, efficiency, and cost of remanufacturing, which is the key to the success of remanufacturing. Based on the different usage and failure modes of the used products, remanufacturing process designers need to formulate corresponding remanufacturing process schemes. Meanwhile, the used products may be upgraded and remanufactured based on the condition of the used products and the individual demands of customers. However, the traditional remanufacturing process scheme design method cannot generate and adjust the scheme in real time based on the product situation and remanufacturing environment, which can affect the efficiency and cost of remanufacturing processes. In order to generate and adjust the remanufacturing process scheme in real time, a remanufacturing process scheme design method based on the DT model is proposed. Firstly, the framework of the virtual model of the remanufacturing process is developed according to the remanufacturing workshop environment. Then, sensors and RFID technology are used to collect the remanufacturing workshop and remanufacturing process data in real time, and the real-time communication of virtual and real data can be realized. Then BPNN algorithm is used to predict the remanufacturing process scheme based on the failure characteristic parameters, product information parameters, and customer demand [25]. Furthermore, the DT model is used to simulate the remanufacturing process and analyzes the size error, performance error, and feasibility of the remanufacturing process. Finally, the remanufacturing process scheme is fed back to the process designer through the data communication module. The product is then processed and upgraded based on the remanufacturing process scheme. Subsequently, the remanufacturing process information (including remanufacturing error, product quality, and performance) is fed back to the virtual space for remanufacturing process scheme adjustment and model revision. Through continuous iterative optimization of remanufacturing process information between physical space and virtual space, the remanufacturing process scheme is continuously improved, which could improve remanufacturing efficiency and quality. The design framework of the remanufacturing process scheme based on DT is shown in Figure 9.

5. Case Study

In order to verify the feasibility of the intelligent redesign method based on DT, the redesign process of a commercial vehicle clutch is taken as a case study. A clutch may fail due to frequent use in service and improper operation by the driver. However, in most cases, clutch failures are caused by the failure of some parts due to wear, ablation, and other failure behaviors, resulting in scrapping of the clutch. In order to retain the value of used clutch parts and improve clutch performance, it is necessary to redesign the clutch.

Firstly, the original product data of the used clutch are collected, including the original size, performance, movement mode, and service environment of the clutch. Meanwhile, the reverse point cloud model of the used clutch is constructed by using a handy scan 300 precision scanner, and then the point cloud data of clutch defects are quickly and effectively registered with the original data using a unit step integral iteration method. The damage point cloud information is obtained through information reconstruction. Finally, the complete clutch product model is generated by point cloud fitting function. The construction process of the clutch model is shown in Figure 10.

The used clutch not only needs to recover its performance but also needs to be upgraded to meet customer demands; therefore, it is necessary to redesign the used clutch. Customer demand information is obtained through online reviews, questionnaire surveys, and customer return visits. In this case, 1000 pieces of customer demand information were supplied by the industry partner, such as improving the wear resistance of the friction plate, improving the rigidity of the clutch cover, and lightening the clutch. VSM is used to classify the demand data, and the Jieba word segmentation plugin [26] is used to process the demand information. Meanwhile, the k-means algorithm is applied to cluster the demand texts and obtain the main customer demand information. The analysis results are given in

Table 1. Clustering number of demand texts in each category.

Serial Number	Demand Text	Cluster Number
1	Clutch coupling smelly, improve high temperature resistance Clutch coupling, ablation phenomenon, change material … Clutch friction plate ablation cohere, difficult separate	325
2	Clutch remanufacturing disassembly, structure go against disassembly, change the connection method Clutch remanufacturing less scheme, parts large waste … Clutch remanufacturing driven plate, bad disassembly change structure	205
3	Clutch separate halfway, possible heavy, reduce material Clutch heavy, hope can lightweight … Clutch material loss, reduce weight	168
4	Clutch abnormal noise, hope provide NVH test Vehicle front abnormal noise, hope cooperate solve … Vehicle put into gear difficulty, cooperate adjust	196

. NVH denotes noise, vibration, and harshness in Table 1.

According to the clustering number of demand texts, the following can be concluded: improving high-temperature resistance, improving the disassembly of the clutch, lightweight clutch, and reducing abnormal noise. These four customer demands account for a large proportion of the 1000 demand texts. To this end, the QFD method is used to analyze the product engineering characteristics of the demand information, and the analysis results are given in

Table 2. Analysis of engineering characteristics of QFD.

Customer Demand	Ablation	Difficulty in Disassemble	Heavy Weight	Abnormal Noise
Weighted value	0.325	0.205	0.168	0.196
Engineering characteristic	Withstand temperature 360 °C	Good disassembly	Clutch weight ≤ 45 kg	Strong NVH detection capability

.

According to the engineering features of customer demand, the core is improved and visualized by the virtual model. The virtual model is shown in Figure 11.

In Figure 11, the material of the friction disk is changed to improve its high-temperature resistance, and the structure and material of the pressure plate are changed to reduce the weight of the clutch and maintain the same performance. The integral waveform piece is modified to a segmented waveform piece which could improve the disassembly of the clutch.

After determining customer demand and engineering features, CBR technology is used to identify the product structure parameters, such as driven disc diameter, friction plate thickness, and heat dissipation hole area, and the DT model is established according to the structure parameters. The designer adjusts the structure model according to the product structure constraints (e.g., clutch maximum size constraint, volume constraint, and weight constraint) until the best product structure is obtained. The specific process is shown in Figure 12.

In Figure 12, the cooling holes are added on the side of the cover, which can not only reduce the weight but also promote heat dissipation. Meanwhile, the strength of the clutch can be analyzed by the ANSYS software, which can simulate the force of the clutch, and dimensional accuracy can be analyzed based on the SolidWorks software, which can simulate the assembly process of the clutch. After obtaining the design parameters of the product structure, it is necessary to develop a suitable remanufacturing process for product reuse. The remanufacturing process design process is shown in Figure 13.

In Figure 13, the BPNN is trained using the historical remanufacturing process data. The trained neural network model is used to predict the clutch remanufacturing process, and the output remanufacturing process is able to meet the corresponding customer demand. However, due to certain errors in the accuracy of the prediction model or the absence of corresponding output in the remanufacturing process database, the designer needs to make process revisions and develop a new process according to the redesign scheme. Meanwhile, process designers need to formulate a reasonable process sequence based on the quality of used products and the performance requirements of remanufactured products. According to the assembly sequence of the clutch, firstly, the old friction plate needs to be disassembled; meanwhile, the new friction material is injected into the friction plate. Then the separable corrugated plate is pressed, and the side of the clutch cover is drilled to increase the heat dissipation hole and reduce the weight of the cover. Finally, all parts are assembled into the clutch, and the clutch is loaded for the NVH test to find out the resonance area. Then, the virtual machining model is used to simulate the remanufacturing scheme; simultaneously, the cost and environment of the remanufacturing process are analyzed and verified, and the simulation results are adjusted. Through the comparison of the physical process scheme and virtual process scheme, iterative optimization is carried out until the optimal remanufacturing process scheme is obtained.

Based on the data comparison and the experience of the designers, the redesign scheme of the clutch satisfies the scope of various evaluation indicators and customer demand. Compared to the traditional redesign method, the DT of the clutch can realize the visualization of the redesign process and improve the efficiency and reliability of the redesign scheme. Meanwhile, the cost of the clutch remanufacturing process and the consumption of energy and materials can be reduced.

6. Conclusions and Future Work

Due to the diversity of failure features, the uncertainty in service conditions, and product structure constraints, the redesign process of used products is complex and time-consuming. For solving these, an intelligent redesign method of used products based on DT is proposed, and the method consists of three parts: “redesign process framework and features analysis”, “intelligent redesign framework based on DT”, and “application of redesign process based on DT”. This method has been verified by the redesign process of the used clutch. In this case, the DT redesign process model based on axiomatic design is used to generate the redesign scheme of the used clutch that satisfies the customer demand rapidly. Meanwhile, the DT-driven redesign scheme evaluation method is used to verify the feasibility of redesign schemes and select the optimal clutch redesign scheme.

Compared with existing redesign methods, the novelty of this proposed method lies in the following three aspects: (1) This method can integrate the redesign process of the used products and simulate each stage through the virtual models, which realizes the visualization of the redesign process and helps designers to analyze the redesign process more intuitively and comprehensively. (2) The DT redesign process model based on axiomatic design can use virtual models to simulate in real time and dynamically adjust the redesign process, which can generate the optimal redesign scheme rapidly. (3) The DT-driven redesign of used products is to realize the decision-making, feedback, and optimization of the redesign process by processing and analyzing full life cycle data and twin data, which can improve the efficiency of redesign and the utilization of product-related data.

The future work requires efforts in the following aspects: (1) It is very important to ensure the security and integrity of DT data in storage and transmission. This paper has not taken into consideration the effective methods to store and transfer data, and direct-attached storage and Modbus for data storage and transmission may be researched in the follow-up study. (2) This paper applies three-dimensional software to construct a virtual model to represent the physical space. However, the accuracy of the mapping between the three-dimensional model and the physical space is unknown. In future research, the mapping accuracy will be researched, and 3D experience or twin builder may be used to improve the mapping accuracy of the virtual model. (3) Real and virtual data are not yet available for real-time feedback and processing, which is a key issue that needs to be addressed in the future.