Digital Twins Enabling Intelligent Manufacturing: From Methodology to Application

In recent years, the scope of research on DT-driven mechanical process systems has expanded significantly, resulting in a wide range of applications. In this section, we examine the utilization of DT-driven mechanical process systems at various stages of the product life cycle, including product design, manufacturing, and service. We utilized the product full life cycle classification standard as a reference to analyze the incorporation of DT at each stage. We present the specific applications of DT technology at each stage, alongside a comprehensive assessment of its strengths and weaknesses. This analysis is presented in Figure 9.4.1. Product Design In the product design stage, the implementation of DT technology can enhance the intuition and accuracy of the design, incorporating user requirements, iterative optimization, and a focus on human-centered product design and green sustainable design goals. Additional benefits include virtual sample processing, product performance verification, and optimization. As an illustration, let us consider the design stage of a semi-industrial combustion furnace. By leveraging Kalman filtering, adjustments to the model’s predictions can be made, taking into account potential uncertainties. This allows for the realization of predictions regarding combustion chamber performance, reliability assessments, and test program optimization, all of which are pivotal in upgrading processing standards and reducing energy consumption [120]. In a separate study, Friederich et al. [121] proposed the application of DT in product design and presented a framework for DT-driven intelligent product design, focusing on key processes and related technologies. They investigated the conceptual theory of product design and described the DT model configuration process based on conceptual design [122,123]. Furthermore, they proposed a DT-driven product design evaluation process and evaluation algorithm [124]. The authors also investigated the application of a DT-driven green design methodology [125]. By expanding the application of DTs to product design, the authors propose a virtual prototype design for CNC machine tools based on DTs and successfully model the coupling relationship of complex electromechanical systems of machine tools [126]. Moreover, they proposed a lean design process for machine tools based on DTs, which allows for the optimization of machine tool feed system parameters [127,128]. In addition, the application of DT technology in the product design stage provides valuable guidance for determining product solutions, analyzing product parameters, and integrating products with user requirements. Tao et al. [62] proposed a novel approach to product design based on the DT method. They analyzed the DT-driven product design framework, developed a DT model for design tasks incorporating task analysis and scenario decomposition, and proved through experiments that it enhanced product quality. Huang et al. [129] proposed a DT data management method for products using blockchain technology to improve the efficiency of data sharing among participants. They created a peer-to-peer network and achieved complexity management in the product design process through product design workload prediction and functional change propagation analysis methods. Wu et al. [130] proposed a DT complex product loop design framework that integrates multidisciplinary collaboration across three phases: conceptual design, detailed design, and virtual verification. This framework enables real-time verification and modification of issues arising from multidisciplinary integration, thereby minimizing the number of iterations and costs in the design process. This study lays a theoretical foundation for bridging the gap between the product design and manufacturing phases. In the product design stage, the application of DTs effectively considers user needs and design goals, making product program development more intuitive and rational. It also provides the design team with additional opportunities for innovation and optimization through virtual machining verification, thus enhancing the efficiency and quality of product design. Moreover, this approach significantly reduces the cost of physical processing verification. 4.2. Product Manufacturing During the product manufacturing stage, the application of DT technology aims to achieve quality data acquisition, processing quality monitoring and analysis, processing quality prediction, and control functions in the mechanical process system. These aspects are crucial for optimizing the product manufacturing process, improving processing quality, and increasing processing efficiency [131,132,133]. This section explores the specific applications of DT-driven machining processes and analyses their research progress and limitations. 4.2.1. Applications in Machining The traditional mechanical process system includes a machining process involving a machine tool system, a workpiece, a tool, and a fixture. In this section, we will examine the concept of the traditional mechanical process system and discuss the specific application of DT-driven mechanical process systems [134]. We will explore the application of DTs in machine tool systems, workpieces, tools, fixtures, and their respective functions. (1) DTs for Machine Systems DT of machine tool systems refers to the application of monitoring the performance status of machine tool systems, including machining parameters, machining errors, modals, and other performance and abnormal fault status processing, based on DT technology. Machine tool condition monitoring is a significant application of the DT of machine tool systems. Traditional monitoring technologies, such as cameras and sensors, may encounter issues such as image delays, delays, and difficulties in determining the source of errors at the physical end. An approach proposed by Liu et al. [135] addresses the fusion of heterogeneous data from multiple sources and the lack of semantic information through DT modeling of CNC machine tools. It enables comprehensive sensing of the operating status of machine tools, thus achieving condition monitoring services. Another study by Guo et al. [128] focused on an improved Gilbert–Johnson–Keerthi distance algorithm to enhance the detection of collision information between a tool and a machine tool during the simulation and monitoring of machine tool motion. This enhancement results in a more realistic display of the workpiece shape. The usability and efficiency of the system are verified through an example of machining a typical shaft part with a CNC machine tool. This remote interactive simulation of the machine tool demonstrates the benefits of the improved algorithm. Overall, the DT modeling of CNC machine tools addresses the challenges of fusing multisource heterogeneous data and addressing insufficient semantic information. It enables effective condition monitoring services by comprehensively sensing the operating status of machine tools and enhances the detection and display of collision information during the simulation and monitoring of machine tool motion. Realizing the prediction process for abnormal states in machine tool systems is another crucial application of machine tool DTs. This includes predicting machine tool performance and component wear degradation. Lv et al. [136] confirmed the effectiveness of a self-built, self-assessed, and self-optimized maintenance system based on a bioinspired DT machine tool. Through bearing fault diagnosis experiments, this system achieved unmanned maintenance of bearings. Yang et al. [137] established a hybrid prediction method for a performance degradation model by constructing a meta-motion DT model, analyzing wear theory, and applying algorithms. This method inherits the advantages of both data-based prediction and model-based prediction, accurately predicting machine tool transmission units. liu et al. [138] proposed a DT approach for motion axes, incorporating a time-varying error model based on heat transfer theory and vision modeling. This method predicts and compensates for time-varying errors, reducing error fluctuations by 69.19%, as demonstrated by experiments on real-time errors in hole spacing. The application of DTs for machine tool systems enables condition monitoring and the prediction of abnormal conditions in machine tools. This significantly promotes the safe and efficient machining of mechanical process systems. Parameter optimization plays a crucial role in optimizing the machining process. Traditional parameter optimization methods rely on manual experience and often involve high levels of uncertainty. The machining process DT facilitates error suppression and optimization of machining parameters, thereby laying the foundation for achieving high-quality and high-level machining. (2) Workpiece DT The term “workpiece DT” refers to the application of DT technology in predicting and optimizing various machining parameters, machining accuracy, surface roughness, and other characteristics of workpieces during the machining process [139]. Ghosh et al. [140] developed a DT structure based on a hidden Markov model to predict the surface roughness generated by continuous grinding operations. This DT structure was then implemented to accurately predict the surface roughness during continuous grinding operations. Zhu et al. [141] proposed a DT-driven manufacturing framework for thin-walled parts. This framework collects and updates DTs of workpieces in different states, providing machine tool operators with real-virtual interaction opportunities. The aim is to make the start-up phase faster and more accurate. The feasibility of this framework was demonstrated in a case study involving leaf disc machining. Wang et al. [142] proposed a DT-driven clamping force control method to improve the machining accuracy of thin-walled parts. By establishing a full-factor information model of the clamping system and incorporating dynamic information from the clamping process, a virtual space model was constructed using finite element simulation and deep neural network algorithms. This method was verified and found to be effective through arithmetic examples. Dai et al. [143] proposed an ontology-based information modeling method for prefabricated parts. This method enables the association and integration of machined part feature information in the process of virtual-real interaction. This study also provides a theoretical foundation for optimizing machining parameters and predicting machining quality. Zhang et al. [144] investigated the interaction between machine tools and milling processes from a system perspective. They developed an integrated model of the ball screw feeding system and the milling process, which enabled multiphysics field simulation of the entire system considering multisource harmonics. This approach is highly significant for studying machine tools and milling processes as a coupled whole. Li et al. [145] presented an aeroengine assembly quality assessment method based on cumulative block information modeling and a process-oriented assembly twin. They established an analytical DT platform that integrates modules for measurement, a digital design model, a geometric deviation analysis model, an information model, and a database. This platform enables quality analysis at the assembly operation stage and lays the foundation for machining key aeroengine components. Wang et al. [146] proposed a novel method for rapidly calculating engine blade strains, aiming to overcome the technical challenges of real-time calculations in DTs and achieve simultaneous mapping of engine blade health states. The accuracy and effectiveness of this method were verified through engine blade strain measurement experiments and numerical simulations. The results demonstrated a significant improvement in real-time performance, with a speed increase of approximately 1444 times compared to that of conventional finite element methods. The minimum run time achieved was 0.91 seconds. Furthermore, the minimum relative error was 0.11%, and the average relative error was less than 0.76%. The application of workpiece DTs facilitates the optimization of machining parameters, the prediction of surface roughness, and the optimization of machining decisions. These functions have a significant impact on the machining of special workpieces, such as aerospace thin-walled parts, aero-engine blades, and case parts. (3) Tool DT Tool DT refers to the application of DT technology in researching and analyzing tool wear monitoring and prediction, selection decisions, and tool service in the machining process [147,148,149]. First, monitoring and predicting tool wear can effectively reduce machining errors and improve efficiency [150]. Qiao et al. [151] proposed a data-driven DT model and a hybrid model prediction method based on deep learning, which demonstrated the accuracy of tool wear prediction through the study of vibration data from a milling machine. In addition, Natarajan et al. [152] proposed a technique that utilizes DTs to construct a balanced virtual instrumentation framework that is perfectly matched to the physical system to achieve exceptional accuracy in inspecting and predicting tool conditions. The tool condition monitoring system deploys the DT model to predict different tool conditions based on sensory data. Deebak et al. [153] presented a deep migration learning-based DT-assisted troubleshooting method for analyzing the operating conditions of machining tools. Furthermore, the system develops an intelligent toolholder that integrates a type K thermocouple and a cloud data acquisition system on a WiFi module. Analytical studies confirm that this intelligent tool holder provides higher accuracy and enables the optimization of milling and drilling operations of cutting tools. Finally, Zhuang et al. [154] proposed a DT-based approach that realizes the physical-virtual symmetry of the DT model by constructing a symmetric virtual tooling system that exactly matches the actual tooling system. This approach accurately maps the real-time state of tool wear. Zhang et al. [155] proposed a framework for model updating based on DTs, which is used to obtain an accurate tool wear model for predicting and managing machining processes. Xia et al. [156] presented a kinematic model and trajectory planning method for the UR10 robot by establishing a DT unit in the inspection system. The authors utilized Unity3D software to create the DT environment for the inspection system. Through socket communication, a synchronous mapping function is established between the robot’s digital model and the physical entity, enabling complex tool edge image acquisition trajectory planning, precise teaching of the virtual scene, digital monitoring of the inspection process, and optimization of the system model. The stability and effectiveness of robot kinematics, trajectory planning, and interactive communication in the tool wear image inspection environment are verified. Chen et al. [157] utilized a bionic digital brain as the intelligent core of the DT double-cutting machining framework, which includes monitoring, prediction, optimization, and control. They demonstrated the powerful information processing capabilities of DTs and presented real-time precision intelligent machining results. Liu et al. [158] proposed a DT-driven method for predicting surface roughness and adaptively optimizing process parameters. When the predicted surface roughness based on real-time data does not meet the machining requirements, the DT system issues a warning and adaptively optimizes the cutting parameters based on the current tool wear prediction. The effectiveness and advancement of the proposed method are verified through the development of a DT system for process optimization and a large number of cutting tests. This approach combines real-time monitoring, accurate prediction, and optimization decision-making in the machining process, resolving the issues of inconsistent quality and efficiency. Song et al. [159] addressed the problem of tool wear-induced vibration and deformation in the milling of thin-walled parts. They proposed a DT tool wear state recognition method using feature vector extraction, hyperparameter optimization, and support vector machine algorithms. The experimental results show a recognition rate of over 90% for the system. However, due to limitations in the support vector machine algorithm itself, the method exhibits weak generalizability and is not applicable to large sample sizes. Additionally, Zhang et al. [160] incorporated a deep migration learning strategy and edge distribution rule into the DT tool model to achieve target domain training with small samples. This approach improves the accuracy and appropriateness of the model and resolves the issue of small samples under variable working conditions. Second, in terms of tooling services, DT-based tooling services enable the efficient utilization of tooling. Xie et al. [161] established a tool DT model consisting of five stages: tool market investigation, research and development, production planning, manufacturing, customer use and service. This model allows real-time status monitoring of cutting tools, visualization of tool information, parameter optimization, maintenance strategies, and virtual maintenance. The application of tool DTs is one of the most widely used and extensively researched applications of DT-driven mechanical process systems. The achievement of tool wear state monitoring and prediction plays a crucial role in reducing machining errors. Research on DT-based tool services, tool selection strategies, and tool change speeds has entered the exploration stage. (4) Fixture DT Fixture DT refers to the application of DT-based technology to the control of the fixture itself and the clamping force control during machining. Wang et al. [142] designed a DT dual-drive clamping force control method by integrating a neural network and finite element simulation to address the problem of machining deformation caused by inadequate fixture clamping in traditional machining processes. Through experiments considering the full-factor information of the clamping system, they proved the effectiveness of the method. Weckx et al. [162] proposed a cloud-based DT for monitoring high-performance composite machining adaptive clamping devices by incorporating computer vision-related technology. This implementation achieved functions such as tool wear monitoring based on the clamping force and the evaluation of clamping device operation. Additionally, they innovatively developed the function of automatically triggering monitoring algorithms for the transmission of machining signals to the cloud. The application of fixture DTs allows for the optimization of clamping parameters, the prediction of clamping stability, and the control of clamping force during machining. These capabilities contribute to improving the stability of clamping and enhancing machining accuracy. By leveraging key technologies such as analysis and prediction, decision-making optimization, and fault processing, the application of fixture DTs significantly improves the efficiency and accuracy of machining. Implementing error prediction in the machining process is an essential step to avoid downtime risks and ensure machining efficiency. Anomaly detection based on DT technology provides a guarantee for improving efficiency and accuracy. The integration of DTs in the machining process enables the suppression of machining errors, effectively improving machining accuracy and ensuring machining safety. 4.2.2. Deficiencies In the product manufacturing stage, the implementation of DT technology has greatly facilitated the transformation and upgrading of mechanical process systems. However, there are still certain limitations that need to be addressed. First, research on DT-driven mechanical process systems requires a substantial amount of data. The existing research predominantly relies on a single method of data monitoring, which incurs high acquisition costs, low accuracy, poor real-time performance, and a limited amount of data [154]. Second, the research on DT-driven mechanical process systems is often limited to specific research objects, processes, working conditions, and machining scenarios, rendering it inapplicable in a universal context [163]. Finally, while there is a significant amount of research on condition monitoring and virtual simulation in DT-driven mechanical process systems, there is a lack of research on algorithms, models, and mechanism analysis. Consequently, the application of algorithms is often constrained by their inherent limitations, which restrict the conditions under which they can be implemented [164]. Therefore, there is a crucial need for improvements in the application of DT-driven mechanical process systems. 4.3. Product Service From the previous discussion on the specific application of DT-driven mechanical process systems and the analysis of DT application in machining processes, it is evident that DT technology enables functions such as machining status monitoring, prediction, and process optimization. Building on this, this section will explore the application of DTs in the product service stage, focusing on fault detection and predictive maintenance of mechanical process systems, machining process evaluation and optimization, and product sustainable manufacturing and operation and maintenance. 4.3.1. Fault Diagnosis and Predictive Maintenance Traditional preventive maintenance, with its high cost and low efficiency [165,166], is being challenged by the emergence of predictive equipment maintenance. This approach, based on DT technology, allows for the prediction of maintenance risks and effectively reduces the chances of equipment failure. By improving maintenance efficiency, it is clear that predictive maintenance is the future trend for mechanical process system maintenance [167,168,169]. To optimize equipment failure monitoring, prediction, and maintenance decisions, He et al. [170] proposed a complex equipment health management approach. A hybrid framework based on DT modeling and DT data, as presented by Luo et al. [56] further enhances this optimization. Building on this framework, a hybrid predictive maintenance algorithm was proposed, and its effectiveness was verified through examples. In the quest for effective DT-based predictive maintenance systems, Van et al. [171] used domain analysis to model key features and synthesize relevant literature. The result was a DT-based predictive maintenance system that derives three views: a user view, an architecture view, and a deployment view. The analysis demonstrated the potential of creating a reference architecture in the field of DT-based predictive maintenance. Zhong et al. [168] summarized the growing research interest in predictive maintenance based on DTs in the field of manufacturing. This paper proposes a gap between DT technology and predictive maintenance technology, emphasizing the importance of utilizing DT technology to achieve effective predictive maintenance. Furthermore, a predictive maintenance approach based on DTs is presented, highlighting the differences between this approach and traditional predictive maintenance. To address fault diagnosis in both the development and maintenance phases, Xu et al. [172] proposed a two-stage DT-assisted method based on deep migration learning. This method identifies potential problems that may not have been considered during the design phase and uses deep neural network-based diagnostic models for fault diagnosis. By employing deep migratory learning, previously trained diagnostic models can be migrated from virtual space to physical space for real-time monitoring and predictive maintenance. This ensures diagnostic accuracy and prevents unnecessary delays. Predictive maintenance based on DTs enables an “ex ante” preventive mode for mechanical process systems, effectively reducing machining losses and improving efficiency compared to the traditional “ex post” repair and fix mode. 4.3.2. Process Evaluation and Optimization The process evaluation of machining processes is crucial for enhancing process execution and reducing product development cycles [173]. Liu et al. [124] proposed a data-driven machining process evaluation method using DTs by aligning machining process data with process design data. This method successfully evaluated the machining process for key components of marine diesel engines. Pereverzev et al. [174] applied DT technology fused with dynamic programming to test and iteratively optimize the grinding cycle of a CNC machine. This optimization ensured consistent quality of machined parts by designing the optimal feed control cycle. To optimize the CNC machining process, Vishnu et al. [175] simulated, predicted, and optimized the workpiece surface roughness in the process planning and machining stages. They developed a surface roughness prediction model based on DTs, providing theoretical support for the development of optimization technology. Zhu et al. [141] established a DT model for the workpiece and utilized algorithmic optimization to provide real-time machining information to the machine operator. By optimizing the tool direction and tool path after each work step, they demonstrated the effectiveness of their method through machining examples. Chen et al. [176] proposed a DT-driven method to suppress delamination damage in real time by analyzing the relationship between the thrust increase caused by tool wear and CFRP delamination. Through extensive drilling experiments, they input cutting parameters and thrust signals into the DT model, Gaussian process regression, and mathematical modeling to predict current tool wear and thrust profiles, respectively. The results showed excellent real-time prediction, with maximum errors of 4.1% and 4.2% for tool wear and exit thrust prediction, respectively. Compared to conventional drilling, DT provided closed-loop feedback on the time-varying critical feed rate for each hole, resulting in no delamination mode I and up to 48.4% suppression of delamination mode III. This intelligent virtual-real linkage in the CFRP drilling process offers important theoretical support for the effective suppression of delamination damage in automated production processes. Optimizing machining processes holds significant value in improving machining performance, accuracy, and efficiency. Utilizing DT technology provides a new approach to mechanical process optimization. Implementing DT-based machining process evaluation and optimization positively impacts reducing machining error rates, formulating optimal machining process routes, and enhancing machining efficiency. 4.3.3. Product Sustainable Manufacturing and O&M Management With the introduction of China’s “Carbon Peak Carbon Neutral” initiative and other major strategic decisions, the green and sustainable product manufacturing mode has become the mainstream approach in the machinery manufacturing industry [177,178,179,180]. The entire life cycle of sustainable manufacturing is supported by a model that includes product material selection, sustainable disassembly, end-of-life recycling, and remanufacturing, all based on DT technology. This model provides technical support for achieving a green and sustainable development model through the monitoring and evaluation of energy consumption during the processing process and visualization of green features [181]. Xiang et al. [182] proposed a DT-based approach for selecting sustainable materials optimally, with the aim of achieving sustainable manufacturing over a specific period. This is done through simulating and evaluating the performance of sustainable materials. Li et al. [183] proposed a DT-driven method for evaluating sustainable performance in intelligent manufacturing and confirmed the effectiveness of this approach through testing. Kerin et al. [184] proposed a product DT model that utilizes data from different instances in the product life cycle to optimize remanufacturing plans. This includes predicting residual product life through neural networks and employing techniques such as bee algorithms for decision-making, ultimately achieving optimal product remanufacturing decisions. Sustainable manufacturing plays a crucial role in building an ecological civilization, and the adoption of DTs in the sustainable manufacturing of products promotes the transformation of the manufacturing industry toward sustainability and intelligence. This is achieved through the monitoring and evaluation of energy consumption during the processing process, the characterization of sustainable features, and the establishment of mechanisms for end-of-life recycling, among other technological means. It is crucial to understand the operation and maintenance of the mechanical process system during the product service stage. The operation and maintenance service system driven by DTs is explained in terms of pattern updating, data application, and system interaction. For instance, this article explores the application of DTs in the operation and maintenance of aviation engines [185], focusing on accurate monitoring, fault diagnosis, performance prediction, control optimization, and other functions. Fu et al. [186] identified the time and cost inefficiencies of traditional design, manufacturing, and maintenance processes as inefficient due to their independent operation and management. To address these inefficiencies, a unified platform is needed for efficient and intelligent design, manufacturing, and maintenance of machinery, equipment, and systems. To achieve this goal, an information-physical combinatorial framework is proposed that enables more accurate design, defect-free manufacturing, smarter maintenance, and advanced sensing technologies. Chen et al. [187] addressed the problems with the traditional “regular maintenance and fault repair” mode for mechanical equipment, including high costs and low efficiency. An intelligent mode of “predictive maintenance” and “condition maintenance” is proposed to achieve predictive maintenance, life prediction of industrial equipment, and improved virtual-realistic interaction and autonomous accurate service. Huang er al. [188] proposed an operation and maintenance service system that included virtual-real space interaction, data-driven knowledge updating, and real-time product diagnosis and maintenance. The effectiveness of this system is verified through a machine tool performance analysis test. It is highlighted that operation and maintenance management based on DTs is significant for mechanical process systems. The establishment of a digital operation and maintenance system reduces the impact of unpredictable factors, such as the aging and wear of equipment and structural deviations. This enables more accurate and efficient operation and maintenance compared to traditional manual experience-based systems [189]. Throughout the product life cycle, the application of DT technology can achieve collaborative optimization of product design and manufacturing, improve production efficiency and product quality, and continuously create value through digital services to meet customer needs and enhance competitiveness. With the continuous development and popularization of digital DT-enabled intelligent manufacturing, its application in the product field is expected to become more extensive, offering greater innovation and development opportunities for enterprises.