Review of Applications of Digital Twins and Industry 4.0 for Machining

Abstract

Digital twins, as part of Industry 4.0, are critical for advanced smart manufacturing processes, including machining. Sensor systems in smart manufacturing allow for real-time tracking of all changes in the machining process as well as simulation of an object’s behavior in the real world. It can also intervene and correct any defects that may arise during the machining process. The current review covers basic concepts for machining processes for the first time in detail, including Big Data, the Internet of Things, product lifecycle management, continuous acquisition and lifecycle support, machine learning, digital twin prototypes, digital twin instances, digital twin aggregates, and digital twin environments. The review article examines digital twins for the most common machining processes, such as turning, milling, drilling, and grinding. This review also highlights the benefits and drawbacks, as well as the prospects for using digital twins in smart manufacturing.

1. Introduction

One of the main trends relevant in terms of the development of the fourth industrial revolution (Industry 4.0) is becoming Digital Twins (DTs) [1]. The basic comprehensive understanding of such a phenomenon as DTs has been described in previous literature in [2]. The development of such areas as Artificial Intelligence (AI) [3] as well as the Internet of Things (IoT) [4] provides many opportunities for the development of DTs. These developments have been most apparent since 2015, as evidenced by the Gartner Hype Cycles for Emerging Technologies, which outlines technology maturity cycles [5].

DTs are a whole complex of digital technologies that implement physical models in a virtual environment [6]. Among such technologies are Machine Learning (ML), optimization, numerical modeling methods, statistical methods, and others [7]. In some cases, DTs can be based on mathematical models that show the behavior of real systems considering physical phenomena, such as the DTs of cylindrical plunge grinding on CNC machines [8].

The possibilities for using DTs are remarkably broad. For example, the use of DTs to create control and recommendation operating modes, as well as their optimization, including in the online mode [3]. The work of Zakharov et al. [9] provides an example of the use of DTs for modeling the assembly of products and determining the error in location and orientation. In the following, a comprehensive overview of the machinability concept is presented and the integration of predictive machining models within digital twins in cyber-physical spaces for in-process monitoring and adaptive control is demonstrated [10]. The following review discusses the applications of DT technologies in the field of machining and optimization in multi-scale quality control [11]. DTs can also function as diagnostic systems, relying on various sensors to determine the presence of deviations in the physical indicators of the process and give appropriate signals about emergencies or resource exhaustion [12]. In addition, DTs can be used in control systems based on the IoT, and based on this data, make control actions [13]. DTs can also be used for the assessment of product quality control and monitoring [14]. Hence, the use of DTs provides good opportunities for modeling and managing a wide variety of phenomena and processes that are inherent in manufacturing [15]. As a result, production processes that use DTs make it possible to track and manage the technological process [16] and eliminate downtime in the production process in unpredictable situations [17].

This paper intentionally focuses on the general application of Digital Twin (DT) technology in machining rather than delving into specific machining processes [18]. By broadening the scope, the study aims to explore how DTs can be integrated across various machining processes such as turning, milling, drilling, and grinding. This approach provides a holistic perspective on the transformative potential of DTs in enhancing manufacturing efficiency, accuracy, and adaptability within the framework of Industry 4.0.

The review addresses a critical gap in the existing literature by examining the integration of DT concepts with Industry 4.0 technologies, including Big Data (BD), Internet of Things (IoT), Product Lifecycle Management (PLM), Continuous Acquisition and Lifecycle Support (CALS), and Machine Learning (ML). It also introduces DT-specific paradigms like Digital Twin Prototypes (DTPs), Digital Twin Instances (DTIs), Digital Twin Aggregates (DTAs), and Digital Twin Environments (DTEs). Through this analysis, the paper highlights the advantages, limitations, and emerging trends associated with DT implementation in advanced intelligent manufacturing systems.

The structure of the rest of the paper is as follows: Section 2 categorizes the foundational concepts of DTs in smart manufacturing. Section 3 explores the general applications of DTs across various machining processes, emphasizing their versatility and adaptability. Section 4 identifies the challenges and future directions for DT implementation, focusing on their potential to enhance machining efficiency and adaptability. Finally, Section 5 concludes the paper by summarizing the key insights and reiterating the importance of DT technology in driving intelligent manufacturing forward.

2. Digital Twin Concepts for Smart Manufacturing

DT technology is a cornerstone in advanced manufacturing, bridging the physical and virtual worlds and enabling advanced control, analysis, and monitoring. It leverages several concepts like BD, IoT, PLM, CALS, and ML to enhance manufacturing processes, ensuring product quality and operational efficiency [19]. Some of these concepts are discussed below.

2.1. Big Data (BD)

BD is integral in handling and interpreting immense quantities and varieties of data, becoming pivotal in day-to-day analytical and operational processes. It plays a critical role in optimizing analytically substantial amounts of data, aiding in precise and informed decision-making and strategy development. Analyzing BD provides an array of benefits such as identifying optimal solutions, cost reductions, time savings, and the inception of new ventures [20]. The evolution of BD started to gain significant traction in the early 2000s. Doug Laney, an industry analyst, conceptualized BD based on three principal dimensions, namely, volume, velocity, and variety, laying the foundation for modern data analytics frameworks. In its early phases, the accumulation and storage of extensive data for analytical purposes were notably time-intensive processes [21]. In contemporary scenarios, the essence of BD is not just confined to the sheer volume of available information but extends to its application and utilization. It allows the extraction of data from diverse data sources, enabling the discovery of solutions that foster innovation and efficiency, particularly when coupled with advanced analytical tools [22]. In the context of smart manufacturing, the application of BD is paramount. When integrated with DT technologies, it brings forth the manifestation of intelligent manufacturing paradigms, maximizing operational productivity and excellence. It offers a transformative approach to manufacturing processes, enhancing their efficiency by enabling real-time data analysis and subsequent optimization of processes [23]. BD’s role in enhancing analytical processes and decision-making has been widely discussed, with significant contributions to fields such as smart manufacturing and investment risk realignment.

The conceptual framework that illustrates the interaction between the data source, data consumer, and the BD ecosystem is presented in Figure 1, highlighting the detailed relationships and dynamics within this ecosystem. At the base layer, data sources include various devices such as RFID systems, CNC machines, sensors, and databases, which generate diverse data types like temperature, motion, and operational metrics. This data can exist either at rest or in motion, with formats ranging from JSON and XML to web-based data, forming the foundation for advanced analytics.

The second layer, the Big Data ecosystem, serves as the core processing hub. It encompasses data storage solutions like HDFS and S3, resource management systems such as YARN and Kubernetes, and a suite of tools for data ingestion, processing, and analysis. Technologies like Kafka and NiFi handle data ingestion, while platforms like Spark and TensorFlow enable real-time and batch processing. Databases, including RDBMS (e.g., MySQL) and NoSQL (e.g., MongoDB), store structured and unstructured data. Additionally, visualization tools like Tableau v2024.2. and Matplotlib v3.8.4. transform raw data into actionable data. This layer is crucial for managing and analyzing large-scale data from manufacturing processes.

The topmost layer represents data consumers, which include various manufacturing applications relying on processed data for decision-making and optimization. These applications span a wide range of functions, such as predictive maintenance, process optimization, real-time monitoring, and quality control. For instance, predictive analytics can forecast tool wear and schedule maintenance to minimize downtime, while real-time monitoring ensures adherence to machining parameters for enhanced product quality. Overall, this ecosystem exemplifies the transformative potential of big data in manufacturing, enabling smarter, more efficient operations and supporting the goals of Industry 4.0.

The interaction of BD with advanced analytics yields substantial benefits in various business dimensions, contributing to organizational success. It enables the instantaneous identification of the underlying causes of failures and discrepancies, allowing for immediate rectification and preventive actions. This interaction also aids in the customization of sales strategies, such as the formulation of personalized discount coupons, based on the purchasing behaviors and preferences of customers, thus contributing to enhanced customer satisfaction and business growth [24]. Moreover, BD analytics facilitate the realignment of risk components associated with new investment portfolios, enabling organizations to make informed investment decisions and mitigate potential financial adversities. It provides an early detection mechanism for operational anomalies, preventing their escalation into significant disruptions and ensuring the continuity and smooth functioning of operations [25].

Figure 1. The conceptual framework of data source, data consumer, and BD ecosystem [26].

Figure 1. The conceptual framework of data source, data consumer, and BD ecosystem [26].

2.2. Internet of Things (IoT)

The integration of the IoT has profoundly simplified the implementation of the DT concept, enabling a seamless merger with machine learning and Artificial Intelligence (AI) technologies to enhance the functional paradigms of Cyber-Physical Systems (CPSs) [27]. This integration is especially notable in industrial engineering, promoting a comprehensive and innovative approach to various applications [28]. The ongoing challenges are primarily associated with achieving consistent awareness of physical processes within virtual settings and utilizing the acquired data for informed decision-making and action implementations [29]. Cyber-physical systems find utility across diverse sectors, including manufacturing, healthcare, energy systems, and consumer services, expanding the possibilities of innovative applications and solutions [30].

IoT’s symbiosis with DT is redefining manufacturing landscapes, offering streamlined, intelligent, and responsive systems, and marking significant advancements toward the realization of Industry 4.0 and smart factories. Figure 2 illustrates the hierarchical evolution of IoT applications, starting with individual smart devices and culminating in their integration within smart cities. At the center, IoT acts as a hub, interconnecting devices and systems to enable seamless communication and functionality. The progression begins with home appliances (1), representing traditional standalone devices. These evolve into smart appliances (2), equipped with IoT capabilities for enhanced connectivity and control.

Moving further, the concept of smart home (3) emerges, where interconnected devices collaborate to create an intelligent living environment, enhancing convenience, energy efficiency, and security. This concept scales up to smart buildings (4), where IoT integration facilitates advanced building management systems, optimizing resource usage and maintenance. Finally, the figure culminates in the smart city (5), showcasing the large-scale implementation of IoT to improve urban infrastructure. This includes applications like 5G networks, smart grids, and intelligent transportation systems, collectively driving sustainable and efficient city management. This exemplifies how IoT bridges the gap between individual devices and large-scale ecosystems, fostering innovation at every level [31].

2.3. Product Lifecycle Management (PLM)

DT technology, in synergy with PLM, is paving the way for innovative solutions addressing various real-world challenges, displaying its versatility across various applications such as analyzing fatigue in marine and wind turbines and evaluating corrosion resistance to enhance efficiency in machining processes. This synergistic integration is instrumental in exploring diverse solutions during the phases of refining production methodologies, prolonging product lifecycles, and fostering product innovation [32]. The incorporation of DTs facilitates a seamless exploration of varied solutions in the developmental and improvement phases of production processes, thereby broadening the spectrum of product lifecycle and development. In this context, financial considerations are paramount. Establishing or testing within a physical environment is notoriously capital-intensive. However, the inception of DTs, reflecting the authentic data of the real-world physical environment, has revolutionized these solutions and procedures, making them highly feasible and cost-efficient [33]. PLM, when harmonized with DTs, provides a comprehensive framework to manage products right from their conceptualization to obsolescence. It amalgamates data, processes, systems, and human resources to ensure a cohesive alignment between the physical and digital counterparts [34]. This alignment is crucial for enhancing product development and quality while minimizing the time-to-market by allowing for the perpetual improvement of products through the utilization of real-time data [35].

This alignment and integration are pivotal for driving operational effectiveness and fostering innovations in product development. The assimilation of real-time data ensures that products are not just conceptualized based on innovative ideas but are also continually refined and optimized throughout their lifecycle. This continuous optimization and refinement process is not just crucial for maintaining product relevance in dynamic markets but also for ensuring that products meet the evolving needs and preferences of the end-users, thereby contributing to enhanced customer satisfaction and business growth [36].

The strategic integration of PLM and DT technology is not just transforming the way products are managed throughout their lifecycle but is also contributing to the realization of more sustainable, efficient, and innovative production ecosystems [37]. Figure 3 illustrates the comprehensive framework of PLM, highlighting its role in harmonizing digital and physical product counterparts to enhance product development and lifecycle management. It is allowing organizations to be more responsive and adaptive to market changes and customer needs, ensuring that products do not just meet the current market standards but are also aligned with future market trends and demands [38].

2.4. Continuous Acquisition and Lifecycle Support (CALS)

CALS technology is pivotal, offering revolutionary potential in diminishing the extensive efforts traditionally involved in design processes. It provides detailed descriptions of numerous components within existing machines, equipment, and systems. This technology proposes a unified format for data interchange, accessible to all users of CALS technology, streamlining complexities inherent in traditional models, as illustrated in Figure 4 [40]. The utility of CALS technology extends to addressing maintenance-related concerns and harmonizing products with varied system types, and its adaptive capabilities ensure its viability in evolving operational landscapes. The success of CALS technology in the domain of complex technical products is believed to be contingent upon several elements, including environmental considerations, the technology’s adaptability, and the specificity of the design organization [41].

Progression in CALS technology is anticipated to facilitate the advent of ‘virtual emergence’. This implies a transformative shift in manufacturing processes that dictates the formulation of software requisites. It enables the installation of sufficient technological tools, both spatially and temporally, required for product realization. It allows collaborative efforts between various entities and autonomous design studios, facilitating the execution of complex design solutions with relative ease [42].

CALS technologies underpin the creation of an extensive, open network focusing on automation design and industrial management. The essential part of implementing CALS technology lies in ensuring uniformity in data representation and interpretation, irrespective of its point of origin. A modular system evolving to achieve universal status combines design, technology, and operation. True efficacy is realized through the standardization of documents and representational languages, separation in time and space, and the utilization of diverse tools by different teams working collaboratively on a singular project, such as CAD/CAM/CAE systems [43]. In such a modular approach, design documents find versatility in their applicability and are adaptable to varied manufacturing scenarios, leading to substantial reductions in overall design and production costs. Moreover, it simplifies the operational aspects of the system, allowing for more seamless and streamlined functioning. In synergy with DT, CALS fosters the acquisition of real-time data and provides continual support across all phases of a product’s lifecycle, enhancing the overall sustainability and efficiency of manufacturing processes. The intertwining of CALS with DTs is instrumental in refining resource allocation, mitigating operational halts, and augmenting the robustness and flexibility of manufacturing systems in the face of environmental variabilities. This complex relationship marks a significant milestone towards evolving manufacturing processes and systems, adapting to the continuous advancements and fluctuations in industrial landscapes [42].

In the context of machining processes, Continuous Acquisition and Lifecycle Support (CALS) technology offers transformative capabilities by streamlining data management and operational efficiency. It enables the integration of design, manufacturing, and maintenance processes through a unified data format [44]. This standardization facilitates seamless collaboration between various teams and systems, including CAD/CAM/CAE tools, ensuring consistent data interpretation and application across the lifecycle of machined components. By addressing maintenance concerns and enabling real-time data acquisition, CALS supports adaptive machining strategies that can respond to dynamic operational conditions [45]. Its modular approach reduces design and production costs, enhances resource allocation, and minimizes downtime, ultimately improving the sustainability and robustness of machining systems [46]. The synergy between CALS and Digital Twin (DT) technologies further optimizes machining processes by providing predictive data and enhancing decision-making in complex manufacturing environments [47].

2.5. Machine Learning (ML)

Machine learning (ML) within DT architectures enables the execution of complex and costly operations, first in a virtual setting, before implementing them in reality, based on derived outcomes. This capability provides a strategic advantage, allowing such operations to be conducted more effortlessly and economically by analyzing and interpreting real-time data in the virtual domain. It provides the facility to conduct multiple simulations to digitally assess innovations before their application in the real world. The integration of sophisticated methods such as artificial intelligence and machine learning facilitates significant enhancements in numerous processes. Observing tangible issues within the DTs before translating them into the actual production space is a more logical and economical approach. The combination of machine learning and artificial intelligence not only aids in the meticulous analysis of the present circumstances but also extends to forecasting impending scenarios [48]. Such predictive capabilities induced by machine learning and artificial intelligence play a significant role in providing substantial cost-related benefits to enterprises. They allow for the identification of underlying patterns, the formulation of informed predictions, and the optimization of overall performance within DT frameworks. The convergence of machine learning models with DTs aids in the prediction of machinery breakdowns refines maintenance timelines and fosters perpetual enhancements by assimilating information from operational data. Figure 5 demonstrates the integration of ML within manufacturing processes, highlighting its role in optimizing operations and predicting outcomes in real-time. This synthesis is instrumental in progressing towards the development of self-reliant and sophisticated manufacturing ecosystems. It augments the capabilities of DTs by enabling them to learn, adapt, and optimize, thereby magnifying their utility and efficacy in manufacturing processes. Such advanced integrations are paving a new era in manufacturing, characterized by heightened intelligence, autonomy, and efficiency, poised to redefine industrial paradigms and contribute significantly to organizational success and innovation [49].

The progressive taxonomy of DT, from prototypes (DTPs) to individual instances (DTIs), aggregated systems (DTAs), and encompassing environments (DTEs), illustrates the comprehensive and layered approach to translating physical entities into digital counterparts. Each layer, with its unique functionalities and applications, contributes to optimizing design principles, operational efficiencies, system interactions, and overall environmental coherence, highlighting the multifaceted and transformative potential of DTs in various industrial applications. This hierarchical structure underlines the intricate and scalable nature of DTs, serving as the backbone for the next generation of intelligent and responsive manufacturing systems.

• Digital Twin Aggregate (DTA)

A DTA combines multiple DTIs, representing either a singular system or a process. This aggregation facilitates a comprehensive examination and refinement of interlinked components, assuring coherence, reliability, and efficacy across the system. DTAs are essential for overseeing complex systems, yielding data into systemic interconnections and dependencies, and enabling the refinement of the entire system’s performance. Whether co-located within a singular entity or distributed across various entities, DTAs have the potential to unveil novel and unexpected data, going beyond the mere summation of individual behaviors [50].

2.

DT Environments (DTEs)

A DTE defines the overarching ecosystem wherein a DT functions, incorporating diverse technologies, platforms, norms, and protocols. DTEs secure the interoperability, scalability, and protection of DTs, enabling their amalgamation into varied operational contexts. They are pivotal for unlocking the extensive potential of DTs, granting them the capacity to interact, adapt, and progress within cohesive and fluctuating operational terrains. It is within DTEs that DTs undergo simulation, modeling, and evaluation, ensuring their applicability and effectiveness in replicating real-world environments and conditions [51].

3.

DT Instance (DTI)

A DTI is the specific digital representation created for every distinct physical entity, facilitating instantaneous observation, examination, and regulation of the respective physical counterparts. DTIs are paramount for sustaining the operational vitality and longevity of individual entities, enabling preemptive maintenance, enhancing operational efficacy, and prolonging the operational life of the physical entity through perpetual synchronization of data between the tangible and virtual worlds. Once established, DTIs keep receiving data from the real world, allowing monitoring, predicting system behavior, and ensuring any modification in one realm is mirrored in the other, maintaining bidirectional coherence [1].

4.

DT Prototype (DTP)

DTP serves as the foundational model, epitomizing the initial conceptualization of a physical object. It plays a vital role in assessing and confirming the design ideologies, operational functionalities, and performance parameters before the commencement of actual manufacturing. DTP is integral in mitigating design discrepancies and assuring the dependability of the product, fine-tuning design elements to conform to predetermined specifications and operational prerequisites. Utilizing Computer-Aided Design (CAD) software, a virtual replica is created, where analog data collected via sensors is digitized and superimposed on the virtual model, enabling simulations that mirror actual operating conditions, especially critical for products developed for sensitive missions [15].

In the machining process context, the integration of Machine Learning (ML) within Digital Twin (DT) frameworks revolutionizes operational efficiency and precision [52]. By simulating complex machining operations in a virtual environment, ML models predict outcomes such as tool wear, surface finish, and material behavior under varying conditions [53]. This capability enables real-time adjustments to machining parameters, optimizing processes while minimizing waste and downtime [54]. Additionally, ML-driven data enhances predictive maintenance by identifying patterns in operational data, forecasting machinery breakdowns, and refining maintenance schedules [55]. Through continuous learning and adaptation, ML empowers DT systems to improve machining accuracy, reduce costs, and foster sustainable manufacturing practices, paving the way for intelligent, autonomous machining ecosystems [56].

3. Applications of DTs in Machining

DTs can be applied in many manufacturing fields, machining included, despite being a new concept [35]. Corallo et al. [57] further states that the technique allows the monitoring, simulation, and improvement of the machining process, especially decision-making, process optimization, predictive maintenance, production plan, equipment lifetime, and fault diagnosis. Furthermore, as stated by He et al. [58], those improvements provided by the DT are a crucial step towards more sustainable manufacturing.

3.1. BD in Machining

Data gathering is crucial in implementing a DT in any process [59]. The success of the implementation can be increased as the volume and variety of data increases, usually collected in real or near real-time and by means of cloud storage. This strategy usually requires high-end computer processing capacity to implement the models and simulations necessary for DT and high-end data transmission, now possible due to 5G technologies. Another critical point is that the use of a vast array of sensor types allows for greater fidelity in the representation of the DT compared to its physical counterpart [58]. The advent of numerical control was one of the main revolutions in manufacturing, especially in machining. It was previously reported that optimizing these parameters can reduce production time and lower tool damage, thus significantly improving process rentability [59]. Indeed, [59] evaluated the use of BD to create a DT model and optimize tool pathing as input variables, which evaluated the tool path, vibration, and cutting forces. Based on analysis and DT implementation, the processing time was reduced by up to 50%. Vishnu et al. [53] reported comparable results using a BD-driven DT for predicting surface roughness in the milling process.

One of the main advantages of BD is the reuse of previous data to accelerate the process planning of new products. Liu et al. [59] investigated this approach using a model of DT process knowledge applied to diesel parts machining. The data was divided into machining feature type, tool access direction, machining face, typology relation, workpiece size, and process equipment information. The experimental tests were conducted oon turning and milling processes under roughing, semi-finishing, and finishing conditions. Using those parameters, the authors developed a BD-driven process-planning methodology, effectively increasing process-planning speed with data reuse. Data construction and management are critical aspects in constructing and improving the reliability and manageability of a BD system. Kong et al. [60] proposed a data construction method applied to a DT machining application composed of the following framework for the data: analysis of functional requirements, hierarchical representation, and characteristic representations; pre-processing of the raw data; customized processing; and large-scale storage and retrieval [61]. Those functions were allocated in data representation, organization, and management modules. Using DT technology, the authors applied the methodology in CNC machining centers and improved the machining conditions regarding tool wear and pre-load conditions, reaching modeling levels with up to 99.7% accuracy using the Random Forest algorithm.

As Wang et al. [62] reported, remanufacturing is an effective approach to conserve resources and promote greener manufacturing practices. The authors proposed a BD-driven hierarchical things predictive remanufacturing framework, summarized in Figure 6, to enhance the efficiency and feasibility of remanufacturing processes.

This framework integrates Big Data (BD) collected through advanced fusion sensor technologies connected via IoT to a centralized database. Using machine learning algorithms, the data is processed and analyzed to create a Digital Twin (DT) model. This DT model predicts potential challenges in the product lifecycle, optimizes remanufacturing operations, and informs future cycles by feeding insights back into the database.

The system is structured around two primary platforms: the Remanufacturing Enterprise Platform (RECPSDT) and the Manufacturing Enterprise Platform (MECPSDT). These platforms are interconnected through a multi-lifecycle decision-making layer powered by Cyber-Physical Systems (CPSs) and DT technology. The RECPSDT focuses on remanufacturing tasks such as cleaning, disassembly, and reassembly, while the MECPSDT handles manufacturing nodes, intelligent equipment, and workshop platforms, enabling real-time data sharing and intelligent negotiation.

Big Data analysis is central to this framework, employing clustering, classification, association, and prediction techniques to process information from IoT-enabled sensors. This data-driven approach supports predictive decision-making and optimizes workshop operations. Additionally, IoT nodes are deployed throughout the system to monitor the status of products in the market, enabling edge computing and seamless communication with the cloud platform.

The integration of advanced technologies, such as PLC, CNC, MTConnect, and EtherCAT, further enhances the system’s ability to manage multi-life cycle information interactively. Visualization tools, including VR/AR, provide real-time insights into the remanufacturing process, offering a comprehensive view of operations. In summary, the BD-driven hierarchical framework represents a robust and sustainable approach to predictive remanufacturing, leveraging Big Data, IoT, and digital twin technologies to optimize operations and reduce uncertainties in production. Figure 6 illustrates the detailed structure and workflow of this innovative framework.

Cellular manufacturing is often the most responsive manufacturing arrangement for industries that follow the Just-In-Time methodology. The pressure to increase cell autonomy was the justification for the Zhang et al. [63] research to develop a DT manufacturing cell framework. According to the authors, data gathering and processing at the speeds necessary to achieve the responsiveness needed for modern applications is possible due to BD, IoT, and deep learning technologies. In their work, the machining cell was composed of milling and turning machines with robots to perform the material handling, further decreasing the processing time and increasing the system responsiveness. However, it was also reported that the complexity of the DT drastically increases, thus the importance of efficient data gathering, storage, and processing. Among the benefits of the DT in this application are increased flexibility and lower production cost allowed by the better process optimization, prediction, and control strategies compared to the conventional controlling strategies. Dai et al. [64] presented a modeling method to virtualize machined parts used in DT modeling, a critical application to implement BD. The method relies upon the design data (dimensional, tolerance, and topological variables), inspection data (mainly from a metrological standpoint), processing data (machining parameters, machining and cutting tool, environment), and other additional data (material supplier, usage). The method uses machine learning to concatenate the data model of the DT part and label it so that its design and production cycle can be recovered and used as a model for similar parts. The authors presented a case study to apply a generic test part manufactured in an aviation plant. The results indicate that the methodology outperforms the usual methods such as 3D scanning, as it provides a digital mode (twin) of the part and can be used to optimize its production lifecycle. A summary of the papers reviewed in this section is presented in

Table 1. Summary of the papers presented in Section 3.1.

Ref.	Type of Machining	Application	Technologies Applied	Key Benefits
[59]	General Machining	Data gathering and real-time representation using DT	Cloud storage, 5G technologies, advanced sensors	Improved fidelity, reduced production time, and tool damage
[65]	CNC Machining	Tool path optimization and vibration analysis	Big Data (BD), DT, numerical control	Reduced processing time by 50%, improved rentability
[53]	Milling	Surface roughness prediction using BD-driven DT	IoT, BD, machine learning	Enhanced surface quality
[60]	CNC Machining Centers	Data construction for DT applications	Random Forest algorithm, BD, hierarchical data representation	Improved tool wear and pre-load condition modeling with 99.7% accuracy
[62]	Remanufacturing	Predictive remanufacturing framework using BD-driven DT	IoT, machine learning, CPS, VR/AR, advanced fusion sensors	Optimized operations, reduced uncertainties, greener manufacturing practices
[63]	Cellular Manufacturing	DT manufacturing cell framework for responsive Just-In-Time (JIT) operations	BD, IoT, deep learning, milling and turning machines, robotic material handling	Increased flexibility, reduced production cost, better optimization strategies
[64]	Aviation Machining	Virtualization of machined parts for DT modeling	Machine learning, dimensional and tolerance data, inspection data	Improved production lifecycle optimization

.

3.2. Internet of Things (IoT) in Machining

The IoT is one of the technologies that present better interaction with DT manufacturing [58]. IoT allows more efficient real-time data gathering from all the sensors attached to the machine, thus enabling a more precise representation of the physical part. The large amount and significant variability of data usually needed for BD processing could only be achieved using IoT, especially when real-time gathering and modeling are needed [66].

Liu et al. [67] developed an IoT-enabled framework to facilitate dynamic control and processing within a Digital Twin (DT) manufacturing unit, as illustrated in Figure 7. The framework is organized into five interconnected layers, each playing a specific role in the system’s operation. The application layer serves as the core of the system, integrating various servers such as ERP, CAPP, and MES to manage quality monitoring, control, and data processing for all subsystems. Supporting this is the data management layer, which provides essential databases including a perception database for machining quality, a process knowledge database for operational insights, and a quality inspection algorithm library for decision-making.

The data transport layer connects all components through wired and wireless networks, ensuring seamless data transmission across layers and devices. The data perception layer includes sensors, RFID readers, bar code scanners, and control units like PLC and CNC systems, which gather real-time data from the manufacturing environment to enable process monitoring and automation. At the foundation is the manufacturing unit layer, comprising the physical elements where machining occurs. This layer captures data related to equipment status, workpiece processing, production progress, and environmental parameters, ensuring accurate interaction between physical processes and the digital system. The integration of these layers highlights the role of IoT in connecting data acquisition, processing, and control functions. The system was validated by monitoring machining parameters such as roundness, symmetry, and perpendicularity during the production of connecting rods, demonstrating its reliability in process control and quality assurance.

Zhang et al. [68] expanded the concept of IoT to information and communication technologies (ICT), composed of IoT, virtual reality, BD, 5G, artificial intelligence, and other interconnection technologies. According to the authors, all these technologies can improve data gathering and processing for DT implementation. The authors present a study case of application in a manufacturing shop floor to process turbine fan blades. They reported that ICT improved the process control and management from the forging of the raw material to the final machining finishing steps. Qiao et al. [69] evaluated using a DT model with IoT as a data gathering and communication tool to achieve a tool condition prediction model. The authors also used a deep learning technique called Deep Stacked GRU to identify and predict the tool wear. The input variables were obtained from vibration, acoustic emission, and force sensors measured on a milling machine. The authors reported that using the DT model with real-time data gathering greatly improved the tool condition compared with standard machine learning methods.

Thin-walled machining is still one of the main challenges in the field due to the low structural resistance of the manufactured part, which poor choice of machining parameters can easily compromise [70]. Those challenges are even more significant for high-complexity parts such as in turbine blades of an automotive turbo-compressor. Researchers evaluated a digital-twin-driven machining process, using IoT to gather real-time data from 2D camera images and convert it into a 3D physical space of the workpiece and machine tool condition and force and vibration sensors. Based on the data, the cutting forces were modeled using an Ansys v2018 R1, enabling cutting parameters and tool path optimization. The optimization of a production line, even for a seemingly straightforward process such as pipe-cutting can be the critical difference in a highly competitive global market, the critical difference in a highly competitive global market [71]. The data gathering and communication between the PLM, enterprise resource planning, DT system, manufacturing operation management, and field control system were achieved using IoT technology for real-time data gathering and processing. The data was stored using a MySQL database and processed using neural networks, leading to an optimized Gantt graph with the production line tasks, increasing the overall process efficiency. In this regard, Xi et al. [72] evaluated the use of IoT, called by the authors IoP (Internet of Production), as the method for real-time data gathering intended for modeling the DT of roughing and finishing milling processes. According to the authors, the difference in the proposed approach is that the data was gathered from built-in sensors as well as the numerical control of the production process. The authors reported that throughout twenty batches, the process quality indicator rose from around 30% to more than 70% for both finishing and roughing processes. A summary of the papers reviewed in this section is presented in

Table 2. Summary of the papers presented in Section 3.2.

Ref.	Type of Machining	Application	Technologies Used	Key Benefits
[58]	General machining	IoT integration with Digital Twin (DT)	Sensors, real-time data gathering	Precise representation of physical parts, enhanced data variability for Big Data (BD) processing
[67]	Milling and quality monitoring	IoT-enabled framework for dynamic control and processing in DT manufacturing	IoT, ERP, CAPP, MES, perception database, process knowledge database, quality inspection algorithm library	Improved process monitoring, quality assurance, and dynamic control during machining
[68]	Turbine blade machining	Expansion of IoT to ICT for turbine blade manufacturing	IoT, Virtual Reality (VR), Big Data, 5G, Artificial Intelligence (AI)	Enhanced process control and management from raw material forging to final machining steps
[69]	Milling	Tool condition prediction model using IoT and DT	Deep Stacked GRU, vibration, acoustic emission, force sensors	Real-time tool wear prediction, improved accuracy over standard machine learning methods
[70]	Thin-walled machining (e.g., turbine parts)	Thin-walled part machining using IoT-driven DT	IoT, 2D image to 3D space conversion, ANSYS simulation, force and vibration sensors	Optimization of cutting forces, machining parameters, and tool paths for high-complexity parts
[71]	Pipe cutting	Pipe-cutting production line optimization	IoT, MySQL database, neural networks	Real-time data gathering and optimized Gantt chart generation, leading to increased process efficiency
[72]	Roughing and finishing milling	IoP (Internet of Production) for roughing and finishing milling processes	Built-in sensors, numerical control data	Process quality improvement from 30% to over 70% through real-time data gathering and modeling

.

3.3. Product Lifecycle Management in Machining

DTs are mostly used in the design, production, and Population Health Management (PHM) [73], as illustrated in Figure 8. In product design, the use of DTs increases the responsiveness of the manufacturing system to the market needs. DTs can be applied to increase the overall reliability of the process since they allow a combination of monitoring and prediction of the process parameters that will be used. DT technology also allows a less empirical and more statistically accurate PHM compared to traditional methods. Tao et al. [73] also remarked that DTs are significantly used in other areas to increase process flexibility.

Comparatively, Liu et al. [74] divide the lifecycle into design, manufacturing, service, and retirement. The design phase was further subdivided into task clarification, conceptual design, embodiment design, and detail design. In all those steps, the DT assists the designer in making the footprint for the processed workpiece by compiling all the data in useful information. In the manufacturing step, the DT can help communicate the digital and physical aspects of the process, enabling increasingly self-aware machines to make more intelligent decisions regarding production control, evaluation, and optimization. The service phase is more out of the control of the manufacturers, and the DT can help differentiate different product batch performances and reliability. Finally, in the retirement phase of the physical product, most of the data from the user is lost, and the DT can be stored as lifecycle information of the product, significantly, if the product is returned to production in the future.

Managing the tool lifecycle is an application that relies heavily upon data collection [75], as illustrated in Figure 9. The process is divided into market analysis, which usually employs BD analysis of customer demand, market share, production capability, and stock management. The tool development relies on function and machining parameter data such as tool geometry and optimal machining parameters, which can be further improved using DTs. Tool manufacturing relies on process and machine-tool data, following a similar pattern described by Tao et al. [73]. The usage and service can be modeled via failure and maintenance data, usually using force, thermal, acoustic emission, electric power, and vibration sensors, and directly inputted into the DT.

Botkina et al. [76] studied the methodology to develop a DT cutting tool based on data collected on the tool lifecycle. The digital tool model is based on ISO 13399, using CAD, CAM, and the software ToolMaker^® to gather data and design the model, coupled with additional data such as tool compensations and machining parameters, using machine learning to compile the data. The methodology was used to model the tool used in the machining process of a cylinder head, increasing productivity by up to 15%. The workpiece clamping and positioning are among the most time-consuming variables in the machining setup, and crucial in the manufacturing lifecycle. Liu et al. [77] studied DT technology to increase the reliability and flexibility of clamping and positioning systems of cylinder heads. This application requires modeling the workpiece and the fixture system; otherwise, the geometric representation will not be accurate. The data flow of the model is composed of acquisition, data fusion, and data decision layers. The input data for the modeling was acquired from pressure sensors used to control the hydraulic cylinders of the positioning system. The authors reported that using a DT decreased from 30 to 20 min in setup time compared to the traditional positioning previously used.

Hänel et al. [78] evaluated the use of a DT to control the production lifecycle of machining stainless steel levers intended for aerospace applications. The input data for DT modeling was gathered from acoustic emission, vibration, and force sensors. Based on the data, the DT was used to model the cutting force for each tool position during the tool path, improving tool life and production time. Cheng et al. [79] reported comparable results for online quality control of marine diesel engine machining. In the authors’ study, the DT was composed of the machining process and the entire machining workshop, leading to an accurate representation of the process with a mean average error of 5.223% regarding tool life prediction. DTs can also be implemented to increase process sustainability and lower carbon emissions [80]. Researchers investigated this possibility by evaluating the machining lifecycle and modeling the process, thus optimizing the machining parameters to be more energy efficient. The presented case study in the milling process of S45C carbon steel was achieved by optimizing the spindle speed, feed rate, cutting forces, and processing time, which successfully decreased the equivalent carbon emissions of the process by 6.1%. Lifecycle management is commonplace in the natural process. Liu et al. [56] evaluated the use of a biomimicry-based digital twin system, where multiple DT-sub models were used to simulate a perceive–stimulate response, such as those in chameleon skin color-change biomimicry, as illustrated in Figure 10. The proposed model differs from the existing one by simulating the machine/tool pair and evaluating the model by geometry, behavior, and context aspects. The authors evaluated the process in a missile air rudder machining case study. The authors reported that, despite the success in modeling the process, the modeling speed and simulation lag were excessive, which can make the approach unsuitable for more complex models. A summary of the papers reviewed in this section is presented in

Table 3. Summary of the papers presented in Section 3.3.

Ref.	Type of Machining	Application	Technologies Used	Key Benefits
[73]	General machining	Product lifecycle management with Digital Twin (DT)	DT, market analysis, process monitoring, predictive modeling	Improved reliability, flexibility, and responsiveness to market needs
[74]	Design and manufacturing	Lifecycle divided into design, manufacturing, service, and retirement stages	DT, task clarification, conceptual design, embodiment design, detail design	Enhanced decision-making, self-aware machines, and intelligent production control
[75]	Tool lifecycle management	Data-driven management of cutting tools	DT, force, thermal, acoustic emission, electric power, and vibration sensors	Optimized tool development, manufacturing, usage, and service; improved failure prediction and maintenance
[76]	Cylinder head machining	Digital twin-based modeling of cutting tools	ISO 13399, CAD, CAM, ToolMaker®, machine learning	Increased productivity (up to 15%), optimized machining parameters, and reduced setup time
[77]	Cylinder head clamping and positioning	Dynamic clamping and positioning system for flexible tooling	DT, pressure sensors, hydraulic cylinders, data acquisition, fusion, and decision layers	Reduced setup time (from 30 to 20 min), improved reliability and flexibility
[78]	Aerospace component machining	Production lifecycle control for machining stainless steel levers	DT, acoustic emission, vibration, and force sensors	Improved cutting force modeling, tool life, and production time
[79]	Marine diesel engine machining	Online quality control of machining processes	DT, workshop-wide process modeling	Accurate tool life prediction (mean average error of 5.223%), enhanced process representation
[80]	Milling of S45C carbon steel	Optimization for energy efficiency and reduced carbon emissions	DT, spindle speed, feed rate, cutting forces, processing time	Decreased carbon emissions by 6.1%, improved energy efficiency
[56]	Missile air rudder machining	Biomimicry-based digital twin modeling	DT sub-models, geometry, behavior, and context evaluation	Simulation of machine/tool pairs, enhanced process modeling, but challenges with simulation lag and modeling speed

.

3.4. Continuous Acquisition and Life Cycle Support in Machining

DT technology has been effectively utilized as a web-based method for monitoring machine tool conditions, with real-time machining forces serving as input data. This approach, as evaluated by Liu et al. [81], has been shown to enhance system reliability by enabling continuous monitoring of the process through a DT accessible on a web page. Tong et al. [82] developed a real-time data application DT-driven machining service called Intelligent Machine Tool (IMT) that uses multisensory fusion to model and optimize the process. The proposed framework for the IMT DT is designed to map the physical system, using as many sensors as possible, such as encoders, torque, current, voltage, accelerometers, acoustic emission, microphones, thermometers, code recognition, energy consumption, and industrial cameras. This data is used to model the DT in real-time, which will give feedback to the physical machine tool. Among the feedback proposed in the model is the correction in the machine parameters to solve contour errors, machine vibration, and poor surface finish. An example of the effect of the feedback in surface finish regarding vibration control in the machining process of a turbine blade is illustrated in Figure 11, where Figure 11a shows the surface of the blade before the finishing process, Figure 11b shows the finishing process under conventional parameters, and Figure 11c show the finishing process with the feedback from the DT. Zhang et al. [83] reported comparable results in the same turbine blade application.

Most information and communication technologies are not readily suitable for industry because of the responsiveness implementation cost. Armenia et al. [84] proposed using DTs to control the milling process in an industrial environment capable of continuously improving the overall equipment efficiency during the entire process lifecycle. The DT was created using a combination of 3D multi-body simulation and the finite element method. Spindle power, torque, and forces were chosen and continuously monitored as input variables for the modeling process, as those variables can be correlated with tool wear. The authors reported that the use of the DT enabled a decrease in machining time by up to 4% and process setup time by up to 11%, leading to an overall 1.1% reduction in tooling costs. Augmented Reality (AR) is one of the most promising modern technologies as it can be applied from product design to process diagnostics. Zhu et al. [85] proposed a method to visualize the DT data using AR, as illustrated in Figure 12. The author reported that to accurately represent the process, both physical and digital parts of the system must be calibrated appropriately, and the data input must be carried out in real-time, otherwise, the AR will only display outdated information. The main advantages of using the AR interface are increased awareness of the process state and increased control of the variables that can be carried out in real-time with voice commands or gestures. Similarly, Liu et al. [86] also investigated the use of DT-driven AR. The augmented reality, using the machining process of a carrier box as a case study as input variables, used the CAD/CAM data from the intended workpiece and tool, force, temperature, and vibration sensors, and machining parameters from the machine tool controller. The authors also reported an increase in system responsiveness due to the real-time display of machine data and more accessible and interactive process control.

An innovative approach using an improved extended short-term memory network has been evaluated as a parameter optimization strategy for thermal error control in DT machining systems. This framework, named the four-terminal architecture by Liu et al. [87], consists of four key components: (1) the machine terminal, equipped with sensors for stress, video, audio, temperature, deformation, lubrication, positioning, process parameters, and electric current to enable continuous data acquisition; (2) the control terminal, featuring the CNC controlling platform for machining parameters; (3) the data terminal, responsible for data storage and fusion management; and (4) the intelligent decision-making terminal, which includes GPU computing centers for data processing and DT generation. The authors reported achieving thermal error predictions exceeding 90% across all evaluated conditions using this methodology.

Luo et al. [88] proposed a hybrid approach to maintaining a machining tool using the DT technology. The framework of the method consists of a combination of data-driven and model-based techniques. Combined with Multiphysics simulation, the authors were able to increase the DT fidelity to the physical model as input data, acoustic emission, vibration, and machining forces were evaluated. The authors reported that using this hybrid approach decreased the deviation of the model compared to a pure data-driven model. Cao et al. [89] studied DT-driven real-time cutting simulation and CNC controlling. The simulation was based on real-time monitoring and CAD/CAM data, implemented in the DT, and monitored and improved machinability in the physical process. According to the authors, the methodology still needs improvements, especially regarding the simulation updating time. A summary of the papers reviewed in this section is presented in

Table 4. Summary of the papers presented in Section 3.4.

Ref.	Type of Machining	Application	Technologies Used	Key Benefits
[81]	Machine tool condition monitoring	Real-time web-based monitoring of machining forces	DT, web-based monitoring	Enhanced system reliability through continuous monitoring
[82]	Turbine blade machining	Intelligent Machine Tool (IMT) for real-time process optimization	DT, multisensory fusion: encoders, torque, current, voltage, accelerometers, acoustic emission, cameras	Improved surface finish, reduced contour errors, and vibration control
[83]	Turbine blade machining	Application of DT feedback for surface finish optimization	DT, real-time feedback	Enhanced machining accuracy and quality
[84]	Milling process	DT for industrial milling process lifecycle optimization	DT, 3D multi-body simulation, finite element method	Decreased machining time (4%), setup time (11%), and tooling costs (1.1%)
[85]	General machining	Visualization of DT data using Augmented Reality (AR)	AR, real-time data calibration	Increased process awareness and interactive control via voice commands or gestures
[86]	Carrier box machining	DT-driven augmented reality for process monitoring	DT, AR, CAD/CAM data, force, temperature, vibration sensors	Improved system responsiveness and interactive process control
[87]	Thermal error control in machining	Four-terminal architecture for thermal error prediction	DT, extended short-term memory network, sensors (stress, audio, temperature, deformation, etc.), GPU computing	Thermal error prediction accuracy exceeding 90%, improved process reliability
[88]	CNC machining tool maintenance	Hybrid predictive maintenance approach	DT, Multiphysics simulation, data-driven and model-based techniques	Increased DT fidelity, reduced model deviation
[89]	CNC machining	Real-time cutting simulation and CNC control	DT, CAD/CAM data, real-time monitoring	Improved machinability monitoring; identified need for simulation update improvements

.

3.5. Machine Learning in Machining

ML is among the most common tools to concatenate and process the data from the plethora of sensors used as input variables in DT modeling [90]. Among the main difficulties of implementing machine learning is the relationship between the model’s accuracy and the volume of data collected; however, this problem can be promptly solved by increasing the machining test repetitions [91]. The use of these techniques allows the correlation of indirect variables to the process, such as electrical power consumption, with some of the main output variables related to machineabilities such as tool wear and surface integrity. Among the most evaluated techniques are the random forest, decision tree, neural network, genetic algorithm, and generalized linear model, applied to both conventional [92] and non-conventional [93] machining processes. For example, Ladj et al. [94] studied the digital shadow as a data collection tool to process data from the physical counterpart and a digital model from the DT, as illustrated in Figure 13. The design of the digital shadow consists of leveraging data and knowledge models. The digital shadow algorithm has data processing and aggregation algorithms that will create the set laws followed by the model, resulting in the DT. The authors presented a case study of applying the digital shadow methodology in aeronautic machining, using the machine tool vibration as the primary input data. The knowledge model used for data analysis is based on the Gaussian mixture model machine learning method. This model is used to estimate the distribution of random variables, in this case, from the accelerometer signal. A tool failure rule was created using this method and data was gathered from previous production cycles. The authors reported that the methodology successfully detected tool failure and addressed chatter and faulty programming conditions.

Liu et al. [14] investigated the use of a multi-scale evolution mechanism to generate knowledge models for DT machining applications. The proposed model’s advantage is its ability to ensure product quality control at the micro, meso, and macro scales of the manufacturing process. The framework of the method includes product data and a knowledge base. The knowledge generation architecture is divided into data pre-processing, classification, fusion, and conversion modules. Input variables are derived from vibration, force, and temperature sensors, using a multi-objective optimization algorithm to process the data and correlate these signals with dimensional accuracy and surface roughness. The results demonstrated that the model successfully mimicked the physical model concerning the chosen output data.

Deeback et al. [95] used Deep Transfer Learning (DTL) as a tool to improve fault diagnosis in the DT-assisted machining process. The technique was applied to the milling and drilling process monitoring to monitor and predict tool wear, chipping, breakage, chatter, and faulty tools. The authors reported that the DTL method improved accuracy (up to 92.33%) compared to traditional deep neural network methods. Ma et al. [96] evaluated using a self-learning digital-twin-based framework to control machining. The framework used long short-term memory and neural network methods for data processing and the Bayesian algorithm for data optimization. Using the methodology, the authors successfully modeled the thermal elongation and temperature of the process. The error prediction model can self-learn and reduce, after more iterations, the incoming data.

Surface roughness, tool wear, and power consumption are the most used machinability standards. Liu et al. [97] evaluated the use of the DT method to improve the surface roughness prediction using adaptative optimization. The framework was composed of three layers: (1) the data acquisition layer that is responsible for the process data acquisition and mapping of the DT; (2) the data processing layer that, as the name suggests, realizes the real-time data transmission, storage, and processing; and (3) the service layer, that is imbued with the task of real-time monitoring and process optimization. The optimization was made using the improved particle swarm optimization-generalized regression neural network, which optimized tool wear and surface roughness based on input data such as electric current and voltage, energy consumption, vibration, and noise. The authors reported that combining the DT approach with the proposed machine learning method to ensure higher accuracy in the process modeling leads to a significant decrease in surface roughness and tool wear. Zhao et al. [98] evaluated a self-learning and adaptation DT-driven roughness stabilization methodology using the pigeon-inspired optimization and support vector machine for this task. Tool geometry, machining parameters, and cutting force were used as input parameters. The experimental tests were performed on the five-axis machining of an aeronautic AL7075 alloy under a factorial design set of experiments, and the average roughness (Ra) was measured before and after each tool pass. The data was modeled into the DT, and the machining parameters were adjusted in the case of unstable variations of the Ra. The authors reported prediction errors below 9% for the proposed methodology.

Sun et al. [99] studied the use of neural networks to model a DT-based federate and machine cycle time prediction method. The system input variables used to train the model were the feed rate, tool acceleration, and toolpath as, according to the authors, most of the models proposed in the literature do not consider those parameters. The authors reported that using neural networks, the DT was able to model the cycle time with up to 90% accuracy. Ward et al. [100] reported comparable results also using adaptative feed rate control in a real-time DT-based approach capable of residual stress simulations and chatter prediction. A summary of the papers reviewed in this section is presented in

Table 5. Summary of the papers presented in Section 3.5.

Ref.	Application	Machine Learning Techniques	Key Benefits
[90]	Data processing for DT modeling	General ML techniques	Improved data concatenation and processing for DT modeling
[91]	Machine test repetitions for model accuracy improvement	No specific technique mentioned	Increased accuracy of ML models by utilizing repeated testing
[92]	Surface roughness prediction	Random forest, decision tree, neural network, genetic algorithm, generalized linear model	Correlation of indirect variables (e.g., power consumption) with machinability standards
[93]	Tool shape prediction in electrical discharge machining	Machine learning techniques (specific type not mentioned)	Enhanced prediction capabilities for tool shape in non-conventional machining
[94]	Digital shadow for aeronautic machining	Gaussian Mixture Model	Effective tool failure detection, chatter reduction, and faulty programming correction
[95]	Fault diagnosis in milling and drilling processes	Deep Transfer Learning (DTL)	Improved fault diagnosis accuracy (92.33%) compared to traditional deep neural networks
[96]	Self-learning DT framework for thermal error control	Long Short-Term Memory (LSTM), neural network, Bayesian algorithm	Self-learning error prediction model with reduced incoming data errors
[97]	Surface roughness prediction and adaptive optimization	Improved Particle Swarm Optimization-Generalized Regression Neural Networks	Reduced surface roughness and tool wear through adaptive optimization
[98]	Roughness stabilization in 5-axis machining	Pigeon-inspired Optimization, Support Vector Machine	Prediction errors below 9%, improved machining parameter adjustment
[99]	Feed rate and cycle time prediction	Neural networks	Accurate cycle time prediction (up to 90%) using feed rate, tool acceleration, and toolpath
[100]	Real-time DT-based machining control	Adaptive feed rate control, residual stress simulation, chatter prediction	Enhanced closed-loop machining control with real-time simulations

.

4. Challenges and Future Trends

The application of Digital Twins (DTs) in machining represents a significant development in the integration between physical and virtual systems. However, the evolution of this technology introduces challenges that limit its implementation and efficiency. This section outlines the main challenges and trends that are expected to influence the future of Digital Twins in machining.

4.1. Challenges

• Technological Challenges
  • Alignment Between Physical and Virtual Systems: Digital models often fail to replicate physical machining processes due to the complexity of operations and the limitations in processing large-scale data in real real-time.
  • Data Processing: Real-time data collection frequently results in incomplete or unstructured datasets, which limit the accuracy of virtual models in reflecting operational conditions such as tool wear and vibrations.
• Infrastructure and Security
  • Implementation Costs: The adoption of DTs requires advanced networks, cloud storage, and computational resources, which may not be accessible to smaller enterprises.
  • Cybersecurity Risks: Storing industrial data in cloud systems increases exposure to cyber threats, requiring robust security measures. As DT systems rely heavily on real-time data transmission, they are particularly vulnerable to potential cyberattacks. Implementing encryption, secure access controls, and regular vulnerability assessments is imperative to mitigate these risks.
• Operational Challenges
  • Integration with Existing Processes: Limitations in real-time problem simulation reduce the practical application of DTs for predictive maintenance and process optimization.
  • Standardization Issues: A critical barrier to broader adoption is the lack of standardization in communication protocols, such as OPC-UA. Standardization is essential to ensure interoperability between machines, sensors, and software, reducing deployment complexity and improving system compatibility.
  • Scalability: Deployment costs and scalability remain significant challenges, especially for Small and Medium-sized Enterprises (SMEs). Tailored, cost-effective solutions such as open-source platforms, government incentives, and flexible subscription models could help overcome these barriers.

4.2. Future Trends

• Hybrid Machining Integration
  • Combining CNC machining with additive manufacturing through DTs can enhance flexibility and quality in production processes.
• Predictive Maintenance
  • Advanced algorithms for fault detection in real-time can reduce downtime and extend equipment life.
• Automation and Robotics
  • Collaborative robots are expected to integrate with DTs more effectively as interfaces and safety mechanisms improve.
• Sustainability
  • DTs can support the optimization of machining practices to reduce material waste and energy consumption, aligning with environmental goals.
• Real-Time Simulations
  • Models capable of handling complex data in real-time can improve the accuracy of simulations and decision-making processes.
• Data Security
  • Enhanced encryption and secure data transmission methods can mitigate risks associated with cyber threats, ensuring safe and reliable DT operations.
• Standardization and Accessibility
  • Establishing clear industry standards for communication protocols will streamline data exchange and facilitate smoother integration across diverse manufacturing environments. Furthermore, scalable solutions tailored to SMEs will enable broader adoption of DT technology, fostering innovation and competitiveness in smaller enterprises.

4.3. Research Opportunities

• Model Development and Data Processing: Artificial intelligence can improve the precision of digital models and their ability to process complex datasets.
• Sustainable Practices: Research into energy-efficient processes and materials can support the integration of sustainability into machining.
• Automation Interactions: Exploring the interaction between DTs and autonomous systems can lead to advancements in manufacturing.

5. Conclusions

This review summarizes the transformative potential of DTs as a central element in Industry 4.0, particularly in the context of smart manufacturing and machining processes. By enabling real-time monitoring and simulation of physical entities, DTs offer a paradigm shift in the management and optimization of manufacturing operations.

• The integration of advanced technologies such as BD, IoT, PLM, CALS, and ML within the DT framework highlights the multifaceted nature and expansive potential of these innovations. Together, these technologies enhance manufacturing processes by optimizing operations, improving product quality, and minimizing downtime through predictive maintenance and real-time process adjustments.
• This comprehensive review highlights the significant advantages of DTs, particularly their ability to create virtual replicas of physical systems. This capability facilitates extensive analysis and allows for the testing and refinement of processes in a cost-effective and risk-free virtual environment, which is especially beneficial in machining operations such as turning, milling, drilling, and grinding, where precision and efficiency are paramount.
• Despite these benefits, implementing DTs comes with several challenges. Chief among these is the need for robust and scalable infrastructure capable of handling vast volumes of data generated and processed in real-time. Furthermore, the seamless integration of diverse technological components is imperative to fully realize the benefits of DTs.
• Looking ahead, the future of DTs in smart manufacturing appears promising, with trends suggesting increased adoption and integration across various industries. The focus is likely to be on enhancing sustainability, reducing operational costs, and increasing the flexibility and adaptability of manufacturing processes. As industries continue to evolve, DTs are poised to play a crucial role in driving innovation and maintaining a competitive advantage.
• In summary, DTs represent a significant advancement in the evolution of smart manufacturing, with the potential to redefine industrial practices and foster continuous innovation. Continued research and technological advancements are essential to overcome existing challenges and fully exploit the capabilities of DTs across diverse industrial applications. By embracing these technologies, manufacturers can achieve enhanced efficiency, resilience, and responsiveness in an increasingly dynamic global market.