Digital Twin Implementation in Additive Manufacturing: A Comprehensive Review

Abstract

The additive manufacturing (AM) field is rapidly expanding, attracting significant scientific attention. This family of processes will be widely used in the evolution of Industry 4.0, particularly in the production of customized components. However, as the complexity and variability of additive manufacturing processes increase, there is an increasing need for advanced techniques to ensure quality control, optimize performance, and reduce production costs. Multiple tests are required to optimize processing variables for specific equipment and processes, to achieve optimum processing conditions. The application of digital twins (DTs) has significantly enhanced the field of additive manufacturing. A digital twin, abbreviated as DT, refers to a computer-generated model that accurately depicts a real-world object, system, or process. A DT comprises the complete additive manufacturing process, from the initial conception phase to the final manufacturing phase. It enables the manufacturing process to be continuously monitored, studied, and optimized in real time. DT has emerged as an important tool in the additive manufacturing industry. They allow manufacturers to enhance the process, improve product quality, decrease costs, and accelerate innovation. However, the development of DT in AM is an iterative and continuous process. It requires collaboration between domain experts, data scientists, engineers, and manufacturing teams to guarantee an accurate representation of the process by the digital twin. This paper aims to provide a comprehensive analysis of the current state of DT for additive manufacturing, examining their applications, benefits, challenges, and future directions.

1. Introduction

The term additive manufacturing (AM), or 3D printing as it is commonly known, refers to a production process that involves adding materials progressively to build parts layer by layer using 3D model data; it is different to subtractive manufacturing techniques, where the final shape is obtained by removing material [1]. The AM process enables the manufacture of complex components using different materials, including polymers, metals, ceramics, and combinations [2,3,4]. AM technology has completely changed the manufacturing industry, by creating complex geometries and customized products with greater efficiency and flexibility [5,6]. Industry 4.0 is a revolution in traditional manufacturing, using advanced technologies to create intelligent, interconnected systems [7]. The convergence of digital technology, including artificial intelligence (AI), big data analytics, the Internet of Things (IoT) [8], and cloud computing, with traditional manufacturing processes, is represented by Industry 4.0 [9,10].

Additive manufacturing is recognized as an innovative and powerful technology of Industry 4.0 which provides greater flexibility and personalization in the development and manufacture of complex parts, enhances manufacturing competitiveness, and reduces production time and waste of materials [11,12,13]. The principles of Industry 4.0 are illustrated in additive manufacturing using real-time data and digital design files to regulate production processes. Data-driven methodologies facilitate meticulous management and improved component quality through continuous monitoring and adjustment of manufacturing parameters. In addition, the integration of additive manufacturing systems with other manufacturing platforms and systems enables the seamless transmission of data throughout the production ecosystem. Interconnectivity facilitates optimized collaboration, communication, and data exchange, ultimately improving the flexibility and efficiency of manufacturing operations. However, as the complexity and variability of additive manufacturing processes increase, there is a greater need for advanced techniques to ensure quality control, optimize performance, and reduce production costs [14]. Further investigations are currently being carried out into digitalization, simulation [15,16,17], and machine learning technologies [18,19,20]. In this case, digital twin (DT) appeared as a powerful tool in the context of additive manufacturing [21,22]. A DT serves as a virtual depiction of the production process, providing a comprehensive perspective that encompasses the entire manufacturing workflow while capturing data in real time [23] from sensors, machines, and additional sources within the manufacturing environment [24], such as product lifecycle management (PLM), supply chain management, and quality management systems.

Digital twins enable the monitoring and control of the additive manufacturing process in real time [23,25], as shown in Figure 1. By integrating sensors and data-acquisition systems with the digital twin, manufacturers can continuously capture data taken through the physical manufacturing environment and compare it with the virtual model [26]. This enables early detection of defects, predictive maintenance, and dynamic adjustment of process parameters to ensure consistent quality and performance [27,28]. The objective of this review paper is to present a succinct and thorough examination of the current state of the art, the challenges posed by DT applications in AM, and the building hierarchy of a DT for the additive manufacturing process. This will serve as a valuable resource for future research efforts focused on DT systems for the AM process. Through a critical investigation focused on the current state of knowledge, key challenges, and prospects, this study aims to contribute to the expanding knowledge body in the field of digital twins and additive manufacturing research. By highlighting the unique prospects and advances provided by the incorporation of digital twins with additive manufacturing, we aspire to offer researchers, practitioners, and industry professionals a valuable resource that goes beyond the scope of existing journals, providing a fresh perspective on the potential synergies between these transformative technologies. The present paper is structured as shown below. After this introduction, Section 2 presents the current state of the art in digital twins and additive manufacturing. Section 3 then describes the hierarchy of DT construction for the additive manufacturing process. Next, the current development of DT for the additive manufacturing process is examined in Section 4, with a conclusion and outlook presented in Section 5.

2. The State of the Art

2.1. Additive Manufacturing

Additive manufacturing (AM) refers to the precise process of creating a tangible object in 3D based on digital data. This process involves adding successive layers of material, which distinguishes it from traditional methods that involve the subtraction of material [13,29]. Numerous AM techniques are available [30], each one characterized by the materials used and the method of layer deposition. The American Society for Testing and Materials (ASTM) subgroup F42, specifically dedicated to additive manufacturing, has categorized AM processes into seven categories [31,32], as shown in Figure 2. Each additive manufacturing technique has its own applications, constraints, and advantages in the field of part manufacturing and prototyping [33].

Material extrusion includes fused filament fabrication (FFF) [34], which is a 3D printing technique that builds layered parts by depositing heated thermoplastic material in a computer-controlled printer [35,36,37].

Material-jetting technology, such as Polyjet–Multijet modeling, uses ink-jet processes [38,39]. In this method, components are generated using a print head accumulating photopolymers. In addition, almost all layers are photocured by exposure to ultraviolet light, which complements spray head technology [40].

Binder jetting is a method of 3D printing that involves selectively bonding layers of powdered materials together using a liquid binder [39,41]. The procedure entails applying a binding agent to the powder layer, which adheres the powder particles together to create the intended form [42].

For directed energy deposition (DED) technology, the required 3D parts are produced by a metal deposition process based on a targeted energy source, using a laser source or electron beam [43,44].

The sheet lamination process builds objects layer by layer using sheets or rolls of material [45]. In this method, individual layers are cut from sheet materials (such as paper, polymer, or metal) and then bonded together to form the final 3D object [46].

For the powder bed fusion (PBF) process, a powder building material is uniformly spread in a thin, even layer [47]. The heat generated by a laser or electron beam source causes partial melting or sintering of the powder material, resulting in the melting of adjacent powder grains [48,49].

The vat photopolymerization technologies employing liquid materials are categorized according to the type of source of light and the materials used [50,51]. The process of solidifying the liquid material within a chamber using a laser beam is termed stereolithography (SLA), while the curing process utilizing a lamp-style light source is referred to as digital light processing (DLP) [52,53].

Figure 2. Classification of additive manufacturing process [54].

Figure 2. Classification of additive manufacturing process [54].

2.2. Digital Twin

The digital twin (DT) represents a virtual model that reflects the dynamics and behavior of physical objects, systems, or processes in real time [55]. It captures not only the physical attributes but also the behavioral aspects of its physical counterpart throughout its entire lifecycle [56]. This virtual model is developed by collecting and integrating data from various sources, including IoT devices, sensors, simulations, and historical records [57]. The DT provides real-time, accurate representation of the corresponding physical entity [58]. A substantial degree of integration occurs during the development of a DT between the virtual and real models via automated bidirectional data flow [3,59], as shown in Figure 1. The notion of DT may be traced back to its initial introduction by the National Aeronautics and Space Administration (NASA), where it was employed to monitor satellite activity. This approach was also used to generate simulations of possible parameter changes [13]. NASA aimed to venture into space exploration using a DT, essentially a virtual model, simulating the physical satellite’s behavior [60].

DT technology is experiencing rapid growth as a core concept within Industry 4.0 [61,62], providing a powerful tool for enhancing productivity, efficiency, and innovation in manufacturing processes [56,63]. DTs enable manufacturers to leverage data, analytics, and simulation to optimize operations, improve product quality, and drive continuous improvement in an increasingly interconnected and digitized manufacturing landscape [24,64]. Based on a recent study, the DT market had a value of USD 6.9 billion in 2022. By 2027, it is expected to grow significantly, reaching a value of USD 73.5 billion [65]. This substantial increase reflects the growing adoption and recognition of the value and potential of DT technology across various industries. As more and more organizations recognize the potential offered by digital twins to enhance their efficiency, optimize processes, and enhance decision making, the market will experience rapid growth over the coming years [66]. Research in the field of DT technology has garnered significant interest from both the academic community and industry. It has become a hot area of study, leading to the development of various tools, frameworks, and architectures aiming to accelerate the implementation of DT applications [12]. Nowadays, DT plays a critical and transformative role in industrial manufacturing [67,68], especially in the context of the AM process [69,70]. For hybrid manufacturing systems as well, digital twin systems are becoming essential in improving the efficiency, excellence, and environmental friendliness of hybrid subtractive and additive manufacturing processes [71]. They create a virtual model that can be used to optimize, control, and make decisions at every phase of the product lifecycle. Manufacturers can use them to improve the process, enhance product quality, reduce costs, and accelerate innovation [72].

3. The Hierarchy for Building a DT System for the AM Process

The process for creating a DT for AM process follows a hierarchical structure consisting of several essential steps and components, as presented in Figure 3. Although the specific hierarchy may differ according to the context and specific objectives of the DT system, Figure 3 provides a general overview.

3.1. Data Acquisition

The initial step is to collect relevant data from different sources, including sensors, monitoring systems, and simulation tools [74]. This includes data on process parameters, material properties, environmental conditions, and machine performance. The real-time aspect is typically included in the data acquisition step of building a DT for the 3D printing process [12,75]. Data acquisition involves the collection and control of data in real time throughout the AM process [76]. This may be achieved using in situ sensors [77] or other monitoring devices that provide instantaneous measurements of process parameters, machine performance, and environmental conditions [23]. By integrating real-time data with historical data and simulation models, the DT can comprehensively understand the process and enable predictive capabilities to optimize process parameters, detect defects, and improve overall performance [14]. Training the digital twin of an AM process requires a substantial amount of data to enhance its accuracy. These data can be sourced from experiments, the academic literature, sensors, and numerical simulations [78]. However, digital twins in additive manufacturing face various obstacles in collecting data. These limitations include the need to accurately model material behaviors, guarantee sensor accuracy, and mitigate potential biases. The accuracy of simulations is affected by these limitations, which in turn affect the accuracy of predictions of additive manufacturing processes.

3.2. Model Development

The model describes system behavior according to the physical process. Both mathematical and numerical models are developed to represent the AM process. These models encompass the physical behavior of materials, process dynamics, and interactions among different components of the system [79,80]. There is a wide variety of additive manufacturing processes, each employing a distinct material composed of filaments and powders of varying dimensions [14]. Numerous models become necessary once the environmental conditions during the printing process are incorporated into the model due to the many possible combinations.

3.3. Calibration and Validation

This step represents a fundamental part of the development and deployment of the DT framework. It involves assuring the accuracy and reliability of DT models and simulations in representing the real-world AM process. AM requires extensive experimentation and destructive testing to examine microstructures and mechanical properties. Prediction of thermal evolution, microstructure, mechanical properties, and deformation of components is essential. Moreover, it is one of the most important elements of a DT for the AM procedure [81].

3.4. Digital Twin Creation

The calibrated and validated models are integrated into a DT framework. The digital twin integrates mathematical models with real-time data from the AM process to generate a digital representation of the physical system [82]. A 3D real-time development platform is utilized in a diverse range of applications, including the creation of virtual human beings. The software offers a sophisticated hardware motor, a powerful system of events, an animated system, enhanced parameters for lighting, and the possibility of incorporating customized software logic using a programming language. Pantelidakis et al. [12] used this platform to build the DT ecosystem through a virtual illustration.

3.5. Real-Time Surveillance and Management

The DT ensures continuous monitoring in real-time of the AM process, where sensor data are continuously fed into the system to update and refine the digital twin’s predictions and simulations [83]. This enables proactive identification of potential issues, process optimization, and control of the AM process in real time. Machine learning and the Internet of Things (IoT) [84,85] are crucial in the hierarchy of building a DT for the AM process [86,87]. IoT devices and sensors collect real-time data from the AM process, capturing parameters, performance, and environmental conditions [14,88]. These data are then integrated into the DT system, where machine learning algorithms [89] can identify patterns, detect anomalies, and optimize process parameters, leading to improved efficiency and quality in the AM process [90]. The use of artificial intelligence (AI) in digital twin technologies for additive manufacturing (AM) is becoming more prevalent, offering numerous advantages in terms of efficiency, optimization, and quality control [91]. Sampedro et al. [92] developed a digital-twin-driven system using the 3D-AmplifAI algorithm to detect defects within a 3D printer. Real-time monitoring and control constraints in digital twins for AM encompass the difficulties of recording rapid process changes, synchronizing between digital and physical systems, and managing possible time lags.

A practical digital twin for fused deposition modeling (FDM) is initially implemented by generating a complex virtual replica of the physical three-dimensional printer. The virtual model effectively represents the structure, components, and functionality of the printer. The digital twin then incorporates real-time data from various sensors physically integrated into the printer. Critical parameters such as nozzle temperature, filament feed speed, build platform position, and environmental conditions are captured by these sensors. The digital twin uses this real-time data to virtually reproduce the complete FDM printing process. The process is accurately emulated, including the extrusion of the thermoplastic material layer by layer, the cooling and solidification of the printed layers, and the movement of the print head. The digital twin incorporates the properties of the thermoplastic filament used throughout the simulation. Various factors are considered, including thermal conductivity, melt flow rate, viscosity, and shrinkage. These variables have a direct influence on the printing process and the mechanical characteristics of the final product. By examining and comparing the simulated printing process with real-time sensor data obtained from the physical printer, the digital twin can detect any irregularities, potential problems, or deviations. This functionality allows operators to take timely corrective action, including the ability to change print parameters or schedule maintenance tasks.

In addition, predictive maintenance is facilitated by the digital twin through the analysis of historical sensor data, which uses patterns and trends to predict potential malfunctions or maintenance needs. By adopting this proactive approach, downtime is reduced, and printer performance is optimized.

4. The Development of Digital Twins for the AM Process

The research on digital twins for AM aims to advance the understanding and capabilities of this technology. It focuses on optimizing processes, improving control quality, and driving creativity in the field of 3D printing [50,93]. The findings from this research have the potential to revolutionize additive manufacturing and further unlock its benefits in a variety of industries [3]. Different digital twins for the AM process were developed in the current literature. Zhang et al. [94] provide a concise survey of the current status, challenges, and prospective potential of DT in additive manufacturing. Kantaros et al. [50] presented a mechanistic approach model for 3D printing processes. This approach involves the use of models and equations based on physics to generate simulations and predictions of production process performance, as illustrated in Figure 4.

Mukherjee et al. [81] demonstrated that a DT of a 3D printer reduces trial and error, achieves desired product features, and makes printed components cost-effective. The digital twin for 3D printing comprises the mechanical model, the detection and monitoring model, and the statistical model, along with machine learning and big data, as illustrated in Figure 5. In their research, they showed that, by using a comprehensive digital twin, it is possible to minimize testing, defaults, and the time from conception to production. In the same context, Knapp et al. [95] introduced an innovative framework of a mechanistic model for forecasting the processes occurring at the melt pool level.

Kabaldin et al. [96] created a system to enable 3D printing using arc welding within CNC machines. This system offers the possibility of connecting high-speed computing modules and training neural networks with feedback. These improve the preparation of inspection programs and components for CNC machines. Additionally, the integration of a DT enhances the capabilities of computer-aided manufacturing systems, allowing for improved preparation of inspection programs and devices for CNC machines. This innovative approach enables advanced control and optimization in 3D printing using arc welding, leading to enhanced productivity and superior output quality.

Haw et al. [97] proposed a potential application of DT for the design of additively manufactured biomedical scaffolding products. DT offers significant benefits during all three stages in the process of product development: design, development, and virtual validation. The application of DT in the design of additively produced biomedical scaffolds increases efficiency, reduces costs, and improves the global quality and safety of the final products. In the same context, Scime et al. [98] introduced a cyber-physical infrastructure aiming to facilitate the production of instance-qualified AM components. The proposed methodology emphasizes scalability and leverages the power of artificial intelligence (AI). A further methodology has been proposed by Arden et al. [99] to analyze, optimize, and manage the spreading of powder in the AM process.

The authors used the discrete element method (DEM) to simulate the PBF coating, which serves as a virtual environment for training and testing their framework. The approach investigates the compromises inherent in the spreading of powder, such as the roughness of the layer, the overlap time, and the addition height of the fused layer. They use a multi-objective Bayesian optimization technique to explore these trade-offs. This optimization process investigates the space of parameters defined by the recoated speed, recoated angle, and platen displacement, thereby revealing new recoating techniques that promote specific desired results. By integrating the DT approach and DEM simulation, the proposed methodology is designed to enhance the performance and reliability of AM powder spreading processes.

In the literature, various researchers have collectively discussed the conception and testing of a DT for quality control and evaluation in different categories of the AM process. Fabio et al. [82] presented the design, implementation, and verification of a novel DT solution for the material extrusion process. The solution is based on three major elements: a central module that contains the simulation motor, an interface to manage incoming data, and a graphic interface for remote control by the user. The developed DT takes the input of process data obtained from different sensors and the G-Code file employed by the actual printer, to provide a range of functions to control the process, monitor condition, and check geometric accuracy in real time. In the same context, Pantelidakis et al. [12] implemented a new digital twin ecosystem designed to test, monitor processes, and remotely manage the fused deposition modeling (FDM) printer using a virtual simulation platform. The DT ecosystem developed consists of two methods. The first is a data-driven method that uses an open-source 3D printer control application for capturing machine status and relevant settings. The second method uses external sensors that approach the real behavior of the 3D printer, ensuring precise harmonization between the virtual and physical printers. This implementation improves the control and monitoring of the AM process, facilitating improved efficiency and accuracy. Castelló et al. [100] emphasize an innovative digital twin for modeling materials which simplifies the composite part creation process using large-format additive manufacturing (LFAM). For the same category of process, an in-depth analysis was carried out by Reddy et al. [101] to examine the application of DT in 3D printers, and to present the recommended framework and method. For example, the process of updating the digital model is time-consuming, and there is a noticeable delay in results at the interaction layer level.

Concerning the directed energy deposition (DED) process, Reisch et al. [102] explored the concept of context awareness within process control applications for the wire arc additive manufacturing (WAAM) process. They integrated three types of contexts, i.e., the machine context, the time context, and the spatial context, using a digital twin as an illustration of the real WAAM part during the manufacturing process. Exception detection is performed using both the machine and temporal environments. In the same context, Haochen et al. [103] presented a comprehensive overview of current advances in the application of DT to AM processes. The objective is to gather comprehensive insights into the existing developments and identify key areas for improvement. The research aims to propose a suitable DT for the WAAM system. Another digital twin for the DED process was developed by Noh et al. [104]. Initially, a digital representation of the DED equipment was created, which included a three-dimensional model of the DED machine. An inertial measurement unit (IMU) was incorporated into the DED machine to coordinate, in real time, the movements of digital components with the corresponding physical counterparts of the machine. Subsequently, the DT of the process is created by categorizing items according to the precise geometry of the deposited components. Ultimately, the DT of both the DED equipment and process are merged to accurately recreate the identical system as the physical one.

For the powder bed fusion category, a new DT or the manufacturing process using metal powder bed fusion was presented in [105]. This advanced numerical model can simulate the dynamic temperature profile at individual points in the process, enabling the detection of potential problems such as superheating. The approach is supported by a machine learning algorithm that has been developed on experimental confirmation of synthesized data derived from finite element analysis (FEA). This algorithm takes into account various factors, including scanning strategy, laser settings, material characteristics, and part geometry. Gaikwad et al. [23] presented the concept of DT applied to AM, to improve the detection of process anomalies. The concept involves integrating physics-based predictions, real-time sensor data, and machine learning techniques. The suggested DT technique, which integrates insights of a theoretical model with sensor data, exhibits enhanced accuracy in identifying process faults as compared to employing either the theoretical model or sensor data independently. The efficacy of this digital twin methodology was confirmed using empirical results derived from metal additive AM processes. For the same category of AM process, Yeung et al. [106] have developed a DT system for monitoring and optimizing the laser powder bed fusion. The framework is constructed using the recently invented state-of-the-art point-wise scan control approach.

In conclusion, this section has presented the development of DT approaches for AM. The studies and research discussed have highlighted the immense power of DTs in enhancing various aspects of the AM process.

Table 1. Comparative study of some digital twins for additive manufacturing processes.

Authors	Contributions	Limitations
Pantelidakis et al. [12]	A digital twin ecosystem designed to test, monitor processes, and remotely manage the fused deposition modeling (FDM) printer using a virtual simulation platform.	The experimental conditions relating to the environment were not fully regulated. Specifically, the room lighting and printer surfaces were left unchanged to optimize the accuracy of the sensor readings.
Arden et al. [99]	An innovative approach to improving the assessment, configuration, and control of powder spreading throughout the AM process.	Surface texture cannot be measured in real time by current additive manufacturing powder bed monitoring systems. To meet this challenge, sophisticated hardware controllers and sensors are needed to update the parameters and measurements of the printing process in real time.
Castelló et al. [100]	A digital twin for material extrusion in large-format additive manufacturing involves the integration of machine learning, real-time data acquisition, and computational modeling to optimize and control the manufacturing process.	Computing requirements and the complexity of modeling processes, prevent real-time implementation and integration in practical contexts.
Noh et al. [104]	A digital representation of the DED machine, including a three-dimensional model of the DED machine, was developed.	The lack of real-life implementation studies for the proposed digital twin.
Yeung et al. [106]	A digital twin system for monitoring and optimizing the laser powder bed fusion process.	Manufacturing requires enhanced incorporation of machine learning and data management.

summarizes some of the previous studies carried out in the DT for the AM process.

Although digital twins for AM provide useful information and efficiency gains, they are not without limitations. Challenges include the complexity of accurately simulating material behavior, effectively representing process dynamics in real time, and the sensitivity of results to the quality of input data. Restrictions are further exacerbated by the complexity of multi-physics modeling, post-processing considerations, and the need for considerable computing resources. The dynamic nature of additive manufacturing technology and the absence of standardized data formats pose ongoing problems. To overcome these restrictions, it is necessary to constantly improve models, guarantee data confidentiality, and enhance operator expertise. Ongoing research is striving to overcome these difficulties and unleash the full capabilities of DT in additive manufacturing.

5. Conclusions

In summary, our examination of the synergy between additive manufacturing and digital twin technology has highlighted key findings that underscore the transformative potential of this integration. Additive manufacturing, with its unrivaled flexibility, facilitates the development of complex customized parts, significantly improving manufacturing competitiveness and simultaneously addressing concerns about manufacturing time and wasted materials. The advent of DT as a robust solution in this sector enables manufacturers to optimize operations, enhance product quality, reduce costs, and accelerate innovation. However, as our study shows, tapping the full potential of digital twins in AM is not without its complexities. Developing a digital twin requires meticulous attention to nuances such as model accuracy, data integration, and scalability. Following a comprehensive analysis of the existing literature, we have identified persistent challenges and focal points in the research field. There is an urgent need for standardized frameworks that consider the continuous technological progress and increasing complexity of AM processes.

Our research underscores the necessity of collaboration between researchers, software developers, and industry experts. Such partnerships are essential to navigate the twists and turns of successfully implementing and deploying digital twin solutions in real manufacturing environments. Going forward, the trajectory of research should be guided by a commitment to refining the digital twin framework, integrating advanced analysis techniques, and exploring innovative avenues for process optimization and control.

By addressing these challenges, the field can lead the way to a more robust, efficient, and adaptive future in additive manufacturing, fulfilling the full potential of digital twin technology.

The future aim is to create a customized digital twin for FDM processes. This digital replica will consist of an improved simulation technique and the integration of real-time sensor data to create a virtual representation of the FDM 3D printer. By simulating the printing process and analyzing relevant parameters such as nozzle temperature, filament feed speed, and build platform position, potential faults and deviations from expected results can be predicted using the digital twin. In addition, using machine learning algorithms, the digital twin will improve the efficiency of print parameters to achieve superior print quality and consistency. The digital twin we have designed for FDM has the potential to increase the efficiency of manufacturing processes and accelerate advances in additive manufacturing.