Integrating Virtual Reality into Welding Training: An Industry 5.0 Approach

Abstract

In the context of Industry 5.0, Virtual Reality (VR) is transforming workforce training by enhancing skill acquisition and smart assistance to improve safety, among other things. This study presents a VR-based welding training simulator, developed in Unreal Engine 5 and deployed on Meta Quest 3, designed to teach and standardize soldering techniques through progressive training stations, interactive guidance, and real-time feedback. A mixed-methods evaluation, combining quantitative metrics (execution time, positioning errors, grip errors) and qualitative user feedback, demonstrated a significant reduction in execution time and elimination of errors in positioning and grip. Additionally, no errors were made in real-world soldering post-training, confirming a high transfer of learning and reduced variability in performance among users. These findings highlight VR as a viable and scalable alternative to traditional training, ensuring process consistency, reduced learning time, and increased training efficiency. Future work will explore enhanced haptic feedback and the integration of generative artificial intelligence to improve real-time analysis.

1. Introduction

Industry 5.0 represents a new era in manufacturing, characterized by collaboration between humans and robots. In a more general context, it is characterized by its human-centricity, that is, not displacing the person but enhancing their skills [1], with the purpose of seeking to integrate advanced digital technologies with human experience [2,3]. Virtual reality (VR), a key technology in this new revolution, is transforming the way operators train in and adapt to increasingly complex work environments [4].

The capacity of VR technology to generate immersive and realistic environments establishes it as an optimal instrument for training, applicable across various industrial sectors [5,6,7]. In contrast to conventional methodologies, VR permits operators to engage in real-world scenarios within a secure and controlled virtual setting [7,8]. This capability offers significant advantages for the manufacturing sector, where errors during training may incur significant financial or even hazardous consequences [6], potentially causing harm to both personnel and machinery. It is crucial to emphasize that in circumstances where operators require extended training periods, VR environments allow for retraining as frequently as necessary without necessitating the presence of a trainer or the withdrawal of machinery from production for training purposes.

In the context of Industry 4.0, VR has been shown to provide considerable benefits in training applications. Simulations of complex processes, such as the reconfiguration of wiring in battery systems [7] or the machining of components [9], facilitate operators’ familiarization with various tasks before interacting with actual machinery. Industry 5.0 further enhances this experience by promoting a human-centric approach.

Industry 5.0 is designed not to replace human labor with robotic equivalents, as has been stated, but rather to enhance human capabilities within the workplace. This is illustrated by studies such as those cited in [1,10], which explore the application of digital twins for monitoring factors such as physical and mental fatigue.

Manufacturing operators encounter a variety of constraints that may impact their performance, safety, and overall well-being. These constraints are exacerbated by the growing complexity of production processes associated with Industry 5.0, characterized by an elevated level of product customization.

Here are some of the constraints operators face and how VR can help overcome them:

• Training Safety Risks: Industrial environments present potential hazards, particularly for novice operators. The operation of heavy machinery, tools, or chemicals involves intrinsic risks. VR allows for practice within a controlled virtual environment, thus mitigating the likelihood of accidents during training [3].
• Elevated costs associated with prototyping and testing: The creation of physical prototypes for the evaluation of new designs or assembly processes, as well as the temporary cessation of machine operations for training purposes, can incur substantial financial expenditures. Alternatively, a virtual reality environment facilitates the development of virtual prototypes and the simulation of their behavior, thereby markedly reducing both the costs and the duration of development [3].
• Increased difficulty adapting to complexity: The advent of new technologies and the mass customization of products in Industry 5.0 require that operators rapidly acclimate to increasingly complex processes. VR aids in the comprehension of these complexities by allowing operators to engage with intricate virtual models [11,12].
• Ergonomics and Operator Well-being: The performance of repetitive tasks or tasks that require awkward postures has the potential to cause injuries and affect the health of operators, which can ultimately result in disability. VR can facilitate the simulation of workstations and the evaluation of design ergonomics, thus helping to prevent injuries and optimize operator well-being [13].
• Constraints on Human–Robot Collaboration: Industry 5.0 aims to advance the integration of human–robot collaboration. However, a significant factor to consider is the lack of familiarity with robotics, which can engender apprehension and impede effective interaction. Virtual reality facilitates the acquaintance of operators with robotic systems within a virtual and secure environment [7].

This research introduces a VR-based method aimed at standardizing soldering iron welding techniques, ensuring that operators perform the procedure consistently and integratively within a sophisticated manufacturing setting. The development of a training station using Unreal Engine 5 creates an immersive and controlled environment in which people can practice and refine the soldering technique under various configurations and conditions. By allowing for systematic repetition of the welding process without incurring physical risks, this solution guarantees that all operators develop and reinforce a uniform work methodology. Furthermore, the incorporation of haptic, visual, and interactive feedback mechanisms within the simulator improves the learning curve, facilitating the adoption of standardized practices that elevate the quality of the final product and increase operational efficiency in the context of Industry 5.0.

According to [14,15], in order to differentiate between Industry 4.0 (I4.0) and Industry 5.0 (I5.0), it is important to be able to contrast their main objectives. The core of I4.0 lies in optimizing industrial processes through digitization and advanced automation, integrating technologies such as IoT, AI, and Big Data to maximize efficiency and productivity, often with limited human interaction. I5.0, on the other hand, leverages the technologies of I4.0, while redefining the core purpose by complementing technological efficiency with a focus on people, its core concept is based on three pillars: human centricity, which places worker well-being, skills, and collaboration at the heart of the production system; sustainability; and resilience, which seeks to strengthen industry’s ability to adapt to change and crisis. Therefore, the key difference lies not only in the technologies employed, but in the fundamental paradigm shift. While I4.0 focuses primarily on technology as a driver of efficiency, I5.0 adopts a holistic view in which technology serves to empower the human being and achieve broader goals of social well-being, environmental sustainability, and industrial resilience.

The rest of this document is organized as follows: Section 2 provides a comprehensive review of the pertinent literature, elucidating the influence of virtual reality on industrial training and its integration with digital twins and intelligent manufacturing systems. Section 3 elaborates on the development of the simulator, detailing the implementation of the environment in Unreal Engine 5, the interaction mechanics, and the integration of Meta Quest 3 controllers, and the results derived from using the simulator, assessing its impact on the acquisition of welding skills and the standardization of procedures. Finally, Section 4 conveys the study’s conclusions, underscoring the benefits of the proposed approach and suggesting potential avenues for future research.

2. Related Work

This section will comprehensively review the principal applications of virtual reality within the manufacturing sector. While the preceding section emphasized the significant role of VR technology in operator training, it is imperative to acknowledge that VR’s utility extends beyond this domain. Notably, its applications encompass areas such as process design and remote assistance, which will be elaborated upon in subsequent sections.

2.1. AI and Digital Twins for Personalized VR Training

The investigation conducted by [16] examines the influence of artificial intelligence (AI) within the metaverse, with particular emphasis on its implementation in the manufacturing sector through digital twins. The study elucidates how AI facilitates the development of virtual representations of production systems, thereby enhancing machine monitoring, fault detection, and the planning of production lines. Furthermore, it identifies critical challenges such as interoperability and network latency, which need to be addressed to fully realize the potential of these technologies in the manufacturing industry within the metaverse.

Researchers, including those referenced in [14], critically examine the potential of the metaverse as a facilitative technology for the advancement of Industry 5.0. Focusing on pivotal aspects such as sustainability, they conduct a systematic review of existing literature to identify opportunities within manufacturing, education, and urban planning sectors, where the metaverse could contribute to reducing the carbon footprint, optimizing industrial processes, and enhancing human–machine interaction.

In the research of [17], the integration of the metaverse in the context of Industry 5.0 is explored, promoting a human-centric approach to industrial processes and presenting a framework that combines enabling technologies such as Digital Twins (DT), VR, and the Industrial Internet of Things (IIoT), to develop an industrial metaverse that facilitates collaborative interaction between humans and cyber-physical systems (CPS). In this environment, humans can interact more immersively and realistically with digital assets, enabling collaborative tasks between humans and machines.

In the research conducted by [8], the application of VR to training in industrial assembly was explored, focusing on a comparison between two VR training modalities—one incorporating visual assistance and the other without such enhancements—and conventional training methods. The experiment utilized the HTC Vive Pro in conjunction with the Unity platform for the development of the system. The findings indicated that those who participated in VR training without visual assistance, characterized by the absence of position markers and automatic part adjustments, demonstrated performance levels analogous to those of the control group, which executed the assembly under the guidance of an instructor in person. These outcomes imply that the reduction of visual aids improves the cognitive load of users beneficially, compelling them to engage in more active information processing and recall of assembly procedures, thus fortifying learning and memory retention.

The research by [4] proposes an MR-based manual assembly training system with virtual sensors, designed to help operators quickly adapt to complex tasks amid product variability and short life cycles. MR enables virtual instructions to be superimposed on the physical environment, enhancing accuracy and reducing errors. The study used Microsoft HoloLens 2 to visualize assembly instructions and employed “shape sensors” and “grid sensors” to monitor and guide user actions in real time.

The integration of novel manufacturing methodologies, such as the production of electric batteries in gigafactories and the application of virtual reality (VR) and mixed reality (MR), facilitates the training of operators. According to [18], Simubat 4.0 Gen-2, a digital twin tool, incorporates VR and MR to provide training for the production of lithium-ion batteries. This tool enables users to engage in learning and practice within a virtual environment safely and efficiently, thereby addressing the intricate nature of battery production within gigafactories. Designed to fulfill the requirement for skilled personnel in this sector, Simubat 4.0 Gen-2 assists trainees in comprehending and managing electrode properties and the manufacturing process.

The research by [19] presents a virtual simulation teaching platform specifically designed for smart manufacturing training, using technologies such as cloud computing, edge computing, and terminal equipment. This platform addresses common challenges in practical engineering education, such as equipment shortages, high material investment, safety risks, and the difficulty of implementing repeatable and observable practices. It focuses on specialties such as electrical automation, mechatronics, and industrial robotics, integrating six main modules, including production line simulation and virtual robot disassembly.

In [20], a comparative study was conducted to evaluate the efficacy of VR versus traditional training methods, which involve paper-based instructions and supervision, for the assembly of electric motors. The research involved 22 participants possessing varying levels of experience. The results indicated that individuals trained through VR completed the assembly task 10% more quickly and without committing errors, whereas the cohort following traditional training methods made two mistakes pertaining to wire connections. VR training demonstrated pronounced advantages for novice individuals, enhancing their performance significantly. In contrast, for experienced workers, the performance metrics were comparable across both methods. However, it is notable that the training duration was longer in the VR context, which underscores its potential to minimize errors and foster learner autonomy.

The manual adjustment of parameters in CNC machines, particularly those with five or more axes, is typically executed at the discretion of an expert, rendering the programming of such equipment both challenging and susceptible to errors. Reference [9] proposes an innovative approach to toolpath planning in computer-aided manufacturing (CAM) systems by leveraging VR and AI. The objective of the study is to enhance the intuitiveness of toolpath planning by allowing users to manually render the toolpath within a virtual reality environment, thus obviating the need for manual entry of extensive parameters. This system utilizes convolutional neural networks (CNNs) to interpret user movements and convert them into precise tool configurations.

In [2], the integration of extended reality (XR) with the Manufacturing as a Service (MaaS) framework is examined within the realm of smart factories and cloud manufacturing. The study proposes a distributed microservices-based XR architecture that facilitates interaction for on-site operators through MR with Microsoft HoloLens 2, as well as for remote operators through VR using HTC Vive. This XR integration enables on-site operators to use HoloLens 2 to obtain real-time data concerning temperature, weld quality, and operational parameters, thus obviating the need for physical documents or external displays. Simultaneously, remote operators can oversee the process through VR, thereby mitigating risk exposure and permitting remote intervention should there be deviations in process quality. The findings of the case study demonstrate that the use of XR improves operator efficiency, optimizes process monitoring, and facilitates real-time data traceability.

Augmented reality (AR) is employed in [21] to enhance human–robot interaction within flexible manufacturing environments. The primary impetus for this research is the necessity to adapt industrial processes to flexible manufacturing paradigms, whereby automated systems must become more user-friendly and reconfigurable without necessitating specialized expertise in robotic programming. The shift towards smart factories and bespoke production has resulted in an escalation of human–robot collaboration; however, existing interaction methodologies remain dependent on intricate interfaces and conventional programming techniques, thereby constraining flexibility and impeding adoption by operators lacking advanced skills.

In a parallel comparison presented in [22], operators trained in KUKA robot programming through VR methods are contrasted with those utilizing conventional training techniques without VR. The findings suggest that VR training enables operators to perform tasks on real robots with a success rate similar to that of traditional methods. However, this is accompanied by increased cognitive demands and marginally extended completion times.

The work of [23] presents the development of a virtual training system to learn industrial robotics, using full simulation and Hardware-in-the-Loop (HIL) techniques to teach the control and programming of the Scara SR-800 robot. The system is designed to provide a safe and accessible environment for students, avoiding the costs and risks associated with the use of physical robots. It includes both a laboratory and an industrial environment, allowing students to observe and evaluate the robot’s behavior in position and trajectory tasks.

A key feature of Industry 5.0 is its emphasis on human-centric approaches. In this context, the advancement of collaborative robots (cobots) and VR simulators for operator training in robotic usage is notable. As detailed in [24], we focus on the development of a simulator designed to facilitate safe interactions between operators and robots, underscoring the significance of safety standards ISO 10218-1 and ISO/TS 15066, which outline risk mitigation strategies in environments where shared spaces are utilized.

In [25], an additional application of VR to enhancing safety and empowerment within human–robot interactions is demonstrated by proposing the implementation of an AR virtual barrier system. This system aims to improve safety and operational efficiency in collaborative human–robot environments within manufacturing settings. Using Microsoft HoloLens 2, ROS, and MoveIt, the proposed system is capable of generating dynamic virtual barriers. These barriers serve to protect both users and surrounding objects by mitigating collision risks, thus eliminating the necessity for physical barriers.

The study presented in [26] investigates safety within the domain of human–robot collaboration by proposing an MR and artificial vision system designed to enhance both safety and efficiency in carpentry operations. This system incorporates the use of Microsoft HoloLens 2, a UR5 robot, and a YOLOv5 vision system. Carpenters can mark wooden pieces with conventional symbols, which the robot autonomously interprets and executes. The integration of a digital twin within the MR framework permits the pre-visualization of trajectories prior to execution, thereby preventing collisions in real-time. The findings illustrate that the utilization of mixed reality and digital twins significantly facilitates robot programming within manufacturing contexts, thereby reducing errors and enhancing efficiency without necessitating physical barriers.

Virtual reality technology is being utilized in the development of simulators for machine training. For instance, as presented in [6], VRWeldLearner is a welding training system that aims to mitigate the shortage of skilled welders within the industry. Its objective is to offer a training platform that is both accessible and cost-effective, thereby enabling users to acquire welding skills in a virtual setting without incurring the risks and expenses associated with physical equipment training. This system incorporates reverse design principles, crafting structured modules that cover critical topics such as equipment setup, maintenance, and the psychomotor skills requisite for performing welding tasks in actual settings.

A comparison presented in [6] describes a scalable MIG welding training simulator using reverse instructional design and integrating visual and haptic feedback through a virtual environment with a physical welding gun. In [18], virtual reality combined with RM enhances lithium-ion battery manufacturing training, providing real-time electrode microstructure visualization and tracking with HoloLens 2.

In comparison, this research proposes a simulator for affordable, immersive, and portable training in fine manual soldering with a soldering iron. It uses the Meta Quest 3 and a series of stations providing visual and audio feedback. While more lightweight, it shares the goal of enhancing accessibility and safety in technical training. This complements existing approaches by applying virtual reality to scenarios where manual skill and compact systems are crucial, addressing a vital aspect of industrial training.

2.2. Virtual Reality and I5.0

In [27], the author elucidates the development of Operator 5.0, a collaborative platform grounded in extended reality (XR) for the enhancement of engineering education within the Industry 5.0 paradigm. Through the integration of VR, AR, and MR, complemented by cloud technologies such as Unity 3D, Playfab, and NoSQL databases, this platform furnishes an immersive setting conducive to training in maintenance, repair, and advanced manufacturing. Empirical evidence from three laboratory-based case studies corroborates the proposition that the implementation of XR in educational settings augments knowledge retention, enhances interactivity in collaborative environments, and facilitates more efficacious training in technical competencies.

Similarly, in the research of [28], the TwinXR framework, which combines the use of digital twins with XR, is introduced in a platform that reduces the need for manual programming. Two implementations with an industrial crane and a UR5e robotic arm demonstrated that TwinXR improves interoperability, customization, and efficiency in smart manufacturing. The TwinXR method proposes an efficient integration between digital twins (DTs) and XR in industrial environments, enabling the creation and scalability of XR applications using DT documents. Based on a three-layer architecture, TwinXR uses Unity, Twinbase, OPC UA, and ROS to link physical machines, semantic data, and XR simulations.

The transition from Industry 4.0 to 5.0 necessitates a thorough evaluation of the interaction between operators and production equipment. To achieve a person-centric approach, it is imperative to scrutinize the methods of interaction with the aim of enhancing communication. As identified in [29], six essential requirements for designing Human–Machine Interfaces (HMIs) include ethical principles, ergonomics and adaptability, user empowerment, intuitive interfaces, effective collaboration, and sustainability. Validation was conducted through two industrial case studies: the first concentrated on wind turbine maintenance utilizing AR and image recognition, while the second examined flexible assembly planning with the integration of touch interfaces and AI. The findings revealed improvements in operational efficiency, reductions in task scheduling complexity, and enhanced safety for operators. In the development of adaptive VR, five key technological pillars are identified: personalized content, intelligent haptic devices, automatic assessment, autonomous agents, and multimodal technologies, which contribute to improved learning efficiency by facilitating scalable and accessible training [30].

An additional foundational element of Industry 5.0, as embraced by VR, is sustainability. Reference [31] introduces an innovative framework that utilizes a Sustainability Digital Twin to enhance sustainable design within automated production lines. This framework integrates VR and AR with industrial data to promote sustainability. The research is situated within the principles of Industry 5.0 (I5.0), which highlight resilience, sustainability, and a human-centered approach.

2.3. Instructional Design Principles and Theoretical Foundations

In designing the simulator proposed in this study, it was decided, as will be elaborated upon in the subsequent section, to incorporate multiple workstations within the virtual environment. The operator is required to navigate through these workstations to complete the training, with each station increasing in complexity. This approach aligns with concepts derived from Cognitive Load Theory (CLT) and the pedagogical scaffolding model, which have been extensively utilized in learning environments enhanced by immersive technologies. As demonstrated in [32], the inclusion of preparatory activities preceding a VR experience has been shown to significantly alleviate extrinsic cognitive load and enhance students’ sense of presence and performance when engaging with the simulator.

The structured progression approach is further validated by the study conducted in [33], wherein a systematized VR training model aids the learner’s acclimatization to hazardous environments, thereby fostering a gradual internalization of knowledge. Similarly, the research presented in [34] indicates that incrementally increasing interactivity within VR environments enhances the Germanic cognitive load—associated with constructive processing—and augments the sense of agency alongside learning outcomes. In alignment with these findings, the simulator developed in the current study is organized into sequential stations that progressively intensify the difficulty of the assigned tasks, thereby not only promoting knowledge retention but also providing a cognitive and pedagogical framework applicable across various levels of complexity and other domains of training.

3. Virtual Welding

3.1. Development

The simulator is engineered with a series of four training stations within a VR environment, strategically structured to progressively advance the user’s learning and retention of the appropriate technique for wielding a soldering iron. In this specific instance, the task involves soldering a metal pipe.

Figure 1 presents the sequential arrangement of the four training stations developed within the VR simulator. Each station serves a distinct role in promoting the incremental acquisition of the soldering iron technique.

• Station 1: Figure 2 illustrates that this initial station is designed to instruct the user how to properly grip the soldering iron and its optimal alignment relative to the metal tube in a fixed linear orientation. It comprises a visual guide, which is a semi-transparent depiction of the soldering iron in the optimal position, indicating the appropriate orientation, position, and duration required for soldering. The instruction is further enhanced by providing the user with haptic and audible feedback.
• Station 2: At the second station, the visual guide is eliminated to enhance the user’s mastery of the technique using an independent application. The metal tube maintains the same straight and fixed orientation as observed in the previous station, thereby necessitating the user to rely on muscle memory and the knowledge acquired, devoid of direct visual assistance. Haptic and auditory feedback continues to be provided.
• Station 3 and 4: These stations slightly mitigate the challenge of improving proficiency in soldering iron handling. However, visual guidance remains unavailable, and the metal tube is subject to rotation along the Z and Y axes, as illustrated in Figure 3 for Station 4. This variability requires the user to actively adjust the grip and positioning of the soldering iron to maintain the correct technique across different spatial orientations, thus assessing their ability to generalize the acquired skill. Haptic and auditory feedback continues to be provided.

The development of this simulator employs the Unreal Engine 5 game engine, with the implementation of VR facilitated through the utilization of Meta Quest 3 viewers. The system is designed to provide instruction on the correct positioning and duration required for soldering iron welding. Within this simulation environment, the object subjected to welding is a metal tube, serving as a representative component within the manufacturing industry. This methodology is particularly pertinent to medical device manufacturing, a sector characterized by numerous manual operations. In these contexts, the accurate execution of welding is a pivotal factor, as it can significantly affect the quality of the final product. Consequently, the simulator aims to enhance the acquisition of suitable welding techniques, tailored to the specific operation type, thereby augmenting the precision and standardization of the operators’ performance.

Figure 2. Virtual workstation developed in UE5.

Figure 2. Virtual workstation developed in UE5.

Figure 3. Workstation 4 where the object to be soldered is rotated compared to the other stations in order to reinforce for the user the correct position of the soldering iron.

Figure 3. Workstation 4 where the object to be soldered is rotated compared to the other stations in order to reinforce for the user the correct position of the soldering iron.

The simulator integrates various feedback modalities to augment the educational experience. Initially, visual feedback is employed, which comprises messages within the virtual environment that denote the appropriate positioning of the soldering iron and the optimal duration of the soldering process. Additionally, visual effects, including particle and smoke animations, are incorporated and activated if the user exceeds the permissible time limit, serving as an auxiliary warning. Conversely, the simulator also encompasses haptic feedback. Once the welding time has been surpassed, in conjunction with the visual cues, the right controller—which the user utilizes to manipulate the soldering iron—commences vibrating, thereby providing tactile feedback that reinforces the error indication and promotes the rectification of the technique in future attempts.

Figure 4 presents a diagram that depicts the commencement of the interaction between the operator and the virtual environment. The system has been configured to enable the user to employ the Meta Quest 3 device controller, thereby receiving haptic feedback. Future work envisions the integration of hand-tracking technology to facilitate a more natural interaction. A primary aim of this research is to offer the user multisensory feedback, including vibrational stimuli, as an integral component of the training system.

To operate the virtual soldering iron, the user is required to depress a specific button on the controller. This functionality is executed through the Enhanced Input system of Unreal Engine 5, which facilitates the allocation of a button to trigger the interaction. It is crucial to emphasize that within the virtual environment, there exists a demarcation between two categories of objects: grabbable objects, which can be lifted and manipulated by the virtual hands, and static objects, with which direct interaction is prohibited.

In the VR scene, each hand and the soldering iron are surrounded by collision spheres that simplify grip detection. If the two spheres overlap when pressing the button configured with UE5’s Enhanced Inputs, an event is triggered that anchors the soldering iron to the hand, temporarily disables its collision, and changes the animation to a grasping pose. When the button is released, the soldering iron is released, and its collision is reenabled, providing a natural and computationally efficient interaction. Figure 5 illustrates the collision spheres associated with the hands.

Finally, the interaction between the soldering iron and the target metal bar, which is central to this study, is regulated by collision detection. A collision sphere is assigned to one end of the metal bar, as illustrated in Figure 6. To facilitate this, a customized blueprint is constructed that includes an event that detects the entry of the soldering iron into the sphere. Upon such a collision, a series of sensory responses are initiated: a visual indicator is presented to the user displaying the time required to complete the welding process, a haptic vibration is engaged on the knob, and an auditory cue is emitted. These responses persist while the soldering iron remains within the collision zone; once removed, all animations and feedback mechanisms are automatically deactivated.

3.2. Results

In the initial phase of experimentation, a test was administered within a controlled laboratory setting involving 23 participants, comprising 12 males and 11 females, all of whom were university students aged between 22 and 25 years. None of the participants possessed prior experience in soldering iron welding or in the utilization of virtual reality simulators for training purposes. This preliminary condition aligns with the research objective, which aims to assess the efficacy of technological tools, such as virtual reality, in fostering or enhancing technical skills among individuals with minimal or no previous experience, within a secure practice environment that can be emulated in professional contexts.

The assessment of this experimental phase was categorized into two complementary methodologies: a qualitative evaluation, purposed to capture user experience during simulator utilization, and a quantitative evaluation, concentrated on the analysis of error reduction. The quantitative evaluation specifically targeted two types of faults: errors in the manipulation of the virtual tool and errors associated with the welding positioning.

3.2.1. Qualitative Evaluation

The qualitative assessment was conducted using a questionnaire administered to each participant, with the questions delineated in

Table 1. Survey questions for qualitative analysis of the VR welding simulator.

ID	Question
User Experience
Q1	How intuitive did you find the VR welding simulator?
Q2	How realistic did the welding simulation feel to you?
Q3	Did you feel that the simulator interface was easy to understand and use?
Q4	How would you rate the comfort of the simulator during the training session?
Q5	Did you find that the simulation responded fluidly to your movements and actions?
Knowledge Transfer and Skill Retention
Q6	How confident did you feel about welding before and after training in VR?
Q7	Do you feel that VR training adequately prepared you for real-world welding?
Q8	How easy was it to recall the correct soldering iron posture in real welding after VR training?
Q9	How many times do you think you would need to practice in VR before attempting real welding?
Training Preferences and Improvements
Q10	If you were to train in VR again, what aspects would you improve or change?
Q11	Which training method do you prefer for learning welding?
Q12	Do you believe VR training reduces the fear or anxiety of making mistakes in real welding?
Q13	Did you feel more or less anxious practicing in VR compared to a real instructor?
Q14	Do you think using VR helps reduce learning time compared to a human instructor?
Feedback and Interaction
Q15	Did the VR simulation provide enough visual and auditory feedback for you to understand if you were welding correctly?

. The survey examined various dimensions, including user experience, transfer, and retention of acquired knowledge, training preferences, and potential improvements to the simulator. In addition, participants were asked about the potential optimization of their use through the inclusion of new features. Finally, the feedback provided by the participants was analyzed, focusing on visual cues within the scene, auditory effects, and haptic feedback from the controls. The survey questions are scored on a Likert scale of 1–5.

The survey responses provided significant insights into the users’ perceptions and the efficacy of the VR welding simulator. Participants highlighted the need for optimization in the grip mechanism of the soldering iron and other virtual components, citing challenges primarily related to depth perception. Initially, users struggled to accurately discern the distance to virtual objects. Additionally, they sought enhancements in the clarity of instructions within the simulator, indicating a necessity for improvements in the interface design and user guidance.

Concerning the preferred training method, 10 participants opted for the utilization of the VR simulator, while six indicated that the optimal approach would be to integrate it with actual hands-on training. This suggests that, although simulation is considered valuable, it is not deemed an absolute replacement for physical practice. In terms of anxiety, 15 out of 18 user participants reported feeling less anxious when practicing in a VR setting compared to with a real instructor, implying that virtual reality offers a safer and less stressful learning environment.

Finally, in terms of learning efficiency, most of the participants acknowledged that VR facilitates a reduction in training duration. However, some participants identified specific limitations and the need for greater human supervision. Taken together, the findings suggest that the VR simulator is a valuable and effective instrument for welding instruction. However, there remain potential areas for improvement, particularly concerning the system’s ergonomics, the quality of visual and auditory feedback, and its synergy with conventional teaching methodologies.

Table 2. Summary of VR simulator survey results: mean scores and positive ratings (4 or 5).

Question ID	Mean Score (1–5)	% Rated
Q1	4.11	88.9%
Q2	4.61	100.0%
Q3	4.11	77.8%
Q4	4.39	94.4%
Q5	4.22	88.9%
Q6	4.00	72.2%
Q7	3.61	38.9%
Q8	4.56	94.4%
Q12	4.39	100.0%
Q15	3.89	61.1%

presents a comprehensive summary of the principal findings derived from the post-training survey conducted with each participant. This survey was designed to assess the participants’ perceptions of the VR welding simulator, utilizing a 5-point Likert scale. The results indicate a substantial level of satisfaction with the usability and realism provided by the VR environment. Questions related to user experience received high scores, with a specific emphasis on simulation realism (Q2) and the ease of recalling correct posture post-training (Q8), achieving mean scores exceeding 4.5 and positive response rates of 100% and 94.4%, respectively. Furthermore, the intuitiveness of the simulator (Q1), the ease of use of the interface (Q3), and the ability to respond to movements (Q5) were favorably assessed, with all items exceeding a response rate of 77% at higher levels (4 or 5).

With regard to knowledge transfer, the users’ perceived effectiveness in preparing for welding in real-world contexts (Q7) yielded a moderate average score of 3.61, with merely 38.9% of responses being positive. This finding implies that while users demonstrated confidence in recalling specific procedures (Q8), a subset of them remains hesitant about applying this confidence to actual welding situations, suggesting a potential gap in perceived transferability. Conversely, perceptions concerning anxiety reduction (Q12) were exceedingly favorable, with 100% of participants affirming that the use of VR contributed to alleviating the fear of making mistakes. The element related to visual and auditory feedback (Q15) achieved a positive mean score (mean = 3.89, with 61.1% favorable responses), indicating opportunities for enhancement regarding the clarity of the instructions and haptic cues provided within the virtual scene.

3.2.2. Quantitative Evaluation

To assess the efficacy of the VR simulator in a quantitative manner, data were gathered from users over two successive attempts. Three primary metrics were evaluated: simulation duration, soldering iron positioning inaccuracies, and grip errors, with the intention of examining the progression and refinement of the technique throughout the training process. The collected data enable the analysis of the influence of practice within the virtual environment on the diminution of errors and enhancement of task execution efficiency.

To quantify the effectiveness of the VR simulator, three main metrics were analyzed: simulation duration, soldering iron positioning errors, and grasping errors.

Table 3. Frequency of errors observed during the first and second attempts.

Error Type	First Attempt	Second Attempt
Iron positioning inaccuracies	6 errors	0 errors
Grip errors	13 errors	0 errors

presents the total frequency of errors recorded in the two attempts made by the participants.

As can be seen, soldering iron positioning errors were reduced from two to zero, while gripping errors, which were more frequent in the first attempt (with a total of 10), also disappeared completely in the second attempt. These results show an improvement in performance after a single practice session, suggesting that the virtual reality environment is effective in facilitating rapid error correction and promoting learning.

The findings demonstrate a marked improvement in welding performance within the VR simulator, particularly in mastering the correct posture of the soldering iron. This improvement suggests a more rapid learning trajectory and a decrease in the variability among user performance. As illustrated in Figure 7, there is a perceptible decrease in simulation time between initial and subsequent attempts, from an average of 2.92 min to 2.36 min, accompanied by a reduction in standard deviation from 0.5 to 0.32 min, as depicted in

Table 4. Paired t-test results comparing simulation times between Attempt 1 and Attempt 2.

Statistic	Value
Number of participants (n)	23
Mean of Attempt 1 (min)	2.92
Standard deviation of Attempt 1	0.45
Mean of Attempt 2 (min)	2.36
Standard deviation of Attempt 2	0.34
Mean difference ($\overline{d}$)	0.56
Standard deviation of the differences	0.31
Standard error of the mean ($s_d/\sqrt{n}$)	0.055
Calculated t statistic	10.25
Degrees of freedom ($df=n-1$)	22
p-value (two-tailed)	$7.67\times {10}^{-10}$

. This reduction signifies not only that participants were able to complete the task more quickly but also that there was increased uniformity in task execution, thus decreasing the variability among users. Such findings suggest that VR training facilitates standardization of the welding process, thereby ensuring that all operators acquire skills in a consistent and efficient manner.

To analyze whether the difference observed in execution times between the first and second attempts in the virtual reality simulator is statistically significant, a paired-sample t-test was applied. The resulting statistic t was $t\left(22\right)=10.25$ with a p value of approximately $7.67\times {10}^{-10}$, indicating that the probability that this improvement occurs by chance is virtually zero. Therefore, it can be stated that participants significantly improved their performance after a single repetition within the virtual environment, supporting the effectiveness of the simulator in promoting rapid skill acquisition.

3.3. Practical Skill Assessment After VR Training

This section presents an analysis of the participants’ performance after using the simulator. For this purpose, a rubric was designed with the objective of evaluating the effectiveness of the transfer of skills acquired in the virtual environment to a practical task, carried out some time after the experience with the simulator.

To evaluate the practical performance of the participants after completing the training in the virtual reality environment, a rubric was developed and is presented in

Table 5. Rubric used to evaluate post-training soldering performance.

Aspect	Optimal	Acceptable	Deficient
Joint quality	All joints solid and well-formed	Some with excess or poor shape	Incomplete joints
Iron positioning	Correct and stable placement throughout the process	Slight misalignment, but adjusted	Frequent mispositioning affecting joint quality
Solder application	Appropriate amount of solder evenly applied at each joint	Slight excess or insufficient solder, but joints remain functional	Excessive, insufficient, or poorly distributed solder compromising joint quality
Execution fluency	Maintains a continuous workflow with confident, autonomous actions	Minor delays or hesitation, but completes task independently	Frequent interruptions, visible uncertainty, or required assistance to complete the task

. The rubric includes four evaluation criteria: joint quality, iron positioning, solder application, and execution fluency. Each criterion is assessed on a three-level scale: Level 3 (Optimal), Level 2 (Acceptable), and Level 1 (Deficient), corresponding to scores of 3, 2, and 1, respectively. The final score for each participant is calculated by adding the values assigned in all criteria, with a maximum possible score of 12 points. This scoring system allowed a standardized and objective comparison of practical performance, capturing the extent to which participants were able to transfer the skills acquired in the virtual environment to a real-world soldering task.

The analysis of the results obtained from the rubric, presented in

Table 6. Individual results obtained from the rubric-based post-training evaluation.

User	Joint Quality	Iron Positioning	Solder Application	Execution Fluency	Total
User 1	2	2	2	2	8
User 2	3	3	2	2	10
User 3	2	3	2	2	9
User 4	2	3	2	2	9
User 5	2	2	2	2	8
User 6	2	2	2	2	8
User 7	3	2	2	2	9
User 8	3	3	3	2	11
User 9	3	3	3	2	11
User 10	3	3	3	2	11
User 11	3	2	3	2	10
User 12	3	2	3	2	10
User 13	3	2	3	2	10
User 14	3	2	3	2	10
User 15	3	2	3	2	10

, reflects the performance of 15 participants who completed the practical soldering task eight days after using the virtual reality simulator. The data reveal an overall positive trend in the transfer of skills from the virtual environment to the real-world task. Most of the participants scored between 9 and 11 points out of a maximum of 12, indicating solid performance in aspects such as iron positioning, solder application, and execution fluency. However, several users demonstrated recurring weaknesses in joint quality, possibly associated with errors in heat distribution or inadequate solder usage, suggesting specific areas for improvement. Three participants achieved only eight points, reflecting a more limited execution that may be linked to factors such as technical insecurity or lower familiarity with the physical tool. It is important to note that the virtual environment does not simulate the actual weight of the soldering iron, whereas in the physical task, this factor may influence the user’s ability to maintain stable positioning and proper control during the procedure.

Despite these variations, no participant scored poorly, reinforcing the effectiveness of the simulator in preparing students for manual welding tasks. These findings support the validity of the assessment and allow us to accurately identify individual strengths and weaknesses, guiding future pedagogical interventions.

4. Conclusions

This research demonstrates that virtual reality (VR)-based training can proficiently accelerate skill acquisition, diminish execution variability, and improve standardization in manual tasks. The simulator, developed using Unreal Engine 5 and implemented on Meta Quest 3, provides an immersive training environment that fortifies accurate soldering techniques through progressive learning modules and multi-modal feedback in real time.

Quantitative results demonstrated a substantial reduction in execution time between training iterations, decreasing from an average of 2.91 to 2.27 min, accompanied by a reduction in variability (standard deviation reduced from 0.50 to 0.32). Furthermore, all positioning and grip inaccuracies were resolved in the second attempt, indicating rapid acquisition of skills and procedural consistency. No practical errors were observed after training, suggesting an effective transfer of skills. Qualitative feedback corroborated these findings, underscoring high levels of user engagement, decreased anxiety, and enhanced confidence.

However, some constraints hinder the generalizability of these findings. The study was carried out under controlled laboratory conditions with a solitary group of participants, in the absence of a control group trained using conventional techniques. This limits the ability to draw causal inferences about the efficacy of the VR-based intervention. In addition, the evaluation was limited to a single soldering task—specifically, welding a metal pipe—and did not encompass long-term follow-up evaluations. The simulator feedback mechanisms, although functional, currently lack sophisticated real-time error analysis or personalized guidance.

To address these limitations, future research will introduce randomized control groups that follow conventional training methods, allowing for more rigorous comparisons. Longitudinal studies will be conducted to evaluate long-term skill retention. The simulator will also be expanded to support a wider range of welding techniques, materials, and task complexities, ensuring applicability in various industrial contexts.

In conclusion, to augment the educational experience, forthcoming advancements will examine the integration of artificial intelligence algorithms to deliver dynamic, real-time, and individualized feedback to trainees. This initiative will facilitate more profound error analysis and adaptive instruction, thereby aligning the training platform more closely with the cognitive requirements and skill levels of individual users.