Virtual‐reality system for elevator maintenance education: Design, implementation and evaluation

With the rapid development of information technology, new educational models using virtual reality technology have received widespread attention from relevant researchers. In the field of vocational education, vocational colleges and training institutions can effectively mobilize students' learning initiative and improve their learning efficiency by using virtual reality technology. This study details the development process and system evaluation of a bespoke virtual reality system that offers a solution to the issues of uncertainty regarding hazards, high teaching expenses, and spatial constraints inherent in the practical training of elevator maintenance. By establishing a virtual environment that is highly reproducible and designing abundant interaction methods, this system facilitates students in attaining mastery over the structural make‐up of elevators, the principles of their operation, and the techniques involved in calibrating elevator governors. The system underwent testing by multiple users, and the satisfaction level of the system was ascertained through a questionnaire study, while the effectiveness of the system was evaluated using independent samples t test for data statistics concerning students' performance. The results of the study indicate that the system gained widespread praise among users, and it notably enhanced the students' learning drive, practical abilities, and on‐site adaptability.


F I G U R E 1 Elevator ownership in China.
these accidents is improper use and management.The main cause of these accidents is improper use and management of elevators.87.10% of the accidents are caused by the lack of professional competence of maintenance personnel and the lack of safety awareness in the operation process.Only 39,000 elevator maintenance personnel in China are licensed, and about 3000 people can successfully obtain the license through industry training every year. 2 However, due to the wide range of theoretical knowledge involved in elevators and the complexity of the knowledge points, it is difficult for trainees to learn and understand.The relevant training and assessment mainly focus on theoretical knowledge, even if the trainee obtains the practicing qualification, he/she still lacks the on-site operation ability, and he/she needs to carry out a long time-on-site internship on the site in order to obtain sufficient on-site operation experience.It is evident that the training efficiency of the elevator industry in China cannot keep pace with the rapid increase in elevator numbers.Maintenance personnel are trained to varying levels, with some lacking the ability to handle emergency situations and lacking theoretical knowledge related to elevators.Take the most important elevator maintenance process in the elevator speed limiter calibration link as an example, in addition to the major brands of elevator companies arranged by professional engineers can be done in about 15 min to complete the specification of the calibration, the rest of the professional elevator maintenance unit maintenance personnel calibrate an elevator speed limiter need to spend 30 min.Training courses for building test beds and using scaled-down models in vocational colleges and universities already exist.However, this traditional teaching mode has several shortcomings.The training efficiency is low, practical training equipment is easily damaged, and there are safety hazards during operation.Furthermore, the operating environment and field environment do not match, which further exacerbates the issues.Limited by human and material resources, local units are unable to provide systematic training for relevant operators.Consequently, some maintenance personnel have limited awareness of safety protocols, fail to adhere to standardized operations, and lack the ability to assess whether the results printed by the elevator governor calibration instrument meet the specified requirements.However, the space of the elevator maintenance site is very narrow, with many mechanical and electrical parts and harsh environment, the lack of safety awareness and professionalism will lead to danger for the maintenance staff.Therefore, this paper uses virtual reality technology to design a virtual training simulation system for elevator maintenance and repair in response to the above problems.Different from the traditional teaching model, this system uses the advantages of virtual reality technology to fully mobilize the learning initiative of the trainees, vividly and vividly describe relevant theoretical knowledge to the trainees, and design for the trainees according to the standardized maintenance procedures in the technical documents, such as routine maintenance test, speed limiter-safety gear joint control test.Through emerging technologies, trainees can deepen their impression of the operating principles, structural composition and safety precautions of elevators, and further standardize their maintenance work, which not only improves their maintenance capabilities but also reduces the risk of injury caused by improper operation.
VR technology makes reasonable use of a variety of human visual features, which makes virtual images realistic and enhances their immersion in the visual observation range, thus providing users with a more realistic visual experience, realizing virtual reality interaction, and deepening the contrast in functionality.By harnessing strong immersion, realism, and interactivity, it allows users to become fully engaged and immersed in the interactive processes of the virtual environment.Fernando's study illustrates that the incorporation of virtual reality technology into engineering curricula can notably augment the educational experiences of students.According to his research results, nearly 90% of students report increased contentment with this pioneering pedagogical method in contrast to conventional teaching approaches. 3herefore, the in-depth integration of virtual reality technology and vocational education, studying high-quality VR training teaching models can improve the quality of vocational education, and has important theoretical significance and practical value.
Training for elevator-related skills is primarily conducted by vocational schools or specialized institutions, typically employing scaled-down models for instruction.For maintenance personnel to acquire the competence for repair work, they must first undergo theoretical knowledge training, followed by a three-month practical apprenticeship on-site alongside two experienced technicians with several years of expertise.This traditional teaching model presents several issues: 1.The equipment available for practical training severely mismatches the number of trainees, resulting in limited hands-on experience, and prolonged training periods.2. Due to trainees' lack of familiarity with equipment operation, the likelihood of errors is high, posing not only a risk of equipment damage, thus increasing instructional costs but also endangering trainee safety.3. The theoretical learning process primarily relies on text-based instruction, diminishing trainees' engagement and training efficiency, with the theoretical content falling short of enhancing trainees' practical skills.4. Constrained by economic costs, the practical training content on the available equipment remains incomplete, and trainees are unable to utilize training instruments for extended periods.
The elevator training virtual reality system offers an effective solution to the aforementioned issues, with various modules within the system designed to achieve the following objectives: 1.The theoretical knowledge module establishes a multimedia-rich environment that immerses trainees through sound, images, text, and tactile feedback, reducing cognitive load-induced distractions and promoting immersive learning.2. The structural cognition module creates an extensive repository of high-fidelity models encompassing all elevator components, including equipment used during maintenance.These models are interactive, allowing trainees to rotate, scale, move, and examine them from various perspectives.Moreover, information regarding the functions, parameters, and usage of each component or device is readily accessible to trainees.3. The training and instructional module designs standardized, efficient operational procedures for daily maintenance and inspection, based on technical documentation provided by the organization.Drawing from constructivist theory, this module employs a range of instructional strategies, such as experiential learning, problem-based learning, and contextual learning. 4These strategies enable trainees to independently select their preferred learning modes, deepening their memory of practical operations, ultimately equipping them with proficient, standardized operational skills.

Virtual reality in the field of education related research
In 1968, Sutherland invented the VR head-mounted display, the closest thing to the modern concept of a VR device, which allows the computer to calculate new graphics in real time to be displayed to the user as their head posture changes. 5ince the 1990s, the emergence of virtual reality (VR) technology led to its application across diverse domains, encompassing military, engineering simulation, education, healthcare, and entertainment. 6In education, Antônio used the UE2 platform to conduct virtual modeling and system construction for nuclear power plants, and the virtual environment visualized the danger degree of radiation quantity. 7Cha et al. apply virtual reality technology to fire knowledge teaching to provide realistic training scenes for inexperienced firefighters. 8According to Tong's research, a significant number of scholars began integrating VR technology with education, forming a novel instructional paradigm, starting from what has been referred to as the "VR Year" in 2016.As of 2022, a cumulative total of 1678 high-quality research papers related to the application of VR technology in the field of education have been indexed in CNKI and WoS databases. 9These articles encompass topics such as the development of VR technology, system design, and evaluation, all within the context of the education industry.With the advent of the COVID-19 era, people have come to realize the significance of online education, as traditional modes of education are constrained by temporal and spatial limitations.In 2011, Hussin conducted a survey to assess educators' perspectives on emerging VR education models, and the results indicated that the majority of educational professionals were open to the use of virtual reality technology in the classroom. 10Marks analyzed and evaluated the course statistics for the Virtual Reality Lab at the University of Sydney, which showed that students were very interested in the lab and that teaching with VR technology helped students learn what they were learning in the classroom. 11In the post-COVID-19 era, virtual reality technology is considered advantageous and well-suited for the needs of higher education. 12,13n the past decade, with the commercial mixed reality head-mounted displays developed by Microsoft entering the market, augmented reality and mixed reality technologies have gradually been used in the field of teaching and training.Despite appearing later than virtual reality technology, these two technologies are widely used in education.Unlike virtual reality, AR and MR technologies do not replace reality, but rather focus on the connection between reality and the virtual, so in some areas AR and MR perform better than VR, especially in medical education. 14However, in the teaching and training of some industries, due to the shortage of real training equipment or the high risk of on-site operation, what the relevant enterprises need is to train the trainees by using the safety and immersion of the virtual reality environment, so that they can gain practical experience and avoid dangerous operation on site. 15In addition, AR and MR equipment is much more expensive than VR equipment and is not suitable for large-scale application.With the rapid advancement of communication and semiconductor technologies in recent years, the cost of using VR devices for education has significantly decreased. 16Virtual reality creates environmental simulations for education, offering learners a means of real-time interaction using their bodies and providing them with a unique perspective for a more immersive learning experience. 17When faced with challenges such as high experimental costs, complex or unobservable 3D structures of the teaching subject, or a high level of risk in real operational environments, these issues can be addressed through virtual modeling techniques, ultimately enhancing teaching efficiency.Wang has developed a virtual reality-based simulation system for close-coupled train couplers, addressing the teaching challenges stemming from the poor visualization of the structural aspects and the inaccessibility of their operational principles. 18Leveraging the excellent immersion capabilities of VR technology, Yu has developed a virtual simulation system for space-based lessons, enabling students to immerse themselves in and experience physics experiments in a space environment. 19Based on the actual training conditions for shipborne missile fire control, Cheng created a virtual model of the entire system and used C++ programming to provide real-time driving for the model.This allowed for training in the virtual missile combat process, resulting in significant cost savings and improved training effectiveness. 20Shen's research substantiates this viewpoint, with the study findings indicating that virtual reality simulation training effectively reduces training duration while providing safety and cost savings for the training process. 21Hamilton's research findings demonstrate that virtual reality technology excels in acquiring declarative knowledge and is particularly effective in obtaining procedural knowledge. 22Kablitz conducted a study on the application of VR technology in apprenticeship training within the retail sector of vocational schools, examining both the acquisition of knowledge and the occurrence of motion sickness.The results indicate that the new educational model significantly enhances knowledge acquisition in specific fields, with a relatively low incidence of motion sickness, and motion sickness does not impair learning success. 23Águeda have developed a VR-based aircraft maintenance system designed to capture the trainees' attention through a gamified learning process.This approach enables learners to more readily accept complex training, making it more accessible to the educated individuals. 24Cutini have addressed the limitations of traditional agricultural practices by developing a simulator suitable for precision agriculture. 25This simulator comprises a complete set of hardware for agricultural machinery and virtual reality software, enabling precise driving simulations of agricultural tractors.To address the issue of poor teaching effectiveness in nuclear accident emergency response, Guo conducted research into the incorporation of immersive virtual reality technology for emergency management training.This approach resolved the problem of incomplete and insufficient emergency equipment models, while also transcending the numerous time and space constraints associated with traditional teaching methods. 26Tran has developed a virtual reality welding system for practicing three welding techniques (SMAW, MIG, and TIG welding processes).This system encompasses instructional elements, welding practice, and assessment of welding outcomes, offering users a realistic welding experience. 27

Research of virtual reality technology in the elevator industry
As China has achieved comprehensive well-off status and improved the quality of life for its citizens, there is a growing demand for retrofitting elevators in older buildings.Additionally, the government provides subsidies for projects like elevator installation, further contributing to the increasing number of elevators in the country.However, the rate of elevator installations far exceeds the pace at which qualified elevator maintenance personnel can be trained.This has resulted in a severe shortage of qualified elevator maintenance workers, with a significant portion lacking practical experience and possessing only limited theoretical knowledge.To enhance the efficiency of employee training in elevator maintenance, it is imperative to embrace new technologies and depart from outdated teaching methods.Virtual reality technology can authentically simulate the elevator maintenance process, visually represent complex theoretical knowledge and operational mechanisms, and enable trainees to practice multiple times, thereby enhancing their practical experience.However, the application of virtual reality technology in the elevator field is still relatively small.Chen built a virtual training system for elevator inspection based on Unity3D, which improved the training efficiency of inspectors, but only developed simple daily inspection content and lacked safety tips in the inspection process. 2 Wang and his team have developed a virtual elevator installation system to enhance students' proficiency in elevator installation.However, the system is currently available only as a PC version and lacks immersion. 28Liu has employed virtual reality technology in the training for the installation and adjustment of elevator control systems.In this approach, VR technology is utilized to provide a virtual representation of the mechanical aspects of elevators, while the operation of the elevator control systems continues to rely on physical components.This hybrid approach effectively reduces the cost of hands-on training. 29Xie has explored the use of VR technology in the practical training for elevator inspections, analyzing the necessary functionalities of related systems.It is believed that this approach can effectively reduce the risk associated with the elevator inspection process. 30o date, researchers have not yet developed a virtual reality simulation system specifically tailored for training in elevator maintenance and inspection.Consequently, the development of a high-fidelity, immersive virtual reality simulation training system for elevator maintenance holds significant importance for the entire elevator maintenance industry.

System requirements analysis
The VR-based elevator maintenance virtual simulation system is designed for a user base that includes university students, learners from specialized equipment inspection organizations, and elevator maintenance personnel.Its primary objective is to familiarize learners with theoretical principles, regulations, and work processes while enhancing their practical skills through a realistic operational environment and standardized training procedures.The system should meet the following requirements: 1. Comprehensive educational content and standardized training processes.
2. Design of user interfaces within the scenes from both artistic and engineering perspectives, ensuring that the interfaces are visually appealing and user-friendly. 31. Real-time recording and updating of behavioral data related to the user's actions, automatically adjusting the system's responses based on the input information.

System function framework design
Based on the system's requirements analysis, the functional framework of the elevator maintenance virtual simulation training system consists of four functional modules: data storage, theoretical learning, structural recognition, and elevator maintenance training: 1. Data Storage Module: This module encompasses the storage of student registration information, student performance data, and a model library.It is responsible for identifying student registration and login details, recording student-learning progress, theoretical knowledge quiz scores, and practical assessment outcomes, and storing various types of governor models in the system.2. Theoretical Learning Module: In this module, students can access and learn various aspects of elevator theory, including elevator structural components, operational principles, control systems, elevator inspection safety regulations, and standardized governor inspection procedures.The theoretical knowledge assessment component randomly selects questions from the database to test and improve students' theoretical knowledge.3. Structural Recognition Module: Students entering this module can choose specific elevator models and types of governors and their overspeed governors from the model library for study.This module supports multi-angle model rotation and scaling, features textual explanations and voice narrations, and allows students to visualize the model's outer shell, as well as inspect its internal structural components.Students can disassemble and reassemble the model as part of their learning process.4. Elevator Maintenance Training Module: As the core component of the system, this module is crucial for enhancing students' practical skills.When students select this training module, the system enters VR training mode and offers standardized training in the inspection process for elevator governors.The system guides students through the steps, providing them with the necessary prompts to enhance their professional knowledge.The assessment module allows students to perform the operations independently, and the system records their actions, presenting a scoring summary based on key points and a scoring system that includes detailed performance evaluations.

System development process
The development process for the VR-based elevator maintenance training and education system is illustrated in Figure 2.
Relevant materials include elevator-related data, governors, governor testers, elevator environmental models, and 2D drawings along with three-dimensional parameters and spatial positioning information obtained from on-site surveys.Additionally, theoretical knowledge from the mobile industry and standardized elevator maintenance technical guidance documents are necessary.Using industrial modeling software Creo and 3D modeling software Maya, the acquired model data are used to create detailed models of the relevant equipment and scenes.Substance Painter is employed for model baking, surface processing, and texture mapping.A virtual reality environment is constructed on the Unity3D platform, with C# scripting for human-computer interaction design.The platform incorporates system features such as collision detection, camera perspective control, interactive animations, and trigger mechanisms.For hardware, the HTC VIVE Pro2 professional kit is used.Additionally, to enhance students' adherence to proper procedures through precise motion tracking, the system includes four additional VIVE trackers to monitor the exact positions of students' hands and legs.
After the system was developed, it was released to the computer through the Unity3D platform, and the system could be run in PC or VR mode freely within the software.

Establishment and optimization of virtual reality models
Virtual reality models are the fundamental components for system functionality, scene operation, and visual representation. 32In the system, the primary virtual reality models can be categorized as governors and testers, safety clamps, main motors, control cabinets, elevators, and maintenance environments.High-precision and complex functional models, such as governors and testers, are parameterized using industrial modeling software Creo, while the remaining models are created using 3D modeling software Maya.Parametric modeling using Creo results in 3D models with extensive datasets containing assembly information, geometric relationships, and high precision.However, the presence of abundant redundant information in these models leads to large model sizes, which can slow down model loading within the system, impair the smoothness of the simulation process, and significantly impact computer performance.Therefore, it is necessary to import the models into Maya for lightweight processing: 1. Traverse the extracted feature information and store it in categories.
2. Calculate the error measure matrix according to the QEM algorithm.
3. Calculate the triangle folding factor of the grid model.4. Using the QEM algorithm to fold the mesh model.
After completing the lightweight 3D model, the model remains in a low-poly state.To ensure high-fidelity performance in the system, model optimization is conducted using Substance Painter.In its baking function, specific information required for mapping to high-poly models is chosen based on the model's needs.Common choices include Diffuse, Specular, Metalness, Normal, Roughness, and Ambient Occlusion.By modifying the properties of this information, realistic texture maps are generated.The 3D modeling process is illustrated in Figure 3.
F I G U R E 3 Establishment and optimization of 3D models.

Key aspects of system development
To implement virtual reality training for elevator maintenance, the first step is to design the system's user interface (UI), which provides guidance through UI prompts.The main interface should include the system name and mode selection.
Based on the requirements, two modes should be designed: the teaching mode button triggers instructional logic, providing step-by-step prompts, and the assessment mode button triggers assessment logic, randomly presenting assessment content based on a question bank.Second, a prompt UI needs to be designed, which, based on the training process, automatically displays the required operational information on the prompt UI to guide the next action.Finally, device screen UI needs to be designed, such as the data display interface on the inspection device.The displayed data is synchronized through an SQL Server database, as shown in Figure 4.
The design of the framework is the most critical and complex logical design of the entire system.For this system, the framework is divided into a three-tier architecture: the data layer, the view layer, and the control layer.
The data layer is primarily used to store data that the entire system requires or shared data and is located in the system's common layer.The system is designed to include state data, with Boolean key-value pairs used for logical evaluations.For example, it includes logic for selecting teaching or assessment mode, whether to use fixtures, whether to start the inspection device, whether to detach the rope sheave, and whether the governor is reset.The data layer also incorporates numeric data with integer or float key-value pairs for storing information such as the number of governor brake cycles and test results.Character data, using string key-value pairs, stores textual content, including the governor verification process, assessment content, and more.
The view layer is mainly responsible for storing instances of relevant UI interfaces, independently managing all UI windows, and providing open interfaces for easy integration into other operational modules.It specifically includes the main interface, device interface, prompt interface, and UI panels for function buttons.These panels contain elements such as title text, various buttons, and the UI for the inspection device screen, which displays information like motor rotation speed and governor rotation speed.
The control layer is primarily responsible for the logical control of all system operations.In the case of the training system, the practical operation process is the most crucial part of the system.The control layer encompasses player behavior control, governor control, inspection device control, and control of the tool for locking between the elevator cables and the elevator car.
F I G U R E 4 Key contents of system development.

Real-time collision detection technology
Collision detection technology is a crucial technique for determining and detecting whether two different objects within the same system or game environment collide at a specific point in time.Developers set collision detection information for different game objects and display it in an output window, providing feedback to users for operational guidance.In a virtual environment, collision detection is necessary to determine whether a responsive event should be triggered.The accuracy of collision detection technology directly impacts the system's interactivity and immersion.In the elevator maintenance training system, a significant amount of interactive device modeling is involved, making the choice of collision detection methods particularly important in this system.With the development and application of virtual reality technology, some scholars have conducted in-depth analyses and research on collision detection techniques in the context of virtual reality.They have proposed high-precision collision detection algorithms for different objects.Existing mature collision detection techniques mainly fall into two categories: time-domain partitioning methods and space-domain partitioning methods.Time-domain partitioning methods include static collision detection methods, discrete collision detection algorithms, and continuous collision detection methods.Space-domain partitioning methods include object-based collision detection and image-based collision detection.Researchers have introduced various collision detection algorithms, highlighting their primary applications and limitations.For example, time-domain collision detection algorithms are mainly suited for regular geometric shapes and scene objects composed of geometric elements.One limitation of these algorithms is that they can have issues with interpenetration between models.Space-domain partitioning collision detection algorithms, on the other hand, face the challenge of how to partition the action space logically.However, they do not require preprocessing and are suitable for multi-soft-body models in complex structures.The classification of collision detection algorithms is illustrated in Figure 5.
In the elevator maintenance training system, the model components are relatively small, and the occurrence of interpenetration between models can affect the system's esthetics and precision.Additionally, the system's required functionality is to detect and determine collisions between moving objects in the virtual scene, and it does not require collision detection in static environments where the spatial relationships between models remain unchanged.Therefore, when considering the advantages and application scenarios of different algorithms, the system was developed using the spatial domain partitioning method with the Hierarchical Bounding Volume Hierarchies (BVH) algorithm.Theoretical analysis of the BVH algorithm reveals five key characteristics: (1) Minimal system memory usage when called.(2) Excellent bounding properties and tightness for scene objects.(3) Simple construction with minimal computational overhead.(4) Ease of adaptation for bounding box changes.(5) Simplified intersection testing during collision detection.Different types of bounding boxes have distinct performance characteristics when applied to the same object within a virtual scene.Common bounding box types include Axis Aligned Bounding Box (AABB), Sphere Bounding Box (Sphere), Oriented Bounding Box, and K-Dops Bounding Box, as illustrated in Figure 6.When students interact with the components of the elevator, performing operations such as translation and rotation, collision detection is essential to prevent interference between these components.The system only needs to perform intersection tests, which is why we chose the Axis Aligned Bounding Box.When components are controlled in motion by the user, the motion quantity can be set as Tran = ( t x , t y , t z ) , When a component in motion, the AABB of the component is updated accordingly, and the vertices are adjusted to reflect the movement.This transformation results in the relationship between the initial vertices (x 1 , y 1 , z 1 ) and the vertices after the movement (x 2 , y 2 , z 2 ).
When a component undergoes a change in position within the virtual scene, the update of the component's AABB involves the translation of the bounding box vertices.The predefined motion vector for the component is represented as T = ( t x , t y , t z ) the transformation formula for the change between the bounding box vertices (x, y, z) before translation and bounding box vertices ( x ′ , y ′ , z ′ ) after translation can be expressed as follows: If a component undergoes rotational motion with angles of rotation around the X, Y and Z axes, respectively, , , can be represented using matrices as: Then the transformation matrix is represented as: According to Equation (1), it can be observed that when the object is undergoing only rotational motion, the relationship is as follows: If the object is undergoing both rotational and translational motion, the relationship is as follows In the elevator overspeed governor training using the Axis Aligned Bounding Box, when a valid collision occurs between objects, the system will move the part to the appropriate position according to the relevant position information of the DoTween plug-in.The assembly constraints are automatically satisfied.It performs collision detection to ensure its effectiveness before execution.The collision detection algorithm process for Axis Aligned Bounding Box is illustrated in Figure 7. From the Axis Aligned Bounding Box collision detection algorithm process diagram, it can be observed that during the initial detection phase, bounding boxes are constructed for various objects in the scene to perform preliminary collision detection.If the bounding boxes of different objects do not intersect, it implies that these objects cannot collide.After the initial detection phase, the bounding boxes are updated, and further collision detection is performed, leading to the precise detection stage.In the precise detection stage, if collisions involving triangle faces are detected, it indicates that objects are indeed colliding.The system provides collision detection information and displays the detection results.

The design and implementation of the theoretical knowledge module
Elevator maintenance personnel require a significant amount of theoretical knowledge to obtain their professional qualifications.The theoretical knowledge module departs from traditional educational methods and adopts a more interactive and immersive approach.The goal of this module is to help students learn about elevator operation and control principles, safety education, and equipment operation related to elevator maintenance.During the learning process, students have the freedom to choose their learning pace.The system provides a learning plan for students, breaking down complex and abstract theoretical knowledge into multiple knowledge components.After completing each knowledge component, there is a theoretical knowledge assessment, and the related scores are recorded in the system's database for teacher review.The system's theoretical knowledge learning functionality is primarily implemented through interface UI interactions.After designing visually appealing panel UI in Photoshop, they are cropped and exported as PNG files.These files are then imported into the Unity3D platform, and the UI interfaces are designed using its built-in UI components.The main steps are illustrated in Figure 8.The assessment module is mainly connected through Unity3D's UI components and C# script interfaces, which allow the program to read data streams, recognize the database, control the Text component to display assessment questions, perform scoring in the system backend, and ultimately visualize the results through the script-controlled UI components.This module retains an update interface, enabling teachers to edit questions and update assessment content and scoring criteria in the required format at a later stage.

Design and implementation of the structural cognition module
The Structural Cognition module consists of three subsystems: Elevator Structure Cognition, Elevator Governor Structure Cognition, and Inspection Equipment Structure Cognition.These three subsystems are available in both PC and VR versions, allowing learners to understand the main structural components of elevators, the specific structure of elevator governors, and the structure of the instruments required for daily maintenance at any time.This module enables observational learning, allowing learners to freely explore within the virtual environment.From different spatial perspectives, learners can gain an intuitive understanding of the structure of various elevator components.Moreover, they can interact with components, such as disassembly, rotation, and zooming, to enhance their structural cognition.This forms the foundation for subsequent practical operations.The Unity3D platform incorporates robust collision detection algorithms.As shown in the figure, after adding Axis Aligned Bounding Boxes to all the elevator governor components, these parts can not only detect collisions with each other but also identify collisions through raycasting.This helps prevent model penetration between objects.In the VR version of the system, a combination of bounding box collision algorithms and raycasting detection algorithms is used to achieve disassembly and installation.In the PC version, Unity3D's Animator Component and the DoTween plugin are used to control the models.
F I G U R E 8 Development process of the theoretical knowledge module.

Design and Implementation of the elevator maintenance module
The daily maintenance module is the most crucial and complex functional module in this system, and it is the key to enhancing the practical skills of the trainees.Based on the technical documentation collected in the theoretical knowledge module, a reasonable design was made for the operation and presentation forms of elevator maintenance.The entire design process is based on constructivist teaching strategies and creates relevant practical scenarios.Before teaching, teachers understand the trainees' situation and determine whether VR environment adaptability experience is needed based on the students' individual circumstances.In the training, the system allows teachers to set up scenarios for students where their operational skills are weaker, offering targeted training.
In the virtual training environment, students' hands-on activities receive real-time behavioral results and environmental feedback, enabling them to master the correct operations.On one hand, students can practice repeatedly in the task scenarios simulated by virtual reality technology, ultimately transferring the operation to real work through muscle memory.On the other hand, students can manipulate virtual objects and observe the results of related behaviors, verifying whether their actions meet the requirements, thus deepening the memory of correct operational steps.
To implement the development of the elevator maintenance module, programming was performed in Unity3D using the C# language for the control logic of relevant operations.Taking the example of the most critical training in the elevator maintenance process, the joint verification of the governor and safety clamp, the logic control content of the verification module includes: 1. Control of the test instrument: When students set the diameter of the governor's rotating head and press the button on the test instrument with a gesture, the test instrument starts, and the motor begins to rotate.The display screen of the test instrument calculates the actual rotations based on the number of motor rotations stored in the system's backend and displays the relevant data in real-time.After the verification is completed, a verification certificate should be printed based on the test results.The control module for the test instrument is represented in part in the program content as shown in Figure 9. 2. Regarding the control of the elevator governor: The student picks up the rotating motor and touches it to the governor rope wheel, determining collisions through the trigger.Both interacting objects are bound to the respective shape of collision bounding boxes.If the bounding boxes make contact, it indicates a collision or trigger.Before this operation, the governor checks whether the rope pulley has been released.Only after receiving the result can the next step be performed.When the governor's rotational speed reaches the system's set limit, it triggers the locking mechanism.
The key logic here involves the condition checks for both electrical signal triggering and mechanical signal triggering.Additionally, after the governor locks, it initiates a reverse movement, and the brake block performs a bouncing animation.The frequency of this bouncing is determined by the angular velocity change of the governor rope wheel, creating a realistic physical effect of the governor locking.Part of the program content for the governor control module is shown in Figure 10. 3. Regarding the control of the clamping tool between the elevator cables and the elevator car body: The operation design of the tool mainly involves picking up and using it.According to the system's workflow requirements, it needs to be held in the correct position after picking up the tool.Picking up the tool requires two conditions to be met: whether  the handle is touching the tool and whether the grip button is pressed.When both conditions are satisfied, the tool becomes a child object of the handle, allowing for grabbing.Additionally, when picking up the tool, there should be a highlight prompt for the corresponding placement position.The highlight is turned on when picking up the tool and turned off when it is placed back.To place the tool in the correct position, you should release the grip.This is also determined through trigger detection while considering the grip state.When the conditions are met, the tool resets to the clamping position, completing the operation.The clamping tool control module's partial program content is depicted in Figure 11.

SYSTEM EVALUATION
In order to evaluate the effectiveness of the elevator maintenance training system based on virtual reality technology, this study conducted two sets of experiments.Participants were volunteers and the researchers gave written informed consent in accordance with the Declaration of Helsinki.All tests were conducted in accordance with the ethical guidelines stipulated in the latest revision of China's Measures for Ethical Review of Life Science and Medical Research Involving Human Beings, dated February 18, 2023, as amended.According to the above approach, the content of this experiment was only required to safeguard the participants' right to information and did not require ethical approval.This paper collects participants' subjective evaluation of the system through a questionnaire survey, and objectively evaluates the effectiveness of the system by analyzing their score changes.Two classes at a vocational college (each with 40 students) with similar skill levels were selected for the study.The research subjects were divided into an experimental group and a control group, and the duration of the experiments was set to 1 week in line with the vocational college's training schedule.The experimental group received training through the use of the VR-based system, while the control group received training through traditional teaching methods.Data related to the questionnaires and changes in assessment scores were collected and analyzed at the conclusion of the training.
TA B L E 1 Questionnaire and results of system satisfaction survey.

Mean Standard deviation
Immersion: The virtual environment of the system has the feeling of a real scene, which can enhance the attention of learning 4.225 0.880

Comfort:
The experience of using the system is comfortable enough 4.025 0.935 Interactivity: The system's interaction is perfect and can be easily mastered 3.725 0.948

Functionality:
The function of the system is complete enough 3.950 0.740

Motivation:
The system is interesting enough, willing to learn independently and willing to continuously use the system to improve technical ability 4.15 0.823

System satisfaction survey
The experimental group conducted a satisfaction survey immediately after the experiment, and the questionnaire was shown in the table.The questionnaire adopts a five-level scoring system, which mainly evaluates the five contents of the system: immersion, comfort, interactivity, functionality, and motivation.The survey results of the system satisfaction are shown in Table 1.
Based on the survey analysis, the questionnaire's overall mean score was 4.015.Scores were ranked from highest to lowest: Immersion (M = 4.225, SD = 0.880), Motivation (M = 4.15, SD = 0.823), Comfort (M = 4.025, SD = 0.935), Functionality (M = 3.950, SD = 0.740), and Interactivity (M = 3.725, SD = 0.948).This suggests a high level of student satisfaction with the system.Among these, Immersion received the highest score, suggesting that students find the system's immersive education practical and effective for learning.The score for Interactivity is relatively lower, highlighting the need for enhancements in the system's interactive features,so providing training on the use of VR equipment is recommended to familiarize students with the system.The Comfort score for the system is moderate, indicating that the majority of students find the VR equipment comfortable to use.Nevertheless, a minority of students reported discomfort, including symptoms like eye strain and dizziness during prolonged use of VR headsets.In addition to the questionnaire survey, the author analyzed the students who had physical discomfort and found that some of them were in contact with 3D environment for the first time.In conclusion, 90% of students expressed confidence in the system's ability to offer realistic, engaging, and immersive training, which enhances their learning efficiency through multisensory stimulation.The majority of students demonstrated a willingness to use the system for learning complex theoretical knowledge and voluntarily participated in practical exercises.

System effectiveness study
To assess the effectiveness of VR-based training compared to traditional training, this study collected performance data from both the experimental and control groups before and after the experiment, in which the scores were composed of 20% for the theoretical knowledge assessment, 20% for the safety knowledge assessment, and 60% for the practical assessment.Figure 12A divides the assessment scores into five levels, with different levels corresponding to different score bands.In addition, Figure 12B shows the scatter plot of students' scores.According to the results of the two graphs, there was a significant improvement in the performance of the experimental group.The number of high-scoring students has notably increased, while the count of low-scoring students has substantially decreased, indicating a superior hierarchical structure in the experimental group compared to the control group.To assess the system's effectiveness more precisely, this study conducted independent samples t tests on the pre-experiment and post-experiment scores of the experimental group and the control group to examine the impact of different training methods on student learning.
First, descriptive statistical analysis was performed on the initial scores of both the experimental and control groups using SPSS software, and the results are shown in Table 2 below.Independent samples t tests assume that the data follows a normal distribution.Therefore, this study employed the Kolmogorov-Smirnov and Shapiro-Wilk tests to determine whether the data exhibited a normal distribution.The results are presented in Table 3.Since the sample size of student  scores was less than 50, the Shapiro-Wilk test is more advantageous.According to both the K-S and S-W test results, which are far greater than 0.05, it can be considered that the student scores in the experiment follow a normal distribution.Subsequently, independent samples t tests were conducted on the pre-and post-experiment scores for both the experimental and control groups to assess the impact of the system on student performance.According to the results of Levene's test in Table 4 (Sig.= 0.645 > 0.5, Sig.= 0.815 > 0.5), the variance of the scores of the two groups was consistent.Under this condition, according to the test principle, the experimental group has a significant difference in performance, while the control group has no significant difference in performance.According to Table 5, the training scores of the experimental group significantly improved (t = 3.209, p < 0.05), with the average score increasing from 67.55 to 74.93.In contrast, the training scores of the control group showed some improvement but were not significant (t = 0.896, p > 0.05), with the average score increasing from 67.15 to 68.95.Therefore, while both methods contributed to enhancing student training scores, students performed significantly better in the VR-supported training environment.The VR-based elevator maintenance training system effectively enhances training efficiency and quality.

CONCLUSIONS
This paper presents an innovative approach to elevator maintenance education by utilizing virtual reality systems as a training tool.The use of VR provides interactive and practice training, which enhances training effectiveness and learning efficiency.This approach can be extended to other fields, such as mechanical maintenance, building safety, medical, or military training.This paper provides a detailed description of the development process of a virtual reality system, including architecture design, model construction, scene production, and interaction design.The methodology used in this development can be applied to other virtual reality systems, for education, training, or other fields.The evaluation of the system is conducted through a combination of questionnaires and quantitative analyses.This evaluation method can be applied to assess other virtual reality systems.It helps researchers evaluate the system's effectiveness and user satisfaction by collecting user's feedback, measuring learning outcomes, and other evaluation indicators.This elevator maintenance education system based on virtual reality technology exhibits the following characteristics: 1.The system has generated highly realistic virtual models of elevator structures and maintenance equipment.These models accurately replicate the colors, materials, and sizes of their real-world counterparts.The system also faithfully recreates the challenging conditions of elevator maintenance, employing environmental collision detection to warn students against dangerous actions.2. It dynamically showcases the structural components and operational principles of elevators, speed governors, and maintenance equipment.Additionally, it utilizes shader effects to illustrate the electrical principles behind the cooperative control functions of elevators.3. The system offers multiple forms of assistance, including text, voice, video, and animations, to aid students in their learning.4. The system supports practical training and assessment, identifying hazardous actions and providing automatic scoring and performance ratings.
This system deeply integrates vocational education with virtual reality technology.It effectively addresses issues such as high costs associated with training equipment, facilities, and faculty.Under the premise of ensuring teaching safety, training institutions can transform the traditional single teaching mode into a unified teaching mode, solving the problems of mismatch between training equipment and the number of students and the excessively long training cycle.The training content is designed according to the principles of constructivist teaching, so that students can practice learning in relevant interactive situations, and the efficiency of training has been significantly improved.Subjective satisfaction surveys and objective analysis of student performance data confirmed that the VR system effectively facilitated training.Due to the gamified approach and engaging nature, students willingly utilize the system during their free time for elevator maintenance practice, leading to a substantial improvement in training quality.

F I G U R E 5
Collision detection algorithm.

9
Partial program content of elevator governor tester control module.

F I G U R E 10
Partial program content of the governor control module.F I G U R E 11 Partial program content of the clamping tool control module.

F I G U R E 12
Macroscopic results of test results: (a) Percentage of students in each score band; (b) Scatterplot of Student Achievement.

statistics Group Statistics Standard error Group Statistics Standard error
Test of skewness and kurtosis.Test of normality.Note: * denotes that SPSS cannot calculate the exact p-value, but the lower limit of the true p-value is 0.200, which is still greater than 0.05.Independent samples t test.Analysis results of independent sample t test.
TA B L E 4Note: According to the principle of independent sample t test, when the variance of two independent samples is consistent, if Sig. (2-tailed) < 0.05, the two groups of data can be regarded as significantly different.TA B L E 5