A VR-based vehicle maintenance assistance system

Abstract

A VR-based vehicle repair and maintenance assistance system includes a vehicle data acquisition module, a vehicle fault database, a virtual fault repair database, a data processing module, a vehicle fault diagnosis module, and a virtual reality demonstration module. The present invention has the following beneficial effects: a BP neural network is applied to vehicle fault diagnosis, and the BP neural network is trained using the vehicle fault type and its corresponding vehicle operation data, thereby establishing a mapping between the vehicle operation data and the fault type. The vehicle fault type can then be diagnosed in real time based on the acquired vehicle operation data; virtual reality technology is applied to vehicle maintenance, and the vehicle maintenance process is reproduced through virtual reality technology. Maintenance personnel can more intuitively understand the maintenance process by watching the virtual maintenance process demonstration, laying a foundation for vehicle maintenance.

Description

Technical Field

[0001] This invention relates to the field of vehicle maintenance and repair, and specifically to a VR-based vehicle maintenance and repair assistance system. Background Technology

[0002] Rapid economic development has significantly improved people's living standards, leading to a surge in demand for vehicles, which in turn has propelled the rapid development of my country's automotive industry. Especially in recent years, with the rapid advancements in technology, automobile architectures have become increasingly complex, resulting in more stringent requirements for vehicle maintenance. The continuous advancement of new technologies has also spurred the constant updating and improvement of vehicle fault diagnosis techniques. Today, artificial intelligence has become a rapidly developing field with numerous practical applications and active research topics. Improving existing vehicle fault diagnosis techniques using artificial intelligence methods is of great significance for enhancing the overall level of automotive fault diagnosis technology.

[0003] Furthermore, vehicles have complex internal structures, and vehicle repair involves numerous repair steps and tools. Applying virtual reality technology to vehicle repair allows repair personnel to more intuitively understand the safety conditions, tool usage, and repair steps involved by watching virtual repair process demonstrations, thus laying a foundation for vehicle repair. Summary of the Invention

[0004] To address the aforementioned problems, the present invention aims to provide a VR-based vehicle maintenance and repair assistance system.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A VR-based vehicle maintenance assistance system includes a vehicle data acquisition module, a vehicle fault database, a virtual fault repair database, a data processing module, a vehicle fault diagnosis module, and a virtual reality demonstration module. The vehicle data acquisition module acquires vehicle operating data reflecting the vehicle's status. The vehicle fault database stores vehicle fault types and corresponding vehicle operating data. The virtual fault repair database stores vehicle fault types and corresponding repair methods. The data processing module normalizes the data from the vehicle data acquisition module and the vehicle fault database. The vehicle fault diagnosis module trains a BP neural network diagnostic model using the normalized data from the vehicle fault database and inputs the normalized vehicle operating data from the vehicle data acquisition module into the trained BP neural network diagnostic model for vehicle fault diagnosis. The output of the BP neural network diagnostic model is the vehicle fault type. The virtual reality demonstration module retrieves the corresponding repair method from the virtual fault repair database based on the fault type diagnosed by the vehicle fault diagnosis module and demonstrates the repair method using virtual reality technology.

[0007] The beneficial effects of this invention are as follows: Applying a backpropagation (BP) neural network to vehicle fault diagnosis involves training the BP neural network with vehicle fault types and their corresponding vehicle operating data, thereby establishing a mapping between vehicle operating data and fault types. This allows for real-time diagnosis of vehicle fault types based on the obtained vehicle operating data. Furthermore, applying virtual reality technology to vehicle maintenance allows for the reproduction of the vehicle maintenance process. By watching virtual maintenance demonstrations, maintenance personnel can gain a more intuitive understanding of the maintenance process, laying a foundation for vehicle maintenance. Attached Figure Description

[0008] The invention will be further illustrated with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the invention. For those skilled in the art, other drawings can be obtained based on the following drawings without any creative effort.

[0009] Figure 1 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0010] The present invention will be further described in conjunction with the following embodiments.

[0011] See Figure 1This embodiment of a VR-based vehicle maintenance assistance system includes a vehicle data acquisition module, a vehicle fault database, a virtual fault repair database, a data processing module, a vehicle fault diagnosis module, and a virtual reality demonstration module. The vehicle data acquisition module acquires vehicle operating data reflecting the vehicle's status. The vehicle fault database stores vehicle fault types and corresponding vehicle operating data. The virtual fault repair database stores vehicle fault types and corresponding repair methods. The data processing module normalizes the data from the vehicle data acquisition module and the vehicle fault database. The vehicle fault diagnosis module trains a BP neural network diagnostic model using the normalized data from the vehicle fault database and inputs the normalized vehicle operating data from the vehicle data acquisition module into the trained BP neural network diagnostic model for vehicle fault diagnosis. The output of the BP neural network diagnostic model is the vehicle fault type. The virtual reality demonstration module retrieves the corresponding repair method from the virtual fault repair database based on the fault type diagnosed by the vehicle fault diagnosis module and demonstrates the repair method using virtual reality technology.

[0012] This preferred embodiment applies a backpropagation (BP) neural network to vehicle fault diagnosis. The BP neural network is trained using vehicle fault types and their corresponding vehicle operating data, thereby establishing a mapping between vehicle operating data and fault types. This allows for real-time diagnosis of vehicle fault types based on the obtained vehicle operating data. Furthermore, virtual reality technology is applied to vehicle maintenance. By reproducing the vehicle maintenance process using virtual reality, maintenance personnel can gain a more intuitive understanding of the maintenance process by watching virtual demonstrations, laying a foundation for vehicle maintenance.

[0013] Preferably, the weights and thresholds of the BP neural network used in the vehicle fault diagnosis module are optimized using a particle swarm optimization algorithm. The fitness function f of the particle swarm optimization algorithm is defined as follows:

[0014]

[0015] In the formula, n represents the number of samples, Y ij Let y represent the j-th ideal output value of the i-th sample. ij Let represent the j-th actual output value of the i-th sample, and p represent the number of output nodes.

[0016] This preferred embodiment combines particle swarm optimization (PSO) with backpropagation (BP) neural network (BPNN). PSO optimizes the weights and thresholds of the BPNN, which helps the BPNN escape local minima and improves the performance of fault diagnosis. The smaller the defined fitness function value, the better the optimization result of the particles.

[0017] Preferably, three different particle update modes are defined in the particle swarm optimization algorithm. The first particle update mode is:

[0018]

[0019] X i (t+1)=X i (t)+V i (t+1)

[0020] The second particle update mode is:

[0021]

[0022] X i (t+1)=X i (t)+V i (t+1)

[0023] The third particle update mode is:

[0024]

[0025] X i (t+1)=X i (t)+V i (t+1)

[0026] In the formula, X i (t) and V i (t) represent the position and step size of particle i after the t-th iteration update, respectively, X i (t+1) and V i (t+1) represent the position and step size of particle i after the (t+1)th iteration update, respectively. This indicates that after the t-th iteration, particle i updates its corresponding inertia weight factor using the first update mode. This indicates that after the t-th iteration, particle i uses the second update mode to update its corresponding inertia weight factor. Let P(t) represent the inertia weight factor of particle i after the t-th iteration update, using the third update mode. `rand()` generates a random number between 0 and 1. `c1`, `c2`, `c3`, `c4`, `c5`, and `c6` are learning factors, with their values ​​set to 2. `B(t)` represents the global optimal solution of the particle swarm after the t-th iteration update. i (t) represents the individual optimal solution of particle i after the t-th iteration update, L i (t) represents the local reference solution of particle i after the t-th iteration update, and θ(Xi (t), X l (t) is a judgment function, and Among them, X l (t) represents the position of particle l after the t-th iteration update, f i (t) represents the fitness function value of particle i after the t-th iteration update, f l (t) represents the fitness function value of particle l after the t-th iteration update, d(X) i (t), X l (t) represents the position X of particle i after the t-th iteration update. i (t) and the position X of particle l after the t-th iteration update. l The distance between (t), R i (t) is the local reference radius of particle i after the t-th iteration update, and X represents the position of particle i after the t-th iteration update. i The minimum distance between (t) and the positions of other particles in the particle swarm, and X represents the position of particle i after the t-th iteration update. i The median distance between (t) and the positions of other particles in the swarm, and K i (t) represents the global reference solution of particle i after the t-th iteration update, and d(P i B(t), B(t)) represent the individual optimal solution P of particle i after the t-th iteration update. i The distance between B(t) and the global optimal solution B(t). X represents the position of particle i after the t-th iteration update. i The maximum distance between (t) and the positions of other particles in the particle swarm, and

[0027] This preferred embodiment defines three different particle update modes for particle swarm optimization (PSO), improving the diversity of the PSO algorithm and preventing it from getting trapped in local optima. The first mode allows particles to learn from the global optimum of the PSO, enhancing its global search capability. The second mode allows particles to learn from their individual optima and particles within their local reference radius whose fitness function values ​​are better than their own, improving its local search capability. The third mode incorporates both the individual optima and the global optimum of the PSO, balancing both global and local search capabilities. Furthermore, the third mode introduces a global reference solution; when the difference between the individual optima and the global optimum is significant, the added global reference solution in the third mode can effectively increase V. i The value of (t+1) can effectively improve the convergence speed of particles and is beneficial for global search.

[0028] Preferably, define This indicates the priority of particle i in selecting the j-th update mode for the (t+1)-th iteration update, and The expression is:

[0029]

[0030] In the formula, This represents the total number of iterations from initialization of particle i to the t-th iteration update using the j-th update mode. This represents the number of iterations corresponding to particle i being updated using the j-th update mode for the l-th time. Indicates that particle i is in the first... The fitness function value after the next iteration update. Indicates that particle i is in the first... The fitness function value after the next iteration update. For comparison functions, when At that time, when At that time, This represents the number of iterations corresponding to the most recent update of particle i using the j-th update mode. Indicates that particle i is in the first... The fitness function value after the next iteration update. Indicates that particle i is in the first... The fitness function value f after the next iteration update. i max (t) and f i min(t) represent the maximum and minimum fitness function values ​​of particle i from initialization to the t-th iteration update, respectively. Let f be the correction factor for the priority of particle i being updated using the j-th update mode after the t-th iteration. i When (t)≤H1(f), then When H1(f) < f i When (t)≤H2(f), then When f i When (t)>H2(f), then Where H1(f) and H2(f) are the given thresholds for the first fitness function and the second fitness function, respectively. Where N is the total number of particles in the particle swarm. Let be the average fitness function value of the particle swarm after the t-th iteration update. For comparison functions, Lan At that time, when At that time,

[0031] Particle i ultimately selects the update mode with the highest priority for the (t+1)th iteration update.

[0032] This preferred embodiment determines which update mode a particle should choose for iterative updates by calculating the priority of selecting the first, second, and third update modes. During iterative updates, the priority of each particle's choice of update mode is first calculated. The priority calculation formula includes an exponential function and a correction factor. The value of the exponential function is determined by the change in the particle's fitness function value after the most recent update using that mode and the success rate of the update using that mode. When a particle updates using that mode, the smaller its fitness function value compared to the previous iteration, the better the optimization result of that update mode, and the larger the value of the exponential function. Conversely, a higher success rate indicates a better optimization result, and the larger the value of the exponential function, meaning a higher priority for that update mode. The correction factor in the priority formula... Considering the current fitness function value of a particle, a smaller fitness function value after the current iteration indicates that the particle has a better search ability. This encourages particles with better search ability to explore better search regions globally, thereby driving the search of other particles. Therefore, the correction factor increases the priority for particles with better search ability to select the first update mode. A larger fitness function value after the current iteration indicates that the particle has a poorer search ability. This encourages particles with poorer search ability to strengthen their local search, i.e., the correction factor increases the priority for particles with poorer search ability to select the second update mode. When the fitness function value after the current iteration is between the first and second fitness function thresholds, it indicates that the particle has a moderate search ability in the swarm, possessing good global and local search capabilities. Therefore, the correction factor increases the priority for particles with moderate search ability to select the third update mode. Finally, the particle selects the update mode with the highest priority, as this selected update mode is more suitable for the particle's own characteristics, improving the particle's search ability.

[0033] Preferably, let This indicates that particle i updates its corresponding inertia weight factor using the j-th update mode after the t-th iteration, where j = 1, 2, 3. The value is determined in the following way:

[0034] (1) Arrange the particles in the particle swarm that selected the j-th update mode after the t-th iteration update and performed the (t+1)-th iteration update according to their fitness values ​​in ascending order to form a set J(t), and obtain the sorting of particle i in set J(t).

[0035] (2) Given the ideal fitness function value F(X), define the population optimization detection factor after the t-th iteration update as β(t). Then the expression for β(t) is:

[0036]

[0037] Where, f(P) i (t) represents the individual optimal solution P of particle i after the t-th iteration update. i The fitness function value corresponding to (t), f max (t) represents the maximum fitness function value of the particle swarm after the t-th iteration update, f min F(t) represents the minimum fitness function value of the particle swarm after the t-th iteration update, and F(X) is the given ideal fitness function value. The value of F(X) can be 0.0001. The position X corresponding to the ideal fitness function value F(X) is regarded as the ideal optimal solution.

[0038] but The value is:

[0039]

[0040] In the formula, ω max and ω min Given the maximum and minimum inertia weight factors, To update the corresponding maximum inertia weight factor using the j-th update mode, and N j Let J(t) be the number of particles in the set J(t).

[0041] This preferred embodiment is used to determine the inertia weight factor corresponding to the particle that selects the j-th update mode after the t-th iteration update in the particle swarm. The set inertia weight factor value is jointly determined by the particle's fitness function value and the swarm optimization detection factor. First, the particles that select the same update mode are sorted into a set J(t) according to their fitness values ​​from smallest to largest. The ranking of the particle fitness values ​​is obtained. The ranking reflects the quality of the particle's optimization result compared to other particles that select the same update mode. The higher the ranking, the smaller the fitness function value, indicating that the particle has a better optimization ability compared to other particles that select the same update mode. For such particles, a larger inertia weight factor is set, so that they can explore a better search area from the global perspective. Conversely, particles that are ranked lower have larger fitness values, indicating that the particle has a poorer optimization ability compared to other particles that select the same update mode. For such particles, a smaller inertia weight factor is set, so that they can converge to a better area more quickly, strengthening the optimization of that area. Local search; the ranking of particle fitness values ​​only reflects the optimization state of the individual particle, not the optimization state of the particle swarm as a whole. When the particle swarm has converged to near the ideal optimal solution, if a larger inertia weight is applied to a particle with better optimization ability, that particle is likely to escape the ideal optimal solution, thus reducing its optimization performance. To avoid this phenomenon, this preferred embodiment introduces a population optimization degree detection factor. The population optimization degree detection factor compares the fitness function value of each particle in the particle swarm with the given ideal fitness function value to determine the optimization state of the particle swarm as a whole. The larger the value of β(t), the closer the particle swarm is to the ideal optimal solution; the smaller the value of β(t), the farther the particle swarm is from the ideal optimal solution. By constraining the particle inertia weight factor based on the population optimization degree detection factor, the phenomenon of setting a large inertia weight factor for a particle with better optimization ability, causing that particle to escape the ideal optimal solution, is avoided, thus increasing the optimization performance of the particle swarm algorithm.

[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A VR-based vehicle maintenance and repair assistance system, characterized in that, The system includes a vehicle data acquisition module, a vehicle fault database, a virtual fault repair database, a data processing module, a vehicle fault diagnosis module, and a virtual reality demonstration module. The vehicle data acquisition module acquires vehicle operating data reflecting the vehicle's status. The vehicle fault database stores vehicle fault types and corresponding vehicle operating data. The virtual fault repair database stores vehicle fault types and corresponding repair methods. The data processing module normalizes the data from the vehicle data acquisition module and the vehicle fault database. The vehicle fault diagnosis module trains a BP neural network diagnostic model using the normalized data from the vehicle fault database and inputs the normalized vehicle operating data from the vehicle data acquisition module into the trained BP neural network diagnostic model for vehicle fault diagnosis. The output of the BP neural network diagnostic model is the vehicle fault type. The virtual reality demonstration module retrieves the corresponding repair method from the virtual fault repair database based on the fault type diagnosed by the vehicle fault diagnosis module and demonstrates the repair method using virtual reality technology.

2. The VR-based vehicle maintenance assistance system according to claim 1, characterized in that, The particle swarm optimization (PSO) algorithm is used to optimize the weights and thresholds of the backpropagation (BP) neural network used in the vehicle fault diagnosis module. The fitness function f of the PSO algorithm is defined as follows: In the formula, n represents the number of samples, Y ij Let y represent the j-th ideal output value of the i-th sample. ij Let represent the j-th actual output value of the i-th sample, and p represent the number of output nodes; The particle swarm optimization algorithm defines three different particle update modes. The first particle update mode is: X i (t+1)=X i (t)+V i (t+1) The second particle update mode is: X i (t+1)=X i (t)+V i (t+1) The third particle update mode is: X i (t+1)=X i (t)+V i (t+1) In the formula, X i (t) and V i (t) represent the position and step size of particle i after the t-th iteration update, respectively, X i (t+1) and V i (t+1) represent the position and step size of particle i after the (t+1)th iteration update, respectively. This indicates that after the t-th iteration, particle i updates its corresponding inertia weight factor using the first update mode. This indicates that after the t-th iteration, particle i uses the second update mode to update its corresponding inertia weight factor. Let P(t) represent the inertia weight factor of particle i after the t-th iteration update, using the third update mode. `rand()` generates a random number between 0 and 1. `c1`, `c2`, `c3`, `c4`, `c5`, and `c6` are learning factors, with their values ​​set to 2. `B(t)` represents the global optimal solution of the particle swarm after the t-th iteration update. i (t) represents the individual optimal solution of particle i after the t-th iteration update, L i (t) represents the local reference solution of particle i after the t-th iteration update, and θ(X i (t), X l (t) is a judgment function, and Among them, X l (t) represents the position of particle l after the t-th iteration update, f i (t) represents the fitness function value of particle i after the t-th iteration update, f l (t) represents the fitness function value of particle l after the t-th iteration update, d(X) i (t), X l (t) represents the position X of particle i after the t-th iteration update. i (t) and the position X of particle l after the t-th iteration update. l The distance between (t), R i (t) is the local reference radius of particle i after the t-th iteration update, and X represents the position of particle i after the t-th iteration update. i The minimum distance between (t) and the positions of other particles in the particle swarm, and X represents the position of particle i after the t-th iteration update. i The median distance between (t) and the positions of other particles in the swarm, and K i (t) represents the global reference solution of particle i after the t-th iteration update, and d(P i B(t), B(t)) represent the individual optimal solution P of particle i after the t-th iteration update. i The distance between B(t) and the global optimal solution B(t). X represents the position of particle i after the t-th iteration update. i The maximum distance between (t) and the positions of other particles in the particle swarm, and definition This indicates the priority of particle i in selecting the j-th update mode for the (t+1)-th iteration update, and The expression is: In the formula, This represents the total number of iterations from initialization of particle i to the t-th iteration update using the j-th update mode. This represents the number of iterations corresponding to particle i being updated using the j-th update mode for the l-th time. Indicates that particle i is in the first... The fitness function value after the next iteration update. Indicates that particle i is in the first... The fitness function value after the next iteration update. For comparison functions, when At that time, when At that time, This represents the number of iterations corresponding to the most recent update of particle i using the j-th update mode. Indicates that particle i is in the first... The fitness function value after the next iteration update. Indicates that particle i is in the first... The fitness function value after the next iteration update. and Let these represent the maximum and minimum fitness function values ​​of particle i from initialization to the t-th iteration update, respectively. Let f be the correction factor for the priority of particle i being updated using the j-th update mode after the t-th iteration. i When (t)≤H1(f), then When H1(f) < f i When (t)≤H2(f), then When f i When (t)>H2(f), then Where H1(f) and H2(f) are the given thresholds for the first fitness function and the second fitness function, respectively. Where N is the total number of particles in the particle swarm. Let be the average fitness function value of the particle swarm after the t-th iteration update. For comparison functions, when At that time, when At that time, Particle i ultimately selects the update mode with the highest priority for the (t+1)th iteration update.

3. A VR-based vehicle maintenance assistance system according to claim 2, characterized in that, it is designed to... This indicates that particle i updates its corresponding inertia weight factor using the j-th update mode after the t-th iteration, where j = 1, 2, 3. The value is determined in the following way: (1) Arrange the particles in the particle swarm that selected the j-th update mode after the t-th iteration update and performed the (t+1)-th iteration update according to their fitness values ​​in ascending order to form a set J(t), and obtain the sorting of particle i in set J(t). (2) Given the ideal fitness function value F(X), define the population optimization detection factor after the t-th iteration update as β(t). Then the expression for β(t) is: Where, f(P) i (t) represents the individual optimal solution P of particle i after the t-th iteration update. i The fitness function value corresponding to (t), f max (t) represents the maximum fitness function value of the particle swarm after the t-th iteration update, f min (t) represents the minimum fitness function value of the particle swarm after the t-th iteration update; but The value is: In the formula, ω max and ω min Given the maximum and minimum inertia weight factors, To update the corresponding maximum inertia weight factor using the j-th update mode, and N j Let J(t) be the number of particles in the set J(t).

Images