Fault Diagnosis Method for Unmanned Mine Sweeping Vehicle

Abstract

The present invention relates to a fault diagnosis method for an unmanned mine-sweeping vehicle. First, the vulture search algorithm is improved by combining the Lévy flight strategy and the simulated annealing mechanism to enhance the local search ability of the vulture search algorithm and the search and utilization ability for the established solution space. Then, the improved vulture optimization algorithm is used to optimize the extreme learning machine to make up for the problem that the classification effect of the model is unstable due to the random selection of the initial input weights of the hidden layer and the bias values of the hidden layer during the fault classification process of the extreme learning machine. Finally, a fault diagnosis model based on the improved vulture search algorithm to optimize the extreme learning machine is established to diagnose the related faults of the unmanned mine-sweeping vehicle. Thus, the present invention significantly improves the maintenance efficiency and maintenance accuracy of the unmanned mine-sweeping vehicle.

Description

Technical Field

[0001] The present invention relates to the field of fault diagnosis, and particularly to a fault diagnosis method for an unmanned mine sweeping vehicle. Background Art

[0002] With the rapid development of information technology and assembly technology, modern control systems have shown many characteristics such as complexity, large scale, and integration. Undoubtedly, these characteristics pose major challenges to fault diagnosis technology. The unmanned mine sweeping vehicle is a newly developed unmanned mine sweeping device. In the mine sweeping and explosive disposal tasks, it can adopt a method combining two modes of remote command and autonomous navigation to independently complete the required tasks, and greatly reduce casualties while ensuring the completion of the tasks. However, the problem faced at the same time is that the overall precision of the unmanned mine sweeping vehicle is high, the structure is complex, and there are many components. Facing new types of weapons and equipment, professional maintenance personnel have less maintenance experience. When most inspections and repairs are carried out, not only is the efficiency low, but it is also difficult to ensure accuracy, and it is even easy to cause secondary damage to the equipment system. Therefore, how to quickly and accurately check the operating status of the equipment system, and timely restore the combat capabilities of the weapons and equipment, and achieve precision and standardization of weapon and equipment maintenance, is of great significance. Summary of the Invention

[0003] In view of this, the present invention proposes a fault diagnosis method for an unmanned mine sweeping vehicle, aiming to solve the problems of low maintenance efficiency and low maintenance accuracy of the current unmanned mine sweeping vehicle.

[0004] The fault diagnosis method for an unmanned mine sweeping vehicle according to an embodiment of the present invention includes:

[0005] S1. Introduce the Lévy flight strategy in the first stage of the Condor Search Algorithm to control the influence of the position change parameter and the random parameter;

[0006] S2. Introduce the simulated annealing mechanism in the second stage of the Condor Search Algorithm, and use the position of the condor population as the initial value of the simulated annealing mechanism for iterative optimization to obtain an improved Condor Search Algorithm;

[0007] S3. Optimize the Extreme Learning Machine according to the improved Condor Search Algorithm, and determine the optimal network structure parameters;

[0008] S4. Determine the fault diagnosis model of the unmanned mine sweeping vehicle according to the optimal network structure parameters;

[0009] S5. Perform fault diagnosis of the unmanned mine sweeping vehicle according to the fault diagnosis model of the unmanned mine sweeping vehicle, and evaluate the effectiveness of the algorithm.

[0010] Further, in the S1, the calculation formula for introducing the Lévy flight strategy to assist position update is:

[0011] Pnew，i = P best + α·r·(P mean - P i ) × Levy

[0012] where α is a constant parameter controlling position change, r is a random number uniformly distributed within the range [0, 1], P best represents the optimal search position of the vulture within the current search area, P i is the position of the i-th vulture, and P mean is the average position of the vulture population after completing the current search task.

[0013] Furthermore, in S2, the iterative optimization adopts the Boltzmann criterion, and the calculation formula is:

[0014]

[0015] where θ is the fitness difference between the optimal solution and the neighborhood solution, and T is a parameter that decreases periodically according to a predetermined rule during the search process.

[0016] Furthermore, S3 specifically includes:

[0017] Optimizing the input weights and biases of the hidden layer of the extreme learning machine according to the improved vulture search algorithm, and during the algorithm iteration process, using the fitness function to evaluate the quality of each solution, so as to determine the optimal network structure parameters;

[0018] The neural network formula of the extreme learning machine is:

[0019]

[0020] where x i is the input sample, n is the number of input layer nodes, ω i is the connection weight from the input layer to the hidden layer, b i is the bias, h(x) is the hidden activation function, β is the connection weight from the hidden layer to the output layer, L is the number of hidden layer nodes, y i is the output value, and m is the number of output layer nodes.

[0021] Furthermore, in S5, the diagnostic accuracy mean and the fitness mean are used to evaluate the effectiveness of the algorithm:

[0022] The calculation formula of the diagnostic accuracy mean is:

[0023]

[0024] The calculation formula of the fitness mean is:

[0025]

[0026]

[0027] Among them, M is the number of algorithm runs, acc(i) is the fault diagnosis accuracy rate at the i-th time, fitness(i) is the fitness value at the i-th time, and y i is the output value of the network model, and y ir is the actual value, and N is the length of the data sample.

[0028] In summary, for the fault diagnosis method of the unmanned mine-sweeping vehicle in the embodiment of the present invention, first, the Levy flight strategy and the simulated annealing mechanism are used to improve the vulture search algorithm to enhance the local search ability of the vulture search algorithm and the search and utilization ability for the established solution space. Then, the improved vulture optimization algorithm is used to optimize the extreme learning machine to make up for the problem that the initial input weights of the hidden layer and the bias values of the hidden layer are randomly selected in the process of fault classification, resulting in unstable model classification effects. Finally, a fault diagnosis model based on the improved vulture search algorithm optimized extreme learning machine is established to diagnose related faults of the unmanned mine-sweeping vehicle. Thus, the present invention significantly improves the maintenance efficiency and maintenance accuracy of the unmanned mine-sweeping vehicle. Description of the Drawings

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0030] Figure 1 is a schematic flowchart of the fault diagnosis method of the unmanned mine-sweeping vehicle in the embodiment of the present invention;

[0031] Figure 2 is a schematic flowchart of the improved vulture search algorithm in the embodiment of the present invention;

[0032] Figure 3 is a schematic flowchart of establishing a fault diagnosis model for an unmanned mine-sweeping vehicle based on an improved vulture search algorithm optimized extreme learning machine in the embodiment of the present invention. Detailed Embodiments

[0033] The description of the embodiments of this specification should be combined with the corresponding drawings, and the drawings should be regarded as a part of the complete specification. In the drawings, the shape or thickness of the embodiments can be enlarged and simplified or conveniently marked. Furthermore, each part of the structure in the drawings will be described separately. It should be noted that the elements not shown or described in words in the drawings are in the forms known to those of ordinary skill in the art.

[0034] In the description of the embodiments herein, any reference to directions and orientations is for the convenience of description only and should not be construed as any limitation on the scope of protection of the present invention. The following description of the preferred embodiments will involve combinations of features, which may exist independently or in combination. The present invention is not particularly limited to the preferred embodiments. The scope of the present invention is defined by the claims.

[0035] As Figure 1 shown, the method for fault diagnosis of the unmanned mine sweeping vehicle according to the embodiment of the present invention includes:

[0036] S1. Introduce the Levy flight strategy (Levy) in the first stage of the Bald Eagle Search Algorithm (BES) to control the influence of the position change parameter α and the random parameter r.

[0037] The calculation formula for the above-mentioned position update assisted by the Levy flight strategy is:

[0038] P new，i = P best + α·r·(P mean - P i ) × Leyy

[0039] where α is a constant parameter for controlling the position change, r is a random number uniformly distributed within the range of [0, 1], P best represents the optimal search position of the bald eagle in the current search area, P i is the position of the i-th bald eagle, and P mean is the average position of the bald eagle population after completing the current search task.

[0040] After step S1, the ability of the population in the Bald Eagle Search Algorithm to select the neighborhood can be improved.

[0041] S2. Introduce the Simulated Annealing mechanism (SA) in the second stage of the Bald Eagle Search Algorithm, and use the position of the bald eagle population as the initial value of the Simulated Annealing mechanism for iterative optimization.

[0042] The iterative optimization adopts the Boltzmann criterion, and the calculation formula is:

[0043]

[0044] where θ is the fitness difference between the optimal solution and the neighborhood solution, and T is a parameter that decreases periodically according to a predetermined rule during the search process.

[0045] After step S2, the solution accuracy of the Bald Eagle Search Algorithm can be enhanced.

[0046] The overall process of improving the traditional Bald Eagle Search Algorithm in steps S1 and S2 can be referred to Figure 2As shown, through steps S1 and S2, the Improved Vulture Search Algorithm (IBES) can be obtained.

[0047] S3. Optimize the Extreme Learning Machine (ELM) according to the Improved Vulture Search Algorithm, and determine the optimal network structure parameters.

[0048] Specifically, according to the Improved Vulture Search Algorithm, optimize the input weights and biases of the hidden layer of the Extreme Learning Machine, and during the algorithm iteration process, use the fitness function to evaluate the quality of each solution, so as to determine the optimal network structure parameters. Among them, the Extreme Learning Machine (ELM) is a machine learning algorithm, which simplifies the process of updating parameters by backpropagation in a single-hidden-layer neural network in principle. When applied to the embodiments of the present invention, it can reduce the proportion of human participation, making the model training faster and more accurate.

[0049] In this embodiment, the neural network formula of the Extreme Learning Machine is:

[0050]

[0051] where x i is the input sample, n is the number of input layer nodes, ω i is the connection weight from the input layer to the hidden layer, b i is the bias, h(x) is the hidden activation function, β is the connection weight from the hidden layer to the output layer, L is the number of hidden layer nodes, y i is the output value, and m is the number of output layer nodes.

[0052] S4. Determine the fault diagnosis model of the unmanned mine sweeping vehicle according to the optimal network structure parameters.

[0053] In this embodiment, since the fault diagnosis model of the unmanned mine sweeping vehicle is established by optimizing the Extreme Learning Machine (ELM) based on the Improved Vulture Search Algorithm (IBES), it can be abbreviated as the IBES-ELM model. The overall process of establishing the IBES-ELM model can be referred to Figure 3 as shown, including:

[0054] S41. Construct a preliminary fault diagnosis model according to the basic parameters of the unmanned mine sweeping vehicle. For example, 3 typical faults of the core power system of the unmanned mine sweeping vehicle (involving a total of 9 sensor signals) can be selected, and the Extreme Learning Machine (ELM) is optimized based on the Improved Vulture Search Algorithm (IBES) to establish a preliminary IBES-ELM fault diagnosis model, and input the overall data sample, data normalization processing, and divide the data sample, where the division of the data sample can be carried out according to a predetermined ratio.

[0055] S42. In the model training stage, the divided data samples are respectively input into the model in step S41 for training. Each data sample can obtain a fitness value. Through continuous iterative optimization, an optimal fitness value is determined, and the network structure parameters corresponding to the optimal fitness value are extracted and assigned to the ELM model, so as to determine the optimal IBES-ELM model. The optimal IBES-ELM model is also the finally determined fault diagnosis model for the unmanned mine sweeping vehicle.

[0056] S5. Perform fault diagnosis on the unmanned mine sweeping vehicle according to the fault diagnosis model of the unmanned mine sweeping vehicle, and evaluate the effectiveness of the algorithm.

[0057] Perform fault diagnosis on the unmanned mine sweeping vehicle according to the fault diagnosis model of the unmanned mine sweeping vehicle obtained in step S4.

[0058] The evaluation of the algorithm effectiveness uses the mean diagnostic accuracy and the mean fitness as evaluation indicators.

[0059] The calculation formula for the mean diagnostic accuracy is:

[0060]

[0061] The calculation formula for the mean fitness is:

[0062]

[0063]

[0064] Among them, M is the number of algorithm runs, acc(i) is the fault diagnosis accuracy at the i-th time, fitness(i) is the fitness value at the f-th time, y i is the output value of the network model, y ir is the actual value, and N is the length of the data sample.

[0065] The fault diagnosis method of the embodiment of the present invention aims at the deficiency of the traditional extreme learning machine in the randomization of the initial hidden layer input weight value and the hidden layer bias in the fault classification process, and proposes to optimize it by using an improved vulture optimization algorithm.

[0066] According to the fault diagnosis method of the embodiment of the present invention, first, 3 typical faults of the core power system of the unmanned mine sweeping vehicle involving 9 sensor signals can be selected to establish a fault diagnosis model based on the improved vulture search algorithm to optimize the extreme learning machine, and convert the fault diagnosis problem into a multi-classification problem; then, the sample data can be divided according to a ratio and input into the improved model for training and testing respectively; finally, the models before and after the improvement are compared, and a comprehensive evaluation is carried out in terms of the fault diagnosis accuracy and the error degree.

[0067] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A fault diagnosis method for an unmanned mine-sweeping vehicle, characterized in that, The method includes: S1. Introduce the Levy flight strategy in the first stage of the condor search algorithm to control the influence of the position change parameter and the random parameter; S2. Introduce the simulated annealing mechanism in the second stage of the condor search algorithm, and use the position of the condor population as the initial value of the simulated annealing mechanism for iterative optimization to obtain the improved condor search algorithm; S3. Optimize the extreme learning machine according to the improved condor search algorithm and determine the optimal network structure parameters; S4. Determine the fault diagnosis model of the unmanned mine sweeping vehicle according to the optimal network structure parameters; S5. Perform fault diagnosis on the unmanned mine sweeping vehicle according to the fault diagnosis model of the unmanned mine sweeping vehicle and evaluate the effectiveness of the algorithm.

2. The method for diagnosing faults of an unmanned mine-sweeping vehicle according to claim 1, characterized in that, In the above S1, the calculation formula for introducing the Levy flight strategy to assist position update is: P new，i = P best + α·r·(P mean - P i ) × Levy Among them, α is a constant parameter for controlling position change, r is a random number uniformly distributed within the range of [0, 1], and P best represents the optimal search position of the vulture within the current search area, and P i is the position of the i-th vulture, and P mean is the average position after the vulture population has completed the current search task.

3. The method for diagnosing faults of an unmanned mine-sweeping vehicle according to claim 1, characterized in that In the above S2, the iterative optimization adopts the Boltzmann criterion, and the calculation formula is: Among them, θ is the fitness difference between the optimal solution and the neighborhood solution, and T is a parameter that decreases periodically according to a predetermined rule during the search process.

4. The method for diagnosing faults of the unmanned mine-sweeping vehicle according to claim 1, wherein, The above S3 specifically includes: Optimize the input weights and biases of the hidden layer of the extreme learning machine according to the improved condor search algorithm, and use the fitness function to evaluate the quality of each solution during the algorithm iteration process, so as to determine the optimal network structure parameters; The neural network formula of the extreme learning machine is: Among them, x i is the input sample, n is the number of nodes in the input layer, ω i is the connection weight from the input layer to the hidden layer, b i is the bias, h(x) is the hidden activation function, β is the connection weight from the hidden layer to the output layer, L is the number of nodes in the hidden layer, y i is the output value, and m is the number of nodes in the output layer.

5. The fault diagnosis method of the unmanned mine-sweeping vehicle according to claim 1, characterized in that, In the above S5, the diagnostic accuracy mean and the fitness mean are used to evaluate the effectiveness of the algorithm: The calculation formula for the diagnostic accuracy mean is: The calculation formula for the fitness mean is: Among them, M is the number of algorithm runs, acc(i) is the fault diagnosis accuracy rate at the i-th time, fitness(i) is the fitness value at the i-th time, and y i is the output value of the network model, and y ir is the actual value, and N is the length of the data sample.

Images