Fault diagnosis method for equipment loading system

Abstract

The invention relates to the technical field of equipment loading systems, and discloses an equipment loading system fault diagnosis method, which comprises the following steps: acquiring fault characteristic parameters of an equipment loading system as original data; a tuna algorithm TSO is improved, including introducing Logistic chaotic mapping in a TSO initialization stage, introducing a nonlinear adjustment strategy in a TSO spiral hunting stage to enhance the global search capability and local development capability of the algorithm, and introducing a variation method in a differential evolution method to avoid the singleness of a population. An improved tuna algorithm ITSO is adopted to optimize the weight of a good and inferior solution distance method TOPSIS algorithm, an ITSO-TOPSIS data processing model is constructed, parameter optimization is carried out on a long short-term memory network LSTM algorithm, an ITSO-LSTM fault diagnosis model is constructed, the performance of the two algorithms is improved, the method can be more efficient and accurate in the fault diagnosis process, and the fault diagnosis efficiency is improved. Technical support is provided for maintenance of the equipment filling system, and the defect of blindness of parameter selection in the training process is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment loading systems, and particularly to a fault diagnosis method for an equipment loading system. Background Art

[0002] As an important part of equipment, the performance and working reliability of the automatic loading system play a key role in enhancing the overall combat ability of the equipment. However, the automatic loading system is a system with highly complex control logic and prone to mechanical failures. Once a failure occurs during its operation, it is often very difficult to identify and locate the failure, which may seriously affect the combat ability of the equipment. As a complex system with both sequential control and motion control characteristics, its multi-level logical relationships and strict timing requirements make comprehensive and efficient fault diagnosis a very challenging research topic. Therefore, how to improve the fault diagnosis ability of the automatic loading system has become one of the core issues in ensuring the combat performance of the equipment.

[0003] Traditional fault diagnosis methods are too dependent on manpower, consuming a large amount of resources and having low diagnostic efficiency, making it difficult to meet the requirements of complex systems. At the same time, the fault confirmation rate of these methods in complex systems is also relatively limited. In recent years, fault diagnosis technologies based on intelligent algorithms have gradually become an important research direction in the field of fault diagnosis for complex systems. In the equipment automatic loading system, the components are complex and the circuit logic is highly coupled. Therefore, during the fault diagnosis process, it is necessary to accurately locate the system at the component level and give specific maintenance suggestions to achieve comprehensive condition monitoring and accurate fault diagnosis of the automatic loading system. Summary of the Invention

[0004] Aiming at the above-mentioned disadvantages and deficiencies in the prior art, the present invention provides a fault diagnosis method for an equipment loading system, which improves the tuna algorithm TSO, optimizes the weights of the TOPSIS algorithm through the improved tuna algorithm ITSO, optimizes the key parameters of the LSTM algorithm, and constructs a fault diagnosis model, making up for the defect of blindness in parameter selection during the training process, including the following steps:

[0005] Step S01: Collect the fault characteristic parameters of the equipment loading system as the original data;

[0006] Step S02: Improve the tuna algorithm TSO, including introducing the Logistic chaotic mapping in the initialization stage of TSO, introducing a non-linear adjustment strategy in the spiral hunting stage of TSO to enhance the global search ability and local development ability of the algorithm and introducing the mutation method in the differential evolution method to avoid the singularity of the population, and obtaining the improved tuna algorithm ITSO;

[0007] Step S03: Use the improved tuna algorithm ITSO to optimize the weights of the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) algorithm, construct an ITSO-TOPSIS data processing model, perform correlation information sorting on the original data collected in Step S01 through the ITSO-TOPSIS data processing model, screen out the key attribute information that affects the equipment loading system, construct a data set, and divide it into a training set and a test set;

[0008] Step S04: Use the improved tuna algorithm ITSO to optimize the parameters of the Long Short-Term Memory (LSTM) algorithm, and construct an ITSO-LSTM fault diagnosis model;

[0009] Step S05: Use the training set in Step S03 to train the ITSO-LSTM fault diagnosis model in Step S04;

[0010] Step S06: Use the test set in Step S03 to test the trained ITSO-LSTM fault diagnosis model in Step S05; Step S07: Use the qualified ITSO-LSTM fault diagnosis model to perform health prediction on the equipment loading system and output the prediction results.

[0011] In Step S02, Logistic chaotic mapping is introduced in the initialization stage of TSO. The improved formula is: ; In the formula: is the initial position of the i-th individual after Logistic mapping; X i is the i-th individual randomly generated before Logistic mapping; ub is the upper boundary of the search space; lb is the lower boundary of the search space; N is the number of tuna groups, and R is the chaotic control parameter; Set the parameter a that determines the degree of following the best individual of the previous individual in the initial stage of the individual, and introduce a non-linear adjustment strategy to dynamically adjust a. The formula is: ; Among them; a max and a min are the initial value and the minimum value respectively; is the non-linear control factor; Substitute the improved a into the original mathematical model to update the weight coefficient α1 that controls the trend of the individual moving forward to the previous individual and the weight coefficient α2 that controls the trend of the individual moving to the best individual: ; ; Among them, t is the current iteration number, tmax is the maximum number of iterations; In the spiral foraging stage, the spiral search reference point used will gradually change from a randomly selected individual to the current optimal individual to update the position. The mathematical model is as follows: ; Where, is the individual at the (t + 1)-th iteration, is randomly generated in the search space as the spiral search reference point, is an intermediate variable, is the current best individual, is the individual at the t-th iteration, and rand is a random number in the range [0, 1]; At this time, the mutation method in differential evolution is introduced to implement individual mutation to generate mutant solutions, and then the mutant solutions and the original solutions are combined to generate crossover solutions, avoiding the singularity of the population. According to the value of the fitness function, it is judged whether to adopt new individuals. The calculation formula is as follows: ; ; ; Where: is the position of the mutant individual; both represent randomly selected individuals in the population; is the orientation of the individual after crossover; F is a scaling factor with a value range of [0.4, 1]; CR is a crossover probability factor with a value range of [0, 2].

[0012] Step S04 uses the improved tuna algorithm ITSO to optimize the parameters of the long short-term memory network LSTM algorithm. The optimized parameters include the weight factor, learning rate, and bias term.

[0013] Compared with the prior art, the present invention has the following beneficial technical effects and advantages: The present invention improves the tuna algorithm TSO by introducing the Logistic chaotic mapping, the non-linear adjustment strategy, and the mutation method in differential evolution, enhances its global search ability, as well as the accuracy and efficiency in local development, avoids the singularity of the population, and can have higher applicability to different algorithm parameters.

[0014] The improved tuna optimization algorithm is used to optimize the key parameters of the technique for order preference by similarity to an ideal solution (TOPSIS) algorithm and the long short-term memory network LSTM, improving the performance of these two algorithms, and can be more efficient and accurate in the process of fault diagnosis, providing technical support for the maintenance of the equipment loading system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flowchart of a fault diagnosis method for an equipment loading system of the present invention. Specific embodiments

[0016] The present invention will be described in detail below, but the protection scope of the present invention is not limited by the accompanying drawings.

[0017] The present invention provides a fault diagnosis method for an equipment loading system, including the following steps: Step S01: Collect the fault characteristic parameters of the equipment loading system as the original data.

[0018] Step S02: Improve the tuna swarm optimization (TSO) algorithm, including introducing the Logistic chaotic mapping in the initialization stage of TSO, introducing a non-linear adjustment strategy in the spiral hunting stage of TSO to enhance the global search ability and local development ability of the algorithm, and introducing the mutation method in the differential evolution algorithm to avoid the singularity of the population, obtaining the improved tuna swarm optimization (ITSO) algorithm.

[0019] Specifically: 1) Initialization stage: Introduce the Logistic chaotic mapping, and the improved formula is: ; In the formula: is the initial position of the i-th individual after the Logistic mapping; X i is the i-th individual randomly generated before the Logistic mapping; ub is the upper boundary of the search space; lb is the lower boundary of the search space; N is the number of the tuna swarm, and R is the chaotic control parameter.

[0020] 2) Spiral hunting stage: Based on the spiral hunting behavior of the tuna swarm in nature and combined with the characteristic that the tuna close to each other will transmit information to each other, construct the original mathematical model as follows: ; ; ; ; ; In the formula: is the current best individual, is the individual at the (t + 1)-th iteration, is the individual at the t-th iteration, is the weight coefficient controlling the trend of the individual moving forward to the previous individual, is the weight coefficient controlling the trend of the individual moving to the best individual, a is a constant determining the degree of the individual following the previous individual and the best individual in the initial stage, is a random number in (0, 1], and l are intermediate variables; t maxis the maximum number of iterations.

[0021] To avoid blindly following the optimal individual and resulting in not finding the prey after a long time of searching. Therefore, the existing optimal individual can be discarded, and a coordinate is randomly selected again in the entire search space as the new optimal individual. This is conducive to foraging and hunting in the entire space. The corresponding mathematical model is as follows: ; In the formula: is a point randomly generated in the search space as a reference for spiral search.

[0022] Since the tuna algorithm adjusts global search and local exploitation according to the value of a, and a is linearly changed in the original algorithm, it is difficult to adapt to different optimization problems. Therefore, a non-linear adjustment strategy is introduced here to dynamically adjust a according to the current number of iterations, so as to more flexibly control the balance between global search and local exploitation. The improved new formula is: ; where; a max and a min are the initial value and the minimum value respectively; is the non-linear control factor; Substitute the improved a into the original mathematical model to update the weight coefficient α1 that controls the tendency of an individual to move forward to the previous individual and the weight coefficient α2 that controls the tendency of an individual to move to the best individual, enhancing the global search ability and local exploitation ability of the algorithm.

[0023] The tuna school algorithm will conduct a large-scale global search at the beginning, and then gradually narrow the range, searching area by area. As the number of iterations of the tuna school optimization algorithm gradually increases, the reference point used in the spiral foraging stage will gradually change from a randomly selected individual to the current optimal individual. The mathematical model is as follows: ; where, is the individual at the (t + 1)-th iteration, is a reference point randomly generated in the search space for spiral search, is the current best individual, is the individual at the t-th iteration, and rand is a random number within the range [0, 1].

[0024] After the above position update, the fitness value of the current position is initialized and compared with the fitness value of the previous position to find the optimal offspring and substitute it into the next iteration. This method of updating the individual position without adding interference is prone to the problem of the algorithm falling into local optimum as the number of iterations increases. Therefore, the mutation method in differential evolution is introduced to implement individual mutation to generate mutant solutions, and then the mutant solutions and the original solutions are combined to generate crossover solutions, avoiding the singularity of the population. According to the value of the fitness function, it is judged whether to adopt a new individual, and its calculation formula is as follows: ; ; ; where: is the individual position after mutation; both represent randomly selected individuals in the population; is the individual orientation after crossover; F is a scaling factor with a value range of [0.4, 1]; CR is a crossover probability factor with a value range of [0, 2].

[0025] 3) Parabolic hunting stage: When the tuna school also hunts in a parabolic shape, it will also search for prey around. Assume that the selection probabilities of these two methods are each 50%, and the mathematical model is constructed as follows: ; ; In the formula: TF is a random number with a value of 1 or -1, and p is an adjustment factor that changes with the number of iterations.

[0026] Step S03: Use the improved tuna algorithm ITSO to optimize the weights of the TOPSIS algorithm for the distance method between superior and inferior solutions, construct the ITSO-TOPSIS data processing model, and sort the correlation information of the original data collected in step S01 through the ITSO-TOPSIS data processing model to screen out the key attribute information that affects the equipment loading system, construct a data set, and divide it into a training set and a test set.

[0027] Specifically, the steps of the TOPSIS algorithm: Construct a decision matrix: Let evaluation objects with parameter indicators to obtain a * decision matrix.

[0028] Standardize the decision matrix: To eliminate the influence of dimensions, standardize the decision matrix to obtain the standardized matrix: ; where is the standardized value, representing the standardized result of the j-th index of the i-th sample; x ij is the original index value.

[0029] Construct a weighted standardization matrix: Here, the improved tuna algorithm is introduced to optimize the weights of each index, and the standardized matrix is multiplied by the optimized weights , to obtain the weighted standardization matrix: ; where is the weight of the j-th index; is the weighted index value.

[0030] Determine the ideal solution and the negative ideal solution: According to the weighted standardization matrix, determine the ideal solution (the best value) and the negative ideal solution (the worst value): ; where: (if it is a benefit-type index) (if it is a cost-type index); (if it is a benefit-type index) (if it is a cost-type index); Calculate the distance of each sample to the ideal solution and the negative ideal solution: Calculate the Euclidean distance of the i-th sample to the ideal solution and the negative ideal solution : ; where: is the distance of the i-th sample to the ideal solution; is the distance of the i-th sample to the negative ideal solution.

[0031] Calculate the relative closeness: Calculate the relative closeness C i of the i-th sample relative to the ideal solution: ; where: is the relative closeness of the i-th sample.

[0032] According to the relative closeness C i sort the samples in descending order to obtain the final ranking of advantages and disadvantages.

[0033] Step S04: Use the improved tuna swarm optimization (ITSO) algorithm to optimize the parameters of the long short-term memory (LSTM) algorithm. The optimized parameters include the weight factor, learning rate, and bias term, and construct the ITSO-LSTM fault diagnosis model.

[0034] Specifically, the steps of the long short-term memory (LSTM) algorithm are as follows: (1) Forget gate calculation The forget gate is used to determine how much information from the previous cell state should be discarded.

[0035] The formula is as follows: ; Where: is the output of the forget gate (ranging from 0 to 1), representing the degree to which each cell state information is forgotten; is the Sigmoid activation function, ensuring that the output value is in the range of [0, 1]; is the weight matrix of the forget gate; is the concatenated input, including the previous hidden state and the current input x t ; is the bias term of the forget gate.

[0036] (2) Input gate calculation The input gate is used to determine how much new information should be added to the cell state. It consists of two parts: the input gate weight i t and the candidate state .

[0037] Input gate weight calculation: ; Where: is the output of the input gate (ranging from 0 to 1), representing the degree of influence of the current input on the cell state; is the weight matrix of the input gate; is the bias term of the input gate.

[0038] Candidate state calculation: ; Where: is the candidate cell state at the current moment, representing the information that can be added to the cell state; is the hyperbolic tangent activation function, used to compress the candidate state value to [-1, 1]; is the weight matrix of the candidate state; is the bias term of the candidate state.

[0039] (3) Update cell state Combine the calculation results of the forget gate and the input gate to update the cell state C at the current momentt : ; Wherein: is the cell state at the current moment; is the cell state at the previous moment; is the output of the forget gate, controlling the cell state information to be retained; is the output of the input gate, controlling the addition of new information; Current candidate state.

[0040] (4) Output gate calculation The output gate determines the hidden state h to be output at the current moment t . It is calculated in two steps: Output gate weight calculation: ; Wherein: is the output of the output gate (the value range is from 0 to 1), indicating which information in the cell state needs to be output; is the weight matrix of the output gate; The bias term of the output gate.

[0041] Hidden state update: ; Wherein: is the hidden state at the current moment, as the output at the current moment; is the current cell state; Compresses the value of the cell state to the range [-1, 1].

[0042] Step S05, training the ITSO-LSTM fault diagnosis model in step S04 with the training set in step S03; Step S06, testing the trained ITSO-LSTM fault diagnosis model in step S05 with the test set in step S03; Step S07, using the qualified ITSO-LSTM fault diagnosis model to perform health prediction on the equipment loading system and outputting the prediction result.

[0043] The present invention improves the tuna algorithm TSO by introducing the Logistic chaotic mapping, the nonlinear adjustment strategy and the mutation method in the differential evolution method, enhances its global search ability, as well as the accuracy and efficiency in local development, avoids the singularity of the population, and can have higher applicability to different algorithm parameters.

[0044] The improved tuna optimization algorithm is used to optimize the key parameters of the TOPSIS algorithm of the distance method between superior and inferior solutions and the long short-term memory network LSTM, so as to improve the performance of these two algorithms, making the process of fault diagnosis more efficient and accurate, and providing technical support for the maintenance of the equipment loading system.

[0045] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Any modifications, alterations, substitutions, and variations made by those of ordinary skill in the art to the above embodiments fall within the scope of the present invention.

[0046] 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 principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A fault diagnosis method for an equipment loading system, characterized in that, It includes the following steps: Step S01: Collect the fault characteristic parameters of the equipment loading system as the original data; Step S02: Improve the tuna swarm optimization (TSO) algorithm, including introducing the Logistic chaotic map in the initialization stage of TSO, introducing a non-linear adjustment strategy in the spiral hunting stage of TSO to enhance the global search ability and local exploitation ability of the algorithm, and introducing the mutation method in differential evolution to avoid the singularity of the population, so as to obtain the improved tuna swarm optimization (ITSO) algorithm; Step S03: Use the improved ITSO algorithm to optimize the weights of the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) algorithm, construct an ITSO-TOPSIS data processing model, sort the correlation information of the original data collected in Step S01 through the ITSO-TOPSIS data processing model, screen out the key attribute information affecting the equipment loading system, construct a data set, and divide it into a training set and a test set; Step S04: Use the improved ITSO algorithm to optimize the parameters of the long short-term memory (LSTM) algorithm, and construct an ITSO-LSTM fault diagnosis model; Step S05: Use the training set in Step S03 to train the ITSO-LSTM fault diagnosis model in Step S04; Step S06: Use the test set in Step S03 to test the trained ITSO-LSTM fault diagnosis model in Step S05; Step S07: Use the qualified ITSO-LSTM fault diagnosis model to perform health prediction on the equipment loading system and output the prediction results.

2. The fault diagnosis method of an equipment loading system according to claim 1, wherein: In Step S02, the Logistic chaotic map is introduced in the initialization stage of TSO, and the improved formula is: ； Wherein: is the initial position of the i-th individual after the Logistic mapping; X i is the i-th individual randomly generated before the Logistic mapping; ub is the upper boundary of the search space; lb is the lower boundary of the search space; N is the number of tuna schools, and R is the chaos control parameter; Set the parameter a to determine the degree of following the best individual of the previous individual at the initial stage of the individual, and introduce a non-linear adjustment strategy to dynamically adjust a. The formula is: ； Where ； a max and a min are the initial value and the minimum value respectively; is the non-linear control factor; Substitute the improved a into the original mathematical model to update the weight coefficient α1 that controls the trend of the individual moving forward to the previous individual and the weight coefficient α2 that controls the trend of the individual moving to the best individual: ； ； where t is the current iteration number, and t max is the maximum number of iterations; The spiral search reference point used in the spiral foraging stage will gradually change from a randomly selected individual to the current optimal individual to update the position. The mathematical model is as follows: ； Among them, is the individual at the (t + 1)-th iteration, is randomly generated in the search space as the spiral search reference point, is an intermediate variable, is the current best individual, is the individual at the t-th iteration, and rand is a random number within the range [0, 1]; At this time, the mutation method in differential evolution is introduced to realize individual mutation to generate mutant solutions, and then the mutant solutions and the original solutions are combined to generate crossover solutions, avoiding the singularity of the population and judging whether to adopt new individuals according to the value of the fitness function. The calculation formula is as follows: ； ； ； Wherein: is the position of the mutated individual; both represent randomly selected individuals in the population; is the orientation of the individual after crossover; F is a scaling factor with a value range of [0.4, 1]; CR is a crossover probability factor with a value range of [0, 2].

3. A fault diagnosis method for an equipment loading system according to claim 1, characterized in that: In Step S04, the improved ITSO algorithm is used to optimize the parameters of the long short-term memory (LSTM) algorithm, and the optimized parameters include the weight factor, learning rate, and bias term.

Images