A Fault Diagnosis Method for Gun Control Box Based on IEGO-RF

By using TOPSIS algorithm and improved eel-grouper algorithm to optimize the random forest model in the gun control box fault diagnosis, the training efficiency of intelligent algorithms in the existing technology under high feature dimensions and large-scale data sets is solved. The fitting problem achieves rapid and accurate fault diagnosis of the gun control box.

CN118709084BActive Publication Date: 2024-11-12SHENYANG SHUNYI TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411206446.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-11-12
Estimated Expiration
2044-08-30

AI Technical Summary

Technical Problem

In the fault diagnosis of gun control box, intelligent algorithms such as support vector machines and neural networks are inefficient in training when processing high feature dimensions and large-scale data sets, and are prone to overfitting, resulting in increased model complexity and cost. .

Method used

Using IEGO-RF-based fault diagnosis method, data preprocessing is performed through TOPSIS algorithm, Logistic chaotic mapping, weight factor and triangle walking strategy are introduced to improve the eel-grouper algorithm, and parameter optimization is performed on the random forest algorithm to construct fault diagnosis Model.

Benefits of technology

It improves the accuracy and stability of fault diagnosis, reduces the variance of the model, avoids overfitting, and realizes fast and accurate fault diagnosis of the gun control box.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118709084B_ABST
    Figure CN118709084B_ABST
Patent Text Reader

Abstract

A method for diagnosing a fault of a gun control box based on IEGO-RF relates to the technical field of diagnosing a fault of a gun control box, and comprises the following steps: step S01, collecting the original data of the gun control box signal; step S02, sorting and normalizing the collected original data of the signal by the TOPSIS algorithm correlation information; step S03, improving the eel-grouper optimization algorithm EGO; step S04, optimizing the parameters of the random forest algorithm by using the improved eel-grouper optimization algorithm; step S05, inputting the test set into the IEGO-RF diagnostic model, diagnosing the fault of the gun control box, and outputting the prediction result. The present invention adopts the random forest algorithm to diagnose the fault of the gun control box, and uses the improved eel-grouper algorithm to optimize the key parameters of the random forest, constructs a fault diagnosis model, and realizes the rapid and accurate diagnosis of the fault of the gun control box.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis of a gun control box, and in particular to a fault diagnosis method of a gun control box based on IEGO-RF. Background Art

[0002] As an important equipment for land warfare, the performance of tanks directly affects the strategic deployment and combat effectiveness on the battlefield. In particular, the fire control system of the tank, as a key component to ensure the accuracy of artillery fire, is of great importance. The gun control box, as the core of the fire control system, is responsible for integrating target information, calculating firing data, and controlling the maneuvering operation of the turret. If the gun control box fails, it may cause the failure of the entire system, thus affecting the combat effectiveness of the tank. With the development of technology, the structure and function of the gun control box are becoming increasingly complex. Fault diagnosis of the gun control box has become one of the key technologies to ensure the optimal performance of the tank.

[0003] In recent years, more and more intelligent algorithms have been applied in the field of fault diagnosis. When processing large-scale data sets, support vector machines (SVMs) have low training efficiency and are extremely sensitive to parameter selection, especially when the feature dimension is high. This may lead to a significant increase in the complexity and cost of model training in practical applications. Although neural networks are powerful, they have many internal parameters, and the training process is not only time-consuming, but also easy to fall into overfitting, especially when the amount of data is insufficient. Summary of the invention

[0004] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a gun control box fault diagnosis method based on IEGO-RF, which adopts a random forest algorithm to diagnose the fault of the gun control box, and uses an improved eel-grouper algorithm to optimize the key parameters of the random forest to construct a fault diagnosis model, thereby realizing rapid and accurate diagnosis of the gun control box fault.

[0005] In order to achieve the above object, the main technical solutions adopted by the present invention include:

[0006] A gun control box fault diagnosis method based on IEGO-RF includes the following steps:

[0007] Step S01, collecting the original data of the gun control box signal and dividing it into a test set and a training set;

[0008] Step S02: sorting and normalizing the collected signal raw data through the TOPSIS algorithm, extracting key attribute information and constructing a data set;

[0009] Step S03, improving the eel-grouper optimization algorithm EGO, including introducing Logistic chaotic mapping in the EGO population initialization stage, introducing weight factors in the EGO exploration stage, and introducing a triangle wandering strategy in the EGO development stage;

[0010] Step S04: using the improved eel-grouper optimization algorithm to optimize the parameters of the random forest algorithm, constructing an IEGO-RF fault diagnosis model, and training the training set;

[0011] Step S05: input the test set into the IEGO-RF diagnostic model, perform fault diagnosis on the gun control box, and output the prediction result.

[0012] Furthermore, the step S02 includes:

[0013] Step S201, constructing a decision matrix: assuming n evaluation objects with m parameter indicators, obtaining an n×m decision matrix, where n and m are both positive integers;

[0014] Step S202, decision matrix forward processing: converting the extremely large index, extremely small index, intermediate index, and interval index of the evaluation object parameter index into extremely large index;

[0015] Step S203, normalizing the forward matrix: normalizing the forward matrix to eliminate the dimension, assuming that the n-dimensional m-group matrix is:

[0016] ;

[0017] The normalized matrix is ​​recorded as Z, and the forward matrix is ​​normalized by the following formula:

[0018] ;

[0019] Where: For the matrix Every element in

[0020] Step S204: Calculate the score and normalize it:

[0021] Define the maximum value for each column With minimum , the formula is as follows:

[0022] ;

[0023] ;

[0024] Calculate the Elements and maximum distance , minimum distance , ;

[0025] ;

[0026] ;

[0027] The score is calculated as follows :

[0028] ;

[0029] Where: , , Respectively i The distance between each element and the maximum and minimum values, , are the maximum and minimum values ​​of each column respectively.

[0030] Furthermore, the step S03 includes:

[0031] Step S301, population initialization: Introduce Logistic chaotic mapping to initialize the population position. The mathematical model is:

[0032] ;

[0033] In the formula, is the value of the current iteration; is the system parameter;

[0034] Step S302: Introduce weight factors in the EGO exploration phase. The model of grouper tracking prey in the exploration phase is as follows:

[0035] ;

[0036] In the formula, is a random position vector, For the i Grouper in t The position at the iteration time, For the t The best position vector of the iteration, is the fitness value, , are the vectors of coefficients C1 and C2 respectively, r 1 , r 2 is a random number between 0 and 1, rand(1) is a random number between 0 and 1, is the shrinkage coefficient, T is the maximum number of iterations;

[0037] After introducing the weight factor, the formula is as follows:

[0038] ;

[0039] In the formula, is the weight factor;

[0040] Step S303: Send a signal to the sea eel. i Eel in the t The iteration position formula is as follows:

[0041] ;

[0042] Where: For the i Eel in the t Iteration position, t is the current iteration number, T is the maximum number of iterations; S r A random number between 0 and 100;

[0043] Step S304: Assuming that the probability of catching grouper and eel is equal, the mathematical model of the development stage is as follows:

[0044] ;

[0045] In the formula, is the attack path model of the eel to its prey, is the attack path model of grouper to prey, For the location of the prey, is a random number between 0 and 1;

[0046] The triangle wandering strategy is introduced in the development stage of cooperative attack on prey. The improved formula is as follows:

[0047] ;

[0048] Where: R is the vector calculated by the triangle walk strategy, rand(1) is a random number between 0 and 1, For the i Grouper in t The position at the iteration time, For the t The best position vector for the iteration.

[0049] The beneficial effects of the present invention are:

[0050] 1. The present invention uses the TOPSIS algorithm to preprocess the data, which avoids the subjectivity of the data, does not require an objective function, can well express the comprehensive impact of multiple influencing indicators, and has no strict restrictions on the distribution of data, sample size, and number of indicators.

[0051] 2. The present invention improves the eel-grouper algorithm by introducing Logistic chaotic mapping, weight factors and triangle wandering strategy, solves the problems of low solution accuracy, poor balance between exploration and development stages, and easy falling into local optimal solutions, thereby improving the accuracy of subsequent prediction results.

[0052] 3. The present invention optimizes the main parameters of random forest RF through the improved eel-grouper optimization algorithm IEGO, reduces the variance of the model, prevents overfitting, and improves the accuracy and stability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The present invention is a flowchart of a process for performing parameter optimization on a random forest algorithm using an improved eel-grouper optimization algorithm in Example 1 of the present invention. DETAILED DESCRIPTION

[0054] In order to better explain the present invention and facilitate understanding, the present invention is described in detail below through specific implementation modes in conjunction with the accompanying drawings.

[0055] The present invention provides a gun control box fault diagnosis method based on IEGO-RF, comprising the following steps:

[0056] Step S01, collect the original data of the gun control box signal and divide it into a test set and a training set.

[0057] Step S02: sort and normalize the collected signal raw data using the TOPSIS algorithm to extract key attribute information and construct a data set.

[0058] Specifically, the step S02 includes:

[0059] Step S201, constructing a decision matrix: suppose there are n evaluation objects with m parameter indicators, and obtain an n×m decision matrix, where n and m are both positive integers.

[0060] Step S202, decision matrix forward processing: convert the extremely large index, extremely small index, intermediate index, and interval index of the evaluation object parameter index into extremely large index. Since the evaluation criteria for the evaluation object parameter index are different in different environments, this step converts the four common indicators, extremely large index, extremely small index, intermediate index, and interval index, into extremely large index, i.e., matrix forward processing.

[0061] Step S203, normalizing the forward matrix: normalizing the forward matrix to eliminate the dimension, assuming that the n-dimensional m-group matrix is:

[0062] .

[0063] The normalized matrix is ​​recorded as Z, and the forward matrix is ​​normalized by the following formula:

[0064] ;

[0065] Where: For the matrix Z Every element in .

[0066] Step S204: Calculate the score and normalize it:

[0067] Define the maximum value for each column With minimum , the formula is as follows:

[0068] ;

[0069] ;

[0070] Calculate the Elements and maximum distance , minimum distance , ;

[0071] ;

[0072] ;

[0073] The score is calculated as follows :

[0074] ;

[0075] Where: , The bigger The smaller, , Respectively i The distance between each element and the maximum and minimum values, , are the maximum and minimum values ​​of each column respectively.

[0076] Step S03, improving the eel-grouper optimization algorithm EGO, including introducing Logistic chaotic mapping in the EGO population initialization stage, introducing weight factors in the EGO exploration stage, and introducing a triangle wandering strategy in the EGO development stage.

[0077] Specifically, the step S03 includes:

[0078] Step S301, population initialization: Introduce Logistic chaotic mapping to initialize the population position. Logistic chaotic mapping is a simple nonlinear dynamic system that exhibits chaotic behavior under certain parameter settings, which makes its output sequence very sensitive to initial conditions, showing randomness and unpredictability. Using it in the population initialization stage of the EGO algorithm helps to improve the diversity of the initial population, thereby increasing the coverage of the global search space. The mathematical model is:

[0079] ;

[0080] In the formula, is the value of the current iteration; is the system parameter;

[0081] Step S302, introduce weight factors in the EGO exploration phase. The grouper looks for the best fitness based on the location of the prey, and the location of the conger eel resting nearby is randomly selected from the initial population. Use two coefficients C1 and C2, whose random values ​​are greater than 1 or less than -1, to move the search range away from the optimal position. Use this random search to update the position of the search individual in the exploration phase, and the search individual is the position of the individual other than the optimal individual. This mechanism and coefficient value enhance the search capability and enable the EGO algorithm to perform a global search. The model of grouper tracking prey in the exploration phase is as follows:

[0082] ;

[0083] In the formula, is a random position vector, For the i Grouper in t The position at the iteration time, For the t The best position vector of the iteration, is the fitness value, , are the vectors of coefficients C1 and C2 respectively, r 1 , r 2 is a random number between 0 and 1, rand(1) is a random number between 0 and 1, is the shrinkage coefficient, and T is the maximum number of iterations.

[0084] In the exploration phase, the weight factor is introduced. The weight factor can dynamically adjust the exploration and development capabilities of the algorithm. The weight can be gradually reduced during the iteration process, so that the algorithm can shift from a broad search to a more detailed local search, thereby accelerating convergence and improving accuracy. That is, a large weight factor is conducive to jumping out of the local optimum. The formula after introducing the weight factor is as follows:

[0085] ;

[0086] In the formula, is the weight factor.

[0087] Step S303, send a signal to the moray eel. Groupers share their attack positions and observations by shaking their heads, especially those hungry groupers. If they find a target that has just escaped into the reef, they will swim to the nearest moray eel's hiding place during the day and perform a strange head-shaking dance while swinging their dorsal fins. In order to mathematically represent this head-shaking dance behavior, assume that the position of each eel is proportional to the position of the grouper, and use the formula of this step to update it. In this formula, the signal of the grouper is affected by the degree of hunger. By increasing the hunger level of the grouper, the moray eel is persuaded to attack the prey. In the formula of this step, is a number between 0 and 100 and increases over time. If the eel has no connections, its position will change to a new random position. i Eel in the t The iteration position formula is as follows:

[0088] ;

[0089] Where: For the i Eel in the t Iteration position, t is the current iteration number, T is the maximum number of iterations.

[0090] Step S304, the moray eel swims on the reef to catch prey, while the grouper swims in the open water near the coral reef to catch prey. According to research, the success rate of two fish catching prey together is five times that of them catching prey alone. Assuming that the probability of catching grouper and eel is equal, the mathematical model of the development stage is as follows:

[0091] ;

[0092] In the formula, is the attack path model of the eel to its prey, is the attack path model of grouper to prey, For the location of the prey, A random number between 0 and 1.

[0093] The triangle walk strategy is introduced in the development stage of cooperative attack on prey. The triangle walk strategy can effectively expand the search space, especially in complex or multi-peak optimization problems, helping the algorithm to explore multiple potential good areas, thereby increasing the possibility of finding the global optimal solution. The improved formula is as follows:

[0094] ;

[0095] Where: R is the vector calculated by the triangle walk strategy, rand(1) is a random number between 0 and 1, For the i Grouper in t The position at the iteration time, For the t The best position vector for the iteration.

[0096] Step S04: Use the improved eel-grouper optimization algorithm to optimize the parameters of the random forest algorithm, build an IEGO-RF fault diagnosis model, and train the training set.

[0097] Building a random forest involves the following steps:

[0098] (1) Initialize the model

[0099] Set the key parameters of random forest: number of trees, maximum number of features, minimum sample split, minimum number of samples per leaf node, etc.

[0100] (2) Build a single tree and repeat the following steps (a)-(d) for other trees in the forest.

[0101] (a) Sampling training data

[0102] Samples are extracted from the original training data set by self-sampling, and these samples are used to build the current tree. Assume that there are 𝑁 samples in the training set, and one sample is selected each time, put back and selected again, for a total of N times.

[0103] (b) Randomly select features

[0104] At each decision point, n features are randomly selected from all features. This randomness is the biggest difference between random forests and ordinary decision trees.

[0105] (c) Finding the best partition

[0106] For each selected feature, calculate the effect of different partitioning points and select the best partitioning point. A commonly used partitioning quality metric is information gain, which is calculated as:

[0107] ;

[0108] Where: IG is the information gain; D is the data set of the current node; H(D) It is a dataset D Entropy of f is the partitioning feature under consideration;D left and D right are the left and right subsets after partition; | D ∣、∣ D left ∣ and ∣ D right ∣ are the number of samples in the corresponding data sets.

[0109] The calculation formula of entropy H(D) is:

[0110] ;

[0111] In the formula, Yes Category i The proportion in dataset D, is the total number of categories.

[0112] (d) Building a decision tree

[0113] Repeat steps (c) and (d) until the preset tree depth is reached, or the number of samples in a node is less than the minimum number of samples that a node must have before continuing to split, or the samples of all nodes belong to the same category.

[0114] like Figure 1 As shown in the figure, the process of optimizing the parameters of the random forest algorithm using the improved eel-grouper optimization algorithm is as follows:

[0115] 1) Initialization parameters, including the number of trees, maximum number of features, minimum sample partitioning, minimum number of samples per leaf node, etc.

[0116] 2) According to Initialize the population, calculate the fitness values ​​of the individuals in the population, and determine the leader.

[0117] 3) Conduct exploration activities according to Update the individual's location information and calculate the individual's fitness value; if , then according to Update the best individual position information.

[0118] 4) If , update the position of the sea eel according to the formula in step S303 to prepare for launching a cooperative attack; otherwise, continue to return to 3).

[0119] 5) Cooperate in attacking prey. If p < 0.5, Update individual positions, otherwise, according to Update individual locations.

[0120] 6) Determine whether the algorithm meets the termination condition, that is, whether the maximum number of iterations has been reached, and output the optimal parameters. Otherwise, repeat 3)-5).

[0121] 7) Input the output optimal parameter combination into RF, build a fault diagnosis model, and perform fault diagnosis on the gun control box.

[0122] (3) Integrate the results of all trees

[0123] In classification tasks, the final prediction result is determined by voting.

[0124] Step S05: input the test set into the IEGO-RF diagnostic model, perform fault diagnosis on the gun control box, and output the prediction result.

[0125] The present invention uses a data set processed by the TOPSIS algorithm and a gradient random forest fault diagnosis model optimized by an improved eel-grouper algorithm to perform fault diagnosis on a gun control box. Compared with the original gradient boosting decision tree model, the unimproved random forest model optimized by the eel-grouper algorithm and other optimization algorithm models, the present invention has improvements and is more efficient and accurate in the fault diagnosis process. This model can perform fault diagnosis on a gun control box more efficiently and provide technical support for the maintenance of the gun control box.

[0126] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. Alterations, modifications, substitutions and variations of the above embodiments by a person skilled in the art are all within the scope of the present invention.

Claims

1. A fault diagnosis method for gun control box based on IEGO-RF, It is characterized in that The steps include: Step S01, collecting the original data of the gun control box signal and dividing it into a test set and a training set; Step S02: sorting and normalizing the collected signal raw data through the TOPSIS algorithm, extracting key attribute information and constructing a data set; Step S03, improving the eel-grouper optimization algorithm EGO, including introducing Logistic chaotic mapping in the EGO population initialization stage, introducing weight factors in the EGO exploration stage, and introducing a triangle wandering strategy in the EGO development stage; Step S04: using the improved eel-grouper optimization algorithm to optimize the parameters of the random forest algorithm, constructing an IEGO-RF fault diagnosis model, and training the training set; Step S05, input the test set into the IEGO-RF diagnostic model, perform fault diagnosis on the gun control box, and output the prediction result; The step S03 comprises: Step S301, population initialization: Introduce Logistic chaotic mapping to initialize the population position. The mathematical model is: ; In the formula, is the value of the current iteration; is the system parameter; Step S302: Introduce weight factors in the EGO exploration phase. The model of grouper tracking prey in the exploration phase is as follows: ; In the formula, is a random position vector, For the i Grouper in t The position at the iteration time, For the t The best position vector of the iteration, is the fitness value, , The coefficients C 1 , C 2 The vector of r 1 , r 2 is a random number between 0 and 1, rand(1) is a random number between 0 and 1, is the shrinkage coefficient, T is the maximum number of iterations; After introducing the weight factor, the formula is as follows: ; In the formula, is the weight factor; Step S303: Send a signal to the sea eel. i Eel in the t The iteration position formula is as follows: ; Where: For the i Eel in the t Iteration position, t is the current iteration number, T is the maximum number of iterations; S r A random number between 0 and 100; Step S304: Assuming that the probability of catching grouper and eel is equal, the mathematical model of the development stage is as follows: ; In the formula, is the attack path model of the eel to its prey, is the attack path model of grouper to prey, For the location of the prey, is a random number between 0 and 1; The triangle wandering strategy is introduced in the development stage of cooperative attack on prey. The improved formula is as follows: ; Where: R is the vector calculated by the triangle walk strategy, rand(1) is a random number between 0 and 1, For the i Grouper in t The position at the iteration time, For the t The best position vector for the iteration.

2. According to claim 1, a method for diagnosing faults of a gun control box based on IEGO-RF, It is characterized in that The step S02 comprises: Step S201, constructing a decision matrix: assuming n evaluation objects with m parameter indicators, obtaining an n×m decision matrix, where n and m are both positive integers; Step S202, decision matrix forward processing: converting the extremely large index, extremely small index, intermediate index, and interval index of the evaluation object parameter index into extremely large index; Step S203, normalizing the forward matrix: normalizing the forward matrix to eliminate the dimension, assuming that the n-dimensional m-group matrix is: ; The normalized matrix is ​​recorded as Z, and the forward matrix is ​​normalized by the following formula: ; Where: For the matrix Every element in Step S204: Calculate the score and normalize it: Define the maximum value for each column With minimum , the formula is as follows: ; ; Calculate the Elements and maximum distance , minimum distance , ; ; ; The score is calculated as follows : ; Where: , , Respectively i The distance between each element and the maximum and minimum values, , are the maximum and minimum values ​​of each column respectively.