AC machine fault diagnosis method based on ISBOA-Adaboost

Abstract

The invention relates to the technical field of artificial intelligence fault diagnosis, and discloses an AC machine fault diagnosis method based on ISBOA-Adaboost, and the method comprises the steps: carrying out the preprocessing of collected data through a VIKOR algorithm, improving the SBOA algorithm, introducing Sinsodic chaotic mapping in an SBOA population initialization stage, improving the population initialization uniformity, and carrying out the calculation of the SBOA algorithm. In the SBOA hunting stage, the positions of the snake egrets are updated according to the current number of iterations of the algorithm; an adaptive weight factor is introduced in an SBOA escape stage to optimize snake egret position updating, an improved snake egret optimization algorithm ISBOA is adopted to perform parameter optimization on an iterative algorithm Adaboost, an ISBOA-Adaboost fault diagnosis model is constructed, and the defect of blindness of parameter selection in the training process is made up.

Description

Technical Field

[0001] The invention relates to the technical field of AC machine fault diagnosis, and in particular to an AC machine fault diagnosis method based on ISBOA-Adaboost. Background Art

[0002] As a key component of the military gun control system, the AC machine is mainly composed of multiple modules such as coaxial electric-generator sets and frequency stabilizers. Its main task is to convert power to provide suitable three-phase AC power for the power transformer of the gun control box. Therefore, the AC machine plays a vital role in the gun control system. By diagnosing the faults of the tank AC machine, the battlefield survivability of the tank armor can be significantly improved, the stability of the overall system can be effectively improved, and the mobility and firepower output on the battlefield can be enhanced to a certain extent, further improving the overall combat effectiveness.

[0003] In the field of AC machine fault diagnosis, time-frequency analysis and feature extraction methods are usually used for processing. The frequency domain features of AC machine vibration and acoustic signals are extracted through short-time Fourier transform (STFT) or wavelet packet decomposition, and mechanical faults (such as bearing wear and rotor imbalance) are predicted by combining threshold judgment rules. However, this method requires a pre-defined fault feature library, and new faults are difficult to identify. In terms of model selection, support vector machines (SVMs) are usually selected to improve classification accuracy. However, when processing multi-channel sensor data of power modules, such as multi-dimensional time series signals such as voltage, current, and temperature, the kernel function selection mechanism of support vector machines (SVMs) lacks adaptability, resulting in inaccurate feature space mapping. Expert systems are limited by the scarcity of fault samples when building knowledge bases. When analyzing the multi-level topology of power modules, the combinatorial explosion problem of the fault tree analysis method causes the computational complexity to increase exponentially.

[0004] The AdaBoost algorithm overcomes the limitations of a single classifier by integrating multiple weak classifiers, especially when dealing with noise and sample imbalance. Compared with support vector machines (SVMs) and decision trees, AdaBoost has higher training efficiency and accuracy, and can effectively improve the generalization ability of the model. By iteratively weighting error samples, AdaBoost can improve the diagnostic ability of complex and dynamic fault conditions. Compared with expert systems and other traditional algorithms, it has obvious advantages in real-time and accuracy, becoming a powerful tool in fault diagnosis. Summary of the invention

[0005] In view of the above shortcomings and deficiencies in the prior art, the present invention provides an AC machine fault diagnosis method based on ISBOA-Adaboost, which uses an improved snake heron optimization algorithm ISBOA to optimize the parameters of the iterative algorithm Adaboost, and constructs an ISBOA-Adaboost fault diagnosis model, which makes up for the blindness of parameter selection in the training process, and includes the following steps: S1, collect AC module signal data; S2. Preprocess the collected data through the multi-criteria compromise solution sorting method VIKOR to screen out the data set that can be used as model input; divide the model input data set into a test data set and a training data set; S3. Improvements to the SBOA optimization algorithm for snake egrets, including the introduction of Sinusoidal chaotic mapping in the SBOA population initialization phase to improve the uniformity of population initialization; updating the position of snake egrets according to the current iteration number of the algorithm in the SBOA hunting phase; and introducing an adaptive weight factor in the SBOA escape phase. The position update of the snake heron is optimized to obtain the improved snake heron optimization algorithm ISBOA; S4, using the improved snake heron optimization algorithm ISBOA to optimize the key parameters of the iterative algorithm Adaboost, constructing the ISBOA-Adaboost fault diagnosis model, and training the fault diagnosis model with the training data set in S2; S5. Input the test data set in S2 into the trained ISBOA-Adaboost model, perform fault diagnosis on the AC machine, and output the prediction result.

[0006] In step S3, the Sinusoidal chaotic mapping is introduced in the SBOA population initialization stage, and the optimized position update formula is: ; Where: For the i +1 snake heron position; For the i Snake Heron Locations; a To control the parameters; Introducing adaptive weighting factors in the SBOA escape phase The formula for updating the position of the snake heron is: ; ; ; In the formula, is the adaptive weight factor; The position of the Secretary Bird during the escape phase; is the optimal position at the current stage; is the random position in the current iteration; and RB are both random numbers between [0,1]; is the initial position of the secretary bird; t is the current iteration number; T is the maximum iteration number; K is a random value 1 or 2; r is a random parameter.

[0007] The improved snake heron optimization algorithm ISBOA in step S4 performs parameter optimization on the iterative algorithm Adaboost, and the optimized parameters include learning rate, number of weak classifiers, decision tree depth and update method of sample weights.

[0008] Compared with the prior art, the present invention has the following beneficial technical effects and advantages: The improved snake heron optimization algorithm ISBOA is used to optimize the parameters of the iterative algorithm Adaboost, and the ISBOA-Adaboost fault diagnosis model is constructed, which makes up for the blindness of parameter selection in the training process and improves the prediction accuracy of the regression prediction model; the Sinusoidal chaotic mapping is introduced in the SBOA population initialization stage to improve the uniformity of population initialization and improve the convergence rate of the algorithm; in the SBOA hunting stage, the snake heron position is updated according to the current number of iterations of the algorithm; in the SBOA escape stage, an adaptive weight factor is introduced Optimize the position update of the snake heron to avoid falling into the local optimum. DETAILED DESCRIPTION

[0009] The present invention is described in detail below.

[0010] The present invention provides an AC machine fault diagnosis method based on ISBOA-Adaboost, comprising the following steps: S1. Collect the signal data value of the AC module through the equipment test bench as the initial data of the experiment.

[0011] S2. Preprocess the collected data through the multi-criteria compromise solution ranking method VIKOR to screen out the data set that can be used as model input; divide the model input data set into a test data set and a training data set.

[0012] Specifically, in step S2, the collected data is evaluated, sorted and normalized by the VIKOR algorithm: S201, construct the initial decision matrix: set an evaluation object with parameter indicators, determine the object set and indicator set , construct a decision matrix ; S202, decision matrix Normalize the decision matrix using the standard 0-1 transformation Normalize and get , The specific calculation formula is: When the indicator When it is a benefit indicator, , (1-1) When the indicator When it is a cost indicator, , (1-2) In the formula is the normalized decision matrix, is the maximum value in the matrix, is the minimum value in the matrix; S203, according to the normalized matrix , determine the positive ideal solution and negative ideal solution : , (1-3) in, , (1-4) In the formula is the maximum value in the normalized matrix, is the minimum value of the normalized matrix.

[0013] S204, calculate weight vector :Determine the index by constructing an optimization model Weight , the weight vector is calculated as follows: Using the positive ideal solution of the decision matrix and negative ideal solution Calculate the initial weight vector : , (1-5) For the initial vector After normalization, we can get: , (1-6) Then the indicator The weight of ; In the formula is the solution normalized to the ideal solution, is the normalized matrix, is a positive ideal solution, is a negative ideal solution.

[0014] S205, calculate each evaluation object The group benefit value 、Individual regret value , compromise value : , (1-7) , (1-8) , (1-9) in, , is the maximum group benefit value; , is the minimum group benefit value; , is the maximum individual regret value; , is the minimum individual regret value; is the normalized weight of each indicator; is a positive ideal solution, is the negative ideal solution, v is the decision mechanism coefficient; For the evaluation object , , These three values ​​are sorted in descending order, resulting in three sorting methods: Sorting, Sorting and Sorting, the final sorting result is based on The value is determined.

[0015] S206, determine the best solution: evaluate the object according to Sort the values ​​and get the solution ,choose The solution with the smallest value is regarded as the optimal solution; If the evaluation object in the program Value Satisfaction , then the compromise solution is obtained by verifying sorting conditions 1 and 2: Sorting condition 1: The smallest and second smallest solutions The value satisfies: , (1-10) Sorting condition 2: The smallest and second smallest solutions Value Sorting Satisfaction ,at the same time Sorting Satisfaction ; If sorting condition 1 and sorting condition 2 are met at the same time, then the solution It is the final compromise solution in the decision-making process; if only sorting condition 1 or only sorting condition 2 is met, the solution and solutions All are compromise solutions. As a final compromise solution; if sorting condition 1 and sorting condition 2 are not met at the same time, , get the maximum J, then the solution These are the final compromise solutions.

[0016] S3. Secretary bird optimization algorithm (SBOA) is a swarm intelligence optimization algorithm. The algorithm is inspired by the survival behavior of secretary bird in the natural environment. Secretary bird needs to constantly search for prey and avoid predators to survive. In order to solve the problems of poor convergence, low exploration efficiency and easy to fall into local optimality in the algorithm, the secretary bird optimization algorithm SBOA is improved, including the introduction of Sinusoidal chaotic mapping in the SBOA population initialization stage to improve the uniformity of population initialization and improve the convergence rate of the algorithm; in the SBOA hunting stage, the position of the secretary bird is updated according to the current number of iterations of the algorithm; in the SBOA escape stage, an adaptive weight factor is introduced The position update of the snake heron is optimized to avoid falling into the local optimum, and the improved snake heron optimization algorithm ISBOA is obtained.

[0017] (1) Initialization phase: In the initial phase of the algorithm, each secretary bird is randomly distributed in the search space, which can be expressed as: , (2-1) Where: is the initial position of the secretary bird; and are the upper and lower boundaries of the algorithm's search range; is a random number between [0,1]; i is a secretary bird individual; j is the dimension of the search problem.

[0018] There is still an initialization unevenness problem in the initialization stage, so the Sinusoidal chaotic mapping strategy is introduced to improve the snake heron optimization algorithm. The Sinusoidal chaotic mapping has the advantages of uniform traversal and fast convergence speed, which can make the positions of the population uniformly distributed during initialization, thereby improving the convergence rate of the algorithm. The improved initialization position update formula is as follows: , (2-2) Where: is the position of the i+1th snake heron; is the position of the ith snake heron; a is the control parameter and is generally taken as 2.3.

[0019] (2) Hunting phase: The hunting phase of the secretary vulture is divided into three small phases: searching, consuming, and attacking prey. Each small phase accounts for 1 / 3 of the algorithm iterations. Let t be the current iteration number; T is the maximum iteration number. when , currently in the search phase, the snake heron position is updated by the following formula: , (2-3) when , currently in the consumption phase, the snake heron position is updated by the following formula: , (2-4) when , currently in the stage of attacking prey, the position of the snake heron is updated by the following formula: , (2-5) In the formula, The position of the Secretary Bird during the hunting phase; and is the random position during the iteration of this stage; and RB are both random numbers between [0,1]; RL is a random number between [0,0.5]; is the initial position of the secretary bird; It is the optimal position at the current stage.

[0020] (3) Escape phase: When a secretary bird encounters a threat, it will activate an escape strategy, which can be divided into quick escape or camouflage. r is a random value; When 0< r When <0.5, the fast escape strategy is executed and the position of the snake heron is updated by the following formula: , (2-6) When 0.5≤ r When <1, the camouflage strategy is implemented and the position of the snake heron is updated by the following formula: , (2-7) In the formula, The position of the Secretary Bird during the escape phase; is the optimal position at the current stage; is the random position in the current iteration; is a random number between [0,1]; is the initial position of the secretary bird; K is a random value 1 or 2.

[0021] When the secretary bird is escaping, its position update will be close to the local optimal solution, but at this time it can only be close to the optimal solution, and thus cannot perform better local optimization. In order to solve the problem that the algorithm is prone to fall into the local optimal solution and has a slow convergence speed, an adaptive weight factor is introduced. It can not only affect the search capability and accuracy, but also speed up the algorithm convergence speed and effectively reduce the situation of falling into the local optimum. The improved position update formula is: adaptive weight factor , (2-8) When 0<r When <0.5, the fast escape strategy is executed and the position of the snake heron is updated by the following formula: , (2-9) When 0.5≤ r When <1, the camouflage strategy is implemented and the position of the snake heron is updated by the following formula: , (2-10).

[0022] S4. Use the improved snake-heron optimization algorithm ISBOA to optimize the parameters of the iterative algorithm Adaboost, build the ISBOA-Adaboost fault diagnosis model, and train the training data set.

[0023] AdaBoost is a classic boosting algorithm that improves the diagnostic ability of the overall model by combining multiple weak classifiers. The core idea of ​​AdaBoost is to improve the overall performance of the classifier by weighted combination of multiple simple classifiers and gradually correcting the errors of the previous classifier. During the training process, AdaBoost calculates the diagnostic error of each sample and adjusts the weight of the sample according to the error rate. The weight of the misclassified sample will increase, so that the subsequent classifier will pay more attention to these difficult-to-classify samples. In each iteration, AdaBoost adjusts the weight of the sample according to the performance of the current classifier, and combines the results of the previous round of classifiers to gradually form a strong classifier. In the classification task, the final model is obtained by weighted average of multiple weak classifiers. The key steps of the algorithm are as follows: Step 1: Initialize the weights of all training samples and set them to be uniformly distributed, that is, the weight of each sample is the same: , (3-1) Where: To represent the weight of sample i in the first round of training, initially, all sample weights are , that is, all samples are treated equally.

[0024] Step 2: Iteratively train multiple weak classifiers: Train the weak classifier under the current weight distribution: use the training set D and the current sample weight Training a weak classifier .

[0025] Calculate the classification error rate : Under the current weight distribution, calculate the weak classifier The weighted error rate is the total weight of the misclassified samples of the classifier under the current weight distribution: , (3-2) Where: is the classification error rate; is a weak classifier; Represents the sample weight of the current iteration; is an indicator function, if 1 if the value is true, 0 otherwise.

[0026] Calculate the weight coefficient of the weak classifier : Use classification error rate To determine the weight of the current weak classifier, the smaller the error rate, the greater the weight coefficient Bigger: , (3-3) Where: is the weight coefficient of the weak classifier; is the classification error rate; if ,but , indicating that the classifier has a certain ability to distinguish; if ,but , the performance of the classifier is equivalent to random guessing.

[0027] Update the sample weights so that the next round of weak classifiers pay more attention to the currently misclassified samples. The new weights are defined as: , (3-4) Where: is the new weight for the t+1th iteration; is the weight of the tth iteration; is the normalization factor, so that the sum of all sample weights is 1; is a weak classifier; is the weight coefficient of the weak classifier; are the training set labels.

[0028] Step 3: Combine weak classifiers After T rounds of training, all weak classifiers are combined into a strong classifier: , (3-5) Where; is a strong classifier; is a weak classifier; is the weight coefficient of the weak classifier; t is the current number of iterations, and T is the maximum number of iterations.

[0029] The final classification results are: , (3-6) In the formula: the sign function represents the sign; is the final classifier output; is the weight coefficient of the weak classifier; t is the current number of iterations, and T is the maximum number of iterations.

[0030] Improve the parameters of algorithm optimization; The AdaBoost algorithm also involves several key parameters in the modeling process, the most important of which include the learning rate (learning_rate), the number of weak classifiers (n_estimators), the decision tree depth (max_depth), and the update method of sample weights. The learning rate controls the adjustment of the weights of the weak classifiers at each iteration. If the learning rate is set too large, it may cause the model to overfit. If it is set too small, the model training will be too slow, or even the performance of the classifier cannot be fully improved. The number of weak classifiers controls the number of iterations. Generally, increasing the number of iterations helps improve the accuracy of the model, but it is also easy to cause overfitting. Therefore, it needs to be reasonably set according to the performance of the training set. The depth of the decision tree (if the decision tree is used as a weak classifier) ​​controls the complexity of each weak classifier. A tree that is too deep may cause overfitting, while a tree that is too shallow may not be enough to capture the complex patterns in the data. The update method of sample weights determines the degree of attention AdaBoost pays to misclassified samples. The weights of misclassified samples will be increased, so that subsequent classifiers will pay more attention to these difficult-to-classify samples. In order to ensure the performance of the model, it is crucial to reasonably adjust these parameter combinations for the training effect and diagnostic accuracy of AdaBoost.

[0031] S5. Input the test data set into the ISBOA-Adaboost prediction model, perform fault diagnosis on the AC machine, and output the prediction results.

[0032] The AC machine fault prediction process of the present invention is as follows: Step 1), input the data set, use the VIKOR algorithm to preprocess it, and divide the training sample and test sample data; Step 2), initialize the control parameters in ISBOA, including population size and maximum number of iterations; Step 3), using Sinusoidal chaotic mapping to generate the initial population position; Step 4), update the position of the snake heron in the hunting phase: determine the current number of iterations, when , at this time it is in the search stage and updates the position of the snake heron according to formula (2-3); , the current stage is the consumption stage and the position of the snake heron is updated according to formula (2-4); when , currently it is the stage of attacking prey and the position of the snake heron is updated according to formula (2-5); Step 5), update the position of the snake heron in the escape phase, when 0< r When <0.5, the snake heron executes a quick escape strategy and updates its position according to formula (2-9); when 0.5≤ r When <1, the snake heron executes the camouflage strategy and updates its position according to formula (2-10); Step 6), update the current best fitness and the corresponding snake heron position; Step 7) Determine whether the termination condition has been reached. If not, return to step 4); Step 8), assign the obtained optimal parameter combination to the Adaboost model; Step 9) Use the training data set to build the ISBOA-Adaboost fault diagnosis model, and finally use the test data set to verify the accuracy of the model.

[0033] 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.

[0034] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An AC machine fault diagnosis method based on ISBOA-Adaboost, characterized in that: The following steps are involved: S1, collect AC module signal data; S2. Preprocess the collected data through the multi-criteria compromise solution sorting method VIKOR to screen out the data set that can be used as model input; divide the model input data set into a test data set and a training data set; S3. Improvement of the snake heron optimization algorithm SBOA, including the introduction of Sinusoidal chaotic mapping in the SBOA population initialization stage to improve the uniformity of population initialization; In the SBOA hunting phase, the position of the snake heron is updated according to the current iteration number of the algorithm; in the SBOA escape phase, an adaptive weight factor is introduced The position update of the snake heron is optimized to obtain the improved snake heron optimization algorithm ISBOA; S4, using the improved snake heron optimization algorithm ISBOA to optimize the key parameters of the iterative algorithm Adaboost, constructing the ISBOA-Adaboost fault diagnosis model, and training the fault diagnosis model with the training data set in S2; S5. Input the test data set in S2 into the trained ISBOA-Adaboost model, perform fault diagnosis on the AC machine, and output the prediction result.

2. The AC machine fault diagnosis method based on ISBOA-Adaboost according to claim 1, characterized in that: In step S3, the Sinusoidal chaotic mapping is introduced in the SBOA population initialization stage, and the optimized position update formula is: ； Where: For the i +1 snake heron position; For the i Snake Heron Locations; a To control the parameters; Introducing adaptive weighting factors in the SBOA escape phase The formula for updating the position of the snake heron is: ； ； ； In the formula, is the adaptive weight factor; The position of the Secretary Bird during the escape phase; is the optimal position at the current stage; is the random position in the current iteration; and RB are both random numbers between [0,1]; is the initial position of the secretary bird; t is the current iteration number; T is the maximum iteration number; K is a random value 1 or 2; r is a random parameter.

3. The AC machine fault diagnosis method based on ISBOA-Adaboost according to claim 1, characterized in that: The improved snake heron optimization algorithm ISBOA in step S4 performs parameter optimization on the iterative algorithm Adaboost, and the optimized parameters include learning rate, number of weak classifiers, decision tree depth and update method of sample weights.

Images