| The technology of armored vehicle fire control system is constantly updated with the development of science and technology.At the present stage,the fire control system is characterized by higher technical content,more complex structure,more advanced control system and more difficult fault judgment.Because of the nonlinear mapping relationship between sensor detection quantity and fault feature,fault feature and maintenance strategy in fire control system,it is difficult to meet the requirements of fault diagnosis and prediction by using sensor data directly.In this paper,the state assessment and fault prediction of fire control system are studied based on machine learning.The signal data collected from the gun control box is analyzed and processed by the method of grey correlation degree,and the state characteristics are extracted.The processed data set of the gun control box is used as the input of the evaluation and prediction model to predict the state of the gun control box.This paper focuses on the important parameters of the impact assessment prediction model,studies the optimization methods of the prediction model by several machine learning algorithms.An evaluation and prediction model based on improved search strategy and gray wolf algorithm optimization is designed to improve the prediction performance of conventional models.Finally,experiments verify the superiority of this method for state evaluation and fault prediction of fire control system.The work in this paper is mainly as follows:Firstly,the required fire control system feature sample data set is constructed.The gun control box is selected as the main research object,and the input and output signals of the gun control box are increased by simulating various peripheral equipment signals of the fire control system on the test bench;The collected signals of gun control box are analyzed by using the method of grey correlation degree,the grey correlation degree of pin signals is calculated,and the correlation degree of signals is sorted.Screen out the pins with high correlation.Complete attribute reduction of the pin signal;According to the selected pin signals,the comprehensive evaluation index of gun control box operation state is summarized through expert experience,and the sample data set of gun control box is constructed.This step can reduce the input dimension of the evaluation prediction model,optimize the topology of the system,and improve the accuracy of the evaluation prediction model.Then,the evaluation and prediction model of gun control box is established.The grey wolf algorithm with improved search strategy is adopted.Adding a new learning strategy in the process of seeking the optimal solution improves the performance of the conventional algorithm,and the generalization ability of the model.The new learning strategy algorithm is used to optimize the important parameters of the evaluation prediction model to improve the ability of the evaluation prediction model.Finally,in the experimental stage,the sample data set of the gun control box is evaluated and predicted,and the data set of the gun control box is classified into training samples and prediction samples.Input the gun control box training samples and gun control box prediction samples into the evaluation and prediction model,and train the gun control box training sample data set.The input gun control box test samples are predicted in the established evaluation and prediction model,and the prediction results and accuracy are obtained.In this paper,four different models are used for state prediction,including the evaluation and prediction model optimized by the improved search strategy and gray wolf algorithm.Through the comparative analysis of four kinds of experiments,it is found that the improved search strategy gray wolf algorithm has the best optimization effect on the evaluation prediction model,and has less dependence on the number of sample training sets than the other three algorithms.The experiment shows that the evaluation and prediction model optimized by the improved search strategy gray wolf algorithm has the best performance in the evaluation and prediction of gun control box data set,and has strong popularization,applicability and engineering application value. |
