Design And Implementation Of Redundant Signal Detection For Sealed Relay Based On Random Forest

With the rapid development of China's aerospace industry,the problem of loose particles in sealed relays has become a major problem limiting the safety performance of aerospace systems.Particle Impact Noise Detection(PIND)is an indispensable experiment for detecting loose particles after the completion of the production of aerospace sealed relays.Due to the traditional method of detecting loose particles is easy to misjudge component or loose particle signals,which poses a great security risk for the development of the aerospace industry.This thesis is based on this situation to carry out research,the main purpose is to build a set of efficient random forest classification model simulation system in the Python programming environment,improve the accuracy of loose particles recognition,solve the problem of low detection accuracy of loose particles.The thesis mainly classifies the data set obtained after the PIND experiment collection and processing,and completes the efficient identification of the loose particles through simulation experiments.In the face of the imbalance of the data set categories obtained by experiments,the LR-SMOTE algorithm is proposed based on the original SMOTE oversampling algorithm,so that the newly generated data is more representative of a few types.By measuring the classification accuracy of a single decision tree and the correlation between decision trees in the random forest algorithm,retaining high-quality decision trees to construct the final random forest model.Then combining the ant colony algorithm and the grid search algorithm,optimize the random forest intrinsic parameters 8)(6?(90)? and 8)(6?1)0)(60),so that the accuracy of the final model to identify the loose particles is the highest.Finally,the optimized random forest model is built and connected with the detection equipment,which is applied to the actual detection.The research in this study effectively improves the problem of misjudgment of loose particle signals,and improves the accuracy of loose particles recognition,which has important research significance for the safe development of China's military and aerospace fields.