Ammunition Coordination Fault Diagnosis Method Based on Wavelet Packet and CPO-LSSVM

Aiming at the problems of difficulty in extracting fault features of ammunition coordinator and ow accuracy of fault classification using traditional machine learning methods, the author of the article proposes a fault diagnosis method for ammunition coordinator based on Crested Porcupine Optimizer (CPO) optimized least squares support vector machines (LSSVM). Firstly, a total of 200 sets of experimental acceleration responses of the ammunition coordinator under five scenarios are taken as samples. The wavelet packet node energy value is obtained by wavelet packet decomposition, and subsequently used as the eigenvalue input to the CPO-LSSVM fault diagnosis model for training. The experimental results show that the fault recognition rate of the fault diagnosis model on the test set is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$100 \%$</tex>; Then, BP neural network, BNN neural network, extreme learning machine and random forest regression prediction model are used for fault identification experiments respectively. The recognition accuracy of the four machine learning classification models on the test set is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 7 . 4 1 9} \%, \mathbf{9 4 . 0 1 7 \% , 9 0 . 5 9 8 \%}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 7 . 8 2 2 \%}$</tex> respectively. The results of pattern recognition show that the CPO-LSSVM fault diagnosis model has higher fault diagnosis rate than the above four, which verifies the accuracy and effectiveness of the proposed method.