Small-sample State Recognition of Ammunition Feeding Mechanism Based on Stacking Ensemble Learning
Aiming at the problem of small sample size and uneven sample size of the motion data of the on-board recorder for testing the ammunition feeding mechanism of artillery,a Stacking ensemble learning based small sample state recognition method for ammunition feeding mechanisms is proposed.The original data are first preprocessed by wavelet filtering and standard deviation normalization,and the improved drosophila optimization grey neural network and the long short-term memory neural network(LSTM)are regarded as the primary learners,and linear regression(LR)is used as secondary learners to form the Stacking ensemble learning model.Based on the real data obtained from the onboard recorder,an experiment is conducted to identify the abnormal state of the limiter in the ammunition supply mecha-nism.The research results show that in a small sample environment,the integrated learning model has higher accuracy and stability than that of a single learning model,and can more effectively identify the abnormal state of the ammunition supply mechanism.
万方数据
维普
