Research on Fault Detection Method of Aircraft Fuel System Based on Transfer Learning
In order to improve the accuracy and reliability of fuel system fault detection,a random forest-based transfer learning method(TRLM)is proposed in this paper.According to the existing data of source domain and target domain,the most suitable classification model is established respec-tively.Taking the public labels of source domain and target domain as the migration hub,the contri-bution matrix of source feature and target feature to the public label is established based on random forest,and the feature mapping from source domain to target domain is constructed.Finally,the Mul-ticlassTrAdaBoost was used to determine the weights of the source domain samples and the target do-main samples after mapping and the final classification was completed.In order to verify the effec-tiveness of this method,a real aircraft fuel system data set was used for experimental evaluation.The experimental results show that the transfer learning method based on random forest achieves excellent performance in the fault detection of aircraft fuel system.Compared with other transfer learning methods,TRLM has achieved significant improvements in both accuracy and reliability.
万方数据
维普
