Application of Transfer Learning in Rudder Fault Diagnosis Under Variable Operating Conditions
In order to solve the low fault diagnosis accuracy of aircraft rudders under complex and dynami-cally changing operating conditions,an NDTL-CNN fault diagnosis method that combined convolutional neural network(CNN)with network-based deep transfer learning(NDTL)is proposed.Firstly,a simula-tion model of aircraft rudder fault is established to collect multi-dimensional sensor data under different op-erating conditions and health states;Then,a CNN is designed to adaptively extract deep features from the fixed operating conditions data,which can effectively capture the fault feature signals of the rudder;Final-ly,the pre-trained CNN under the fixed operating conditions is fine-tuned and transferred to variable operat-ing conditions for fault diagnosis.The experimental results show that the proposed method improves the di-agnostic accuracy of the CNN under variable operating conditions by 15％in a short time and the final di-agnostic accuracy of the NDTL-CNN reaches 97.7％,which is capable of accurately recognizing the rudder's health state under complex and dynamically changing operating conditions.
Aircraft rudderFault diagnosisConvolutional neural networkDeep transfer learning
NETL
NSTL
万方数据
