Development and Validation of Multi-model Fusion Fault Diagnosis Model for Air-conditioning Packs of Civil Aircraft Environmental Control System
Aiming at the difficulty to obtain fault data and the unbalanced fault data composition of the air-conditioning packs in the civil aircraft environmental control system,the method of dynamic simulation to study the occurrence of faults,and obtains the fault data set by manually injecting fault conditions is adopted in this paper.The white noise in the original timing signals of the air-conditioning packs is effectively suppressed using sliding average and empirical modal decomposition,which reduces the risk of misjudging normal signals as abnormal.A fault diagnosis model integrating isolated forest,spectral residual,and autoencoder is developed,and the accuracy of the combined model is analyzed for different combinations of algorithms.The results show that,the developed model can effectively diagnose air-conditioning packs faults in the dynamic simulation dataset with a diagnostic precision of 100%;for the open dataset of mechanical faults,the trained model combined with the developed signal processing method can achieve a diagnostic accuracy of 91.12%,and the recall rate reaches 98.66%,which verifies the effectiveness and ubiquity of the solution.
万方数据
维普
