A Deep-Learning Based Fault Diagnosis Method for Electrical Motor of Electro-Hydrostatic Actuator

An electro-hydrostatic actuator (EHA) is becoming a key component in aircraft flight control. During flight, the reliability of the electrical motor is fundamental in energy transformation. Recently, focuses are given to the fault diagnosis method for working state classification of an EHA. In this work, efforts are made in establishing a fault diagnosis model of motor in working EHA. A deep-learning based approach using deep convolution neural network and the attention network is devised. The proposed model has the ability in extracting both the local and global features from the given inputs. Experiments are performed on a dataset of various working states. The proposed model outperforms the baseline in the classification accuracy. Our method accomplishes an average accuracy of 84.8%, which sets a strong evidence of a satisfying working performance.