Operation Mode Recognition With Multiple Sensing Data for Electro-Mechanical Actuator Based on Deep-Shallow Fusion Network

In next-generation aircraft, electro-mechanical actuators (EMAs) are increasingly used. But the safety of EMA is not sufficient for the primary flight control actuation of aircraft. One effective way to improve EMA safety is to develop EMA prognostics and health management (PHM). However, variable operation modes make it difficult to implement high-performance EMA PHM. Thus, EMA operation modes need to be recognized, but the high similarity of sensing data between different operation modes makes it challenging. Thus, a new deep-shallow fusion network with convolutional neural network, self-attention mechanism, and Bayesian network (CSBN) is proposed for operation mode recognition, which can overcome the challenge of high similarity between EMA multiple sensing data. In the proposed CSBN-based recognition method, the statistical features of EMA multiple sensing data are first extracted and discretized. Then, the recognition is conducted with the discretized statistical features based on CSBN. Finally, the output of CSBN is used as the recognition results. To validate its effectiveness, experiments utilizing the practical data of EMA are implemented. Experimental results demonstrate that CSBN is suitable for EMA operation mode recognition.