Fault location and separation method of Distributed Inertial Measurement Units based on IAC

With the development of unmanned technology, Distributed Inertial Measurement Units (DIMU) play an increasingly important role in Unmanned Aerial Vehicles (UAV). The large number of sensors and high data redundancy of DIMU bring huge challenges to its fault location and separation. However, the fault calculation complexity of the conventional methods increases with the number of sensors. Existing intelligent fault diagnosis methods usually determine the fault type for a single Inertial Measurement Units (IMU), and there are relatively few studies on fault location for DIMU. In this paper, we propose a DIMU fault location and separation method based on Improved Attention-CNN (IAC). First, we extract temporal and spatial nonlinear features of all sensor-measured data by the IAC encoder. Then, we complete the location and separation of fault sensor in DIMU by decoding the extracted fault features. We can better extract the spatio-temporal correlation features of data between sensors in DIMU to improve the accuracy of fault location and separation by the attention mechanism in IAC. IAC is not limited by the number of sensors because it is a data-driven fault diagnosis method. IAC fills the research gap of DIMU fault location and fault separation. The IAC can be used to identify the location number and fault type of the faulty IMU in the DIMU. Through simulation experiments, the fault location accuracy rate of our proposed IAC method reaches 95% and the fault separation accuracy rate after a fault occurs reaches 99%.