A novel fault detection method for flight data based on wavelet transform algorithm

Abstract

With the rapid increase of the flight parameters, fault detection for the aviation data tends to be more complex and less efficient. Therefore, it becomes prevalent to compress the flight data for transmitting and detection, which could reduce the information storage and make the fault detection more efficient. However, there are some limitations with the conventional compression algorithms, such as complicated compression process and low reconstruction accuracy, which would fail to retain the characteristic of data. To overcome such boundaries, a novel fault detection method for the flight data in Flight Data Recorder was proposed in this paper, which utilized the wavelet transform matrix for the compression. With the property of sparse matrix, the wavelet transform matrix could be efficient for fault detection, as it could detect the abnormal area without reconstruction. Also, to increase the reconstruction accuracy, the adaptive threshold function was used in this method. To confirm the proposed approach, Yaw angle parameter was taken as the trial data. Preliminary results show that the proposed method could achieve a low relative error with the same compression ratio, also it could reduce the computation burden of the fault detection. If Daubechies order 4 Wavelet for four levels decomposition is applied in compressing the flight data, with the adaptive threshold function, the compression ratio could exceed 13, whereas the relative error is still below 1%. Furthermore, the method could locate the detection area accurately and quickly, with the detection area falling to 3.71% of the raw data, therefore the performance of fault detection would improve significantly.