Segmented Adaptive Singular Value Decomposition for Data Compression of IGBT
Jin Qian, Dong Liu · 2022 IEEE 11th Data Driven Control and Learning Systems Conference (DDCLS) · 2022
As Insulated Gate Bipolar Transistor (IGBT) is widely used in industrial practice, a compression algorithm based on two-dimensional block adaptive threshold singular value decomposition is proposed to solve the problem of storing massive data when IGBT is switched on. In order to more effectively improve the compression ratio and reduce the error of data, this method is to arrange the set of one-dimensional IGBT data into a two-dimensional matrix normalizing to form two-dimensional gray image processing according to each part of the image characteristic of regional block. After adjustment by adaptively presetting threshold characteristic values of singular values, we use compression sensing to complete data compression. Multiple sets of original data can not only ensure the mean square error, but also meet the requirements of high compression ratio, high precision and low distortion, which significantly reduces the amount of data stored. The experimental results show that when the mean square error is about 0.05% and the mean absolute error is about 0.015, the compression ratio of 3% and the signal-to-noise ratio of 41.5dB can be achieved, which verifies the feasibility of the algorithm and has certain practical value.