Real-Time Adaptive Data Compression for Permanent Magnet Motor Control System

Permanent magnet synchronous motor (PMSM) control system processes massive electrical and mechanical data sets which are important for various data-driven methods of big data analysis or failure data recording. However, as the control frequency going up, motor data processing faces more challenges of the limited computational resource of a motor controller. This paper proposes a real-time data compression method based on connected piecewise linear regression (CPLR) and compressed sensing (CS), so the motor operational data can be recorded with a minimum memory storage space. The study demonstrates the parameter boundary of CPLR under the condition that samples obtained by digital sensors. On this basis, adaptive-CPLR is proposed, which tunes parameter automatically with the given active function and has parameter reset mechanism. Further, CS is implemented using sparsity binary matrix (SBM) for space and time saving and triggered by adaptive-CPLR aimed to preserve frequency domain information conditionally. Experiment results is presented to demonstrate the real-time characteristic of the proposed compression method in resource limited motor controller STM32F407. Adaptive-CPLR and CS are also evaluated, showing that the proposed method successfully achieves desired performance of high compression ratio and low percentage root mean square residues.