Adaptive Embedded Learning-Control Architectures for Reconfigurable Sensorless Motor Drive Platforms

Abstract

Sensorless motor drive systems are becoming very popular in embedded and industrial application because of cheap hardware cost, higher reliability and small implementation. Nevertheless, the operation of the traditional sensorless control methods is usually constrained by the parameter variations, nonlinear motor dynamics and external disturbance conditions especially during low speed and variable load operation conditions. To overcome these issues, this paper will suggest an adaptive embedded learning-control architecture based on the reconfigurable hardware platforms to operate sensorless motor drives. The suggested methodology consolidates an online learning mechanism, real-time control and sensorless approximation loops in a single, hardware-aware system. A modular architecture is built on an embedded system using an FPGA that is able to execute control, estimation, and learning at the same time with deterministic timing limits. Hardware reconfigurability is used to dynamically allocate computational resources and run time states, to improve system scalability and real time performance. An embedded learning module is able to keep updating the parameters of the observer and the controller according to real time error feedback such that the system may be able to adjust to modeling uncertainties as well as changes in the operating conditions without requiring offline training. The results of experimental validation based on the proposed approach under different speed and load conditions prove that the proposed solution has a better estimation accuracy, quicker dynamic response, and greater robustness with respect to the conventional fixed-parameter sensorless controllers. The findings endorse the fact that embedded learning coupled with reconfigurable hardware integration is a viable and effective solution to intelligent motor drive systems. The literature provides a scalable platform upon which future generation adaptive motor control systems can be built in resource-granted embedded systems.