Automatic Model-based PID Tuning of a Servo Axis

This thesis considers the automatic parameter tuning of a servo axis. The target servo axis of this thesis is controlled with an industrial servo drive that utilizes PID controllers in its internal structure. In particular, the focus of this study is on the functional behavior and performance of common controller auto-tuning methods when used in conjunction with the OptoFidelity OptoDrive servo drive. The purpose of this thesis is to address the state of robot setup methodology where a trained professional is responsible for the manual tuning of each servo axis, and to build a basis for PID auto-tuning with the OptoDrive. Manual tuning takes a considerable amount of time and has proven to be problematic in large scale applications, such as mass-production environments. The research effort on auto-tuning methods in this thesis is motivated by the desire to eliminate the need of manual controller tuning altogether. The initial results suggest that automatic controller parameter tuning is feasible for the OptoDrive’s axis velocity and position controllers. Four controller auto-tuning methods were tested, from which setpoint overshoot, closed-loop SIMC, and Ziegler-Nichols methods were found to be suitable for tuning the velocity controller of the OptoDrive. The Ziegler-Nichols was found to be the only method that is suitable for tuning the position controller of the OptoDrive. Therefore, the Ziegler-Nichols autotuning method was studied further, and effort was put into automatically optimizing the mathematical formulae utilized by the method to achieve the best possible auto-tuning results. After optimization, it produced the best performance results recorded in the experiments of this thesis. The performance surpassed the average tuning performance of humans by significant margins, and even more importantly, with a significantly better standard deviation of performance. The tests were conducted on a single linear servo axis, whose mass was varied in three configurations to emulate three systems with different inertia. The results obtained in this thesis suggest that automatic tuning should be implemented when the best possible robot performance is desired with the OptoDrive. If the desire is to use an other servo drive than the OptoDrive, the performance tests should be rerun to be valid. As for the impact of this thesis, a patenting process on the aforementioned optimization method is ongoing, and OptoFidelity is moving to use PID auto-tuning with the OptoDrive