Tuning PID Controllers: From process experiments, general linear state space models, and tuning controllers via Process Reaction Curve Methods

Abstract

In today’s industrial landscape, the necessity and usefulness of computer systems are undeniable. And with the need for computer systems and computer controller machinery, the interest in optimizing the efficiency of the computer system itself is highly emphasized. This project is centered around testing and comparing the more popular forms of control, mainly process reaction curve methods. The objective of this thesis to present the reader with general information about the control process as well as presenting the theory behind processes. In addition, these forms of control should be tested on the random state-space models in order to view each controllers results and compare the controllers against each other.

This testing was done through the usage of MATLAB using seven different controllers for comparison. These controllers included the Ziegler-Nichols open-loop method, “megatunerplain”, “megatuner”, and two versions of both the “pidstd”- and “pidtune” methods. “megatuner” and “megatunerplain” being created by Christer Dalen.

With varying results from the different controllers, the Ziegler-Nichols open-loop method and the PID-version of the “pidstd” showed unstable and poor results. The PI -version of “pidstd” and “megatunerplain” showed overall stable results with “megatunerplain” being slightly better overall. “Pidtune” and “megatuner” returned high stable performance across the different sampling times and system orders. “Megatuner” was the only method to not experience a crash/unstable control across the project and its testing.