An analysis of Model Predictive Control with Integral Action applied to Digital Displacement Cylinders

Abstract

This article aims to analyze Model Predictive Control (MPC) for the control of multi-chamber cylinders. MPC with and without integral action has been introduced. Three different algorithms have been used to solve the optimization problem in the MPC. The different algorithms have been compared with an industrial solver. The influence of changing mass, choosing a different middle line pressure, system delays, signal noise, velocity estimation, and changing pressure levels has been investigated. It is concluded that for the small prediction horizon used in the paper a simple algorithm such as A* can produce results as good as the previously used Differential Evolution algorithm in less than half the time. It is further concluded that unknown software delays and unknown changes in mass have the largest effect on system performance.