Testing and implementation of a backlash detection algorithm

Abstract: Backlash is a well documented and rather common non-linearity which is present in most mechanical and hydraulic systems. The amount of backlash varies depending on the dynamics of the system, for example a gear box in a car needs some space for heat expansion and consequently some amount of backlash in order to function. It is of interest to detect and estimate the amount of backlash automatically, since the time between the manual inspections often is long. This thesis covers the method described in 'Automatic on-line estimation of backlash in control loops' written by Tore Hägglund and published in the Journal of Process Control in January 2007. One of the main objectives of this master thesis is to extend the safety net to guarantee the robustness of the backlash estimation procedure described in the article. The purpose of the safety net is to make sure that backlash estimates are only calculated when the system behavior resembles backlash. Conditions for period time, curve form and setpoint are discussed and investigated. The backlash estimation procedure is implemented as a control module in a prototype library in ABB Control Builder, a software used for PLC programming in an ABB 800xA environment. Finally, industrial field tests are performed to confirm the robustness of the method.