Development and Implementation of a Constraint Detection Tool for Industrial Process Data

Abstract

Saturation occurs in the process industry for a variety of reasons. On the one hand, saturation can be caused by poor actuator dimensioning, external disturbances, and poor controller tuning. On the other hand, saturation can also be a deliberate way of plant operation. It becomes visible in the data captured from the process, in particular in the process variable and controller output signals. This thesis introduces a clear definition and distinction between saturation in the controller output and constraints in the actuator, sensor and setpoint. Two previously published detection methods for constraints and saturation in control loop data are discussed and a novel detection method is introduced. All three detection methods were tested on industrial process data where the newly developed method showed the highest robustness, a low computation time and high comprehensibility. With the newly developed method, a constraint and saturation detection tool was developed using Python as a programming language. The tool has a user-friendly interface, can process several hundred signals at once, and produces easy-to-interpret analysis results which makes it suitable for the use by engineers for control loop performance monitoring. The tool is publicly available on GitHub and PyPI. On PyPI, the tool can be found as a Python package with the name “seeq-constraintdetection” and can be installed as an add-on in the data analytics software Seeq.