A Modular Soft Sensor for Centrifugal Pumps

Abstract

Soft sensors experience an increasing interest in recent years, as they can replace expensive hardware meters and the required embedded computing hardware has become cheap and powerful. We report results for the implementation of a soft sensor for the flow rate estimation in centrifugal pumps that achieves root mean square errors of about 5%. The proposed soft sensor is based on generic models for the drive and hydraulic part of the pump to ensure widespread applicability. We show the soft sensor and the models it is based on can be parametrized with simple measurements. All theoretical considerations are corroborated with measurements on a real industrial pump in a laboratory setup. The results show that the proposed soft sensor is capable of providing reliable flow rate estimates in spite of plant model mismatch and uncertain hardware components.