Machine Learning-Based Modeling of Valve Solenoids

Abstract

Abstract Software-based functionalities including virtual sensors and model-based control are becoming increasingly important in fluid power. Such functionalities require computationally efficient and accurate valve models. This article investigates data-driven modelling of proportional solenoids to address one of the key challenges in hydraulic valve modelling. First, the valve solenoid and the measurement setup used are described, followed by an investigation to identify static and dynamic system characteristics. The analysis reveals the complex nonlinear behavior of the solenoid, which is difficult to represent using traditional models. A modelling approach is presented. Generated test signals are used to record data to train, validate and test the model. The evaluation of the developed model shows a good representation of both static and dynamic states. Overall, the described data-driven modelling concept offers a promising approach to meet the growing demand for computationally efficient and accurate valve models.

Authors 2

1. Simon Hucko Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University , ,

2. Katharina Schmitz Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University , ,