Reality-driven simulation for estimating unmodeled hydraulic cylinder friction

Abstract

In recent years, the simulation of hydraulically actuated cylinder has become essential for modern engineering design. However, nonlinearity and complexity of friction in hydraulic cylinders make it a complicated task. On the other hand, traditional friction models that are based on physics are computationally expensive and difficult to parameterize accurately which leads to limit their effectiveness in real-time applications. Estimating the friction force in a hydraulic cylinder is one of the solutions to address these challenges without relying on an explicit friction model by using a reality-driven simulation framework. In current study, an indirect extended Kalman filter (eEKF) is employed for the mentioned approach, where the friction force is treated as a state variable to be estimated. The estimator uses two simulation which one of them works as a ground-truth model and the other as estimation model. A single degree of freedom hydraulic actuator is used to investigate the method. The ground-truth model uses the Brown-McPhee friction model to act as realistic dynamic behavior, while the estimation model operates with known parametric errors. The results demonstrate that the eEKF can accurately tracks the true system states including piston position, velocity and chamber pressures. Although minor estimation lags were observed during rapid zero-velocity crossings, the filter exhibited robust convergence and stability.