Identification and modeling of 3-DoF mechanism with nonlinear friction effects by combination of LuGre and neural network modeling
Karel Kraus, Petr Kočandrle, Zbyněk Šika
Abstract
Many mechanical systems suffer from strong passive effects due to presence of various imperfect kinematic joints and bearings. Such effects are more crucial for free oscillation systems or vibration absorbers since their ideal design should be as friction free as possible to work properly. To be able to deal with such inconveniences, it is first necessary to describe passive effects acting to moving bodies of the mechanism. However, such forces are usually far from linear function of velocity and incorporates hysteresis, backlash, stick–slip or nonlinear functions of velocity as well as of position. Therefore, state-full friction model is needed. This paper deals with passive effects identification present in multi degrees of freedom absorber demonstrator’s kinematic joints. Such absorbers are suitable for absorbing spatially complex unwanted vibrations of robots and other complex mechanisms. For such identification purpose, the well known LuGre dynamical model of dissipative forces is firstly tested. Finally the absorber mechanism model with optimized LuGre models is extended by trained neural network improving the whole mechanism modeling in order to reach sufficient simulation corresponding to measured behavior of absorber.