Online identification of a mechatronic system with structured recurrent neural networks

In this paper, we present an online identification method for mechatronic systems consisting of a linear part with unknown parameters and a nonlinear system part with unknown static nonlinear characteristics (systems with isolated nonlinearities). A structured recurrent neural network is used to identify the unknown parameters of the known signal flow chart. In this paper we present the successful identification of a typical motion control environment consisting of a driving machine connected by an elastic shaft to the load. The presented identification algorithm uses only the speed of the driving machine for parameter adaption. Besides the detailed steps to develop the structured recurrent network, we present simulation results as well as measurement results. The identified linear parameters are the inertias of the driving machine and the load, the spring and damping constant of the elastic shaft. Identification results for the nonlinear friction characteristics are also derived. The novelty of this approach is the simultaneous identification of the parameters of the linear part and the nonlineatity. Due to the use of this approach physical interpretation of the identification results is possible. It is possible to use the identification results in order to optimize nonlinear observers and state space controllers.