A novel data-driven model based parameter estimation of nonlinear systems

In practice, it is usually difficult to estimate the characteristic parameters of a system due to the system nonlinearity and uncertainty. To address this problem, in this study, the characteristic parameters of a nonlinear system are identified by using a data driven method based on the best discretization of the Nonlinear Differential Equation (NDE) model of the system. The best discretization of a NDE model is firstly determined, and then the discretized model, known as the Nonlinear Auto Regressive with eXegenous input (NARX) model, is determined by using the Least Squares (LS) algorithm from the input and output data of system. A case study is discussed to validate the proposed system parameter identification method, where the characteristic parameters of a rotating blade-casing system are evaluated under a bandwidth rub impact in the horizontal direction with noise. The result shows that the identified model can be used to describe the characteristics of the underlying system accurately, which provides a reliable model for the dynamic analysis, control of rotating blade-casing system. (C) 2019 Elsevier Ltd. All rights reserved.