On the sensitivity of the equivalent dynamic stiffness mapping technique to measurement noise and modelling error

Abstract

The objective of this study is to investigate the sensitivity of the Equivalent Dynamic Stiffness Mapping (EDSM) identification method to typical types of inaccuracy that are often present during the identification process. These sources of inaccuracy may include the presence of noise in the simulated/measured data, expansion error in the estimation of unmeasured coordinates, modelling error in the updated underlying linear model, and the error due to neglecting the higher harmonics in the nonlinear response of the system. An analytical study is performed to identify the structural nonlinearities of two nonlinear systems, a discrete three-DOF Duffing system and a cantilever beam with a nonlinear restoring force applied to the tip of the beam, considering the presence of all the aforementioned sources of inaccuracy. First, the EDSM technique is utilized to identify the nonlinear elements of two example systems to verify the accuracy of the EDSM technique. Finite Element modelling, the Modified Complex Averaging Technique (MCXA), and arc-length continuation are exploited in this study to obtain the steady state dynamics of the nonlinear systems. Numerical models of the two systems are then simulated in MATLAB and the numerical results of the simulation are used to identify the unknown nonlinear elements using the EDSM technique and investigate the effect of different sources of error on the outcome of the identification process. The nonlinear response of the system has been regenerated using the identified parameters with the sources of error present and the generated response has been compared to the simulated response in the absence of any noise or error. The EDSM technique is capable of identifying accurately the nonlinear elements in the absence of any source of inaccuracy although, based on the results, this method is highly sensitive to the aforementioned sources of inaccuracy that results in significant error in the identified model of the nonlinear system. Finally, an optimization-based framework, developed by the authors, is utilized to identify the nonlinear cantilever beam and the results are compared with the results of the EDSM technique. It is shown that by using the optimization method, the inaccuracy due to different sources of noise and error is significantly reduced. Indeed, by using the optimization method, the necessity to use an expansion method and consider the higher harmonics of the response is eliminated.