Force prediction can basically be done by two methods: direct methods and optimization methods. Direct methods use the inverse of the forward system model to calculate the excitation directly from the measured responses. Optimization methods use a forward model in an optimization loop wherein the input to the forward model is adjusted until the model responses matches the measured responses. In practice, a direct method using an experimentally obtained Frequency Response Function (FRF) is generally used. The direct method can be applied iteratively to enable convergence towards an excitation signal when dealing with nonlinear systems. Previous research of the authors, applying such a code to a highly nonlinear multibody quarter car model, showed an acceptable match between the calculated and the original excitation. However, the iterative process takes many steps and needs user interaction to reach overall convergence, like the manual exclusion from the time signal of excitation peaks that cause divergence. The test case is a representative benchmark for real life problems. This paper focuses on the improvement of speed and robustness of force prediction methods when dealing with nonlinear systems. Contrary to commercial codes, where the system model is treated as a black-box, we use a-priori knowledge of the system dynamics obtained from parametric modeling.We set out from the direct method using the inverse FRF. A simple demonstrator has been built consisting of a beam, clamped at one side and the other side subjected to different end conditions: free and supported by a repulsing magnet. The demonstrator has also been modeled in a multibody code supporting flexible bodies, to enable preliminary research and to compare experiments and simulations. This paper is restricted to the reconstruction of harmonic forces acting at known locations with different amplitudes and frequencies.
展开▼
