Abstract
In order to reduce machine vibration, it is useful to estimate the excitation force as well as grasp the vibration characteristics of machines. In this study, we investigate the method of estimating the excitation force in the time domain. Tikhonov regularization and the L-curve method are employed to solve the inverse problem for the input-output relation expressed in the convolutional form and the excitation forces are estimated in time domain. When the excitation forces cannot be measured directly, the excitation force is estimated from the acceleration response of other places. Generally the excitation points and the response measurement points are different positions. However, the estimation accuracy is significantly degraded when the measurement point is near the excitation point. In this study, we prove that this reason is the effect of the direct term from the excitation force to the acceleration. The estimation results of the input-output relation reconstructed in the convolutional form are compared with ones obtained by deriving the state equation from the realization problem. We investigate the influence of the direct term on the estimation results. As an example, we perform an excitation force estimation experiment on a five-layer structure and a cantilever beam, and investigate the influence of the direct term when the measurement points of the excitation point and the acceleration response are near each other. It is found that the degradation of the estimation accuracy is small even if the response points and the excitation points are close when the direct term is considered.
