A COMPUTATIONAL INVERSE TECHNIQUE FOR RECONSTRUCTION OF MULTISOURCE LOADS IN TIME DOMAIN
韩旭,刘杰,李伟杰,赵卓群
DOI: https://doi.org/10.3321/j.issn:0459-1879.2009.04.017
2009-01-01
Abstract:The knowledge of the dynamic load acting on the structure is always required and important in many practical engineering problems, such as structural strength analysis, health monitoring and fault diagnosis, and vibration isolation. However, it is difficult to directly measure the dynamic load on a structure in some situations, such as the wind load on the tall building, the exciting force from road on the vehicle, etc. Meanwhile, the dynamic response measurement is correspondingly easy and accurate on a structure. Therefore, it is necessary to develop some inverse analysis techniques for load identification based on the measured dynamic responses.With the linearity and time-invariant suppositions, the loads are firstly expressed as a series of kernels of impulse functions or step functions in time domain and the total response of the system can be obtained using the product of the convolution integral of the kernel response and the loads. Through the discretization of convolution integral, the forward model for load identification is established. In fact, the inverse analysis for the load identification is to solve a deconvolution problem, but the deconvolution is an ill-conditioned problem in which the noisy responses and high condition numbers of the kernel matrix will induce the amplified errors in the identified load. Therefore, it is difficult to obtain a stable and accurate solution for such inverse problems. To deal with ill-condition of load reconstruction from the noisy responses, zero-phase digital filter, several regularization methods and optimized strategy for stable load identification are discussed. Through general filter, the noisy response signal will be smooth. But, it has a phase delaying compared with the original signal, and the errors will also be amplified in the identified load. The zero-phase digital filter, whose phase error is zero in the curve of phase-frequency characteristic, is realized through reversing the time serials of the signal. Moreover, a new extension algorithm is applied to improve the performance of the filter. Comparing with the common difference filter, this zero phase digital filter can not only avoid phase delaying, but also improve the wave aberration of the start and end section. After the investigation the ill-posedness arising from the inverse problem of load reconstruction, Tikhonov regularization, truncated singular value decomposition and total least squares method are adopted to provide efficient and numerically stable solution of the desired unknown load, and the L-curve method is proposed to determine the optimal regularization parameter. In order to avoid the inverse operation of the matrix, many optimized methods can be available and here the conjugate gradient method is adopted. In the numerical example, the reconstruction of dynamic loads from two sources with the noisy responses in the hood structure is investigated. The result indicates that the presented computational inverse technique is effective and stable for the load identification with the noisy response in time domain.