Abstract
The distribution of the sum of the principal stresses was obtained by the infrared stress analyzing system. The unknown boundary-values were estimated from the internal point information of an infrared stress image by a boundary-value inverse problem analysis, and the stress separation was carried out by a direct problem analysis using estimated boundary values. The truncated singular value decomposition (TSVD) method and the Tikhonov regularization were used as the optimization technique of boundary-value inverse problem analysis. As an optimization parameter determining method, the norm minimizing method was employed. Then, the stress separation was performed by the boundary-value inverse problem analyses based on error evaluation, the least squares method, and the 2-dimensional hybrid method proposed by Nishioka et al. Comparing the results obtained by three different methods to those of the 3-dimensional finite element analysis, each error became about 4%, 5%, and 5% by TSVD, the Tikhonov regularization, and the 2-dimensional hybrid method, respectively. The result of TSVD was the most accurate. However, it is considered that the 2-dimensional hybrid method is effective as the stress separation method from the viewpoint of the accuracy and calculation time.
