Analysis of 3D In-situ Stress Field and Query System's Development Based on Visual BP Neural Network

In-situ stress field is very important in the numerical simulation and stability analysis as well as in the engineering design and construction. So it requests an effective analysis method. The typical in-situ stress analysis method is multivariate regression, which is based on 3-D FEM direct modeling, but it may be not so accurate to obtain the distribution of in-situ stress field of the project, because when the multivariate regression is carried out, the cross influence of the main factors of in-situ stress field is ignored, and the relationship of in-situ stress field and the main factors is too complicated to express by subharmonic multivariate function. This paper presents an improved method back propagation (BP) neural network back analysis of in-situ stress field, based on the calculated results of multivariate linear regression analysis. For in the ANN (Artificial Neural Networks) method, the training samples are difficult to generate, while combining the linear regression method with ANN method this problem can be solved perfectly. Linear elastic multivariate regression is carried out to determine the general bounds of optimized paraments, and uniform design method is carried out to settle different combinations of factor levels. Training samples are gained by FEM analysis. And the visual neural network by using ActiveX technology is realized in this paper, by taking full advantages of strong capability in computing of Matlab and features of friendly interface of VB perfectly. In fact, one needs to know the status of the in-situ stress field at some important location in the engineering, by using neural network method, intelligent expression method of the in-situ stress field is presented, and the developed man-computer interface has the property of easy use in the actual engineering.