Abstract: The geostress field is the basic data for the excavation and support design of deep mine roadways. Aiming at the large random errors of geostress measurements in the deep strata of the coal mine, a geostress field inversion method based on support vector regression(SVR) optimization algorithm is proposed. The optimization algorithm takes the feature vector structure risk minimization as the principle, making the empirical risk and confidence range minimized to reduce the interference of random errors of measured geostress on parameter inversion. A three-dimensional finite element calculation model is established based on the geological and geostress data of Pingmei No.1 mine. The lateral pressure coefficient is introduced and the uniform design table test method is used to construct the stress boundary condition. The learning samples are created based on the calculation results. Then the support vector regression algorithm is used to optimize the boundary condition parameters, and the optimal stress boundary condition is determined. The optimal stress boundary condition is applied to the model to obtain the geostress field of the whole model. The comparisons between the inversion value and the geostress measurement point show that the errors are within a reasonable range, and the inversion values are less random than the measured values. The distribution of geostress field is consistent with that of inversion results in Pingdingshan mining area, revealing that the inversion results are accurate and reliable.

Key words: geostress field, inversion, support vector regression(SVR), optimization algorithm, finite element