Abstract: Due to such problems as local minima，slow convergence，and dependence on initialized values arising by the back propagation (BP) neural network algorithms, it is the first time to adopt this new algorithm called the spline weight function artificial neural network.Based on the analysis on effects of stability in fracture surrounding rocks, the neural network model is adopted to select an optimal parameters in forepoling bolt for the fractures.The study shows that the supporting parameters with 0.4~0.6 m in space, 0.4~0.5 m in rank, 10°~20° at angle and 20~22 mm in diameter are optimal for forepoling bolt.Based on this, how to select the parameter for forepoling bolt and how to improve the stability of rock body can be investigated, then providing a theoretical basis for safety construction and supporting strengthening in local place for the fracture rocks.

Key words: Spline weight Function Artificial Neural Network, Forepoling Bolt, Surrounding rock in engineering, Supporting Parameters