Underground works relying on the geological conditions of the rock types (wall rock) is directly related to project precise mouth site, project main building outline, means of construction and engineering span, and affect the protective performance evaluation on the project. In order to accurately assess the stability of surrounding rock, based on the BP neural network model, optimize the operations engineering in front of face surrounding rock stability assessment model. By adding the momentum factor, use these two methods of adaptive adjustment of the transfer function to be optimized to improve the neural network convergence speed. For some non-linear, multi-model, multi-objective function, optimization problem is inherently implicit parallelism and better ability of global optimization. Use a probability-based optimization method that can automatically access and guidance to optimize the search space and adaptively adjust the search direction do not need to determine the rules. Optimize the initial weights and thresholds of BP neural network, compare them with the characteristics of the surrounding rock by the force, it can make accurate rapid assessment of the stability of surrounding rock, save human resources, and improve work efficiency.