Abstract: It studied the structure of BP neural network and the steps of L-M learning algorithm. Through the analysis of coal mine ventilation system, the index system of coal mine safety evaluation which was controlled by many factors was established. On this basis, it compiled BP network program with Matlab, and trained the network by mentor signal, thus it set up safety evaluation model of coal mine ventilation system. The simulation results showed that the safety level of the sample was in accordance with the actual situation, the convergence rate of L-M algorithm could meet the requirements.