The coal mine safety evaluation involves many uncertain information. It is difficult to ensure the accuracy and reliability of finall consequence using traditional evaluation method. Therefore it uses RBF neural net to design the coal mine safety evaluation model, according to the practical situation of our country’s coal mine. Through the accident tree analysis(FTA)and event tree analysis(ETA), this paper summarized coal mine accident risk factors and influenced of mine production safety factors. Meanwhile in order to overcome the neural net easy to fall into the local minimum, neural net model of the right value (threshold) is optimized using quantum genetic algorithm. The method is used in fuxin mining company of a subordinate mine. Results show that the model can accurately evaluate coal mine safety in production, for coal mine safety evaluation provided a new way, having important significance for coal mine safety in production.