Abstract: Multiple procedures/sections, complex operation process, abrupt time/space change and severe working conditions feature are in the production of a coal mine. So, its safety system is typically a nonlinear one, and the safety estimation of a coal mine is inevitably of more and more importance nowadays in China. A safety estimation model is therefore developed via wavelet neural network, where the primary function of wavelet, which is based on retract and translation factors, is introduced instead of sigmoid transfer function, etc., with 23 indices picked out as input nodes. The model is able to make certain of network parameters automatically, thus avoiding the trouble in which the artificial intervention is needed to modify the structural parameters of network if using conventional neural network. It is exemplified that the mean absolute error of the estimation results via WNN network as proposed is 0.425%, while that via BP network is up to 3.1%. It means that the generalizability of WNN is much better than that of BP. So, the model is highly applicable.