Most current common cause failure models are static models in which failure rate is assumed constant. The assumption is unrealistic in numerous situations, especially for mechanical systems. Based on the idea of shock models, this paper modifies unreasonable assumptions in conventional models and describes common cause failure process with non-homogeneous Poisson process, power-exponential process. A dynamic model is obtained. The parameters are estimated with neural network. The model presented in this paper, which is based on time-related power-exponential failure rate like bathtub curve, has more advantages and wider application than the model based on constant failure rate. Examples are given to illustrate its feasibility and precision of computation.