The relationship between the process parameters and bonding quality is very complicated due to the features of non-linear and coupling relation. It’s therefore hard to express their relationship in an exact mathematic model and very difficult to obtain an reasonable parameters setting to get high bonding quality. In this paper, a wire bonding process model based on BP(Back Propagation) neural network is proposed, and based on this model a computer-aided process design system for bonding quality improvement is developed through VC++6.0. Based on the training data from design of experiments (DoE), the BP network is built. The system is validated by test data and the testing result shows that the predicted error is samll and reasonable and the system developed can be used for parameter optimization and quality improvement.