In order to address the problem of quality control faced in multi-type and small-batch manufacturing mode, the method based on Bayesian Network (BN) is proposed. The building, learning and evolving method as well as the quality prediction and diagnosis method of BN model are described in this paper. The combination of BN model and Shewhart control chart is also mentioned. The model building and evolving method was conducted in PCB micro-drilling process as example, verifying that the prediction accuracy increases with the evolved model. The drilling quality prediction was compared with that obtained through regression analysis and artificial neural network. The advantage of BN model in advanced manufacturing is proved.