Expected product quality is affected by multi-parameter in complex manufacturing processes. Product quality prediction can offer the possibility of designing better system parameters at the early production stage. Many existing approaches fail at providing favorable results duo to shallow architecture in prediction model that can not learn multi-parameter's features insufficiently. To address this issue, a deep neural network (DNN), consisting of a deep belief network (DBN) in the bottom and a regression layer on the top, is proposed in this paper. The DBN uses a greedy algorithm for unsupervised feature learning. It could learn effective features for manufacturing quality prediction in an unsupervised pattern which has been proven to be effective for many fields. Then the learned features are inputted into the regression tool, and the quality predictions are obtained. One type of manufacturing system with multi-parameter is investigated by the proposed DNN model. The experiments show that the DNN has good performance of the deep architecture, and overwhelms the peer shallow models. It is recommended from this study that the deep learning technique is more promising in manufacturing quality prediction.