Abstract: In the context of Industry 4.0, product quality prediction is critical for efficient operation and resource savings in intelligent manufacturing lines. However, existing methods often face reduced accuracy due to accumulated errors in process parameter predictions, leading to waste. Therefore, a residual correction-based method for product quality prediction is proposed. Firstly, the product quality prediction model is constructed based on the deep learning model, and the correlation model between process parameter prediction residual and product quality prediction residual is constructed. Secondly, the unprocessed process parameters are predicted to complete the information of production process parameters, and the product quality prediction results are obtained by using the quality prediction model. Thirdly, the residual of process parameter prediction is calculated, and the residual of product quality prediction is obtained by the correlation model. Finally, the residual of product quality prediction is used to modify the preliminary prediction result, so as to obtain the final product quality prediction result. Case analysis shows that this method can effectively reduce the error of product quality prediction and improve the accuracy of prediction. It can find the potential unqualified products on the intelligent manufacturing production line in time and stop the subsequent production process, so as to achieve the goal of resource saving.