| Sustaining healthy development is an enduring objective for enterprises,making product quality a critical factor.Quality control during the product manufacturing process is an important means of ensuring product quality.The field of intelligent manufacturing has put forward higher requirements for the real-time and accuracy of product quality information.The development of advanced information technologies such as industrial big data and artificial intelligence has provided new means for real-time quality prediction and diagnosis of multi-process products,becoming the current research hotspot.This thesis studies the data-driven real-time quality prediction and diagnosis of multi-process products using deep learning and data mining technologies,which is of great theoretical significance and practical value in enhancing the level of product quality control.A framework for quality control of multi-process products that considers data integration management and real-time quality control was established.The manufacturing quality formation process of multi-process products was analyzed,and a distributed Internet of Things architecture was adopted.The big data technology was integrated into the platform layer,the edge computing framework was introduced,and a data-driven framework for quality control of multi-process products was constructed.Critical technologies,including edge data processing,quality prediction and diagnosis,and quality control visualization,were analyzed and studied.A data-driven multi-process product quality prediction model was developed by combining whale optimizer algorithm,convolutional neural network,long short-term memory network,and attention mechanism.Considering the high-dimensional,highly correlated,multi-feature,and time-series characteristics of the quality data in multi-process product manufacturing process,kernel principal component analysis was used to reduce the dimensionality of the manufacturing process data.Deep features were extracted by convolutional neural networks and the quality information transmission between upstream and downstream processes was modeled by long short-term memory networks based on attention mechanism.The data-driven multi-process product quality prediction model was established.The whale optimizer algorithm was applied for model optimization,realizing real-time prediction of product quality characteristics and verifying the accuracy and effectiveness of the prediction model through instances.A data-driven multi-process product quality diagnosis model was developed by combining clustering algorithm based on PAM and frequent pattern growth algorithm.The complex correlation between process data in multi-process products was analyzed,and PAM algorithm was applied to conduct cluster analysis on quality data.Based on the clustering results,frequent pattern growth algorithm was applied to rapidly mine the massive data,and the implicit correlation between process characteristic data and quality indicator state was discovered,forming a quality correlation rule base,which realized the diagnosis of abnormal factors in multiprocess products.The effectiveness of the model was verified through examples.A data-driven prototype system for multi-process product quality control was developed using a front-end framework consisting of Vue and ElementUI,along with programming languages including Python and C#.The system realized functions such as product information management,quality prediction,and diagnosis,providing a technological basis for achieving real-time quality control of multi-process products. |
