Process Knowledge Discovery And Roughness Prediction Based On Data Mining Approaches

Manufacturing system and manufacturing process are full of complexity and dynamics,and knowledge discovery means such as data mining and machine learning can be applied in all stages of manufacturing process,which is an important method and technology to cope with manufacturing challenges.Combining these technical means with manufacturing can not only fully explore the process knowledge contained in enterprise data,but also bring enterprises faster response speed and more profound insight.At the same time,the powerful computing power of machine learning is used to analyze and optimize the process,which is conducive to improving the design and production efficiency,reducing costs and improving quality.Data mining and machine learning methods have attracted more and more attention because of their high accuracy and good usability.This paper studies the application of knowledge discovery in manufacturing process,and proposes corresponding algorithms to solve different specific problems.The specific research content is as follows:(1)Aiming at the problem that the accumulated historical process data of the enterprise cannot be fully and effectively utilized,this paper proposes an association rule mining method based on the chaotic gravity search algorithm,introduces the chaotic gravity map to improve the update of the gravity constant,thus improving the search ability of the algorithm.Multiple evaluation measures are considered comprehensively,and similarity evaluation indexes are added to maintain rule diversity.This method can effectively mine process knowledge and improve the quality of mining results compared with other classical methods.(2)In view of the limitation that artificial neural network cannot explicitly reveal the nonlinear relationship between process parameters and production quality,a dragonfly algorithm based rule extraction algorithm is developed to generate accurate and understandable classification rules from machining process,for expressing causal relationship between process parameters and surface roughness.Furthermore,a hybrid learning model based on KBANN and classification rules is proposed for surface quality prediction.Experiments show that the model can discover the knowledge in the process of machining,guide the adjustment of process parameters,and finally achieve the goal of improving the quality of machining.(3)In view of the current situation that the models used for surface roughness prediction mainly adopts the traditional machine learning method,this paper introduces the deep learning methods to achieve the feature extraction of unsupervised learning,and uses multiple deep neural networks to model and predict the surface roughness based on tool wear.Compared with other methods,the fitting ability and generalization ability of deep neural network are good,which can help engineers to improve the quality of processing.This paper proposes several methods which can effectively improve the utilization rate of production data,the ability of process quality control,the stability of manufacturing process,production efficiency and product quality by mining process knowledge.At the same time,the application of data mining and machine learning to process knowledge discovery and surface quality monitoring also adds new content to the traditional manufacturing process control and optimization theory.