Intelligent Diagnosis Of Manufacturing Process Based On WSVM

In today's global competitive market,product quality is one of the key factors for success.The quality of a product is determined by product design and manufacturing.The final quality of products is based on the control of process quality.Statistical Process Control(SPC)plays an important role in product process quality control.As the most important statistical process Control technology,control charts have been widely used to monitor the manufacturing process is under control.Although various control charts of statistical process control play an important role in the process of quality control,the control chart still belonging to the lag control,can only monitor the abnormal fluctuation of the process,and can not indicate what the abnormal is and where the anomaly occurs.The detection and diagnosis of abnormal manufacturing process is often done by manual quality inspection,but manual quality inspection,a time-consuming and tedious task,needs to be accomplished by strictly trained people.The quality control method of manufacturing process relies solely on statistical process control and the abnormal diagnosis of artificial manufacturing engineering poses a great chal enge.Currently,Machine learning is used to intelligent monitoring and diagnosis of manufacturing process quality.Many researches are done based on the balanced data,but we often face the imbalanced classification problems in reality.For example,the samples of normal control chart pattern are much bigger than abnormal control patterns.How to achieve a diagnosis effect of better quality under the condition of imbalanced data samples is a problem that needs to be solved in the process of manufacturing quality diagnosis,which is also the emphasis of this research.By improving the traditional support vector machine algorithm,this paper sets different weights for different kinds of samples to improve the correct recognition rate of classifier to the abnormal pattern of few classes,which help realizing the intelligent diagnosis of process quality.This paper focuses on the intelligent diagnosis of the single-variable process and the uneven sample of the multivariate process.(1)A single variable procedure.This paper summarizes the control chart model which represents the normal process and the process abnormity,and introduces the relationship between the control chart pattern recognition and the process quality diagnosis.Then,according to the more actual situation,the normal control chart pattern and abnormal control chart patterns are highly unbalanced and data sets are not balanced.The weighted support vector machine(WSVM)model is based on intelligent algorithm to recognize normal control chart pattern and abnormal control chart pattern.So it can achieve intelligent process quality diagnosis.The control chart pattern recognition process of WSVM model based on Intelligent Algorithm is:Firstly,the sample data including normal control chart pattern and abnormal control chart pattern are obtained.The best way to obtain sample data is to obtain a large amount of sample data directly from manufacturing processes.But this method is difficult to achieve for various reasons.Therefore,this paper obtains the controlled data from the actual manufacturing process,and gets the sample mean,standard deviation and other parameters.Then,the Monte Carlo simulation method is used to generate samples for training and testing the WSVM model.Secondly,preprocessing and feature extraction of the sampled data.The purpose of preprocessing is to make the model effective for different processes or changes,and can be independent in the specific processes of the distribution parameters.The purpose of feature extraction is to enhance the difference between different patterns and improve the recognition performance based on WSVM model.Thirdly,there're following advantages to use WSVM Gauss radial basis kernel function is the kernel function:to classify nonlinear and high-dimensional data;to reduce the complexity of the operation,to achieve higher classification only needing to adjust the penalty factor and kernel function parameter C,g.Therefore,the preprocessing and feature extraction training samples and test samples are input into the WSVM model,and the C and g parameters of the WSVM model are optimized by genetic algorithm and particle swarm optimization respectively.After the optimal fitness function is obtained by training samp les,the pattern recognition of the test set is obtained by using the obtained parameters.Fourthly,based on the comparative analysis of the experimental results of the test set,the classification accuracy is selected as the basis of the selection.This paper chooses the weighted support vector machine based on genetic algorithm for pattern recognition.Finally,the features play a vital role in the classification of digital signal types.In order to investigate the effectiveness of the selected features,we have used features that have been introduced in some references.The other simulation setups are the same.The results imply that the proposed features and PSO-SVM have effective properties incontrol chart pattern representation.(2)Multivariate process.In the field of quality diagnosis,we can transform the quality diagnosis problem into a classification problem when there is a corresponding anomaly classification or anomaly pattern.The multivariate control chart can help find the abnormal process,but failed to find the reason for the abnormal process which is caused by abnormal variables or other variables;and when there is a correlation between the quality characteristics of all kinds of variables;imbalanced classification problems of multivariate manufacturing process data sets.We propose a particle swarm algorithm and WSVM based model to diagnose faults in out-of-control conditions and to help identify aberrant variables when Shewhart-type multivariate control charts based on Hotelling's T~2 are used.Then,we can find which variable or variables would lead to the process out of control.And then we could take measures to make the process under control.