Data Mining Method And Application Research In Pressure Vessel Quality Process Control

With the development of data mining technology,data mining method applied to the quality control attracts more and more attention.In the quality control of pressure vessel,the solution to quality classification prediction problems is gradually changed from the traditional statistical process control to the data mining technology.According to the process control of pressure vessel quality problems,this paper focuses on the classification of weld quality process control of pressure vessel quality prediction model,mainly from the decision tree model,neural network model and Logistic regression model,which provide guidance and reference for the production process the pressure volume optimization of enterprises.First of all,for the quality prediction model based on decision tree algorithm of date mining,through the analysis of the data of pressure vessel weld quality,this paper choose 20 attribute variables as input variables of the decision tree model,discretize these variables,optimize the relevant parameters of decision tree algorithm in the classification process,and establish reasonable and effective quality classification model.Secondly,this paper makes the same original sample data through the data ready be the input variables which change into 0-1 variables,then the use of multi-layer training method establishes BP neural network model with 2 hidden layers.Then,based on the two regression analysis equation,a multivariate Logistic regression analysis model was established for predicting the weld quality of pressure vessels.Finally,the three data mining classification prediction models are compared and analyzed,the best prediction model is found to solve the problem of quality process control of pressure vessel.The comparison of these models is from the aspects of correct rate and accuracy of the model of quality classification and prediction.And from gain map,response figure and improve figure the rule generalization ability,rule confidence and the ability to capture the characteristics of the sample covers of the models are evaluated.We can get that the decision tree model has obvious advantages in the comparative perspective by comparing the models.In the process of modeling,Z angle,thickness,allowable deviation of outer perimeter and developed length are main factors affecting the weld quality and classification of pressure vessel.