Research On Online Error Evaluation And Machining Process Quality Control Of Valves’ High-precise Inner Hole Based On Support Vector Machine

Valve parts are important components of mechanical products.Whether the machining accuracy of valve holes can be ensured is very important for the product quality and reliability.With the continuous raising of machining precision requirements for the inner holes of valve parts,the quality control method based on conventional process methods and sampling inspection can not meet the production requirements well under mass production conditions.It is great significance to achieve fast and accurate on-line evaluation of hole parts and to conduct quality monitoring and adjustment of the machining process in real time based on the evaluation results.Relies on the specific engineering projects,this paper presents a method for the fast evaluation of cylindricity based on the fusion features and SVR,a pattern recognition and abnormal parameter estimation method of machining process control chart based on SVM online updating,and studies the anomaly tracing method for the processing of hole parts,in order to provide the basis for the effective monitoring and adjustment of the machining process.The main research work is as follows:(1)A method of measuring the cylindricity error of the inner hole based on PC A feature reduction and SVR is proposed.By combining the measurement data of diameter of the inner hole with the roundness,taper and straightness as the fused feature set,the feature set is reduced by the reduction algorithm,and the cylindricity error of the inner hole is evaluated with the SVR as the evaluation tool.The PC A feature reduction algorithm is used to eliminate the redundant feature components in the feature convergence for the feature fusion of the inner hole measurement data.The control parameters of the PCA reduction algorithm and the S VR’s penalty parameters and the kernel function parameters are realized by the particle swarm optimization(PSO)and the cross validation.It can be optimized synchronously.Through the comparative analysis with the data of conventional measurement methods and cylindricity instrument.The results show that the evaluation accuracy of this method is obviously better than that of conventional method,and it can well meet the requirements of real-time detection.(2)A chart pattern recognition algorithm to the process control chart based on online updating of SVM is proposed,and the parameters of the identified anomaly patterns are estimated with the SVR parameter estimation method.By establishing the control chart of the machining process and updating SVM training samples based on monitoring the process statistical parameters,the process parameter threshold is designed,when the parameters of the quality data distribution exceed the threshold,the off-line simulation training data will be regenerated and the SVM model will be retrained to meet the dynamic production process.The online updating of SVM identification model is realized.Through the simulation experiments adapted to the actual machining process,the results show that the recognition algorithm can achieve good abnormal pattern recognition effect.At the same time,the parameter estimation algorithm can obtain a good parameter estimation of the size of the abnormal fluctuation.(3)An anomaly tracing method is proposed,which combines rule knowledge rules and SVR reasoning mechanism.By using knowledge rules that describe the abnormal information and abnormal reasons as feature sets,the SVR model is trained to get the traceability model that can be used for abnormal reasoning,and the extension of knowledge rules is realized.The concrete rules are expressed as the confidence degree of abnormal reasons based on control charts pattern category,process parameters and abnormal state information.Examples show that the rule knowledge retraining and utilization through the integration of SVR reasoning mode,improve the accuracy of dynamic process of abnormal traceability and effectiveness.(4)The quality intelligent evaluation and control system of the multi-channel valve inner hole is developed,and the system structure,process and function are designed.The function modules include the general configuration module,the quality detection and intelligent evaluation module,the SPC analysis and pattern intelligent recognition module,the abnormal tracing module.The system has been applied in multi-channel valve manufacturing enterprises and achieved good results.