Research On The Processing Quality Prediction Method Of Turbine Die-forging Blade

Since the national strategy “Made in China 2025” had been put forward,researches on Intelligent Manufacturing and development of digital manufacturing emerged constantly.In April 2018,the "Trade War",which was provoked by the United States,pushed the independent innovation ability of China’s high-tech to an "urgent" position.Under this background,as a major equipment,the demand for digitalization and quality management of process becomes increasingly prominent.As the “heart” of the steam turbine performance,the blade has been still used in the traditional methods to control quality.As the fact of low control-effect,high quality-cost and long assembly-cycle time,it is necessary to utilize the intelligent methods to predict and control the machining quality of the turbine blade.In this paper,the processing quality of the die-forging blade of a steam turbine factory has been taken as the research object.The following research is carried out to realize the effective prediction and control of the processing quality of the die forging blades:Firstly,considering the processing quality problem of turbine die forging blades,data mining algorithm and parameter optimization algorithm are studied and compared based on analyzation of quality management,turbine blade quality control,processing quality prediction and key process identification at home and abroad.A quality prediction model construction method based on PSO-SVR algorithm is proposed,and a quality control system based on the quality prediction model is constructed with the SPC control theory.Carding the overall plan of the research on the quality prediction and control of the blade processing;Researching and contrasting data mining algorithm and parameter optimization algorithm system,radiate a method of establishing quality prediction model based on PSO-SVR algorithm;and establish the overall research framework of quality control combining SPC control theory.Secondly,the current existing processing quality error sources basing on the quality fluctuation theory is analyzed.The influence mode and strength of the error source on the quality of the workpiece are clarified.The processing technology rules,processing system parameters,the quality of the workpiece and other information in the present situation are discussed.A method to describe processing state based on the processing quality transmission network model is proposed.Based on the graph theory,the quality transfer network model is processed with the Gephi graph processing software to realize the analysis of the process state.Thirdly,the principles and applications of Support Vector Regression(SVR),Particle Swarm Optimization(PSO)and K-fold method are studied.Focused on processing quality transfer network model,a PSO-SVR based quality prediction model of single process is constructed,and the quality prediction model of multi-stage process is constructed as well through topology merging.Combined with SPC anomaly pattern recognition,a quality control model based on quality prediction mode is presented which extended the traditional processing quality control paradigms.Finally,based on Visual Studio 2013 development platform,using the C# language,Microsoft SQL Sever 2012 and MATLAB numerical analysis tools,the turbine dieforging blade processing quality prediction system is designed and developed.The system is implemented in a stream turbine factory and verified by a machining example of die forging blade.Through the quality transfer network model to analyze the process state;built the quality prediction model based on the regression technology to predict the quality,thus realizing the quality control.Made up the shortage of the quality control of the blade production process in the steam turbine industry,it has of great theoretical and practical significance to realize the intelligibility of the blade production process.