Research On Prediction Method Of Engine Quality Characteristics Under The Influence Of Multiple Quality Parameters

Engine assembly is the important link in the whole engine production process. Assembly quality directly determines the performance of the engine, multivariate quality factors affects engine performance in the form of a complex nonlinear in this process. According to effect of quality factors on quality characteristics is difficult to quantify in the machining and assembly process, As prediction of quality characteristics under the influence of multiple quality factors the object of study, put forward the quality characteristics prediction model which base on PMI-Elman network, in order to realize the prediction and diagnosis of engine qualityFirstly, through analysis multivariate quality parameters effect on the quality characteristic in different ways, expounds the correlation analysis theory, introduce mutual information to analysis correlation; introduce parameters based on partial mutual information method, avoid computing the distribution function, use Copula function to estimate the partial mutual information, determine the key quality parameters. This paper introduces the prediction method of the non-feedback neural network and the feedback neural network, analysis the advantages and disadvantages of the BP neural network, and introduce the Elman network to forecast and diagnose the quality characteristics, so as to make the theoretical preparation for modelingSecondly, by analyzing the characteristics of engine assembly process and machining process, find out the factors that affect the quality of assembly process and machining process. Introduce the relationship between testing process and performance index, determine the quality characteristics of the exhaust gas pressure as the research of this paper. Through the statistical method, analysis the characteristics of the partial parameters of the exhaust pressure quality characteristic, describe the distribution of the parameters and the degree of dispersion using mathematics figure, prepare for the study of correlation.Thirdly, use the method of probability transformation to carry out the non-dimensional treatment of quality parameters By means of partial mutual information theory and Copula entropy, calculate the correlation between quality characteristics and quality factors, select the input variables of neural network. by Elman network, establish the prediction and diagnosis model, by comparing with the common BP neural network, confirm the validity and accuracy of the model. Establish the predictive value of the output variable exhaust pressure quality characteristic, establish the control chart diagnosis model, the prediction and diagnosis are combined together to provide a mathematical basis for engine performance prediction and diagnosis.Finally, taking an engine factory as an example, put forward the prediction and diagnosis system of engine quality. Through definite the system requirements, software and hardware conditions, modeling system; analysis system function module, develop visual interface to help forecast and diagnose; through the Elman neural network training sample, predict quality properties and the analysis through SPC chart, realize active prevention and early warning control of the engine assembly quality.