Research And Application Of Multivariate Correlation Between Engine Cold Test And Processing Data

With a wide range of application on Engine Cold Test and the rising requirements of technology for automotive engine manufacturing quality and consistent overall performance. The research on how to make better use of Engine Cold Test for rapid diagnostic before delivery is getting more and more engineering and academic attention. Engine Cold Test is the last quality testing for engine before leaving the factory; the effective analysis of test not only can ensure the quality of engine and improve the overall performance of the machine, but also is objective basis of rapid diagnosis of the failure engine. In this paper, we start the research on Engine Cold Test focus on finding the relationship between engine cold test data and the key parts manufacturing data. And apply it to cold test fault diagnosis. The main purpose is to establish an accurate and practical rapid engine cold test fault diagnosis method; with less fault diagnosis blindness occurs, in order to improve the diagnostic efficiency.In this paper, the research is based on the philosophy of modern quality control, "Data-driven Quality", and then starts an exploratory research on some key issues addressed in the mainline of "detection-analysis-Control" in engine manufacturing quality control and the "normalized" approach in data correlation analysis. We give an introduction of Copula on measurement of data between engine cold test and processing test, establish a method for engine cold test data analysis and fault diagnosis based on correlation matrix. The research provides new methods and ideas for future process control and complex evaluation index of data dependence, in the same time, it helps to improve the efficiency of fault diagnosis, and makes better service overall performance consistency control. The main research work in the paper was focused on several areas as follows:(1)Inductive analysis of engine testing data characteristics from two types of data sources. Give a comprehensive analysis of all kinds of engine test data from three aspects:central tendency, the degree of dispersion and the evaluation of distribution shape. Indicate the advantages of Copula dependence structure in data analysis, in combination with the lack of conventional correlation analysis methods.(2)Give a detailed analysis about the advantages and the weakness of various methods combined with some types of data normalization methods and the multi-source data heterogeneous characteristics. Starting from the definition of data probability density function, explore a method of data standardization based on the probability integral transform.(3)Use the residual control charts to improve the engine cold test monitoring, for the autocorrelation characteristics reflected in data, in order to achieve better process control and anomaly identification.(4)Give an introduction of Copula function method which can process various types of correlation indexes "normalized". Starting from the definition of experience distribution function, explore a more efficient and effective method for optimal Copula selection in different distribution function family. Give an introduction of method based on Euclidean. Use rank correlation coefficient to do fast fault diagnosis; use tail correlation coefficient to do monitoring and diagnostic for data fluctuations exception. This method is applied to a multivariate analysis of engine test data, and solves practical engineering problems in the manufacturing department.All in all, this paper focuses on the engineering problem of engine cold test failure diagnosis and the three types of theoretical issues:multi-source heterogeneous data standardization, auto correlated process control diagram optimization, complex non-normal distribution of data correlation analysis. Study a method to determine the quantitative relationship of data between different parametric and put it into use in rapid diagnostic of cold test in SGMW.