Engine Cold Test Exhaust Test Fault Detection And Diagnosis Method

With increasing requirements to manufacturing quality and overallperformance of automobile engine, research in making use of the cold testtechnology to accurately detect and rapidly diagnose engine failure ispaid more and more attention by domestic and foreign automotiveindustry. Engine cold test technology Engine cold test technology is akind of fast-paced, high-quality and low-cost detection method withoutcombustion process which can reflect the overall assembly performanceand achieve online100%measurement. The main purpose is to analyzeand judge engine condition by detecting data. Exhaust test is an importantproject of cold test, including exhaust pressure peak, valve opening angle,valve opening leakage and other parameters and the unqualified samplestake up1/4-1/3of the total. But there exists following problems in faultdetection and diagnosis: Firstly, design of exhaust test parameter controllimit lack scientific basis and parameter correlation is ignored, leadingsome fault omission and false detection; on the other hand, fault repairmainly depends on subjective experience to traversing check anddetection information is underutilized, resulting in low efficiency andeven failure. To solve these problems mentioned above, this article constructs aset of engine cold test fault detection and diagnosis method based on avariety of statistical theory, focusing on key issues about single variablecontrol limit determination, multivariable control limit design, exhausttest fault source diagnosis, intake and exhaust valve clearance failuremode separation, developing data processing and analysis software andgetting effective application in manufacturing quality control of a certaintype of engine. The main research work in the paper was focused onseveral areas as follows:(1) Exhaust test single parameters control limit determinationmethod.To solve the problem that single variable control limit determinationlack scientific basis, propose building tail distribution to get control limit.Firstly, calculate tail distribution model from average excess function;Then get model quantile to replace normal distribution one as parametercontrol limit; Finally, estimate parameter confidence interval according toposterior distribution parameters to characterize the data state as the baseto judge whether to adjust the control limit.(2) Exhaust test multi-parameter control limit design method.Given conventional single variable control limit monitoring methodsignore parameter correlation, combine with support vector datadescription to achieve multivariate monitoring. Firstly, take tail dependence as related metrics to select multi-variable parameter; Thenintroduce support vector data description, screen sample, improveparameter optimization method combining with false positives and falsenegatives; Finally, deal with learning problem of large samples byincremental learning strategies, false positives will reduce and stabilizewith the learning process.(3) Exhaust test fault source diagnosis network modeling method.For fault source diagnosis mainly depends on engineeringexperience, and conclusion presents uncertainty, propose combiningBayesian network theory with cold test diagnostic process. Firstly, extractfault event network input and output nodes, construct three-tier networkarchitecture according to the characteristics of engine failure; Thenexpress complex causal relationship of fault source and the fault symptomwith probability model according to historical inspection data; Finally,use cold test and field test data as multi-source evidence, reason potentialfault source of the and update fault network.(4) Intake/exhaust valve clearance fault mode separation method.In consideration that intake/exhaust valve clearance fault has similarfault sign and distinguish only by traversing check, extract fault featurefrom exhaust pressure wave to separate fault source. Firstly, makemulti-scale decomposition of exhaust pressure wave with empirical modedecomposition; Then remove trend term and noise term, get characteristic energy vector; Finally, establish corresponding relation of exhaustpressure wave characteristics and valve clearance status, construct valveclearance fault source multi-class classification.(5) Software development and engineering applications.On the basis of above study, develop exhaust test module in dataprocessing system, achieve the function of single parameter control limitdetermination, cold test fault diagnosis, apply in several cases likeexhaust pressure peak control limit adjustment and exhaust test faultdiagnosis network establishment, verify the method of this paper.In summary, this paper focuses on the engineering problem of faultdetection and diagnosis in cold test and studies a set of engine cold testexhaust test quality control method, taking full advantage of cold testdetection data, providing basis to design parameter control limit andguidance to cold test repair. The method can be applied not only inexhaust test data processing, but also to promote to the torque test，ignition test and other cold test project.