Knock Diagnosis Of Gasoline Engine Based On Dynamic Knock Sensitive Window Width And Statistics Method

Further research on Knock recognition and intensity evaluation is really needed for its strongly negative influence on power performance and fuel economy of gasoline engine. Knock diagnosis methods based on engine's combustion pressure signal, which were direct knock detection resource ,were used more and more in industrial field, while some limitation still exit. Three new knock diagnosis methods based on engine combustion signal were brought in this paper.Knock detection with dynamic knock sensitive window(KSW for short) in angle-domain focus on integral calculation of vibration energy of high pass cylinder combustion signal and knock intensity evaluation . The identification parameterΔE, which is the subtractions between accumulative energy of high-pass cylinder combustion pressure signal in a KSW and in a reference window, was introduced to relatively denote the knock energy compared with the noise energy. The dynamical identification method of signal window width in angle-domain according to the identification parameterΔE was presented so that reasonable knock recognition and intensify evaluation under various engine operating conditions were possible. Moreover, an experimental analysis system ,constructed by Simulink subsystem in Matlab,was developed for practical application of knock diagnosis based on this dynamic KSW in angle-domain method. Results show this dynamic KSW method performed well and overcome the short point of constant KSW length in VDO knock diagnosis method.Random characteristic of knock events was the basement of knock detection method with probability distribution performance of knock factors .D-test, Lilliefors-test and other statistic parameters were used to test the probability distribution characteristic of knock factors in load case with and without knock events. Results show that: knock factors of load case without knock events abbey normal distribution while knock factors of knock load case not, future more ,this knock factors samples with knock events can be normalized by stephase,an outlier detecting method ,and these outliers were just knock events of the load case. Knock threshold ,strongly depended on one's engineering experience, was not necessary in this knock diagnosis method, furthermore, it was not sensitive to KSW width. Results of knock diagnosis gasoline engine combustion experiment show this statistic method performed well and made a more reasonable use of knock factors. System identification method ,such as ARMA time series analysis, was brought into knock detection method based high-pass combustion vibration signal.100,150 samples before and after peak value of combustion pressure was chosen as the KSW , data, basement of parameters vectors'estimating in ARMA model, was chosen by Monte-Carlo random method, and mahalabobis distance of this parameters vector's mean value was figured and used as knock identification factor. Threshold ,with 99% degree of confidence, was computed by Monte-Carlo method ,using load case without knock events. One of the most charming advantages was that , degree of confidence for threshold can be well chosen to make mahalanobis distance of no-knock load case far from this threshold ,results of this analysis showed mahalanobis distance of knock cycles were much bigger than no-knock ones, this helped a lot in knock detection .Crank angle, which was not a necessary for time series analysis, was not needed to test with combustion signals simultaneously , that helped to simplify combustion test.Results of knock detection by this three knock diagnosis methods show they all can make knock events identified accurately, clearly and suitably, furthermore ,have much advantage than current knock detection method.