Research On Gasoline Engine Knock Feature Extraction And Intensity Evaluation Using Vibration Signal

Knock has become a key factor that stops the gasoline engine from downsizing and improving efficiency. Accurate knock detection can not only keep the engine run at a proper condition, but also improve the power and fuel economy. Using the vibration signal of the cylinder block to detect the knock is a widely used solution on a gasoline engine. However, compared to methods based on cylinder pressure, the accuracy of the result acquired from vibration signal cannot be guaranteed because of its low signal-to-noise ratio. In this paper, the research is focused on the data processing method of the vibration signal. After the featured knock frequency band was evaluated using power spectrum density estimation, the knock characteristic was extracted from the original signal according to the ensemble empirical mode decomposition(EEMD). The extracted result was then used to detect the knock onset and evaluate the knock intensity.Power spectrum density(PSD) expresses the change of power versus frequency of a specific signal, in this paper, a model-based PSD method was applied to accurately verify the featured knock frequency band. EEMD was used to decompose the original signal to acquire intrinsic mode functions(IMFs). With the chosen mother wavelet, continues wavelet transform(CWT) was used to analyze the time-frequency characteristic of each IMF. The IMF that contains the featured knock frequency band was considered as the knock characteristic and then used to detect the knock onset and evaluate the knock intensity.The timing of the first obvious peak in the knock characteristic was identified as the knock onset. To prove the accuracy, it was compared with the result using cylinder pressure signal.To improve the accuracy and calculating process of the knock intensity detection, a new knock intensity factor K which indicates the average power inside the knock window was brought up in this paper. After the identification of the knock onset timing, it was consider as the start of the knock window. The influence of different window width on the knock intensity evaluation result was studied, based on which the best window width was chosen. After the knock window was determined, indicator K of different knock cycles was calculated. The result was also approved by comparing to the result obtained from VDO method on cylinder pressure signal.