Experimental study of knocking phenomenon in different gasoline–natural gas combinations with gasoline as the predominant fuel in a SI engine

Abstract

The disadvantages of the separate use of gasoline and CNG pose great challenges to the development of spark ignition (SI) engines. One of the major obstacles to achieving higher thermal efficiency in gasoline engines is the knocking phenomenon. One solution to tackle this problem is the use of a combination of gasoline and CNG. Gasoline can be used as the predominant fuel due to having higher burning velocity. In the current study, different fuel mixtures containing 100%, 90%, 80%, and 70% gasoline and the rest NG (G100, G90, G80, and G70, respectively) were investigated in a single-cylinder SI research engine. The aim was to investigate the knocking features of the mixtures at the stoichiometric equivalence ratio, the compression ratio of 11, and the engine speed of 1800 rpm. After capturing data for each mixture at every spark advance and analyzing 400 subsequent cycles, the required processes were performed to determine the knocking features of the combinations at Optimum Spark Advances (OSA). At the OSA of G100, the average maximum amplitude of pressure oscillation $(\overline{\text{MAPO}}_{\text{Tot}} )$ and the knocking cycle percentage (%KC) were found to be 0.289 bar and 5.3%, respectively, and the distance between OSA and the impending knock limit advance (Δθ) was found to be less than 1 °CA. With the increase in NG proportion in the fuel mixture, $\overline{\text{MAPO}}_{\text{Tot}}$ and %KC decreased significantly, as their values in G80 reached 0.156 bar and 0.5%, respectively. Furthermore, as NG fraction increased, Δθ also increased significantly as its value reached 4.5 °CA in G70.

Appendix

Fourier transfer (FT) of a time domain function explains the concept of the function in the frequency domain. To calculate FT of a discrete Fourier transfer (DFT), the following equation can be used [23]:

$$\hat{x}_{{\rm m}} = \sum\limits_{n = 0}^{{\text{N}} - 1} {x_{{\rm n}} e^{{ - 2{\text{j}}\frac{\uppi {\text{mn}}}{{\text{N}}}}} } ,\quad m = 0,1, \ldots ,N - 1$$

(4)

where N is the number of data undertaken to transfer, x_n and $\hat{x}_{{\rm m}}$ are the values of the function in the time and frequency domains, respectively, and j is the character of the imaginary part of a complex number. The equation can be simply rewritten as follows:

$$\hat{x}_{{\rm m}} = \sum\limits_{n = 0}^{N - 1} {x_{{\rm n}} W_{{\rm n}}^{nm} }$$

(5)

in which $W_{{\rm N}} = e^{{\frac{ - 2\pi j}{N}}}$. After applying the required filtration in the frequency domain detailed in Ref. [24], the filtered data can be transferred to the time domain using inverse discrete Fourier transfer (IDFT) as follows:

$$x_{{\rm n}} = \frac{1}{N}\sum\limits_{n = 0}^{N - 1} {\hat{x}_{{\rm m}} W_{{\rm N}}^{-nm} }$$

(6)

The most common method to calculate fast DFT or IDFT, which are, respectively, the same as FFT or IFFT, is Cooley–Tukey method which is described in greater detail in Refs [25] and [26].