Investigation on online perception method of gas injection characteristics for the high-pressure natural gas direct injection engine

To realize the online perception of the gas injection process for high-pressure natural gas direct injection engines. This paper presents an innovative perception method for gas injection characteristics. The injection characteristics are divided into time characteristics and quantify characteristics. The gas inlet pressure is used to be the sensing signal. Firstly, time characteristics are highlighted on the frequency domain signal by short-time Fourier transform, and all kinds of time characteristics are defined in the time domain. To improve the real-time performance of the algorithm, the mean instantaneous frequency (MIF) was selected as the representative of the frequency at each moment to reduce the dimension of the time-frequency signal. Based on the zero and local maximum points of the MIF signal, a recognition method of the injector time characteristics is proposed. The neural network is used to build the prediction model of the quantified characteristics. The results show that: In all the ranges of the injector working conditions, the error of the time characteristics is less than 5%. The root mean square error of the injection mass is 2.68 mg, and the regression coefficient between the predicted value and the measured value is 0.99878. The method proposed in this paper has relatively high accuracy.

Automated summary

This paper presents an innovative perception method for gas injection characteristics in high-pressure natural gas direct injection engines, using short-time Fourier transform and neural network modeling for online perception of the gas injection process.