Through a simulation analysis on multiple injections

it was found that

in the second injection

the period of injection quantity fluctuation was the same as that of the initial injection pressure fluctuation at the inlet of an injector

both were 1.2ms. A fluctuation rate was defined to judge the severity of injection quantity fluctuation

so that the relevant laws of fuel injection quantity fluctuation were obtained. The two injection strategies of pilot-main injection and main-post injection were compared

and it was found that the injection quantity fluctuation law in the second injection changed. When the main injection pulse width reached 1.4ms

there occurred the shift of the phase of the injection quantity fluctuation curve of the main-post injection

while the phase of the injection quantity fluctuation curve of the pilot-main injection did not shift. Finally

a combined prediction model of fuel injection quantity was established using the BP neural network prediction model and the Fourier approximation function prediction model and using the reciprocal variance method to determine the weights of the two prediction models. Results show that the combined prediction model can reduce the system error of prediction and make the prediction result more accurate

and root mean square error is only 0.71mg.