Fuel Injection Quantity Fluctuation Prediction of Common Rail System Based on Bond Graph Model

Abstract

Common rail injection system (CRIS) is an advanced technology which meets the stringent emission standards of diesel engines and satisfies consumer demand for better fuel efficiency and increased power. The coherence of fuel injection quantity is the key injection characteristic for CRIS to match diesel engines successfully. As a critical component for CRIS, the variation of injector characteristic parameters has significant influence on the coherence of fuel injection quantity of the system. In this paper, combining numerical modeling and design of experiments, the response predicted relation between fuel injection quantity fluctuation of CRIS and its influence factors had been investigated. A numerical model of common rail injector was presented for the purpose of creating a tool for simulation experiments. The model is developed using power bond graph method, which is a modeling method that has shown its superiority in modeling systems consisting of sub-models from several energy domains in a unified approach. Experiments were conducted at the same model conditions to validate the model. The results are quite encouraging and in agreement with model predictions, which imply that the model can accurately predict the fuel injection quantity of CRIS and it can be used to simulation and design experiments. Experiments were designed using D-optimal method in which the characteristic parameters of common rail injector were chosen as design factors and the fuel injection quantity fluctuation was selected as the response. The fuel injection quantity fluctuation responses at different design factor levels were obtained using the developed numerical model which had been validated. A regressive prediction model of fuel injection quantity fluctuation was suggested according to the simulation experiments by means of partial least-squares regression (PLR) analysis. Analysis of variance, normal distribution of standardized residuals and relation between observed and predicted fuel injection quantity fluctuation for the regressive prediction model were analyzed which demonstrate the favorable goodness of fit and significance of the regressive model to predict fuel injection quantity fluctuation of the system. By changing design factor levels, the comparisons between numerical results of the bond graph model and the predicted fuel injection quantity fluctuation of the regressive prediction model were conducted. Results show that the regressive prediction model can reliably predict the fuel injection quantity fluctuation caused by the variation of characteristic parameters of common rail injector. Research results of this paper can provide novel ideas to predict fuel injection quantity fluctuation and a theoretical guidance for design and parameters optimization of CRIS.