Neural Network Nonlinear Modeling of a Common Rail Injection System for a CNG Engine

In this paper, nonlinear dynamical black-box models of a common rail injection system for a CNG engine are developed. In particular, the common rail pressure dynamics is modeled on the basis of three input signals, easily and cheaply measurable on board a vehicle. The nonlinear model is identified by means of Multi Layer Perceptron neural networks. Both non-autoregressive (NMAX) and autoregressive (NARMAX) models have been developed, showing satisfactory performance.