Neural network-based air handling control for modern diesel engines

Abstract

Complex air handling systems, featuring technologies such as exhaust gas recirculation (EGR) and variable geometry turbochargers (VGTs), are commonly used in modern diesel engines to meet stringent emissions and fuel economy requirements. The control of diesel air handling systems with EGR and VGTs is challenging because of their nonlinearity and coupled dynamics. In this paper, artificial neural networks (ANNs) and recurrent neural networks (RNNs) are applied to control the low pressure (LP) EGR valve position and VGT vane position simultaneously on a light-duty multi-cylinder diesel engine. Intake manifold pressure (IMP) and air-fuel equivalence ratio, or $\lambda$ , are selected as the control objectives since they directly impact engine emissions and cylinder power output. Meanwhile both signals are available on production engines, so no additional hardware costs for measurement systems will be introduced. Both transient and steady state experimental data are separately used to train the two categories of neural networks (NNs). The NNs with minimum mean square error (MSE) for the training data sets are compared to conventional proportional-integral (PI) control and model predictive control (MPC). The most accurate NN controller has almost no overshoot during the transient process while the steady state error $\lambda$ and IMP are at most 7.0% and 3.8%, respectively, under a wide range of engine speeds from 2500 to 4000 rpm, thus showing the potential of this approach.