Grey-box modeling and control of HCCI engine emissions

Real-time model based control of Homogeneous Charge Compression Ignition (HCCI) engines faces a critical challenge of maintaining a perfect balance between model accuracy and computational load. In particular, currently available HCCI emissions models in the literature are highly computationally expensive for control applications. This paper develops a computationally efficient grey-box HCCI engine model for predicting Total Hydrocarbon (THC), Carbon Monoxide (CO), and Nitrogen Oxides (NO_x). The grey-box model consists of a feed forward Artificial Neural Networks (ANN) model in combination with physical models for estimating combustion phasing and Indicated Mean Effective Pressure (IMEP). The emission model is experimentally validated over a large range of HCCI engine operation including 208 steady state test conditions. The validation results show that the grey-box model is able to predict NOx, CO, and THC with average relative errors less than 10%. Using a Genetic Algorithm optimization method along with the developed emission grey-box model, an optimum CA50 trajectory is obtained for every given load trajectory in order to minimize THC and CO emissions. A model-based controller is designed and tested on the grey-box virtual engine model for tracking IMEP and the optimum CA50 trajectories, while indirectly minimizing the engine emissions. Control results show that the developed grey-box model is of utility for real time HCCI control applications.