Grey-box modeling of HCCI engines

Abstract Homogenous Charge Compression Ignition (HCCI) holds promise to increase indicated thermal efficiency and reduce Nitrogen Oxides (NOx) and Particulate Matter (PM) emissions from internal combustion engines. Lack of a direct means to initiate the combustion and high levels of Total Hydrocarbon (THC) and Carbon Monoxide (CO) emissions are major drawbacks associated with HCCI engines. Control of combustion phasing for optimum indicated thermal efficiency and minimizing emissions are vital for putting HCCI engines into practice. One major challenge is to develop accurate models for understanding engine performance, as those models can run real-time for HCCI control. This paper develops the first computationally efficient grey-box model for predicting major HCCI engine variables. The grey-box model consists of a combination of physical models and three feed-forward artificial neural networks models to estimate six major HCCI variables including combustion phasing, load, exhaust gas temperature, THC, CO, and NOx emissions. The grey-box model is experimentally validated over a large range of HCCI engines operation including 309 steady state and transient test conditions. The validation results show that the grey-box model is able to predict the HCCI engine outputs with average relative errors less than 10%. Performance of the grey-box methodology is tested for two different HCCI engines and the verification results show that the developed six-output grey-box model can be successfully used for performance modeling of different HCCI engine applications.