A Comparative Study of Physics Based Grey Box and Neural Network Trained Black Box Dynamic Models in an RCCI Engine Control Parameter Prediction

Reactivity controlled compression ignition (RCCI) engines are considered as a potent solution to realize near zero nitrogen oxides (NO_x) and soot emission with higher thermal efficiency. However, operational control in RCCI engines is challenging, as events such as ignition and combustion phasing etc. are mostly decoupled from hardware induced start of injection. In modern control architecture, these real time data are internally computed using signals from cylinder pressure sensor (CPS). Lately, physics based control models or grey box models in RCCI engines were considered as a cost competitive and smart alternative to hardware signal source. In this work, an attempt was made to develop and compare physics based grey box model with data based neural networks, trained through supervised learning (or the black box models) to accurately predict dynamic combustion control parameters across five engine loads and incremental premix energy share not exceeding 60%. Chosen control parameters for the study were the start of combustion (Ɵ_soc), 50% mass fraction burnt crank angle (Ɵ₅₀) and combustion peak pressure (PP). The model predictions were compared with real time experimentally measured data and comparative model quality study was carried out. Results indicated that the grey box model regression (R²) values for Ɵ_soc, PP and Ɵ₅₀ were 0.7054, 0.9535 and 0.7658 respectively that increased to 0.7637, 0.9654 and 0.8854 respectively for black box models. The overall tracking error for Ɵ_soc was 2.442 and 2.078⁰CA for grey and black box models, respectively. For Ɵ_50, the error was 4.262 and 2.904⁰CA in the same order. Finally, for PP, the error was 6.420 and 5.475 bar respectively for grey and black box models. Grey box models exhibited higher skewness and lower kurtosis values in error distribution compared to black box models.