A real-time model-based approach for control of the Brake Mean Effective Pressure (BMEP) has been developed and assessed for a Euro 6 1.6L GM diesel engine. The model provides the fuel quantity necessary to achieve a desired BMEP target. The engine features complex injection patterns, including pilot, main and multi-after injections.
The approach is based on the use of feed-forward artificial neural networks (ANNs), which have been trained using virtual tests simulated by a previously developed, low-throughput, mean-value, physical combustion model.
The physical combustion model is based on an improved version of the accumulated fuel mass approach and is capable of predicting the heat release and the in-cylinder pressure. The latter quantity is in turn used to extract the Indicated Mean Effective Pressure (IMEP). The BMEP is then obtained from the IMEP while taking into account the friction and accessory-related contributions. A novelty of this study is the assessment of the low-throughput physical combustion model for complex injection patterns, including not only the pilot and main shots but also multi-after-injection pulses.
The physical model and the ANN-based model have been assessed considering experimental data acquired at the General Motors—Global Propulsion Systems (GM-GPS) facilities under steady-state and transient conditions over several driving cycles.
