Real-time modeling of ringing in HCCI engines using artificial neural networks

Abstract

Intense ringing operation is one of the major drawbacks associated with homogeneous charge compression ignition (HCCI) engines at high load conditions that limits HCCI operation range and can damage engine parts. This study uses HCCI experimental data to investigate combustion noise and ringing operation in a 0.3�L converted-diesel HCCI engine. A novel method is utilized to operate the HCCI engine in the ringing region by periodic variation in injected fuel amounts. Combustion noise level (CNL) is investigated along with main HCCI combustion parameters and emissions. A strong correlation is found between the CNL and variation of in-cylinder pressure at 10, 15, 20 CAD aTDC (P10, P15> and P20) and maximum in-cylinder pressure (Pmax). These experimental findings are then used to design an artificial neural network (ANN) model to predict CNL for identifying normal and ringing regions. A large amount of experimental data at cyclic and steady-state operating conditions are used to evaluate the ANN noise level (ANL) model. The results indicate that the real-time ANL model can predict CNL with less than 0.5% error for the HCCI engine. The ANL model is of utility to identify engine operating limits to avoid the ringing operation.