Optimized ANN-GA and experimental analysis of the performance and combustion characteristics of HCCI engine

HCCI (Homogeneous Charge Compression Ignition) engine has the benefit of operating at high thermal efficiency and low emissions of NOx and soot. However, it has challenges of complex combustion phase controlling and low operating range. This research work investigated the performance and combustion characteristics of HCCI engine with numerical simulations on ANSYS FLUENT and neural network models. The numerical and neural network results were validated by experimental observations with different fuel properties and reduced valve lifts for trapping of the exhaust gases. Experiments were performed on a SMART engine for different speeds and inlet air temperature, with various reference fuels (PRF30, PRF50, PRF70) and methanol to validate the CFD and ANN-GA observations. The engine performance was analyzed for IMEP, ISFC and thermal efficiency, which were found to be 8.2 bar, 205 g/kWh and 44.5% respectively as the optimum performance with PRF-70 fuel. The trapping of the residual gases was performed with various fuel blends in order to overcome the cyclic variations and to improve the operating zones near the knock boundary. The heat release rate was significantly reduced with trapped exhaust gases, and operating region was improved with the use of methanol fuel. Overall the trapping of the hot residual gases resulted in the maximum increase in the operating region by 12% and reduced cyclic variations by 15% for methanol fuel. The exhaust emissions were analyzed and ultra-low emissions of NOx at lean operating conditions were observed with the reduced valve lifts. The study results indicated thermal NO emissions on an average were decreased by 7.8%, CO emissions reduced by 6% and HC emissions increased by 9%. Methanol had ultra-low emissions of HC and CO, but higher emissions of NO and PRF30 had lower emissions of NO. However, ANN-GA model gave satisfactory combustion characteristics and emissions with respect to experimental results. Thus, CFD simulations, Neural Network methods and experimental study gave valuable thoughts of trapped residual gases approach on performance, combustion and emission characteristics of HCCI with PRF's and methanol fuel.