Hydrogen and dual fuel mode performing in engine with different combustion chamber shapes: Modelling and analysis using RSM-ANN technique

Abstract

This study investigates the impacts of hydrogen (H2) induction along with injected liquid honne biodiesel (BHO)/uppage biodiesel (BUO) as secondary pilot fuel in diesel engine. The effects of compression ratio (CR), hydrogen fuel flow rate (HFR) and different combustion chamber shapes in dual fuel (DF) mode were investigated. In the first phase of experiments, the effects of three different CR (15.5, 16.5, and 17.5) on engine efficacy and emission were presented. In the second phase, the effects of three HFR (0.1, 0.17, and 0.24 kg/h) on engine efficacy and emission, as well as the maximum possible HFR were reported. In the last phase, performance with different combustion chambers i.e., Hemispherical Combustion Chamber (HCC), Toroidal Reentrant Combustion Chamber (TRCC), and Toroidal Combustion Chamber (TCC) at maximum possible CR and HFR was highlighted. The study revealed that for knock free operation of the DF engine, the highest probable HFR was 0.24 kg/h at a CR of 17.5, fuel IT of 27obefore top dead center (bTDC) and injector opening pressure (IOP) of 250 bar. The toroidal re-entrant combustion chamber (TRCC) shape yielded 8%–12% better brake thermal efficiency (BTE) with lower emissions but 20–29% higher oxides of nitrogen (NOx) at 80% load in DF mode as contrasted to the single CI mode. Both peak pressure (PP) and heat release rate (HRR) were 12–15% higher. Response surface methodology (RSM) was used to design the experiments and to carry the optimization process. Artificial Neural Network (ANN) was used to forecast the performance and emission behaviors of the test engine. The findings demonstrated that RSM and ANN were excellent modelling techniques with good accuracy. In addition, ANN's prediction performance (R2 = 0.975 for BTE) was somewhat better than RSM's (R2 = 0.974 for BTE). Both the techniques were found to be successful in terms of agreement with experimental findings with ratios varying from 95% to 98% respectively. The prediction of BTE and NOx was also carried using different machine learning algorithms. It can be seen that R2 value for these models were slightly lower than ANN and RSM models indicating good predicting capability of ANN modelling.