Prediction of emission characteristics in a CRDI engine fitted with a Zeolite catalytic converter using deep learning algorithms

The prediction of emission characteristics in a Common Rail Direct Injection (CRDI) engine equipped with an in-house-made Zeolite Socony Mobil-5 (ZSM5) catalytic converter, employing deep learning algorithms, is attempted. This study employs a dataset comprising emission data extracted from experimentation on a CRDI engine by varying the applied load (3–15 kW), engine speed (1300–2000 rpm), and injection pressure (300–900 bar). By employing deep learning techniques such as artificial neural networks (ANN), dense neural networks (DNN), and recurrent neural networks (RNN), carbon monoxide, hydrocarbon, carbon dioxide and nitrous oxide emissions were predicted. Of the three attempted models, the DNN model provided a closer prediction with a coefficient of determination (R2) of 0.9577, a mean squared error (MSE) of 142.77, a root mean square error (RMSE) of 2.16, and a mean squared logarithmic error (MSLE) of 0.02. The deep learning models can be a protocol to facilitate real-time emission control and aid the automotive industry in designing environmentally friendly engines and developing efficient emission control systems.