NOX Emission Prediction of a dual-fuel (Diesel + CNG) Compression Ignition Engine Using the DCNN model

Main Article Content

Yasser Niknam, Davood Mohammad Zamani, Mohammad Gholami Pareshkoohi

Abstract

One of the viable solutions for utilizing natural gas in diesel engines involves adopting dual-fuel technology. This study converted the MT440C Compression Ignition (CI) engine into a dual-fuel engine, combining Diesel and Compressed Natural Gas (CNG) as its primary fuel. This investigation aims to assess the impact of natural gas as the primary fuel and diesel fuel as the spark ignition source on a 4-cylinder CI engine to reduce diesel fuel consumption. To address the challenge of accurately predicting Nitrogen Oxide (NOX) emissions from a dual-fuel Compression Ignition Engine (utilizing Diesel and CNG) under transient operating conditions, a NOX reduction prediction model is proposed based on the Deep Convolutional Neural Network (DCNN) architecture. Convolutional Neural Networks (CNNs) represent a more intricate class of neural networks than conventional artificial neural networks. The primary advantage of CNNs, owing to their deep architecture, lies in their capability to discern and extract distinct and heterogeneous features at various levels of abstraction. The experimental data utilized in this study were gathered from the Engine Research Center of Tabriz Motorsazan Company in Iran. Experiments were conducted under stable conditions at engine speeds of 1200, 1400, 1600, 1800, and 2000 revolutions per minute (rpm). Each experimental run was repeated three times to ensure comprehensive data collection. To effectively predict NOX emissions from the dual-fuel engine during training, the proposed DCNN initially extracts features from multidimensional data and subsequently employs these features to establish the relationships between different time steps. The experimental results demonstrate that the proposed DCNN model exhibits superior accuracy, convergence, and robustness in predicting transient NOX emissions from the dual-fuel engine. This is evident through a root mean square error (RMSE) of 21.70 and a fitting coefficient (R2) of 0.997, affirming the high precision and effectiveness of the proposed DCNN model in managing multidimensional time-series datasets.

Article Details

Section
Articles