Prediction of Emissions from a Dual-Fuel Compression Ignition Engine Using a Deep Convolutional Neural Network | ||
| مدل سازی در مهندسی | ||
| Articles in Press, Accepted Manuscript, Available Online from 25 November 2025 | ||
| Document Type: Computer Article | ||
| DOI: 10.22075/jme.2025.36595.2792 | ||
| Authors | ||
| Farnaz Hoseini* 1; Yasser Niknam2 | ||
| 1Department of Computer Engineering, Technical and Vocational University (TVU), Tehran, Iran | ||
| 2Department of Mechanical Engineering, Technical and Vocational University (TVU), Tehran, Iran | ||
| Receive Date: 17 January 2025, Revise Date: 25 June 2025, Accept Date: 25 November 2025 | ||
| Abstract | ||
| One of the promising approaches to reducing pollutant emissions in diesel engines is the application of dual-fuel combustion using compressed natural gas (CNG) alongside diesel fuel. In this study, a conventional compression ignition (CI) engine (MT440C model) was structurally modified to operate under a dual-fuel mode without the need for a spark-ignition system. The primary objective is to investigate the feasibility of using CNG in CI engines and to compare key operational and emission characteristics—including engine power output and exhaust emissions—under different engine speeds (1200, 1400, 1600, 1800, and 2000 rpm). To enable accurate and real-time prediction of nitrogen oxides (NOx) emissions, a novel deep convolutional neural network (DCNN) architecture was proposed. The model is designed to extract high-dimensional temporal-spatial features from the multi-variable time-series dataset and model complex nonlinear dependencies in dual-fuel combustion. Experimental results demonstrate superior predictive performance, achieving a root mean square error (RMSE) of 21.70 and a coefficient of determination (R²) of 0.997, significantly outperforming existing baseline models in the literature. The outstanding accuracy and robustness of the proposed DCNN model underscore its applicability for integration into real-time smart engine control systems aimed at optimizing emissions in hybrid combustion platforms. | ||
| Keywords | ||
| Convolutional Neural Network; DCNN; Compression Ignition Engine; Nitrogen Oxide; Dual-Fuel Engine | ||
|
Statistics Article View: 89 |
||