Estimation of NOx pollutants in a spark engine fueled by mixed methane and hydrogen using neural networks and genetic algorithm

Mansour Keshavarzzadeh; Rahim Zahedi; Reza Eskandarpanah; Sajad Qezelbigloo; Siavash Gitifar; Omid Noudeh Farahani; Amir Mohammad Mirzaei

doi:10.1016/j.heliyon.2023.e15304

Estimation of NOx pollutants in a spark engine fueled by mixed methane and hydrogen using neural networks and genetic algorithm

Heliyon. 2023 Apr 7;9(4):e15304. doi: 10.1016/j.heliyon.2023.e15304. eCollection 2023 Apr.

Authors

Mansour Keshavarzzadeh¹, Rahim Zahedi², Reza Eskandarpanah³, Sajad Qezelbigloo⁴, Siavash Gitifar⁵, Omid Noudeh Farahani⁶, Amir Mohammad Mirzaei⁷

Affiliations

¹ Department of Mechanical Engineering Science, University of Johannesburg, Johannesburg, South Africa.
² Department of Renewable Energy and Environmental Engineering, University of Tehran, Tehran, Iran.
³ Department of Energy Systems Engineering, Islamic Azad University, Tehran, Iran.
⁴ School of Automotive Engineering, Iran University of Science and Technology, Tehran, Iran.
⁵ Faculty of Mechanical Engineering, Iran University of Science and Technology, Tehran, Iran.
⁶ Faculty of Computer Engineering and Information Technology, Faran Mehr Danesh University, Tehran, Iran.
⁷ Faculty of Materials Engineering, Tarbiat Modares University, Tehran, Iran.

Abstract

Nowadays, due to stricter pollution standards, more attention has been focused on pollutants emitted from cars. As a very dangerous pollutant, NO_x has always triggered the sensitivity of the related organizations. In the process of developing and designing the engine, estimating the amount of this pollutant is of great importance to reduce future expenses. Calculating the amount of this pollutant has usually been complicated and prone to error. In the present paper, neural networks have been used to find the coefficients of correcting NO_x calculation. The Zeldovich method calculated the value of NO_x with 20% error. By applying the progressive neural network and correcting the equation coefficient, this value decreased. The related model has been validated with other fuel equivalence ratios. The neural network model has fitted the experimental points with a convergence ratio of 0.99 and a squared error of 0.0019. Finally, the value of NO_x anticipated by the neural network has been calculated and validated according to empirical data by applying maximum genetic algorithm. The maximum point for the fuel composed of 20% hydrogen and 80% methane occurred in the equivalence ratio of 0.9; and the maximum point for the fuel composed of 40% hydrogen occurred in equivalence ratio of 0.92. The consistency of the model findings with the empirical data shows the potential of the neural network in anticipating the amount of NO_x.

Keywords: Genetic algorithm; Internal combustion engine; NOx; Neural network.