Politecnico di Torino (logo)

Prediction of the laminar combustion velocity in methane-air mixtures by means of deep learning algorithms

Davide Fittipaldi

Prediction of the laminar combustion velocity in methane-air mixtures by means of deep learning algorithms.

Rel. Daniela Anna Misul, Mirko Baratta. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Meccanica (Mechanical Engineering), 2021

PDF (Tesi_di_laurea) - Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives.

Download (2MB) | Preview

The need to reduce environmental pollution has led to a search for alternative, low-emission fuels. In fact, internal combustion engines, are one of the main source of pollution in the urban environment. Methane is one of the best alternatives to the main and most common fossil fuels, because it is one of the fuels with the lowest CO2 and hydrocarbon emissions. In addition to being one of the cleanest fuels, it has remarkable chemical and physical properties, such as high anti-knocking, which would allow high efficiency in monovalent methane-fuelled heat engines with an high compression ratio. However, compared to petrol, methane burns slowly, which leads to a variation in efficiency and not complete stability cycle by cycle, reducing power and increasing fuel consumption. Low flame front propagation speed and poor burning capacity in poor mixture conditions can be improved by the addition of hydrogen, due to its higher burning speed. The aim of this thesis is to develop a deep learning algorithms for predicting the laminar speed of combustion in methane/air mixtures, providing as input the conditions of the mixture (pressure, temperature, percentage of EGR, equivalent ratio). The objective of this methodology is to see if it is possible to construct a neural network capable of understanding the non-linear relationships of the phenomenon under analysis, allowing a faster simulation, by virtue of the reduction of computational costs, but at the same time must be accurate, respecting what is the physics and kinetics of the combustion process. The results obtained by training our neural network separately on the combustion tables obtained with the main chemical kinetics mechanisms, i.e. Aramco 2.0 and GRImech 3.0, were also compared with the laminar flame speed experimental data available in the literature. Another important objective of this thesis is to continue research in the field of heat engines, in particular to generate combustion tables using LOGEresearch, software for simulating the chemical kinetics of the combustion process, for air/methane mixtures with and without the addition of hydrogen. The work focuses on finding the mixture dosage, with a value that differs from one, to observe how the value of the laminar speed varies, compared to the other input values, which are used to start the simulation.

Relators: Daniela Anna Misul, Mirko Baratta
Academic year: 2020/21
Publication type: Electronic
Number of Pages: 86
Corso di laurea: Corso di laurea magistrale in Ingegneria Meccanica (Mechanical Engineering)
Classe di laurea: New organization > Master science > LM-33 - MECHANICAL ENGINEERING
Aziende collaboratrici: UNSPECIFIED
URI: http://webthesis.biblio.polito.it/id/eprint/18555
Modify record (reserved for operators) Modify record (reserved for operators)