Application and comparison of multiple machine learning techniques for the calculation of laminar burning velocity for hydrogen-methane mixtures

Sven Eckart*, René Josef Prieler, Christoph Hochenauer, Hartmut Krause

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

In the present discussion of transition the energy supply and sector coupling processes, hydrogen and hydrogen/natural gas mixtures will play an important role in future gas usage as gaseous energy carrier mainly natural gas is widely used in industrial combustion systems, combustion engines as well as domestic heating systems. Combustion properties of hydrogen differ completely from natural gas. Therefore, numerical modelling of combustion phenomena is an important task due to development and optimization of innovative combustion systems or for safety issues. In this area laminar burning velocity (LBV) is one of the most important physical properties of a flammable mixture. LBV is one of the parameters used for assessment and development of detailed chemical kinetic mechanisms and burners as well. The goal of this work is to develop models by using machine-learning algorithms for predicting laminar burning velocities of methane/hydrogen/air mixtures at different states. Development of the models is based on a large experimental data set with over 1400 data points collected from the literature after 2005. The models are developed in Python taking into account (i) generalized linear regression model (GLM), (ii) support vector machine (SVM), (iii) Random Forest (RF) and (iv) artificial neural network (ANN). The influence of the number of hidden layers and neurons per layer were investigated to find the best possible solution for an ANN. The performance of the developed models was evaluated with one widely used detailed chemical reaction mechanisms. Therefore the GRI 3.0 DRM was used for this purpose in the numerical simulations. The main advantage of developed models is the much shorter computational time compared to the solving procedures for detailed chemical reaction mechanism.
Original languageEnglish
Article number101306
Number of pages11
JournalThermal Science and Engineering Progress
Volume32
DOIs
Publication statusPublished - 1 Jul 2022

Keywords

  • artificial neural network (ANN)
  • laminar burning velocity (LBV)
  • Machine learning, hydrogen-methane mixture

ASJC Scopus subject areas

  • Fluid Flow and Transfer Processes

Fields of Expertise

  • Sustainable Systems

Fingerprint

Dive into the research topics of 'Application and comparison of multiple machine learning techniques for the calculation of laminar burning velocity for hydrogen-methane mixtures'. Together they form a unique fingerprint.

Cite this