Novel DNNs for Stiff ODEs with Applications to Chemically Reacting Flows

Chemically reacting flows are common in engineering, such as hypersonic flow, combustion, explosions, manufacturing processes and environmental assessments. For the number of reactions can be significant (over 100) due to very large CPU requirements chemical 99%) a flow combustion problems presently beyond capabilities even largest supercomputers.Motivated by this, novel Deep Neural Networks (DNNs) introduced approximate stiff ODEs. Two approaches compared, i.e., either learn solution or derivative these These DNNs applied multiple species chemically flows. Experimental results show that it is helpful account for physical properties while designing DNNs. The proposed approach shown generalize well.