A Hybrid Predictive Reduced Order Model for Laminar Flames
Please login to view abstract download link
This work proposes a new hybrid reduced order model (ROM) based on physical principles, which combines modal decomposition and neural networks. The ROM is applied to predict the temporal evolution of reactive flows. The algorithm is composed of two steps. In the first step, a dimensionality reduction is performed via Proper Orthogonal Decomposition (POD), which extracts the main patterns of the flow. In the second step, a deep learning model is introduced to predict the temporal coefficients of the POD modes. Results show that this novel algorithm is capable to predict the evolution in laminar flames with relative error smaller than 3%.