Spatially Resolved Modeling of the Nonlinear Dynamics of a Laminar Premixed Flame With a Multilayer Perceptron - Convolution Autoencoder Network

Abstract

Abstract This work presents a multilayer perceptron-convolutional autoencoder (MLP-CAE) neural network model, which accurately predicts the two-dimensional flame field dynamics of an acoustically excited premixed laminar flame. The obtained architecture maps the acoustic perturbation time series to a spatially distributed heat release rate field, capturing the flame lengths and shapes. This extends to previous neural network models, which predicted only the field-integrated value of the heat release rate. The MLP-CAE comprises two sub-models: a fully connected MLP and a CAE. The key idea behind the CAE network is to find a lower dimensional latent space representation of the heat release rate field. The MLP is responsible for modeling the flame dynamics by transforming the acoustic forcing signal into this latent space, enabling the decoder to produce the flow field distributions. To train the MLP-CAE, computational fluid dynamics (CFD) flame simulations with a broadband acoustic forcing were used. Its normalized amplitude was set to 0.5 and 1.0, resulting in a nonlinear flame response. The network was found to accurately predict the perturbed flame shapes — both under broadband and harmonic forcing. Additionally, it conserved the correct frequency response characteristics as verified by the global and local flame describing functions. The MLP-CAE provides a building block towards a potential shift away from a purely ‘0D’ analysis with the assumption of acoustic compactness of the flame. When combined with an acoustic network, the generated flame fields could provide more physical insight in the thermoacoustic dynamics of combustion chambers. Those capabilities do not come at an additional significant computational cost, as even the previous nonspatial flame models had to train on the CFD data, which readily included field distributions.