A self-organizing-map approach to chemistry representation in combustion applications

Several alternative techniques have been proposed in the literature in order to avoid the CPU-intensive numerical integration of the thermochemical equations in the simulation of combustion processes. The present paper introduces a new approach, which is based on two artificial neural-network (ANN) paradigms, namely the self-organizing map (SOM) and the multilayer perceptron (MLP). The SOM is first employed for the automatic partitioning of the thermochemical space into subdomains. Then, a specialized MLP is trained in order to fit the thermochemical points belonging to a given subdomain. The presented strategy is tested on a partially stirred reactor (PaSR) with a reduced methane-air mechanism, and encouraging results are reported. The relatively modest CPU-time and memory requirements of the method make the SOM-MLP approach a promising technique for the inclusion of large chemical mechanisms in the context of complex applications, such as the multidimensional simulation of combustion.