A neural network approach to analyzing multi component mixtures

A novel approach to determining the individual chemical concentrations in a mixture of chemicals is described. This approach uses fusion and an artificial neural network to learn the relationships between the outputs from chemical sensors and the individual chemical concentrations in the mixture. The chemical sensors used are of a biologically motivated design, and a multitude of sensors are used in simulation. An artificial neural network is trained on a subset of reaction space and it is tested for its ability to generalize to all reaction space. Research has shown that sensor fusion with an artificial neural network is able to learn to accurately map from sensor outputs to the actual input chemical concentrations. The sensor fusion results are also compared to the results of a more traditional mathematical technique of solving the same problem.<<ETX>>