Neural networks based prediction and optimization applied to siloxane-siloxane copolymers synthesis

Direct and inverse neural network modeling was applied to siloxane-siloxane copolymers synthesis, in heterogeneous catalysis. Feed-forward neural networks with one or two hidden layers, easy to design and train, supply accurate predictions for conversion and copolymer composition (direct modeling) or for reaction conditions that lead to an imposed copolymer composition (inverse modeling).