Design of Fuel Additives Using Neural Networks and Evolutionary Algorithms

It is difficult and challenging to design high-performance fuel additi®es in an industrial-design setting where data are sparse and noisy, and fundamental knowledge is often limited. An automated framework is presented for the design of such fuel-additi®e molecules that minimize the intake-®al®e deposit in the automobile. A hybrid model that combined functional descriptors from a first-principles degradation model with a statisticalrneural-network model was de®eloped to predict additi®e performance, gi®en the additi®e structure. The results of the predicti®e model are discussed for different real industrial case studies. An e®olutionary method using specialized representation and constrained operators to enforce formulation constraints was used to generate optimal additi®e molecules that meet desired performance criteria. The e®olutionary design strategy in combination with the hybrid prediction model was successful in identifying no®el additi®e molecules that also possess good synthesis potential.