A Neural Network Structure for Prediction of Chemical Agent Fate

This work presents the development of a multi-input, multi-output neural network structure to predict the time dependent concentration of chemical agents as they participate in chemical reaction with environmental substrates or moisture content within these substrates. The neural network prediction is based on a computationally or experimentally produced database that includes the concentration of all chemicals presents (reactants and products) as a function of the chemical agent droplet size, wind speed, temperature, and turbulence. The utilization of this prediction structure is made userfriendly via an easy-to-use graphical user interface. Furthermore, upon the knowledge of the time-varying environmental parameters (wind speed and temperature that are usually recorded and available), the time varying concentration of all chemicals can be predicted almost instantaneously by recalling the previously trained network. The network prediction was compared with actual open air test data and the results were found to match.