The overall objective of this research was to develop a neural network solution to resource-constrained scheduling problems. The first phase of this project allowed us to:

1. build a computer simulation of a chemical process;
2. use the results of trial runs of the simulator to provide training data for a neural network that we designed; and
3. produce with 100 percent accuracy process simulation results using the neural network.

Later a Kohonen network was integrated into the existing back propagation structure to serve as a preprocessor for the batch processing data. The Kohonen network acted as a feature detector and allowed the system to work better in the presence of incomplete data and uncertainty. A user friendly neural network tool was developed using Windows and Motif.