Home | People | Research | Publications | Links 

Research > Finished Projects>

Process Optimization with Neural Networks and Genetic Algorithms

Abstract -- A novel method for optimization of manufacturing process with the combined use of neural network and genetic algorithm was proposed. The neural network was trained with the existing datasets. Once trained on the exsting process variables the neural network could predict the output based upon the given input datasets. This output was then provided to the genetic algorithm which used generated the offsprings based upon the fitness criterion evaluated from the trained neural network. The genetic algorithm could then optimize these datasets to predict the process variable which would result in the optimal output. Another salient feature of the approach was the ability of the method to generate additional datasets based upon the existing inputs. The method was first applied to a mathematical surface and later to melt spinning and film extrusion with considerable success.