Developing a Neural Network for Prediction Applied in Materials Science and Engineering

This paper describes how a back-propagation neural network is built for predicting the properties of engineering materials from their other properties, compositions or processes. A predicted result of the impact toughness of a 12Cr-2Ni-1.5Mo stainless steel from its tempering temperature is presented. Furthermore, the combination of this neural network for prediction is able to supplement the unpredictability of the expert system outside its domain of expertise. An integrated system may perform more intelligently and solve more complex problems as well as human expert, in some case even better.