Output- and Input-Response Surfaces Generated from an Artificial Neural Network for an Emperical to Semi-Mechanistic Model: A Heuristic Approach
William A. Young, Gary R. Weckman · ASME Press eBooks · 2007
Artificial Neural Networks (ANNs) are primarily used in academia, due to their ability to model complex non-linear systems. Though ANNs have been used to solve practical problems in industry, they are not typically used in non-academic environments. It is believed that ANNs are not used because they are not very well understood, hard to implement, or the fact that they have a reputation of being a "black-box" model. Surfaces can be analyzed to provide evidence that the relationships "learned" by ANNs are in fact valid and can be used to gain additional knowledge of the system. In addition, equations can be derived that represents the constructed surfaces. This paper extends this form of ANN knowledge extraction technique to derive a semi-mechanistic equation.