Neural networks for providing ‘on-line’ access to discretised modelling techniques
| Autor: | T.W. Usherwood, J.E.J. Wadsworth, D.C.J. Farrugia, M. Patel, A.J. Trowsdale |
|---|---|
| Rok vydání: | 1998 |
| Předmět: |
Sequence
Artificial neural network business.industry Computer science Bar (music) Metals and Alloys Machine learning computer.software_genre Industrial and Manufacturing Engineering Finite element method Computer Science Applications Range (mathematics) Modeling and Simulation Line (geometry) Ceramics and Composites Artificial intelligence Limit (mathematics) business computer Algorithm Plane stress |
| Zdroj: | Journal of Materials Processing Technology. :475-480 |
| ISSN: | 0924-0136 |
| DOI: | 10.1016/s0924-0136(98)00150-2 |
| Popis: | This paper describes the use of a commercial neural network package for developing fast ‘real-time’ interpolative models from finite element (FE) generated data. The work describes the testing of this technique on simple plane strain and plane stress compression models and its extension to the prediction of deformation in a round–oval–round bar rolling sequence. The paper describes how neural network (NN) predictions can be improved and how targeting of the FE runs can be used to limit the requirement for large quantities of training data. This technique should be applicable to a wide range of modelling activities when rapid predictions are of paramount importance, for example, on-line systems. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect