Inverse Problem Analysis Using Negural Network
| Autor: | Kozukue, Wakae, Miyaji, Hideyuki, Hagiwara, Ichiro, Shi, Qinzhong |
|---|---|
| Jazyk: | japonština |
| Rok vydání: | 1998 |
| Předmět: | |
| Zdroj: | 神奈川工科大学研究報告.B,理工学編. 22:1-4 |
| ISSN: | 0916-1902 |
| Popis: | application/pdf The application of a holographic neural network (HNN), which is new algorithm of neural network (NN), to inverse problems such as a structural identification problem is investigated. For a simple beam model the training of NN is carried out by setting eigenmodes of the beam as input data and a element number and Young's modulus of a defect contained in the beam as output data. After the training of NN the result for output is obtained from the testing of NN by using some input data which are not learned by NN. As for the method for obtaining the element number of a defect, the analog method and the degital method are tested to obtain the output. It is shown that the analog method has better acruracy than the digital method. As a result it is concluded that the HNN has the capability to solve inverse problems such as a structural identification problem. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|