Abstract
This thesis concerns the generation of material data using neural nets. On the basis of polyamide it is examined, if a standardised procedure can be used for the characterization of the material performance against different influencing variables. The parameters of a function for the description of the stress-strain characteristic have been generated by bending tests. The tests were performed at different temperatures, humidities and testing velocities. The determined parameters were used as target values for the neural nets. The results show that the amount and the quality of the training data is crucial for the successful creation of neural nets. If all the available data is used for the training, the material performance can be described very well by the neural nets. Only in regions, where the data from the measurements were not of the best quality some problems occurred. A reduction of the data is not possible. If not all data is made available to the neural nets, no appropriate results can be achieved.
Translated title of the contribution | Application of Neural Nets for the Generation of Material Data using the Example of Polyamide |
---|---|
Original language | German |
Qualification | Dipl.-Ing. |
Supervisors/Advisors |
|
Award date | 29 Jun 2007 |
Publication status | Published - 2007 |
Bibliographical note
embargoed until nullKeywords
- neural nets material modelling polyamide bending tests