A CO2 Laser Weld Shape-Predicting Neural Network
We describe two artificial neural networks (ANN) which predict CO2 partial penetration laser welds on grade 304 stainless steel. Given the laser irradiance and travel speed, one ANN (direct) predicts the resulting weld's depth, width, overall shape, energy transfer efficiency, melting efficiency and porosity likelihood in the weld fusion zone. Given the weld size and shape, the second ANN (inverse) predicts the irradiance and travel speed necessary to provide such a weld. The ANNs used 3 nodal layers and perception-type neurons. For the first ANN, with 2 inputs and 17 outputs (12 for shape, and 5 for size, efficiencies and porosity predictions), 12 to 17 intermediate layer neurons were necessary, while for the second, with 14 inputs and 2 outputs, 25 were necessary. Besides their description, data interpretation and weld schedule development via the ANNs will be shown.
- Research Organization:
- Sandia National Laboratories (SNL), Albuquerque, NM, and Livermore, CA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- AC04-94AL85000
- OSTI ID:
- 748
- Report Number(s):
- SAND98-1378C; ON: DE00000748
- Resource Relation:
- Conference: 17th International Congress on Applications of Lasers and Electro Optics; Orlando, FL; 11/16-19/1998
- Country of Publication:
- United States
- Language:
- English
Similar Records
Weldability of irradiated Stainless Steel 304 materials harvested from the National Research Universal (NRU) reactor
Artificial neural network-based transient stability regions of power systems