In this work, we attempt a quantitative estimation of the type IV rupture stress for welds in ferritic power plant steels containing 9 Ð 12 wt % chromium, using a neural network in a Bayesian framework. This article describes the methodology that was used in creating and evaluating the neural network model. The sensitivity of the rupture stress to the test conditions, the composition of the steel and the heat treatment schedule, as perceived by the model, appears to be consistent with engineering experience and known metallurgical effects. It has also been possible, for the first time, to infer the dependence of the stress on welding parameters. The rupture stress increases with the preheat and interpass temperature, consistent with a widening of the heat-affected zone and consequent reduction in the triaxiality of stress in the susceptible region. The heat input has a relatively insignificant effect, as expected from heat flow theory.
International Conference on Trends in Welding Research, ASM International, Atlanta, Georgia, 2005, in press.
Download PDF file of related paper (ISIJ International).
Photographs of Australia, where this work was carried out.
Tempered martensite Fe-9Cr-1Mo weld metal.
|
|
|
| Superalloys | Titanium | Bainite | Martensite | Widmanstätten ferrite |
| Cast iron | Welding | Allotriomorphic ferrite | Movies | Slides |
| Neural Networks | Creep | Mechanicallly Alloyed | Theses | Retained Austenite |
| PT Group Home | Materials Algorithms |
|
|