PREDICTION OF SOLIDIFICATION MODE IN SUPER AUSTENITIC STAINLESS STEEL WELDS
Keywords:
Super Austenitic stainless steel (SASS), Probabilistic Neural Network , Solidification mode ,Austenite ,Martensite, Ferrite content, Alloy composition, Constitution DiagramAbstract
Super Austenitic OR Nickel alloy. Group of stainless steels having Fe- Ni-Cr-Mo alloys contents. The best
known material : 904L (20Cr,25Ni,4.5Mo) offers Superior corrosion resistance providing they are welded carefully with
low heat input (less than 1 KJ/mm recommended) and fast travel speeds with no waving.[1](3)(5) This is because of the
reason that fusion welding of SASS often destroy the chemical homogeneity of the weld metal composition by developing
unavoidable micro segregation of the tramped elements in the solidified weld structure. Which leads to poor corrosion &
Mechanical properties[1]. This has been discovered through a research study. Source [1] done about the influence of
Molybdenum on the solidification mode of high Mo bearing, Fe- Ni-Cr-Mo alloys, although SASS fusion welding best
practices recommending, each run of weld , not to be started until the metal temperature falls below 100ºC. But a nonuniform distribution of alloying elements always remains a possibility. As It has been already discovered by the
researchers[1] that various solidification mode (A, F, AF, FA) and solid state phase transformations will not be a only
function of Cr eq/Nieq but also Mo concentration, specifically due to the transformation of ferrite into eutectoid γ + σ in
high-Mo alloys. So, it becomes very necessary to understand the possible solidification mode and the very effect of various
elements on various solidification modes.
This problem can be overcome by Neural Network analysis, as through well trained model, it is also possible to
establish the relationships between the elements & the different transformation products. The Neural Network classification
method has been approached to solve this problem. The database collected from the research paper has been used to develop
& train the Probabilistic generalized classification Neural Network (PNN) model to meet the overall objective of prediction
of the multifaceted solidification mode of SASS alloys in appropriate welding process as a function of chemical composition,
in order to understand the mechanical and corrosive properties of the weld material for use in service applications.
