Christopher Reynolds1,Andrew Hoffman1,Indranil Roy1,Sandipp Ravi1,Subhrajit Roychowdhury1,Bojun Feng1,Rajnikant Umretiya1
GE Global Research1
Christopher Reynolds1,Andrew Hoffman1,Indranil Roy1,Sandipp Ravi1,Subhrajit Roychowdhury1,Bojun Feng1,Rajnikant Umretiya1
GE Global Research1
One of the primary challenges of using FeCrAl as a nuclear fuel cladding material is the formation of α’-precipitates that can cause brittleness in the alloy. The precipitation causes hardness change during thermal aging which is sensitive to both alloy composition and experimental conditions (i.e., temperature and time of heat treatment). While this behavior is well known for Al free ferritic stainless alloys, in FeCrAl alloys the effects of Al and Mo additions include both influence on kinetics and thermodynamics. To better understand the influence of composition on this age hardening behavior, artificial intelligence and machine learning have been utilized to interpret a large data set of aging experiments. A Gaussian process regression model was built on the hardness data collected at General Electric Research. Subsequently, for the first time, SHapley Additive exPlanations (SHAP) is leveraged as an explainable artificial intelligence (XAI) tool to understand the effect of feature values in driving the hardness change. This new approach broadens the capabilities of machine learning and artificial intelligence. Several key insights are affirmed and obtained from a material science perspective.