Uncertainty quantification for Bayesian active learning in rupture life prediction of ferritic steels

Abstract

Three probabilistic methodologies are developed for predicting the long-term creep rupture life of 9–12 wt%Cr ferritic-martensitic steels using their chemical and processing parameters. The framework developed in this research strives to simultaneously make efficient inference along with associated risk, i.e., the uncertainty of estimation. The study highlights the limitations of applying probabilistic machine learning to model creep life and provides suggestions as to how this might be alleviated to make an efficient and accurate model with the evaluation of epistemic uncertainty of each prediction. Based on extensive experimentation, Gaussian Process Regression yielded more accurate inference ($$Pearson\;correlation\;coefficent> 0.95$$ P e a r s o n c o r r e l a t i o n c o e f f i c e n t > 0.95 for the holdout test set) in addition to meaningful uncertainty estimate (i.e., coverage ranges from 94 to 98% for the test set) as compared to quantile regression and natural gradient boosting algorithm. Furthermore, the possibility of an active learning framework to iteratively explore the material space intelligently was demonstrated by simulating the experimental data collection process. This framework can be subsequently deployed to improve model performance or to explore new alloy domains with minimal experimental effort.

Uncertainty quantification in multivariable regression for material property prediction with Bayesian neural networks

“…The authors employed GPR to calculate point estimates and uncertainty estimates in the predicted rupture lire. Differently from the work by Mamun et al 9 , we develop approaches for point regression and UQ in predicting creep rupture life based on BNNs, and demonstrate that Bayesian deep learning models present a promising framework for this task.…”

“…Among the conventional ML models with an inherent ability for UQ in regression tasks are Quantile Regression (QR) 6 , Gaussian Process Regression (GPR) 7 , and Natural Gradient Boosting (NGBoost) 8 . QR and NGBoost have shortcomings due to the lack of closed-form parameters estimation and are prone to overestimating the uncertainty level in data 9 . In most related works, GPR is generally reported as the state-of-the-art approach for UQ in multivariable regression and it stands out for its predictive accuracy and uncertainty estimates.…”

“…Although prior works have explored the application of ML approaches for predicting material properties 20 – 26 and uncertainty estimates in the predictions 16 , 27 – 29 , our proposed approach introduces novel concepts related to BNNs with incorporated physics priors for UQ in material property prediction. Specifically, the proposed work was inspired by Mamun et al 9 , who proposed an approach for predicting the creep rupture life of ferritic steels using conventional ML methods. The authors employed GPR to calculate point estimates and uncertainty estimates in the predicted rupture lire.…”

Convoluted filtering for process cycle modeling

“…The augmented dataset was most recently used for developing a variety of ML approaches for explaining compositional segmentation, quantifying uncertainty, and others. 19,20 The initial 82 alloy compositions along with processing parameters were fused with the latent variable datasets (such as the upper-critical-point and start-of-martensite-formation temperatures, Cr and Ni equivalent concentrations, and the microphase volumes) for pre-training DeepFreG to learn empirical domain knowledge. 3 Most of the input data were ingested and preprocessed, batched and re-shaped in TensorFlow Python environment similar to prior work.…”

“…The 82‐alloys dataset for the demonstration of CoFi functionality was compiled as part of the U.S. Department of Energy's Advanced Alloy Development program 18 and was later augmented by information about additional 17 alloys. The augmented dataset was most recently used for developing a variety of ML approaches for explaining compositional segmentation, quantifying uncertainty, and others 19,20 …”