Multimodality data fusion for probabilistic strength estimation of aging materials using bayesian networks

Accurate strength estimation and uncertain quantification for aging materials is critical for the time-dependent reliability assessment. Destructive testing is very expensive and may even be impossible to perform. Inference methods are needed for this purpose by multimodality surface material measurements from nondestructive testing, such as micro indentation, chemical composition, microstructure quantification, and surface hardness measures. Bayesian network modeling is utilized to integrate the information from various types of surface measurements for a more accurate bulk mechanical property estimation. To improve the approximation of the actual underlying model and avoid the risk of overfitting, Bayesian model averaging (BMA) of Bayesian networks is implemented to account for model uncertainty. The models are weighted according to the posterior model probabilities. Markov Chain Monte Carlo sampling is used to numerically compute the marginal likelihoods, which are essential for obtaining the posterior model probabilities. The predictive performance of single best model and BMA are compared by logarithmic scoring rule. The predictive capability of the proposed method is evaluated. It is shown that the Bayesian network and model averaging approach can provide more reliable results in statistical sense for predicting the bulk mechanical properties.