A Deep Learning Framework for Efficient Global Sensitivity Analysis and Shap Values Calculations Applied to Eddy Current Testing Problems

Abstract

Abstract In the context of the nondestructive testing and evaluation research community, global sensitivity analysis (GSA) methods are widespread tools for quantifying the sensitivity of measurements concerning the variation of inputs over the whole design space. For parameter ranking purposes, GSA methods have been historically preferred by NDT&E scholars compared to feature importance (FI) techniques such as random forest-based, SHAP values, among others. For practical applications, GSA and FI can face limitations when the number of evaluations of the physical model is very high. The main issues that one needs to face with GSA and FI in practical problems are the low computing time efficiency of the numerical solver and/or the high cardinality (i.e., the number of inputs) of the problem considered. This paper targets two main goals. First, we propose to tackle the problem of an efficient GSA and FI procedure relying on a tailored deep neural network to be employed as a metamodel (or surrogate model) to replace the less efficient numerical model. Second, we compare GSA (i.e., Sobol’ indices and 𝛿-sensitivity measure) indices and FI (i.e., SHAP) method for parameter ranking purposes. In particular, we describe a generative deep neural network framework to be straightforwardly applied to GSA and FI studies. The numerical experiments in this communication correspond to an eddy current testing inspection problem where multiple arbitrarily oriented cracks lie in a conductive planar multilayered structure.