The role of data analysis in uncertainty quantification: case studies for materials modeling

Abstract

In computational materials science, mechanical properties are typically extracted from simulations by means of analysis routines that seek to mimic their experimental counterparts. However, simulated data often exhibit uncertainties that can propagate into final predictions in unexpected ways.Thus, modelers require data analysis tools that (i) address the problems posed by simulated data, and (ii) facilitate uncertainty quantification. In this manuscript, we discuss three case studies in materials modeling where careful data analysis can be leveraged to address specific instances of these issues. As a unifying theme, we highlight the idea that attention to physical and mathematical constraints surrounding the generation of computational data can significantly enhance its analysis. I ncreasingly, the computational materials science community is embracing uncertainty quantification (UQ) as a necessary component of any modeling workflow that aims to provide actionable information for industry.1-4 In many instances, this change has been driven by the observation that computational predictions, while less expensive than their experimental counterparts, are affected by uncertainties not typically encountered in laboratory settings. As a result, there has been significant work focused on identifying and quantifying uncertainties associated with simulation tools per se. 5-7 Despite this, however, much less effort has been devoted to understanding the impact of these uncertainties on subsequent analyses and usage models. Critically, such issues must be addressed if the benefits of computational approaches are to be fully realized. Within this greater context, data analysis remains a fundamental and sometimes overlooked task that can contribute to uncertainties in ways that are difficult to quantify, if

Best Practices for Quantification of Uncertainty and Sampling Quality in Molecular Simulations [Article v1.0]

“…12 It is important to recognize that time-series data usually displays a certain amount of autocorrelation in the sense that the numerical values of nearby points in the series tend to cluster close to one another. Intuition dictates that correlated data does not reveal fully “new” information about the quantity-of-interest, and so we require uncorrelated samples to achieve meaningful sampling [14]. 13 …”

“…12 to estimate the uncertainty. Strictly speaking, the standard uncertainty should be estimated using the true standard deviation of x (e.g., σ x ); given that the true standard deviation is unknown, the experimental standard deviation is used in its place as an estimate of σ x [14]. …”

Dynamic Measurement of Nanoflows: Analysis and Theory of an Optofluidic Flowmeter

“…The inset shows the errors in these measurements relative to the HCS, which remarkably agree to within 6 % for all measurements. write 0 ≤ x, y ≤ a ± a (33) for some constant a and its associated uncertainty a . For our purposes, however, it is more convenient to express the upper bound in terms of the relative uncertainty…”

“…Now, in general there is no unique convex function that fits a set of data by minimizing an objective function such as the sum-of-squared-differences. [33] However, the piecewise linear function that interpolates theÎ i is an upper bound on all such convex functions. Moreover, in the case of a dense grid of dosages ξ i , the variation between all such curves is negligible.…”

Utilizing SciPy and other open source packages to provide a powerful API for materials manipulation in the Schrödinger Materials Suite

“…Recently we implemented convex analysis of the stress strain curve (as described here [PKD18]). scipy.optimize.minimize is used for a constrained minimization with boundary conditions of a function related to the stress strain curve.…”