In this paper, we address the construction of a prior stochastic model for nonGaussian deterministically-bounded positive-definite matrix-valued random fields in the context of mesoscale modeling of heterogeneous elastic microstructures. We first introduce the micromechanical framework and recall, in particular, Huet's Partition Theorem. Based on the latter, we discuss the nature of hierarchical bounds and define, under some given assumptions, deterministic bounds for the apparent elasticity tensor. Having recourse to the Maximum Entropy Principle under the constraints defined by the available information, we then introduce two random matrix models. It is shown that an alternative formulation of the boundedness constraints further allows constructing a probabilistic model for deterministically-bounded positive-definite matrix-valued random fields. Such a construction is presented and relies on a class of random fields previously defined. We finally exemplify the overall methodology considering an experimental database obtained from EBSD measurements and provide a simple numerical application.Key words: Micromechanics; Heterogeneous materials; Apparent elasticity tensor; Mesoscale modeling; Random field; Non-Gaussian. $ J. Guilleminot, A. Noshadravan, R. Ghanem and C. Soize, A probabilistic model for bounded elasticity tensor random fields with application to polycrystalline microstructures,

We propose a methodology for conducting robust comparative life cycle assessments (LCA) by leveraging uncertainty. The method evaluates a broad range of the possible scenario space in a probabilistic fashion while simultaneously considering uncertainty in input data. The method is intended to ascertain which scenarios have a definitive environmentally preferable choice among the alternatives being compared and the significance of the differences given uncertainty in the parameters, which parameters have the most influence on this difference, and how we can identify the resolvable scenarios (where one alternative in the comparison has a clearly lower environmental impact). This is accomplished via an aggregated probabilistic scenario-aware analysis, followed by an assessment of which scenarios have resolvable alternatives. Decision-tree partitioning algorithms are used to isolate meaningful scenario groups. In instances where the alternatives cannot be resolved for scenarios of interest, influential parameters are identified using sensitivity analysis. If those parameters can be refined, the process can be iterated using the refined parameters. We also present definitions of uncertainty quantities that have not been applied in the field of LCA and approaches for characterizing uncertainty in those quantities. We then demonstrate the methodology through a case study of pavements.

This study aims to improve post-disaster preliminary damage assessment (PDA) using artificial intelligence (AI) and unmanned aerial vehicle (UAV) imagery.In particular, a stacked convolutional neural network (CNN) architecture is introduced and trained on an in-house visual dataset from Hurricane Dorian. To account for the ordinality of damage level classes, the cross-entropy classification loss function is replaced with the square of earth mover's distance (EMD 2 ) loss. The trained model achieves 65.6% building localization precision and 61% (90% considering ±1 class deviation from ground-truth) classification accuracy. It also exhibits a positive accuracy-confidence correlation, which is valuable for model assessment in situations where ground-truth information is not readily available. Finally, the outcome of damage assessment is compared with the literature by examining the relationship between building size and number of stories, and severity of induced disaster damage.

No abstract