Search for: All records

Liquefaction under cyclic loads can be predicted through advanced (liquefaction-capable) material constitutive models. However, such constitutive models have several input parameters whose values are often unknown or imprecisely known, requiring calibration via lab/in-situ test data. This study proposes a Bayesian updating framework that integrates probabilistic calibration of the soil model and probabilistic prediction of lateral spreading due to seismic liquefaction. In particular, the framework consists of three main parts: (1) Parametric study based on global sensitivity analysis, (2) Bayesian calibration of the primary input parameters of the constitutive model, and (3) Forward uncertainty propagation through a computational model simulating the response of a soil column under earthquake loading. For demonstration, the PM4Sand model is adopted, and cyclic strength data of Ottawa F-65 sand from cyclic direct simple shear tests are utilized to calibrate the model. The three main uncertainty analyses are performed using quoFEM, a SimCenter open-source software application for uncertainty quantification and optimization in the field of natural hazard engineering. The results demonstrate the potential of the framework linked with quoFEM to perform calibration and uncertainty propagation using sophisticated simulation models that can be part of a performance-based design workflow. 

Observations of the dynamic loading and liquefaction response of a deep medium dense sand deposit to controlled blasting have allowed quantification of its large-volume dynamic behavior from the linear-elastic to nonlinear-inelastic regimes under in-situ conditions unaffected by the influence of sample disturbance or imposed laboratory boundary conditions. The dynamic response of the sand was shown to be governed by the S-waves resulting from blast-induced ground motions, the frequencies of which lie within the range of earthquake ground motions. The experimentally derived dataset allowed ready interpretation of the in-situ γ-ue responses under the cyclic strain approach. However, practitioners have more commonly interpreted cyclic behavior using the cyclic stress-based approach; thus this paper also presents the methodology implemented to interpret the equivalent number of stress cycles, Neq, and deduce the cyclic stress ratios, CSRs, generated during blast-induced shearing to provide a comprehensive comparison of the cyclic resistance of the in-situ and constant-volume, stress- and strain-controlled cyclic direct simple shear (DSS) behavior of reconstituted sand specimens consolidated to the in-situ vertical effective stress, relative density, and Vs. The multi-directional cyclic resistance of the in-situ deposit was observed to be larger than that derived from the results of the cyclic strain and stress interpretations of the uniaxial DSS test data, indicating the substantial contributions of natural soil fabric and partial drainage to liquefaction resistance during shaking. The cyclic resistance ratios, CRRs, computed using case history-based liquefaction triggering procedures based on the SPT, CPT, and Vs are compared to that determined from in-situ CRR-Neq relationships considering justified, assumed slopes of the CRR-N curve, indicating variable degrees of accuracy relative to the in-situ CRR, all of which were smaller than that associated with the in-situ cyclic resistance. 

Model calibration measures the agreement between the predicted probability estimates and the true correctness likelihood. Proper model calibration is vital for high-risk applications. Unfortunately, modern deep neural networks are poorly calibrated, compromising trustworthiness and reliability. Medical image segmentation particularly suffers from this due to the natural uncertainty of tissue boundaries. This is exasperated by their loss functions, which favor overconfidence in the majority classes. We address these challenges with DOMINO, a domain-aware model calibration method that leverages the semantic confusability and hierarchical similarity between class labels. Our experiments demonstrate that our DOMINO-calibrated deep neural networks outperform non-calibrated models and state-of-the-art morphometric methods in head image segmentation. Our results show that our method can consistently achieve better calibration, higher accuracy, and faster inference times than these methods, especially on rarer classes. This performance is attributed to our domain-aware regularization to inform semantic model calibration. These findings show the importance of semantic ties between class labels in building confidence in deep learning models. The framework has the potential to improve the trustworthiness and reliability of generic medical image segmentation models. The code for this article is available at: https://github.com/lab-smile/DOMINO.