-
A Scalable Algorithm for Multi-Material Design of Thermal Insulation Components Under Uncertainty
Authors:
Jingye Tan,
Danial Faghihi
Abstract:
This work presents a scalable computational framework for optimal design under uncertainty with application to multi-material insulation components of building envelopes. The forward model consists of a multi-phase thermo-mechanical model of porous materials governed by coupled partial differential equations (PDEs). The design parameter (material porosity) is an uncertain and space-dependent field…
▽ More
This work presents a scalable computational framework for optimal design under uncertainty with application to multi-material insulation components of building envelopes. The forward model consists of a multi-phase thermo-mechanical model of porous materials governed by coupled partial differential equations (PDEs). The design parameter (material porosity) is an uncertain and space-dependent field, resulting in a high-dimensional PDE-constrained optimization under uncertainty problem after finite element discretization. The robust design framework uses a risk-averse formulation consisting of the mean and variance of the design objective to achieve target thermal and mechanical performances and mitigate uncertainty. To ensure the efficiency and scalability of the solution, a second-order Taylor approximation of the mean and variance and the low-rank structure of the preconditioned Hessian of the design objective are leveraged, which uncovers the low effective dimension of the high-dimensional uncertain parameter space. Moreover, a gradient-based optimization method is implemented using the Lagrangian formalism to derive expressions for the gradient and Hessian with respect to the design and uncertain parameters. Finally, approximated $\ell_0$ regularization functions are utilized via a continuation numerical scheme to promote sparsity in the designed porosity. The framework's accuracy, efficiency, and scalability are demonstrated with numerical examples of a building envelope insulation scenario.
△ Less
Submitted 3 April, 2023;
originally announced April 2023.
-
Technical Note: PDE-constrained Optimization Formulation for Tumor Growth Model Calibration
Authors:
Baoshan Liang,
Luke Lozenski,
Umberto Villa,
Danial Faghihi
Abstract:
We discuss solution algorithms for calibrating a tumor growth model using imaging data posed as a deterministic inverse problem. The forward model consists of a nonlinear and time-dependent reaction-diffusion partial differential equation (PDE) with unknown parameters (diffusivity and proliferation rate) being spatial fields. We use a dimension-independent globalized, inexact Newton Conjugate Grad…
▽ More
We discuss solution algorithms for calibrating a tumor growth model using imaging data posed as a deterministic inverse problem. The forward model consists of a nonlinear and time-dependent reaction-diffusion partial differential equation (PDE) with unknown parameters (diffusivity and proliferation rate) being spatial fields. We use a dimension-independent globalized, inexact Newton Conjugate Gradient algorithm to solve the PDE-constrained optimization. The required gradient and Hessian actions are also presented using the adjoint method and Lagrangian formalism.
△ Less
Submitted 15 February, 2023; v1 submitted 13 February, 2023;
originally announced February 2023.
-
Goal-Oriented Adaptive Modeling of Random Heterogeneous Media and Model-Based Multilevel Monte Carlo Methods
Authors:
Laura Scarabosio,
Barbara Wohlmuth,
J. Tinsley Oden,
Danial Faghihi
Abstract:
Methods for generating sequences of surrogates approximating fine scale models of two-phase random heterogeneous media are presented that are designed to adaptively control the modeling error in key quantities of interest (QoIs). For specificity, the base models considered involve stochastic partial differential equations characterizing, for example, steady-state heat conduction in random heteroge…
▽ More
Methods for generating sequences of surrogates approximating fine scale models of two-phase random heterogeneous media are presented that are designed to adaptively control the modeling error in key quantities of interest (QoIs). For specificity, the base models considered involve stochastic partial differential equations characterizing, for example, steady-state heat conduction in random heterogeneous materials and stochastic elastostatics problems in linear elasticity. The adaptive process involves generating a sequence of surrogate models defined on a partition of the solution domain into regular subdomains and then, based on estimates of the error in the QoIs, assigning homogenized effective material properties to some subdomains and full random fine scale properties to others, to control the error so as to meet a preset tolerance. New model-based Multilevel Monte Carlo (mbMLMC) methods are presented that exploit the adaptive sequencing and are designed to reduce variances and thereby accelerate convergence of Monte Carlo sampling. Estimates of cost and mean squared error of the method are presented. The results of several numerical experiments are discussed that confirm that substantial saving in computer costs can be realized through the use of controlled surrogate models and the associated mbMLMC algorithms.
△ Less
Submitted 6 March, 2019; v1 submitted 6 August, 2018;
originally announced August 2018.
