Abstract:It has been widely recognized that structural materials are better designed simultaneously with the target component so that their properties can be tailored to the service loading conditions to meet component performance. This concurrent approach helps avoid over-strengthening the materials and components through unnecessary reinforcements, leading to lightweight, low-cost, yet durable designs. Recent advances in Integrated Computational Materials Engineering (ICME) have enabled researchers and engineers to predict a structure’s performance using its geometry, material (e.g. constituent selection and volume fraction), and processing conditions; however, two challenges arise when existing design optimization algorithms are applied to ICME: (1) The design problem involves mixed-type variables, including continuous, ordinal (e.g. low or high for material processing temperatures), and categorical variables (e.g. steel or plastic for type of materials), and (2) Evaluation of design performance considering manufacturing-induced local material microstructure evolution is computationally expensive. In this chapter, we present a machine-learning-based method to conquer both challenges through constrained Bayesian optimization (BO) with latent-variable Gaussian process (LVGP) modeling. LVGP allows Gaussian process (GP) regression with mixed-variables by mapping all non-continuous variables to a continuous latent space. It is integrated with BO, a method using (LV)GP’s predicted responses and prediction uncertainty to guide the optimal design search, to explore and exploit the disjoint and non-differentiable design spaces spanned by the non-continuous variables. BO is also shown to be highly efficient for optimization of expensive black-box functions. We demonstrate our approach using two design examples, thin-walled hat section components and short-fiber reinforced polymer (SFRP) composites. The former illustrates concurrent structure and material design while the latter focuses on concurrent material and processing optimization. In both examples, we first establish their respective ICME workflows, i.e. the integrated part-scale manufacturing processes and structural performance simulations coupled with material-scale microstructure and behavior models. Then, we show the benefits of using LVGP as a mixed-variable surrogate model to emulate the physics-based simulations, and the performance of constrained BO (CBO) based on LVGP, particularly on how the search is conducted among different levels of categorical variables. Results indicate that the LVGP-CBO framework can effectively design multi-material structural components for ICME through concurrent material, product, and process optimization.

Descriptor Aided Bayesian Optimization for Many-Level Qualitative Variables with Materials Design Applications

Bayesian Optimization for Materials Design with Mixed Quantitative and Qualitative Variables

Data-Centric Mixed-Variable Bayesian Optimization for Materials Design

Shape, sizing optimization and material selection based on mixed variables and genetic algorithm

Process Quality Control Through Bayesian Optimization with Adaptive Local Convergence

Bayesian Quality-Diversity approaches for constrained optimization problems with mixed continuous, discrete and categorical variables

High-dimensional Bayesian optimization using low-dimensional feature spaces

Multi-fidelity Bayesian Optimization in Engineering Design

Uncertainty-aware mixed-variable machine learning for materials design

Hierarchical Gaussian Process-Based Bayesian Optimization for Materials Discovery in High Entropy Alloy Spaces

Bayesian optimization with adaptive surrogate models for automated experimental design

Bayesian Optimization in Materials Science: A Survey

A Sequential Sampling-Based Bayesian Numerical Method for Reliability-Based Design Optimization

Bayesian optimization of variable-size design space problems

Data-Driven Design via Scalable Gaussian Processes for Multi-Response Big Data With Qualitative Factors

Materials Design using an Active Subspace Batch Bayesian Optimization Approach

Bayesian Optimization for Multi-Objective High-Dimensional Turbine Aero Design

Accelerated Development of Multicomponent Alloys in Discrete Design Space Using Bayesian Multi-Objective Optimisation

High-dimensional mixed-categorical Gaussian processes with application to multidisciplinary design optimization for a green aircraft

Mixed-Variable Concurrent Material, Geometry, and Process Design in Integrated Computational Materials Engineering

Accelerating material discovery with a threshold-driven hybrid acquisition policy-based Bayesian optimization