VIP20

本课程属于：028-CAE Data Science & AIML English Technical Courses · 共 20 门相关课程，组合学习更完整。

查看所属VIP全部课程 →

← 返回所属目录咨询 / 购买

Machine Learning Surrogates and Robust CAE Optimization

30 Chapters

View Course Samples Online

Course introduction

role of surrogates in CAE, motivation for robustness, and overview of the surrogate-based optimization workflow.

CAE simulation fundamentals for surrogate modeling

governing equations, discretization, boundary conditions, and output quantities of interest.

Design of experiments for CAE

sampling strategies (Latin hypercube, Sobol, Halton), space-filling criteria, and handling of constrained design spaces.

Data generation and management

running batch simulations, extracting results, cleaning data, and structuring datasets for surrogate training.

generationmanagement

Surrogate model families

polynomial regression, radial basis functions, Gaussian process regression, and support vector regression for CAE outputs.

Surrogatefamilies

Neural network surrogates

fully connected networks, convolutional networks for field data, and physics-informed neural networks for CAE.

surrogatesconnected

Dimensionality reduction for high-dimensional CAE outputs

proper orthogonal decomposition, autoencoders, and latent space surrogates.

Dimensionalityreduction

Surrogate model training and validation

loss functions, cross-validation, hyperparameter tuning, and overfitting mitigation.

Surrogatetraining

Error metrics and uncertainty quantification for surrogates

RMSE, R², prediction intervals, and confidence bounds.

uncertaintyquantification

Sensitivity analysis and feature importance

global sensitivity methods (Sobol indices, Morris screening) and surrogate-based attribution.

Sensitivityanalysis

Robust design concepts

uncertainty types (aleatory and epistemic), robustness metrics, and formulation of robust optimization problems.

uncertaintyaleatory

Surrogate-based optimization

infill criteria, expected improvement, and efficient global optimization with CAE solvers.

Surrogate-basedoptimization

Multi-objective robust optimization

Pareto front generation, trade-off analysis, and surrogate-assisted evolutionary algorithms.

Multi-objectiveoptimization

Reliability-based design optimization

probabilistic constraints, reliability index, and surrogate-based reliability analysis.

Reliability-basedoptimization

Handling noise and variability in CAE

stochastic simulation, metamodeling of noisy responses, and robust surrogate training.

Handlingvariability

Active learning and adaptive sampling

query strategies, uncertainty sampling, and balancing exploration and exploitation.

adaptivesampling

Surrogate model calibration and validation against CAE

validation metrics, model discrepancy, and iterative refinement.

Surrogatecalibration

Integration of surrogates with CAE workflows

scripting, automation, and coupling with commercial solvers (e.g., ANSYS, Abaqus).

Integrationsurrogates

aerodynamic shape optimization using surrogate models and robust design against operating uncertainties.

aerodynamicoptimization

structural optimization under material and load uncertainties with surrogate-based reliability analysis.

structuraloptimization

thermal management design using machine learning surrogates and robust optimization.

managementsurrogates

crashworthiness optimization with surrogate models and variability in impact conditions.

crashworthinessoptimization

Surrogate model selection and ensemble methods

stacking, bagging, and model averaging for improved robustness.

Surrogateselection

Transfer learning and multi-fidelity surrogates

combining low- and high-fidelity CAE data for efficient robust optimization.

Transfermulti-fidelity

Scalability and deployment

surrogate model export, real-time prediction, and integration into design dashboards.

Scalabilitydeployment

Verification and validation of surrogate-based robust designs

confirmation runs, uncertainty propagation, and robustness assessment.

Verificationvalidation

Practical implementation

building a surrogate-based robust optimization pipeline with Python and CAE automation.

implementationbuilding

Advanced topics

Bayesian optimization for robust design, deep Gaussian processes, and generative models for design exploration.

AdvancedBayesian

Industrial challenges

data scarcity, high-dimensionality, solver noise, and organizational adoption of surrogate workflows.

Industrialchallenges

Capstone project

end-to-end surrogate modeling and robust optimization for a selected CAE application, including documentation and presentation.

Capstoneend-to-end