What are Modeling Agents?

Most engineering design problems require simulation experiments to evaluate the objective and constraint functions when different design parameters are used. For example, to find the best wing shape, the airflow around the wing is often simulated for different shape parameters (length, curvature, material, etc.). For many practical problems, a single simulation can take minutes, hours, or even days to complete. Therefore, tasks like design optimization, design space search, sensitivity analysis, and what-if analysis that require thousands or even millions of simulations, and it is impossible to directly solve the original model.

most

The main challenge of this method is how to use as few high-precision model solving as possible to build a proxy model that is as accurate as possible. This process consists of three interwoven steps.

• Sampling selection (also known as sequential design, optimal experimental design (OED), or active learning)

• Establishing a proxy model and optimizing model parameters (deviation variance tradeoffs)

• Evaluation of proxy model accuracy

The accuracy of the proxy model is related to the number and location of sampling points (high-precision simulation experiments) in the design space. different

The most reused proxy models are polynomial response surface method, kriging method, gradient enhanced kriging method (GEK), support vector machine, space mapping, and

Recent evolutionary algorithms based on comparison proxy models (such as sorted support vector machines), such as CMA-ES, allow some consistency of the optimizer assisted by the proxy model to be preserved:

• The consistency of the function's monotonic transformation (

  Design optimization and design space approximation (also become

  • Linear approximation

  • Response surface method

  • Kriging

  • Gradient Enhanced Kriging (GEK)

  • OptiY

  • Spatial mapping

  • Agent endpoint

  • Proxy data

  • Adaptive approximation

  • Computer experiment

  • model