Multi-Fidelity Optimization Surrogate Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High fidelity aerodynamic simulations are computationally expensive, and existing optimization methods face challenges when dealing with multi-fidelity optimization problems involving non-smooth or discontinuous objective and constraint functions, especially when derivatives are unavailable or unreliable.
Innovation Solution
A computer-implemented search and poll method that constructs and optimizes surrogates of objective and constraint functions, using lower fidelity simulations in ascending order to reduce the number of trial points, and evaluates them with high fidelity simulations to efficiently solve multi-fidelity optimization problems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high fidelity simulation is used to evaluate objective and constraint functions, then solution accuracy is improved, but computational cost increases
Solution Approach 1:
The patent creates surrogate models (copies) of the high-fidelity simulation that approximate its behavior at lower computational cost. These surrogates are trained on a subset of high-fidelity data and then used to evaluate trial points, replacing expensive direct high-fidelity calls while maintaining solution accuracy.
Solution Approach 2:
The patent performs preliminary high-fidelity simulations to train the surrogate models before the optimization process begins. This preliminary action creates a knowledge base that enables subsequent low-cost evaluations, avoiding the need to run high-fidelity simulations for every trial point during optimization.
2Productivity
If lower fidelity simulation is used to reduce computational cost, then speed is improved, but solution accuracy deteriorates
Solution Approach 1:
The patent introduces surrogate models as intermediaries between low-fidelity simulations and the optimization process. These surrogates are trained on high-fidelity data but can be evaluated at low cost, acting as a mediator that provides high-fidelity accuracy without high-fidelity computational expense.
Solution Approach 2:
The patent uses a partial approach by evaluating only a subset of trial points with high-fidelity simulations (those identified as promising by the surrogate models). This partial action maintains accuracy for critical evaluations while achieving speedup through selective high-fidelity usage.
3Adaptability or versatility
If search and poll method is applied to multi-fidelity optimization problems with more than two fidelity levels, then versatility is improved, but device complexity increases
Solution Approach 1:
The patent develops a universal search and poll framework that handles any number of fidelity levels (not just two). The method uses a common surrogate modeling approach and trial point generation strategy that works across multiple fidelity levels, making the solution versatile and adaptable to different problem configurations.
Data Source
AI summary
A computer-implemented search and poll method can iteratively solve a multi-fidelity optimization problem including an objective function and any constraints. A search step of the method includes constructing and optimizing surrogates of the objective function and any constraints to identify a new set of trial points, and running lower fidelity simulations in ascending order to reduce a number of trial points in the new set. The search step further includes evaluating the reduced number of trial points with the objective function and any constraints using a high fidelity simulation.