Finite Element Model Calibration Using Synthetic Simulation Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional design processes in engineering and condition monitoring face challenges in efficiently analyzing and calibrating finite element models due to the need for detailed analysis of numerous design variants and the interaction of many parameters, often relying on reduced metadata which hinders the use of previous valuable design decisions and increases computational complexity.
Innovation Solution
A method for calibrating finite element models using measured data points involves generating simulations based on varying input parameters, calculating mathematical operators to preserve distance metrics, projecting simulations into a lower-dimensional space using eigenvectors, and iteratively generating synthetic simulations that fit measured data points, reducing computational complexity and effort.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional design processes use reduced metadata for analysis, then the analysis process is simpler, but the ability to perform detailed analysis of simulations and learn from previous designs is inhibited
Solution Approach 1:
The patent creates synthetic simulations that replicate the behavior and characteristics of actual simulations without requiring the full computational burden. These synthetic copies preserve the essential information needed for calibration and analysis, enabling detailed examination without processing the complete original dataset.
Solution Approach 2:
The patent introduces synthetic simulations as an intermediary between the original complex simulations and the calibration process. These synthetic representations serve as a mediator that retains sufficient detail for meaningful analysis while being computationally efficient to handle and process.
2Measurement precision
If thousands of finite element design variants are analyzed in detail, then comprehensive design evaluation is achieved, but the computational complexity and time required increase significantly
Solution Approach 1:
Instead of analyzing all thousands of simulations in detail, the patent generates a smaller set of synthetic simulations that capture the essential behavior and variability. These synthetic copies enable comprehensive evaluation at a fraction of the computational cost and time required for full detailed analysis of all variants.
Solution Approach 2:
The patent performs partial analysis by focusing computational resources on generating and analyzing a representative subset of synthetic simulations rather than exhaustively processing all design variants. This selective approach achieves sufficient comprehensiveness for calibration purposes while dramatically reducing computational time.
3Manufacturing precision
If iterative calibration of finite element models is performed using conventional methods, then accurate fitting to measured data is achieved, but the computational effort and complexity increase
Solution Approach 1:
The patent uses synthetic simulations as proxies for actual simulations during the iterative calibration process. These synthetic copies maintain the essential characteristics needed for accurate fitting to measured data while being computationally simpler to generate and evaluate, thereby reducing the overall computational complexity of the calibration workflow.
Data Source
Figure 1
Figure 2a~2b
Figure 3
AI summary
A method for evaluating a plurality of data aggregations is provided. A data aggregation comprising a plurality of data points. The method comprises calculating a mathematical operator for one of the plurality of data aggregations based on a metric for the data points of the one data aggregation. The operator defines a transformation that preserves the metric. Further, the method comprises calculating eigenvectors of the mathematical operator, and projecting the remaining data aggregations of the plurality of data aggregations into a space spanned by the eigenvectors.