Pre-Execution Simulation Cost Estimation with Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current numerical simulation tools lack the ability to accurately estimate computational cost prior to execution, especially for complex systems with billions of degrees of freedom, hindering the effective deployment of service-based simulation tools.
Innovation Solution
A machine learning model is trained using a database of initial settings and performance metrics from varied simulations to predict computational cost characteristics, such as core hours, based on model geometry and simulation metadata, enabling accurate cost estimation before simulation execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional numerical simulation tools are used to estimate computational cost, then accuracy of cost estimation is improved, but the time required for estimation increases to the extent that it takes as long as the simulation itself
Solution Approach 1:
The patent creates a digital twin or surrogate model of the simulation system using machine learning. This surrogate model copies the essential computational behavior of the full simulation but runs much faster, allowing rapid cost estimation without executing the actual expensive simulation. The surrogate model is trained on historical simulation data to replicate computational cost characteristics.
Solution Approach 2:
The patent performs preliminary training of the machine learning model using a database of historical simulation data before actual cost estimation is needed. This preliminary action establishes the surrogate model in advance, so that when new simulations need cost estimation, the pre-trained model can quickly provide accurate estimates without requiring time-consuming real-time analysis.
2Ease of manufacture
If service-based simulation tools are deployed without accurate cost feedback, then tool deployment is simplified, but the ability to optimize resource allocation and simulation settings is reduced
Solution Approach 1:
The patent implements a feedback mechanism that provides accurate computational cost estimates to users before simulation execution. This feedback loop allows users to see the estimated cost (in terms of core hours or computational resources) and adjust their simulation settings accordingly, optimizing resource allocation. The feedback is provided by the machine learning surrogate model which quickly predicts costs based on simulation inputs.
Solution Approach 2:
The patent performs preliminary cost estimation using the trained machine learning model before the actual simulation runs. This preliminary action allows service-based tools to deploy with cost transparency, enabling users to make informed decisions about resource allocation and simulation settings in advance, thereby improving overall productivity without complicating tool deployment.
Data Source
AI summary
Estimating the computational cost of simulation using a machine learning model. An example method includes inputting a feature data set into a machine learning model. The feature data set includes model geometry metadata and simulation metadata. The method further includes predicting, using the machine learning model, a computational cost characteristic for a simulation process.