Trained Model Inference for Mesh Simulation Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Simulations using fine mesh data require high computational complexity and time, and often lack a corresponding coarse mesh model, making it inefficient to infer results directly from fine to coarse mesh data without pre-existing coarse mesh models.
Innovation Solution
A computer program that generates a coarse mesh model from fine mesh data using a trained model to reduce the number of edges, allowing for faster simulation and inference of physical quantities using a second trained model, thereby reducing computational complexity and enabling fast inference of simulation results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fine mesh data is used for simulation, then analysis accuracy is improved, but computational complexity and time increase
Solution Approach 1:
The patent introduces a trained model as an intermediary between coarse mesh simulation results and fine mesh simulation results. The model learns the mapping relationship from coarse to fine mesh results through training, then uses this learned relationship to infer fine mesh results without actually performing computationally expensive fine mesh simulations. This intermediary approach resolves the contradiction by providing accurate fine mesh-like results through the trained model rather than direct fine mesh computation.
Solution Approach 2:
The patent creates a coarse mesh model that copies the essential structural and geometric features of the fine mesh model but with reduced complexity. By performing simulations on this simplified copy and then using a trained model to infer the corresponding fine mesh results, the system achieves accurate analysis without the computational burden of direct fine mesh simulation.
2Measurement precision
If fine mesh data is used for simulation, then analysis accuracy is improved, but simulation time increases
Solution Approach 1:
The patent performs preliminary training of the model using pairs of coarse and fine mesh simulation results. Once trained, the model can rapidly infer fine mesh results from coarse mesh simulations without requiring actual fine mesh simulations for each new case. This preliminary action of model training resolves the time contradiction by establishing a reusable inference mechanism that avoids repeated expensive fine mesh computations.
Solution Approach 2:
The trained model serves as an intermediary that translates coarse mesh simulation results into accurate fine mesh results. Instead of performing time-consuming fine mesh simulations directly, the system uses the trained model to infer the desired fine mesh results, dramatically reducing simulation time while maintaining accuracy.
3Device complexity
If coarse mesh data is used for simulation, then computational complexity is reduced, but analysis accuracy deteriorates
Solution Approach 1:
The trained model acts as an intermediary that enhances coarse mesh simulation results to achieve fine mesh-level accuracy. The model learns the discrepancy patterns between coarse and fine mesh results during training, then applies this learned knowledge to correct and refine coarse mesh results, resolving the accuracy contradiction while maintaining low computational complexity.
Solution Approach 2:
The patent changes the parameter representation by using a trained model to transform coarse mesh results into fine mesh-equivalent results. Rather than changing the mesh resolution itself, the system changes the output parameters through model-based inference, achieving high accuracy from low-resolution input data.
4Loss of time
If coarse mesh data is used for simulation, then simulation time is reduced, but analysis accuracy deteriorates
Solution Approach 1:
The trained model serves as an intermediary that rapidly transforms coarse mesh simulation results into accurate fine mesh results. This approach maintains the speed advantage of coarse mesh simulations while eliminating the accuracy disadvantage through model-based inference, effectively resolving the time-accuracy trade-off.
Solution Approach 2:
The patent performs preliminary training of the inference model using paired coarse and fine mesh data. After this preliminary action, the system can quickly infer accurate fine mesh results from coarse mesh simulations without sacrificing accuracy, resolving the contradiction between simulation speed and accuracy.
Data Source
AI summary
A memory stores a first trained model for determining features of mesh data, which includes a plurality of nodes and a plurality of edges connecting them, by removing some of the edges from the mesh data. A processor generates, from second mesh data, first mesh data having a smaller number of edges than the second mesh data by use of the first trained model. The processor generates, by running a simulation using the first mesh data, first simulation result data that indicates a physical quantity of an object represented by the first mesh data. The processor infers, from the first mesh data and the first simulation result data and by use of a second trained model, second simulation result data that would be obtained by running the simulation using the second mesh data.