Back to results
US 2026/0278215 A1Published application
Methods and systems for the estimation of the computational cost of simulation
AssigneeSynopsys, Inc.
InventorsKyle Kosic, Anil Sehgal, Scott McClennan, Joshua Oster-Morris, Ryan Diestelhorst, David M. Freed, Ian Campbell
Life of this patent
Estimating the term…
Abstract
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.
Claims20
A method for estimating the computational cost of simulation comprising, by one or more processors:
- (a) creating a model feature dataset based on a geometry metadata training dataset and a simulation metadata training dataset, wherein the geometry metadata training dataset and the simulation metadata training dataset correspond to inputs, outputs, and configurations of a plurality of previously performed simulations;
- (b) using the model feature dataset:
- (i) training a first machine learning model based on the geometry metadata training dataset;
- and
- (ii) training a second machine learning model based on the simulation metadata training dataset;
- (c) for a present simulation:
- (i) creating a geometry feature dataset based on a present geometry metadata associated with the present simulation;
- and
- (ii) creating a simulation feature dataset based on a present simulation metadata associated with the present simulation;
- (d) inputting the geometry feature dataset into the first machine learning model, and inputting the simulation feature dataset into the second machine learning model;
- and
- (e) predicting, using the first and second machine learning model, a computational cost characteristic associated with performing the present simulation, wherein the computational cost characteristic is selected from a group consisting of a memory usage, a total runtime, a time for meshing, a time for preprocessing, a time for solve, and a time for post-processing, and wherein the present simulation is a computer-aided engineering process for simulating characteristics of a physical object or physical system represented by a virtual counterpart within the computer-aided engineering process.
9 more elements
A system for estimating the computational cost of simulation comprising one or more processors configured to:
- (a) create a model feature dataset based on a geometry metadata training dataset and a simulation metadata training dataset, wherein the geometry metadata training dataset and the simulation metadata training dataset correspond to inputs, outputs, and configurations of a plurality of previously performed simulations;
- (b) using the model feature dataset:
- (i) train a first machine learning model based on the geometry metadata training dataset;
- and
- (ii) train a second machine learning model based on the simulation metadata training dataset;
- (c) for a present simulation:
- (i) create a geometry feature dataset based on a present geometry metadata associated with the present simulation;
- and
- (ii) create a simulation feature dataset based on a present simulation metadata associated with the present simulation;
- (d) input the geometry feature dataset into the first machine learning model, and input the simulation feature dataset into the second machine learning model;
- and
- (e) predict, using the first and second machine learning model, a computational cost characteristic associated with performing the present simulation, wherein the computational cost characteristic is selected from a group consisting of a memory usage, a total runtime, a time for meshing, a time for preprocessing, a time for solve, and a time for post-processing, and wherein the present simulation is a computer-aided engineering process for simulating characteristics of a physical object or physical system represented by a virtual counterpart within the computer-aided engineering process.
9 more elements
A method for estimating the computational cost of a present simulation comprising, by one or more processors, providing a present geometry metadata and a present simulation metadata to a remote system over a network, and receiving a computational cost characteristic associated with performing the present simulation from the remote system over the network;
- wherein the computational cost characteristic was created by the remote system by:
- (a) creating a model feature dataset based on a geometry metadata training dataset and a simulation metadata training dataset, wherein the geometry metadata training dataset and the simulation metadata training dataset correspond to inputs, outputs, and configurations of a plurality of previously performed simulations;
- (b) using the model feature dataset:
- (i) training a first machine learning model based on the geometry metadata training dataset;
- and
- (ii) training a second machine learning model based on the simulation metadata training dataset;
- (c) for a present simulation:
- (i) creating a geometry feature dataset based on a present geometry metadata associated with the present simulation;
- and
- (ii) creating a simulation feature dataset based on a present simulation metadata associated with the present simulation;
- (d) inputting the geometry feature dataset into the first machine learning model, and inputting the simulation feature dataset into the second machine learning model;
- and
- (e) predicting, using the first and second machine learning model, a computational cost characteristic associated with performing the present simulation, wherein the computational cost characteristic is selected from a group consisting of a memory usage, a total runtime, a time for meshing, a time for preprocessing, a time for solve, and a time for post-processing, and wherein the present simulation is a computer-aided engineering process for simulating characteristics of a physical object or physical system represented by a virtual counterpart within the computer-aided engineering process;
- and
- wherein the computational cost characteristic is selected from a group consisting of a memory usage, a total runtime, a time for meshing, a time for preprocessing, a time for solve, and a time for post-processing, and wherein the present simulation is a computer-aided engineering process for simulating characteristics of a physical object or physical system represented by a virtual counterpart within the computer-aided engineering process.
12 more elements
Dependencies are inferred from the claim text.