Physics Simulator Memory Footprint Reduction via Time Reversal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional design techniques for electromagnetic devices are often inefficient, relying on guesswork and requiring adjustment of numerous parameters, which becomes impractical as device complexity and feature sizes decrease, leading to a need for optimized design methods that can handle large state sizes and simulation time steps within reduced memory footprints.
Innovation Solution
A physics simulator that performs first-principles based design and optimization using operational and adjoint simulations, recording attenuated field values at absorbing boundaries to reduce memory footprint, allowing for the recovery of field responses without storing every point in time and space, and utilizing compressed representations and subsampling to minimize storage requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional design techniques are used with guess and check method, then device design can be performed, but the design efficiency is low and becomes impractical as device complexity increases
Solution Approach 1:
The patent replaces conventional mechanical guess-and-check design methods with a physics-based simulation system that uses first-principles calculations to automatically optimize device parameters. The system substitutes iterative manual adjustment with automated computational optimization, achieving high efficiency even for devices with billions of parameters.
2Measurement precision
If full field response data is stored for adjoint simulation, then accurate gradient calculation is achieved, but memory footprint becomes prohibitively large
Solution Approach 1:
The patent extracts and stores only the essential boundary field data needed for adjoint simulation, rather than storing the complete field response throughout the entire simulation domain. By taking out only the critical boundary information, the system achieves accurate gradient calculation with dramatically reduced memory requirements.
Solution Approach 2:
The patent inverts the conventional approach by performing adjoint simulation with time-reversed boundary conditions. Instead of forward propagating full field data and storing it, the system backward propagates adjoint fields from boundaries, recovering necessary field responses on-demand without storing complete time-space data.
3Quantity of substance
If time-reversed boundary fields are replayed to recover field response, then memory usage is reduced, but computational steps are increased
Solution Approach 1:
The patent performs preliminary recording of boundary field data during the forward simulation phase. This pre-captured boundary information is then reused in the adjoint simulation, eliminating the need to recompute or store extensive field data. The preliminary action of recording boundaries enables efficient recovery of field responses without proportional increases in computational time.
Data Source
AI summary
A system, apparatus, and method for optimizing structural parameters of a physical device are described. The method includes receiving an initial description of the physical device describing the structural parameters within a simulated environment. The method further includes performing a simulation of the physical device in response to an excitation source to determine a performance metric of the physical device. The simulation environment includes one or more absorbing boundaries for attenuation of an output of the excitation source during the simulation. The method further includes recording attenuated field values of the simulated environment associated with the attenuation during the simulation. The method further includes determining a loss metric based on a difference between the performance metric and a target performance metric, backpropagating the loss metric using the attenuated field values, and generating a revised description of the physical device by updating the structural parameters to reduce the loss metric.