Fast Spectral Formulations of Thin Plate Laser Heating with GPU Implementation
Egileak: Oscar Ruiz
Data: 14.01.2020
Abstract
In the context of numerical methods, the problem of frequency-domain (spectral) simulations is crucial for the solution of Partial Differential Equations. Fast Fourier Transform (FFT) algorithms significantly reduce the computational cost of such simulations and enable parallelization using Graphics Processing Units (GPUs). In the particular subdomain of laser heating/cutting of rectangular metal plates, fast simulation is required for tool path planning, parameter optimization and additive manufacturing. The currently used methods include frequency-domain analytic solutions for single-beam and multi-beam laser heating. However, the problem of formulating these spectral problems in terms of Fourier methods and implementing them in efficient manner remains. To overcome these limitations, this article presents two different schemes that translate the problem of laser beam heating of metal plates into equivalent FFT problems. The results show significant improvements in terms of executions times, being 100× faster than current state-of-the-art algorithms. Future work needed involves the inclusion of stress analysis, complex plate geometries and non-constant material properties for the plate.
BIB_text
title = {Fast Spectral Formulations of Thin Plate Laser Heating with GPU Implementation},
pages = {133-140},
keywds = {
Laser beams;Heating systems;Discrete Fourier transforms;Mathematical model;Computational modeling;Heat transfer;Laser modes;Fast Fourier Transform;Discrete Sine Transform;heat transfer;laser heating;thin metal plate;GPU
}
abstract = {
In the context of numerical methods, the problem of frequency-domain (spectral) simulations is crucial for the solution of Partial Differential Equations. Fast Fourier Transform (FFT) algorithms significantly reduce the computational cost of such simulations and enable parallelization using Graphics Processing Units (GPUs). In the particular subdomain of laser heating/cutting of rectangular metal plates, fast simulation is required for tool path planning, parameter optimization and additive manufacturing. The currently used methods include frequency-domain analytic solutions for single-beam and multi-beam laser heating. However, the problem of formulating these spectral problems in terms of Fourier methods and implementing them in efficient manner remains. To overcome these limitations, this article presents two different schemes that translate the problem of laser beam heating of metal plates into equivalent FFT problems. The results show significant improvements in terms of executions times, being 100× faster than current state-of-the-art algorithms. Future work needed involves the inclusion of stress analysis, complex plate geometries and non-constant material properties for the plate.
}
date = {2020-01-14},
}
