Articles
GPU computing for simulation of explicit droplet movement in orchard spray practices
Article number
1395_51
Pages
387 – 392
Language
English
Abstract
Computer virtualisation has been expected as an alternative to physical objects for studying agricultural spray practices.
Some computational models allow the simulation of explicit movement of individual droplet from nozzles to targets.
However, the utilisation of such simulations has been limited due to programming and computing constraints.
The conventional serial programs can only use one compute core at a time, regardless of the availability of multiple cores in a computer (including powerful supercomputers with thousands of cores). Consequently, the complexity and details of simulation are often compromised for efficient computing, or vice versa.
The oversimplification of virtual experimentation results in reduced accuracy and reliability, while the prolonged implementation times are also intolerable by end users.
There have been studies using commercial CFD (computational fluid dynamics) software tools equipped with robust computing capabilities for simulation of spray processes, but such platforms are usually very costly and require programming skills and/or good knowledge of fluid dynamics, which prevents its adoption by the majority of agronomists and growers.
In this research, we developed a GPU-based high performance computing approach for spray simulation where the explicit movement of every droplet can be tracked in detail.
Using a commodity-level GPU card with an ordinary desktop computer, the approach has been capable of completing an indivisible simulation task within an hour, which would otherwise require about 20 h by a serial program based on the same model.
The approach not only enables good efficiency and low variations, but also has been tested on an ordinary desktop computer with commodity components, to ensure its affordability and accessibility.
In future, the technology can also be deployed on high-end computing clusters to further boost the acceleration and enable remote access for large-scale applications.
Some computational models allow the simulation of explicit movement of individual droplet from nozzles to targets.
However, the utilisation of such simulations has been limited due to programming and computing constraints.
The conventional serial programs can only use one compute core at a time, regardless of the availability of multiple cores in a computer (including powerful supercomputers with thousands of cores). Consequently, the complexity and details of simulation are often compromised for efficient computing, or vice versa.
The oversimplification of virtual experimentation results in reduced accuracy and reliability, while the prolonged implementation times are also intolerable by end users.
There have been studies using commercial CFD (computational fluid dynamics) software tools equipped with robust computing capabilities for simulation of spray processes, but such platforms are usually very costly and require programming skills and/or good knowledge of fluid dynamics, which prevents its adoption by the majority of agronomists and growers.
In this research, we developed a GPU-based high performance computing approach for spray simulation where the explicit movement of every droplet can be tracked in detail.
Using a commodity-level GPU card with an ordinary desktop computer, the approach has been capable of completing an indivisible simulation task within an hour, which would otherwise require about 20 h by a serial program based on the same model.
The approach not only enables good efficiency and low variations, but also has been tested on an ordinary desktop computer with commodity components, to ensure its affordability and accessibility.
In future, the technology can also be deployed on high-end computing clusters to further boost the acceleration and enable remote access for large-scale applications.
Authors
J. Cao, D. Abramson, D. Carr, L. Han
Keywords
digital agriculture, digital twins, orchard management, pesticide, sprayer
Groups involved
- Division Temperate Tree Nuts
- Division Vine and Berry Fruits
- Division Temperate Tree Fruits
- Division Precision Horticulture and Engineering
- Division Plant-Environment Interactions in Field Systems
- Division Tropical and Subtropical Fruit and Nuts
- Working Group Precision Management of Orchards and Vineyards
Online Articles (59)