Journal
PHYSICS OF FLUIDS
Volume 35, Issue 8, Pages -Publisher
AIP Publishing
DOI: 10.1063/5.0161301
Keywords
-
Categories
Ask authors/readers for more resources
The analysis of interactions between micro-particles and carrier gas is crucial for studying micro-particle behavior, especially in fuel spray and spray cooling. However, optical imaging techniques have limitations that challenge existing testing methods in achieving high capturing capability and a large field of view for micro-particles simultaneously. This study proposes a technique that integrates dual-view lenses and a cGAN-ResNet joint algorithm to accurately measure micro-particles in a large field of view. The results show that the proposed method outperformed traditional methods and significantly increased the field of view.
The analysis of interactions between micro-particles and carrier gas is a critical aspect in the study of micro-particle behavior, particularly in fuel spray and spray cooling. However, optical imaging techniques face inherent limitations that pose challenges to existing testing methods in achieving high capturing capability for micro-particles and a large field of view simultaneously. The current study proposes a Dual-view Wide-field High-precision Particle Sizing Technique that integrates hardware (dual-view lenses) and software (cGAN-ResNet joint algorithm). It aims to achieve accurate measurements of micro-particles in a large field of view. Our innovative approach involves simultaneous capture using dual-view lenses, where the smaller view lens aims to achieve high-resolution images. By employing machine learning techniques, we establish correspondences between droplets within the overlapping region of the two different-resolution images. This allows us to reconstruct the droplet information with high resolution for all droplets within the larger field of view, enabling accurate measurement of droplet characteristics across a wide field. We created synthetic datasets using the microSIG program to emulate real-world scenarios and validate our algorithm's accuracy and generalization. The results indicate that our method outperformed traditional adaptive threshold methods and significantly increased the field of view by several folds. Our algorithm has a key feature of real-time learning, thereby allowing it to be adaptable to datasets other than those used in this study and their derivatives. Our study shows that the fusion of multiple deep learning techniques is promising for accurately reconstructing and rapidly measuring micro-particles with a large field of view.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available