What is it about?
In our previous study, we trained an AI to classify images of splashing and nonsplashing drops. We discovered that other than ejected secondary droplets, the AI also checks the contour of the drop’s main body. If the contour is high, the AI will classify the drop as a splashing drop. Such findings have not been reported prior to our previous study. In this study, we trained the AI to classify image sequences of splashing and nonsplashing drops and checked the classification process. The results show that the AI relies on the images of earlier timing to identify a splashing drop because the ejected secondary droplets accumulated around the drop at the earlier timing and became scattered all over the place at the later timing.
Featured Image
Photo by Mohan Murugesan on Unsplash
Why is it important?
This study is a continuation of our previous study which the goal is to use AI to observe splashing and nonsplashing drops and find new perspectives from it. From this study, we understand the bias that could occur when we study a phenomenon using an AI, which we want to avoid as we continue with our study.
Perspectives
AI is a very brilliant way of programming. It is very fascinating to analyze and to understand how an AI works. Although it is still the early stage of its development, I believe it has the potential to be the third main approach to study fluid dynamics after experiment and numerical simulation.
Dr Jingzu Yee
Tokyo Noko Daigaku
Read the Original
This page is a summary of: Features of a Splashing Drop on a Solid Surface and the Temporal Evolution extracted through Image-Sequence Classification using an Interpretable Feedforward Neural Network, June 2022, American Institute of Aeronautics and Astronautics (AIAA),
DOI: 10.2514/6.2022-4174.
You can read the full text:
Resources
Contributors
The following have contributed to this page
