Automatic classification of hydrodynamic phenomena using their acoustic signature: The example of bubble entrainment during a drop impact. (July 2022)

• Record Type: Journal Article
• Authors: Gillot, G.; Simon, L.; Génevaux, J.-M.; Benyahia, L.
• Is Part Of: Applied acoustics. Volume 196(2022)
• Publication Date: 2022-07
• Subjects: Drop impact -- Bubble entrainment -- Classification -- Acoustics -- Hydrodynamics -- Signal processing; Acoustical engineering -- Periodicals; Periodicals; 620.2
• DOI: 10.1016/j.apacoust.2022.108868
• Languages: English
• ISSNs: 0003-682X

Abstract

Highlights: Acoustic signature of bubble entrainment mechanisms during drop impact are studied. Specific features are extracted from the acoustic signatures to discriminate them. A machine learning algorithm is used to classify the bubble entrainment mechanisms. An accuracy of 82% is obtained. Abstract: When a liquid drop impacts on a liquid surface, bubbles can be entrained in the liquid by several hydrodynamic mechanisms, generating an acoustic signal in the liquid and in the air. Each of the five mechanisms reported displays a typical acoustic signature, both in the liquid and in the air. Even if the differences between these signatures are subtle, it is possible to distinguish them. A detailed study of a labeled database made of acoustic signals from different bubble mechanisms allows to extract relevant descriptors of the signatures. A classification method is then proposed, using a KNN algorithm and shows great performances to classify the acoustic signals. Thus, the method can be used to study the bubble entrainment hydrodyanmics during drop impacts, only from the acoustic signals and thus without the help of a high-speed camera.