Hybrid Machine Learning Methods for Cavitation Erosion Measurements & Predictions
Abstract
Cavitation erosion causes significant damage to hydraulic and naval systems, from spillways of dams, to pump impellers, liquid pipes, bearings, as well as ship propeller and rudders. The erosion is caused by pitting of surfaces by high speed micro-jets developingas cavitation bubbles collapse rapidly near solid surfaces. The erosive power is augmented significantly when an entire cloud of bubbles collapses synchronously. Design of systems for which cavitation erosion is predictable requires understanding of: the extent and type of erosive cavitation phenomena and their relationships to the flow and pressure fields involved; the location, concentration, and size distribution of collapsing bubbles; the forces applied by these bubbles individually and as a cloud; the response of different materials to these forces after prolonged exposure; and feasible strategies for mitigating the adverse effects of the erosion. Owing to the complexity of the processes and multiple fields involved, prior research related to cavitation erosion focused on elements of the entire picture. However, the increasing range of applications of machine learning (ML) brings up the possibility of adopting this methodology for characterizing and developing mitigation strategies for cavitation erosion. The underlying idea is to develop statistical algorithms that would be trained by available or new data and use them to analyze the entire cavitation erosion process. Some of the training data and ML tools might already be available and others are still missing. Hence, we propose to host a workshop that will bring together experts from academia government, and industry with relevant experience in cavitation, material erosion, and machine learning to discuss and recommend potential courses of actions needed for implementing ML in the cavitation erosionfield. The expected number of participants will be in the 30-50 range, with 30 of them serving as invited speakers, including 15-20experts in the different forms of bubble dynamics and cavitation, 5-7 experts in material erosion, and 7-13 experts in machine learning. This workshop will include both keynote introductions to the different fields involved, followed by detailed technical presentations, and discussions about specific topics. The proposed topics include: (i) introduction machine learning aimed at introducing this field to the cavitation community; (ii) introductions to bubble (and bubble cloud) dynamics and cavitation erosion, (iii) introduction to material erosion; (iv) required machine learning algorithms; (v) type and format of data required for training of ML algorithms to predict cavitation erosion; and (vi) missing data that needs to be collected experimentally or numerically in bubble dynamics, cloud cavitation, detection methods, forces generated by bubble or cloud collapse near boundaries, response of materials to these forces including effects of prolonged exposure, and data integration. The two-day workshop will take place in August 2024 in the new Johns Hopkins University Bloomberg Center located Washington DC. Each series of presentations will be followed by discussions, which will be then summarized by the session chairs, and the results will be integrated into a report. These summaries will be made available to the participants for download from a dedicated workshop web site. Approved for public release
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 15, 2024
- Source ID
- N000142412238
Entities
People
- Joseph Katz
Organizations
- Johns Hopkins University
- Office of Naval Research
- United States Navy