Resilient Maintenance of Maritime Unmanned Surface Vehicles Using Machine Learning and Optimization
Abstract
Maritime Unmanned Surface Vehicles (USVs) have been used for a wide range of applications, including protection, surveillance, and s,earching on the sea. One key advantage of these vehicles is that they are crewless. This means that these vehicles must be self-awar,e of their health state, able to anticipate and respond effectively without human intervention when and if there is an expected degr,adation in their operation or health condition. We propose to construct a hybrid approach for resilient maintenance of USV systems u,sing machine learning, causal models, and optimization techniques. We posit that such a hybrid approach can significantly improve th,e detection of precursors to the systems damage and provide better prognosis of the health state of the system. We propose to use t,he notion of resilience to indicate a condition-based maintenance scheme that can anticipate and respond to hazardous conditions b,y reconfiguring systems operational conditions in order to extend its remaining useful life. Our approach will integrate the underl,ying reliability engineering knowledge obtained by the widely known physics-based models of the system combined with publicly availa,ble data, including simulated sensor data, and when and if available, human inspection data, to gain insights into the USV systems,health conditions. The expected outcome of this effort will be a computer-based decision support tool that determines USV systems h,ealth conditions and automatically make corrective actions (reconfigure) for operational maintenance when system deterioration is de,tected.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 08, 2022
- Source ID
- N000142212459
Entities
People
- Shapour Azarm
Organizations
- Office of Naval Research
- United States Navy
- University of Maryland