Status
Principle Investigator Contact Information
Name: Chris Bassett
Address: Senior Mechanical Engineering, University of Washington Applied Physics Laboratory
Email: cbassett@uw.edu
For commercial inquiries, please contact James Joslin at MarineSitu (james@marinesitu.com).
For scientific or regulatory inquiries, contact Brian Polagye (bpolagye@uw.edu) or Chris Bassett (cbassett@uw.edu).
Description
The project developed and demonstrated variants of the Adaptable Monitoring Package (AMP). The AMP integrates active acoustic, passive acoustic, and optical sensors into a single instrumentation package that can be cabled to shore or operated autonomously. By simultaneously observing rare, but potentially significant, interactions between marine life and marine energy converters with multiple sensor modalities, detection and interpretation of such events is likely to be improved. Automatic detection and classification algorithms now allow the system to make continuous observations without incurring a “data mortgage” and automatic sensor control allows such observations to occur without biasing marine animal behaviour.
Video Archive: https://www.youtube.com/channel/UCqR-J-6LOLjsHCjO285jBbA/
Funding Source
US Department of Energy (Water Power Technologies Office), US Department of Defense (Naval Facilities Engineering Command)
Location of Research
Multiple; cabled system testing has been conducted at the University of Washington and Pacific Northwest National Laboratory’s Marine Science Lab in Sequim, WA, while autonomous system testing has been conducted at PacWave (previously Pacific Marine Energy Center South Energy Test Site) off Newport, OR and at the U.S. Navy Wave Energy Test Site in Kaneohe, HI. A related system, not developed by the project, was deployed with a small-scale tidal turbine in Sequim Bay, WA.
Project Aims
To develop and demonstrate an integrated instrumentation package that can be used in cabled or autonomous modes to study the interactions between marine life and marine energy converters.
Study Progress
- Endurance trial completed for cabled system in May 2016 (> 90% uptime over four-month period for prototype cabled system), with subsequent improvements over three month deployment in 2017.
- Automatic real-time detection and classification of “rare” targets (seals, diving birds, fish schools) in multibeam sonar data with high true positive rates (> 80%) and low false positive rates (< 20%).
- Post-processing identification of fish in optical camera data with acceptable true positive and false positive rates.
- Integration of PAMGuard with the system to automatically detect fish tags and simulated marine mammal vocalizations.
- Development and initial deployment of autonomous lander with duty cycle and ability to “wake up” in response to the presence of Vemco fish tags.
- Integration of an AMP with a wave energy converter. The “WAMP” draws power from the WEC to operate the sensor package and achieved an 84% uptime over a 3.5 month deployment, with an average power draw of 600 W.
- The 3G-AMP was integrated with the Oscilla Triton-C but was recovered after multiple years in the water before the WEC was deployed.
- A related AMP system was deployed on a small-scale, cross-flow tidal turbine in the inlet to Sequim Bay, WA for 141s in 2023-2024. It achieved an uptime greater than 95%.
Key Findings
- Cooperative target testing with drifting or towed objects at known position is effective at establishing sensor ranges and diagnosing sensor functionality.
- Passive acoustic detection of fish tags is likely to occur within the range of active acoustic instruments (e.g., multi-beam sonar, acoustic camera).
- Multibeam sonars capable of detecting marine mammals, fish schools, and individual fish to a range of 10 m. Different sonars have different detection capabilities and some are more easily interpretable by human reviewers, but machine learning classification outcomes are similar (i.e., computers perceive objects differently than humans).
- Sensor fusion across instruments on the platform helpful to improve manual and automatic classification.
- Without real-time target detection, it is unlikely that sufficient training could be collected for automatic tracking and classification algorithms without incurring a large data mortgage.
- Active sonars can produce sound at lower frequencies than their characteristic operating frequencies. This is unlikely to cause harm to marine animals, but could be detectable by animals and should be considered in study design.
- Data with fish, birds, and seals interacting with a tidal turbine when it was not moving were captured and analyzed. Seals and fish were also captured interacting with the moving rotor, including several collisions with fish.