MIT's Intelligent Drag-and-Drop Pool

2019-12-11

The Fluid Mechanics Laboratory of the MIT Sea Grant College Program has developed an automated experimental device called the Intelligent Towing Tank (ITT), which can autonomously, adaptively execute, analyze, and design experiments exploring vortex-induced vibrations (VIV). This is crucial for the engineering design of offshore structures such as marine drilling risers that connect underwater oil wells to the surface. Under the guidance of active learning, ITT can perform a series of experiments: the computer selects the parameters for the next experiment using an "explore-exploit" approach, significantly reducing the number of experiments required to explore and control the complex forces of VIV. Vortex-induced vibration is an inherent nonlinear motion that occurs in irregular cross flows and affects structures. Researchers report that the number of experiments conducted by the International Telephone and Telegraph Company is comparable to the total number of VIV experiments conducted worldwide to date. Blindly measuring each parameter 10 times in an eight-dimensional parameter space would require 100 million experiments. With ITT, researchers can explore this problem in a broader parameter space than before, which would have taken 950 years to complete, but now only takes a few months. The robotic system periodically and automatically conducts a series of initial experiments at a constant speed. During this process, ITT plays a role in partially controlling the parameters for the next experiment by minimizing and quantifying the appropriate acquisition function of uncertainty, while achieving a series of objectives such as reducing resistance. Earlier this year, researchers from ITT received the De Florez Mechanical Engineering Award from MIT, and ITT was recognized as an outstanding example of revitalizing aging oil fields using new methods. In addition, the study identified many other scientific applications of automation and artificial intelligence: at the California Institute of Technology, a robot scientist named "Adam" generates and tests hypotheses; at the Defense Advanced Research Projects Agency, the Big Mechanism project reads thousands of research papers to generate new models. Overall, this new system demonstrates a potential paradigm shift in research, where automation and uncertainty quantification can greatly accelerate scientific discoveries. Researchers assert that the machine learning methods described in the paper can be applied to fluid mechanics and other experimental fields.

RECENT AI NEWS

RECENT AI TOOLS