<p class="MsoNormal" style="text-indent: 15.75pt">Fluid dynamics problem is one of the research hotspots in the offshore technology.&nbsp; As the experiment methods on researching offshore technology require high economic cost, numerical simulation methods bloom to solve these problems. However, numerical simulation methods are different to mesh a complex geometry and also very time consuming due to the billion degrees of freedom that need to be calculated. In the recent years, deep learning (DL) methods, due to its capability of handling the strong nonlinearity and high dimensionality, are proposed deal with the fluid dynamics. Nevertheless, traditional deep learning methods are seemed as black box which to build up a surrogate model by input (e.g. space coordinate and time) and output (e.g. velocity and pressure). It means that an accurate surrogate model is constructed by a large number of data. In this paper, a physics-informed neural network is proposed to solve the fluid dynamics. A fully-connected neural network embedded with the physical laws (conservation equations and N-S equations) as loss function is built up to predict the flow past cylinder. The different structures of the neural network and different training samples are compared and discussed. The results demonstrate that the deep learning constrained by physical laws is able to predict the flow information accurately under limited dataset and simpler network structure.</p>

Deep Learning Method Based on Physics-Informed Neural Network for Solving Flow Past Cylinder

Paper Proceedings: https://asmedigitalcollection.asme.org/OMAE/OMAE2021/volume/85116

OMAE 2021

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.