Underwater cylinder turbulent flow partition flow field prediction method based on deep learning

Abstract

The invention relates to an underwater cylinder turbulent flow partition flow field prediction method based on deep learning. The method comprises the following steps: firstly, acquiring a large amount of underwater cylinder turbulent flow field data through Fluent software; performing spatial partitioning by using a spatial clustering algorithm to obtain a plurality of same proton regions; secondly, preprocessing the data, dividing the data into a training set and a test set, and performing time sequence modeling on any point target in the same proton region; and finally, constructing a prediction model for the same proton region based on a long-short-term memory network (LSTM) according to the characteristics of the shunt field of each part. Therefore, the method is applied to underwater cylinder turbulent flow field prediction. Compared with a traditional whole flow field sharing one set of neural network, the underwater cylindrical turbulent flow field is divided into four areas based on incoming flow velocity comparison division, and models are established according to data characteristics and data volume of each area, and a better prediction effect is expected to be achieved.

Description

technical field [0001] The invention belongs to the field of fluid mechanics and the application of neural networks, and relates to a deep learning-based method for predicting flow fields in sub-divisions of underwater cylinder disturbances. Background technique [0002] When the fluid flows around the cylinder, the flow section shrinks, the flow increases along the process, and the pressure decreases along the process. Due to the existence of viscous force, the separation of the boundary layer around the cylinder will occur, forming a cylinder turbulence. The problem of turbulence of underwater cylinders is also very common in engineering practice, such as the effect of water flow on bridges, offshore drilling platform pillars, offshore transportation pipelines, etc., and the effect of wind on tower equipment, chemical tower equipment, high-altitude cables, etc. Has a significant engineering application background. Therefore, it is not only of theoretical significance, but...

AI Technical Summary

Problems solved by technology

The early neural network only had two layers of input and output layers, which could not handle complex logic, which limited its applicability, while the deep neural network added a hidden layer between the input and output layers, and changed the number of layers and neurons in the hidden layer. number, theoretically can approximate any function

Images