Artificial Neural Network Code For Prediction and Field Reconstruction of Vortex Structures

Published: 27 January 2025| Version 1 | DOI: 10.17632/6bwjyytm37.1
Contributors:
Adarsh Vasa, Kaustav Chaudhury

Description

Velocity field reconstruction using artificial neural networks has been employed on the RANS model data. With the help of opensource Python code, the model is trained over spatial and velocity components as input data and velocity swirling strength as output. Upon training, this code allows the user to perform inverse modeling, to find location and the state of flow at any given value of swirling strength. Furthermore, velocity field reconstruction at any given plane can be seamlessly achieved with this model. The code used in this research is limited to smooth pipe bends, in which friction factor is analysed using a steady state incompressible fluid flow. Program Summary: Program title: VortexPrediction_Code Licensing provisions: Standard CC by 4.0 licence, https://creativecommons.org/licenses/by/4.0/ No. of lines in distributed program, including test data, etc.: 210 No. of bytes in distributed program, including test data, etc.: 175000 Distribution format: .zip Programming language: Python Computer: Any workstation or laptop computer running Google Colab, Anaconda, Jupyter, pandas, NumPy, Microsoft Azure, Alteryx, Tensorflow and MATLAB. Operating system: Windows and Mac OS, Linux. Runtime: The artificial neural network produces results within a span of 50 seconds for two-dimensional geometry, using the allocated free computational resources of Google Colab cloud-based computing system.

Files

Steps to reproduce

1. Copy the code into a Python based platform 2. Load 'Cas90451.csv' 3. Compile the program 4. Give user input for swirling strength when prompted 5. Download trained model file "Swirl_ANN.h5"

Institutions

National Institute of Technology Rourkela

Categories

Artificial Neural Network

Licence