PSO-tuned support vector machine metamodels for assessment of turbulent flows in pipe bends

Computational fluid dynamics (CFD) technique is the most commonly used numerical approach to simulate fluid flow behaviour. Owing to its computationally, cost-intensive nature CFD models may not be easily and quickly deployable. In this regard, this study aims to present a support vector machine (SVM)-based metamodelling approach that can be easily trained and quickly deployed for carrying out large-scale studies.,Radial basis function and e^*-insensitive loss function are used as kernel function and loss function, respectively. To prevent overfitting of the model, five-fold cross-validation root mean squared error is used while training the SVM metamodel. Rather than blindly using any SVM tuning parameters, a particle swarm optimisation (PSO) is used to fine-tune them. The developed SVM metamodel is tested using various error metrics on disjoint test data.,Using the SVM metamodel, a parametric study is conducted to understand the effect of various factors influencing the behaviour of the turbulent fluid flow in the pipe bend with CFD simulation data set. Based on the parametric study carried out, it is seen that the diametric position has the most effect on dimensionless axial velocity, whereas Reynolds number has the least effect.,This paper provides an effective PSO-tuned SVM metamodelling approach, which may be used as a significant cost-saving approach to quickly and accurately estimate fluid flow characteristics that, in general, require the use of expensive CFD models.

[1]  Hossein Bonakdari,et al.  Numerical Analysis and Prediction of the Velocity Field in Curved Open Channel Using Artificial Neural Network and Genetic Algorithm , 2011 .

[2]  Hossein Bonakdari,et al.  Flow Variables Prediction Using Experimental, Computational Fluid Dynamic and Artificial Neural Network Models in a Sharp Bend , 2016 .

[3]  Dominique Laurence,et al.  Large eddy simulation of a T-Junction with upstream elbow: The role of Dean vortices in thermal fatigue , 2016 .

[4]  N. Fujisawa,et al.  Mass and momentum transfer characteristics in 90° elbow under high Reynolds number , 2018 .

[5]  William J. Pitz,et al.  An Approach for Formulating Surrogates for Gasoline with Application toward a Reduced Surrogate Mechanism for CFD Engine Modeling , 2011 .

[6]  Ushasta Aich,et al.  Modeling of EDM responses by support vector machine regression with parameters selected by particle swarm optimization , 2014 .

[7]  Seungjin Kim,et al.  Characteristics of Secondary Flow Induced by 90-Degree Elbow in Turbulent Pipe Flow , 2014 .

[8]  Sondipon Adhikari,et al.  A Response Surface Modelling Approach for Resonance Driven Reliability Based Optimization of Composite Shells , 2016 .

[9]  Yuji Hattori,et al.  Searching for turbulence models by artificial neural network , 2016, 1607.01042.

[10]  Amir Hossein Zaji,et al.  A new hybrid decision tree method based on two artificial neural networks for predicting sediment transport in clean pipes , 2017, Alexandria Engineering Journal.

[11]  Michele Ciofalo,et al.  Numerical prediction of turbulent flow and heat transfer in helically coiled pipes , 2010 .

[12]  Sondipon Adhikari,et al.  Stochastic natural frequency of composite conical shells , 2015 .

[13]  Marziyeh Hajialyani,et al.  Application of artificial neural network and genetic algorithm approaches for prediction of flow characteristic in serpentine microchannels , 2015 .

[14]  Masaaki Tanaka,et al.  Numerical Investigation on Large Scale Eddy Structure in Unsteady Pipe Elbow Flow at High Reynolds Number Conditions with Large Eddy Simulation Approach , 2012 .

[15]  Amir Hossein Alavi,et al.  Detection of fatigue cracking in steel bridge girders: A support vector machine approach , 2017 .

[16]  Elia Merzari,et al.  The three-dimensional structure of swirl-switching in bent pipe flow , 2017, Journal of Fluid Mechanics.

[17]  H. Kamide,et al.  Influence of elbow curvature on flow structure at elbow outlet under high Reynolds number condition , 2011 .

[18]  Ricardo Vinuesa,et al.  Secondary flow in turbulent ducts with increasing aspect ratio , 2018 .

[19]  S. Datta,et al.  Design of novel age-hardenable aluminium alloy using evolutionary computation , 2017 .

[20]  Suad Jakirlić,et al.  Comparative computational study of turbulent flow in a 90° pipe elbow , 2015 .

[21]  Hassan Ghassemi,et al.  Using computational fluid dynamic and artificial neural networks to predict the performance and cavitation volume of a propeller under different geometrical and physical characteristics , 2018 .

[22]  Shubhabrata Datta,et al.  Incorporation of prior knowledge in neural network model for continuous cooling of steel using genetic algorithm , 2017, Appl. Soft Comput..

[23]  Hisashi Ninokata,et al.  Numerical investigation of bent pipe flows at transitional Reynolds number , 2011 .

[24]  Vassili Toropov,et al.  Multi-objective Computational Fluid Dynamics (CFD) design optimisation in commercial bread-baking , 2013 .

[25]  Masaaki Tanaka,et al.  Unsteady Flow Characteristics in a 90 Degree Elbow Affected by Developed, Undeveloped and Swirling Inflow Conditions , 2012 .

[26]  N. Ganesh,et al.  Robust metamodels for accurate quantitative estimation of turbulent flow in pipe bends , 2019, Engineering with Computers.

[27]  Laszlo Fuchs,et al.  Swirl switching in turbulent flow through 90° pipe bends , 2015 .

[28]  Large-eddy simulations of 90° pipe bend flows , 2001 .

[29]  Yongmann M. Chung,et al.  Direct numerical simulation of a turbulent 90° bend pipe flow , 2018, International Journal of Heat and Fluid Flow.

[30]  P. Alfredsson,et al.  Turbulent Flows in Curved Pipes: Recent Advances in Experiments and Simulations , 2016 .

[31]  B. P. Swain,et al.  Genetically optimized diamond-like carbon thin film coatings , 2019, Materials and Manufacturing Processes.

[32]  K. Kalita,et al.  PECVD process parameter optimization: towards increased hardness of diamond-like carbon thin films , 2018, Materials and Manufacturing Processes.

[33]  J. Cotton,et al.  Measurement of the flow field characteristics in single and dual S-shape 90° bends using matched refractive index PIV , 2016 .

[34]  P. Schlatter,et al.  Evolution of turbulence characteristics from straight to curved pipes , 2012 .

[35]  Amir Hossein Zaji,et al.  Improving the performance of multi-layer perceptron and radial basis function models with a decision tree model to predict flow variables in a sharp 90° bend , 2016, Appl. Soft Comput..

[36]  Mauricio Zambrano-Bigiarini,et al.  A model-independent Particle Swarm Optimisation software for model calibration , 2013, Environ. Model. Softw..