A Comparison between MLP and SVR Models in Prediction of Thermal Properties of Nano Fluids

Desirable thermal properties of nanofluid is the vital reason for using nanofluid. There is an exemplary development in various applications using nanofluid. Mathematical and experimental models were developed to predict the thermal properties of nanofluids, the models are tiresome and expensive and...

Full description

Saved in:
Bibliographic Details
Published inJournal of applied fluid mechanics Vol. 11; no. SI; p. 7
Main Authors Kavitha, R, Kumar, P C Mukesh
Format Journal Article
LanguageEnglish
Published Isfahan Isfahan University of Technology 01.10.2018
Subjects
Clustering
Comparative analysis
Computational fluid dynamics
Datasets
Mathematical models
Nanofluids
Predictions
Regression analysis
Regression coefficients
Regression models
Root-mean-square errors
Soft computing
Software
Support vector machines
Thermal conductivity
Thermal properties
Thermodynamic properties
Online AccessGet full text

Cover

More Information
Summary:Desirable thermal properties of nanofluid is the vital reason for using nanofluid. There is an exemplary development in various applications using nanofluid. Mathematical and experimental models were developed to predict the thermal properties of nanofluids, the models are tiresome and expensive and have discrepancies between them. Soft computing tools are most useful in prediction, classification and clustering the data with good accuracy and with less expensive. In this paper, comparative analysis of Multi Layer Perceptron (MLP) model and Support Vector Regression (SVR) model were done by using various evaluation criterions. The two models developed to predict the thermal conductivity ratio of CNT/H2O and the results were compared. The present modeling has been carried out using MATLAB 2017 b. In both the models, the experimental values and predicted values possess good accordance. Regression coefficient value (R2) for overall data is 0.99 and 0.98 for MLP and SVR models respectively. The Root Mean Square Error (RMSE) value is less in MLP model when compared with SVR model, RMSE values are 0.01578 and 0.01812 respectively. The prediction is best in MLP model but with limited experimental data set, it fails to address over fitting problem, whereas SVR model is ideal with limited data set, it overcomes over fitting problem and possess better generalization than MLP model.
ISSN:1735-3572
1735-3645
DOI:10.36884/jafm.11.SI.29411