Back to Search
Start Over
Prediction of surface tension of liquid normal alkanes, 1-alkenes and cycloalkane using neural network
- Source :
- Chemical Engineering Research and Design. 137:154-163
- Publication Year :
- 2018
- Publisher :
- Elsevier BV, 2018.
-
Abstract
- In the light of artificial neural network (ANN) model advantages, a predictive ANN model is proposed to correlate the surface tension of common hydrocarbons including normal alkanes (i.e. n-C4–n-C40), linear alkenes (i.e. 1-C4–1-C40), and cycloalkanes (C4–C20) in a wide range of temperatures. The most important advantage of the current proposed network is its low number of input variables which are only temperature of the system as well as carbon number and critical temperature of components utilized to differentiate among the different components. The obtained results revealed that a model trained by the Levenberg–Marquardt algorithm with hyperbolic tangent and linear transfer functions for the hidden and output layers, respectively, comprised of 27 hidden neurons is the optimum structure. In sum up, the obtained results demonstrated that the proposed ANN model is capable to satisfactorily predict and correlate the 5461 surface tension data points of normal alkanes, linear alkenes, and cycloalkanes as a function of temperature with maximum deviation of 0.47, 0.40 and 0.43 mN/m, respectively, just using three inputs parameters considering testing data subset.
- Subjects :
- Materials science
Artificial neural network
General Chemical Engineering
Hyperbolic function
02 engineering and technology
General Chemistry
Function (mathematics)
021001 nanoscience & nanotechnology
Transfer function
Surface tension
Cycloalkane
chemistry.chemical_compound
Data point
020401 chemical engineering
chemistry
Range (statistics)
Physics::Chemical Physics
0204 chemical engineering
0210 nano-technology
Biological system
Subjects
Details
- ISSN :
- 02638762
- Volume :
- 137
- Database :
- OpenAIRE
- Journal :
- Chemical Engineering Research and Design
- Accession number :
- edsair.doi...........f216bdf2ff5798031ae64c5e0aaab888