Premium
Modeling tensile modulus of (polyamide 6)/nanoclay composites: Response surface method vs. taguchi‐optimized artificial neural network
Author(s) -
Moghri Mehdi,
Seyed Shahabadi Seyed Ismail,
Madic Milos
Publication year - 2016
Publication title -
journal of vinyl and additive technology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.295
H-Index - 35
eISSN - 1548-0585
pISSN - 1083-5601
DOI - 10.1002/vnl.21416
Subject(s) - taguchi methods , materials science , response surface methodology , ultimate tensile strength , composite material , polyamide , modulus , design of experiments , artificial neural network , computer science , machine learning , mathematics , statistics
Tensile modulus is an important mechanical property of polymer/nanoclay nanocomposites. In this study, response surface method (RSM) and Taguchi‐optimized artificial neural network (Taguchi‐optimized ANN) were used to model tensile modulus as a function of nanoclay content, melt temperature, screw speed, and feeding rate for (polyamide 6)/nanoclay nanocomposites prepared in a twin‐screw extruder. The comparison between Taguchi‐optimized‐ANN‐ and RSM‐generated plots showed that predictions made by both models were in agreement in general. Coefficient of determination, R 2 , showed that the RSM model can explain the variation with the accuracy of 0.768, indicating there was no strong correlation. However, from ANOVA, the p value for the RSM model was less than 0.05, signifying that the obtained model could be considered statistically significant. In addition, further assessment in terms of data fitting and prediction capabilities demonstrated the superiority of a properly trained Taguchi‐optimized ANN model in characterizing the nonlinear behavior of a response‐factors relationship. The Taguchi‐optimized ANN model R 2 for training data and testing data were 0.965 and 0.902, respectively. Also, the Taguchi‐optimized ANN model was developed by using 20% less data in comparison to the RSM model. J. VINYL ADDIT. TECHNOL., 22:29–36, 2016. © 2014 Society of Plastics Engineers