![Ajouter à votre panier](https://e-campus.itech.fr/pmb/opac_css/images/basket_small_20x20.png)
[article]
Titre : |
Correlation between injection molding parameters, morphology and mechanical properties of PPS/SEBS blend using artificial neural networks |
Type de document : |
texte imprimé |
Auteurs : |
C. Lotti, Auteur ; R. E. S. Bretas, Auteur |
Année de publication : |
2007 |
Article en page(s) : |
p. 105-116 |
Note générale : |
Bibliogr. |
Langues : |
Anglais (eng) |
Index. décimale : |
668.9 Polymères |
Résumé : |
The objectives of this work were to identify the injection molding processing variables with the greatest effect on the morphology and mechanical properties of an injection molded blend made of poly (p-phenylene sulphide), PPS and block copolymer styrene-ethylene-butadiene-styrene ‘SEBS’. Artificial Neural Networks, ANNs, are used as an alternative method to constitutive and empirical models, to predict morphological features and mechanical properties from the injection molding conditions, and to predict mechanical properties from the morphological features. The quantification of SEBS dispersion in the PPS matrix was done using a dispersion function. Mold temperature and flow rate were the processing variables with the highest influence on the entire morphology, while the holding pressure influenced mainly the inner layers. Impact strength and toughness were most influenced by mold temperature, holding pressure and the outer layers. The flexural modulus was influenced by all processing variables and the intermediate layers. Three different ANNs were evaluated: one (ANN-1) to predict morphology from processing conditions and another two to predict mechanical properties from morphology and from processing conditions (ANN-2 and ANN-3, respectively). These latter ANN models had similar results, indicating that both inputs could be successfully used to predict mechanical properties, as the mean residuals were close to experimental errors. On the other hand, ANN-1 showed a lower performance, with a mean error smaller than the experimental error, suggesting that ANNs could overtake some inherent uncertainties. In this case, it was concluded that the distribution of data along output domain was more important than a high number of training data in the ANN’s performance. |
DOI : |
10.3139/217.0991 |
En ligne : |
https://drive.google.com/file/d/115Z9gMooN7hT1N3if7T4SEz2XX7zHG4u/view?usp=drive [...] |
Format de la ressource électronique : |
Pdf |
Permalink : |
https://e-campus.itech.fr/pmb/opac_css/index.php?lvl=notice_display&id=2774 |
in INTERNATIONAL POLYMER PROCESSING > Vol. XXII, N° 1 (03/2007) . - p. 105-116
[article]
|