نوع مقاله : پژوهشی
نویسندگان
1 تهران، دانشگاه تهران، دانشکدگان فنی، دانشکده مهندسی شیمی، گروه مهندسی پلیمر
2 تهران، دانشگاه تهران، دانشکدگان فنی، دانشکده مهندسی شیمی، آزمایشگاه تحقیقاتی مواد و فرایندهای پیشرفته پلیمری
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Hypothesis:
The combination of impedance spectroscopy with artificial intelligence models can predict the composition of polymer blends containing conductive nanoparticles. This method with the capabilities of machine learning provides a possibility for more precise and faster analysis for the microstructure of conductive polymer blend composites and can replace traditional time-consuming methods.
Methods:
Polypropylene/ poly(ethylene-co-vinyl acetate) blends in composition ratios of 60/40, 50/50, 40/60, 30/70, and 10/90 containing 4 vol% conductive carbon black were prepared by melt-mixing method, and their composition percentages were evaluated using impedance data in the laboratory and through measurement of frequency, real impedance, imaginary impedance, and phase shift parameters. Artificial intelligence models, including Random Forest, k-Nearest Neighbor, Support Vector Regression, Decision Tree, and XGBoost, were used for prediction of the percentage of each polymer in the blend.
Findings:
The results show that polypropylene (PP) containing conductive carbon black, compared to poly(ethylene-co-vinyl acetate) (EVA), has a lower impedance level and a higher critical frequency; this phenomenon is attributed to the higher crystallinity of polypropylene. By blending EVA and PP, the impedance decreased compared to EVA containing carbon black particles. Among the blends, the 60/40 blend had the lowest impedance (40 Ω) and the highest critical frequency (106 Hz), which was attributed to phase refinement in the co-continuous morphology. In addition, the artificial intelligence results showed that the Random Forest model, with a mean absolute error of 4, had better performance than other models in predicting the composition percentages of the blends. This study suggests that combining impedance spectroscopy methods and artificial intelligence can be used as a novel and accurate method for predicting the percentages of each polymer in conductive polymer blends.
کلیدواژه‌ها [English]
)