Carbon additives to improve polymer performance in energy applications using machine learning
The development of polymer composites enhanced with carbon-based additives has been investigated to reinforce their applicability in energy-related systems. Conductive fillers, such as graphene, Carbon Nano Tubes (CNTs), and Short-cut Carbon Fibers (SCFs), were incorporated into a Polydimethylsiloxa...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-12-01
|
| Series: | Case Studies in Construction Materials |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2214509525008976 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The development of polymer composites enhanced with carbon-based additives has been investigated to reinforce their applicability in energy-related systems. Conductive fillers, such as graphene, Carbon Nano Tubes (CNTs), and Short-cut Carbon Fibers (SCFs), were incorporated into a Polydimethylsiloxane (PDMS) matrix to enhance electrical conductivity and thermal performance. Experimental evaluations included four-probe electrical conductivity testing and self-heating measurements under varied input voltages (8–12 V). Results showed that composites containing 6 mm Carbon Fibers (CF), (Long-CNT) achieved the highest electrical conductivity of 1.8 S/m, significantly outperforming the CNT-only control (Base-CNT, 0.1 S/m). Correspondingly, the Long-CNT samples exhibited the fastest thermal response, with heating time constants (τg) as low as 70.01 s and peak surface temperatures exceeding 150 °C under 12 V input. To guide composite optimization, a hybrid Machine Learning (ML) framework combining Random Forest Regression (RFR) and Support Vector Regression (SVR) was developed. This stacked model was trained on 60 samples and achieved high predictive accuracy across all key outputs, including Coefficient of Determination (R²) = 0.985 for conductivity and R² > 0.95 for heating rate (Hr+c), τg, τd, and temperature. Feature importance analysis revealed that carbon fiber length and input voltage were the dominant factors influencing thermal-electrical performance. The model was also used to simulate untested CF–voltage configurations, identifying optimal engineering windows (e.g., 4.5 mm CF at 11 V) that balanced high heating efficiency with manageable power consumption. The integration of data-driven modeling with experimental validation enabled the accurate prediction and strategic tuning of composite properties. This work provides a scalable framework for designing high-performance, self-healing polymer nanocomposites for thermal management, sensing, and energy conversion applications. |
|---|---|
| ISSN: | 2214-5095 |