Effect of bending load on electrical conductivity of carbon/epoxy composites filled with nanoparticles using design of experiment and artificial neural networks

• Published in: Results in engineering Vol. 25; p. 104174
• Main Authors: Sadollah, Ali, Razavi, Seyed Morteza, Al-Shamiri, Abobakr Khalil
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.03.2025; Elsevier
• Subjects: Artificial neural networks; Carbon black; Carbon nanotubes; Electrical conductivity; Extreme learning machine; Carbon nanotubes; Electrical conductivity; Carbon black; Extreme learning machine; Artificial neural networks
• ISSN: 2590-1230; 2590-1230
• DOI: 10.1016/j.rineng.2025.104174

•Investigating bending load effects on electrical conductivity in carbon-epoxy composites.•Identifying optimal CB (25 %) and CNT (10 %) nanoparticle weight percentages.•Utilizing DOE, ANNs, and ELM methods to predict and compare response variables.•Minimal conductivity loss in CNT composites; CB composites show a higher reduction.•Applications include composite electrodes for crude oil tank electrostatic desalting. In this study, we investigate the impact of bending load on the electrical conductivity of carbon-epoxy composites containing various nanoparticles. The samples must meet the bending strength requirements and comply with the electrical conductivity standards set by the U.S. Energy Institute for electrode manufacturing. We use carbon black (CB) nanoparticles, carbon nanotubes (CNTs), and epoxy resin to create these samples. Using the four-point resistance method, we determine the optimal weight percentages of CB and CNTs incorporated into the carbon/epoxy composite and establish the electrical conductivity threshold. Afterward, we subject the samples to bending loads with several transverse displacements, measuring the electrical conductivity during loading and unloading. We analyze the input factors and employ prediction methods such as the design of experiments (DOE), artificial neural networks (ANNs), and extreme learning machines (ELM) to forecast the response factors. The ANNs and ELM models prove effective in accurately predicting data, and the model generated by DOE is statistically valid with a confidence level exceeding 95 %. We then compare the forecasted responses with the experimental results. Our experimental findings indicate that the decrease in electrical conductivity due to bending is minimal in carbon/epoxy samples containing CNTs and most significant in samples containing CB. Additionally, we determine the bending strength of the specimens using a three-point bending method. We examine the distribution pattern of nanoparticles in the samples through scanning electron microscope images. The results of this study carry significant implications for the manufacturing of composite electrodes subjected to bending loads in industrial applications.