Scholarlink

<p>The rapid advancement of nanoelectronics demands materials with exceptional electrical and mechanical properties to support the development of high-performance, miniaturized devices. Graphene nanoplatelets (GNPs), a promising nanomaterial, have demonstrated significant potential in enhancing materials' electrical and structural characteristics at the nanoscale. This study explores the influence of GNPs on the electrical conductivity (EC) and compressive strength (CS) of nanoelectronic components, leveraging experimental investigations and advanced deep learning (DL) models, including non-autoregressive recurrent neural networks (NARNNs), verifiable convolutional neural networks (VCNNs), and Tsukamoto type-2 fuzzy inference system (TT2FIS). Experimental results revealed that the incorporation of GNPs at concentrations of 0.05% and 0.1% improved EC by 28.7% and 35.2%, respectively, while enhancing CS by 18.4% and 22.6%. These findings highlight the potential of GNP-enhanced materials for use in nanoelectronic devices that demand both high EC and mechanical reliability under thermal conditions. DL models demonstrated outstanding accuracy in predicting the properties of GNP-enhanced materials, with VCNNs achieving the highest performance. For EC predictions, VCNNs achieved a correlation coefficient (<i>R</i>) of 0.989, outperforming NARNNs (<i>R</i> = 0.976) and TT2FIS (<i>R</i> = 0.963). For CS, VCNNs exhibited an <i>R</i>-value of 0.993, compared to NARNNs (<i>R</i> = 0.982) and TT2FIS (<i>R</i> = 0.970). Error analysis further validated the superiority of VCNNs, as the mean square error (MSE) for EC predictions was 15.4% lower than NARNNs and 48.7% lower than TT2FIS. Similarly, for TS predictions, VCNNs achieved an MSE reduction of 12.8% compared to NARNNs and 51.3% compared to TT2FIS. SHapley Additive exPlanations analysis identified GNP concentration as the dominant factor influencing both EC and TS, followed by curing conditions. These results highlight the possible of DL-driven methods, particularly VCNNs, in optimizing GNP-enhanced materials for nanoelectronic applications, offering a fast and cost-effective pathway to design advanced materials for next-generation electronic devices.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A hybrid artificial intelligence framework for predicting electrical and thermal properties of graphene nanoplatelet-enhanced nanoelectronic materials

  • 期刊论文

  • 发表日期：

    2025年9月23日

查看全文
  • Bhaskarrao Yakkala,
  • M. Raja,
  • V. Elumalai,
  • B. Muthuraj,
  • L. Umasankar

摘要

The rapid advancement of nanoelectronics demands materials with exceptional electrical and mechanical properties to support the development of high-performance, miniaturized devices. Graphene nanoplatelets (GNPs), a promising nanomaterial, have demonstrated significant potential in enhancing materials' electrical and structural characteristics at the nanoscale. This study explores the influence of GNPs on the electrical conductivity (EC) and compressive strength (CS) of nanoelectronic components, leveraging experimental investigations and advanced deep learning (DL) models, including non-autoregressive recurrent neural networks (NARNNs), verifiable convolutional neural networks (VCNNs), and Tsukamoto type-2 fuzzy inference system (TT2FIS). Experimental results revealed that the incorporation of GNPs at concentrations of 0.05% and 0.1% improved EC by 28.7% and 35.2%, respectively, while enhancing CS by 18.4% and 22.6%. These findings highlight the potential of GNP-enhanced materials for use in nanoelectronic devices that demand both high EC and mechanical reliability under thermal conditions. DL models demonstrated outstanding accuracy in predicting the properties of GNP-enhanced materials, with VCNNs achieving the highest performance. For EC predictions, VCNNs achieved a correlation coefficient (R) of 0.989, outperforming NARNNs (R = 0.976) and TT2FIS (R = 0.963). For CS, VCNNs exhibited an R-value of 0.993, compared to NARNNs (R = 0.982) and TT2FIS (R = 0.970). Error analysis further validated the superiority of VCNNs, as the mean square error (MSE) for EC predictions was 15.4% lower than NARNNs and 48.7% lower than TT2FIS. Similarly, for TS predictions, VCNNs achieved an MSE reduction of 12.8% compared to NARNNs and 51.3% compared to TT2FIS. SHapley Additive exPlanations analysis identified GNP concentration as the dominant factor influencing both EC and TS, followed by curing conditions. These results highlight the possible of DL-driven methods, particularly VCNNs, in optimizing GNP-enhanced materials for nanoelectronic applications, offering a fast and cost-effective pathway to design advanced materials for next-generation electronic devices.