Scholarlink

<p>The excellent mechanical properties that triply periodic minimal surfaces (TPMS) can achieve encourage their applications for structuring optimized components. Moreover, the great manufacturability of TPMS scaffolds with additive manufacturing technologies boosted their investigation for producing lightweight structures that demand high energy absorption capacity. Within this context, the present work aimed to investigate the compression behavior of gyroid structures with a varying number of cells and graded volume fractions manufactured by fused deposition modeling (FDM). Based on experimental results, a forecast model of the stress–strain curves was proposed through a recurrent neural network (RNN) machine learning technique named Long Short-Term Memory Networks (LSTM), including the controlled scaffold variables. The conducted experimental analyses demonstrated that varying volume fractions and the number of cells of these scaffolds affected their mechanical strength, with a higher contribution of the volume fraction. Finally, the developed model with tuned network layers and epochs demonstrated a great potential to fit stress–strain curve behaviors, considering coefficients of determination higher than 0.99 and error metrics below 0.22&#xa0;MPa. So, the study contributed an approach for modeling and predicting gyroid mechanical behaviors according to their tailored structural parameters. The proposed strategy can speed up the scaffolds design selection depending on their applications after the additive manufacturing processing.</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

Modeling the compression behavior of gyroid scaffolds produced by fused deposition modeling using machine learning

  • 期刊论文

  • 发表日期：

    2025年4月29日

查看全文
  • Leandro Marques Benasse,
  • Amanda Rossi de Oliveira,
  • Erik Gustavo Del Conte

摘要

The excellent mechanical properties that triply periodic minimal surfaces (TPMS) can achieve encourage their applications for structuring optimized components. Moreover, the great manufacturability of TPMS scaffolds with additive manufacturing technologies boosted their investigation for producing lightweight structures that demand high energy absorption capacity. Within this context, the present work aimed to investigate the compression behavior of gyroid structures with a varying number of cells and graded volume fractions manufactured by fused deposition modeling (FDM). Based on experimental results, a forecast model of the stress–strain curves was proposed through a recurrent neural network (RNN) machine learning technique named Long Short-Term Memory Networks (LSTM), including the controlled scaffold variables. The conducted experimental analyses demonstrated that varying volume fractions and the number of cells of these scaffolds affected their mechanical strength, with a higher contribution of the volume fraction. Finally, the developed model with tuned network layers and epochs demonstrated a great potential to fit stress–strain curve behaviors, considering coefficients of determination higher than 0.99 and error metrics below 0.22 MPa. So, the study contributed an approach for modeling and predicting gyroid mechanical behaviors according to their tailored structural parameters. The proposed strategy can speed up the scaffolds design selection depending on their applications after the additive manufacturing processing.