Scholarlink

<p>Rapid implementation of Additive Manufacturing (AM) for corrosion-resistant components demands understanding complex process-structure-property relationships without costly characterization. Here we introduce CorrosionOptiMap, a physics-informed process-to-property framework that maps laser powder bed fusion (LPBF) parameters directly to corrosion potential (<i>E</i><sub><i>c</i><i>o</i><i>r</i><i>r</i></sub>), corrosion rate (<i>C</i><i>R</i>), and pitting potential (<i>E</i><sub><i>p</i><i>i</i><i>t</i><i>t</i><i>i</i><i>n</i><i>g</i></sub>). Engineered physics descriptors feed Pareto-selected stacking ensembles with an ElasticNet meta-learner, trained on potentiodynamic data from 77 SS316L specimens, the largest experimental LPBF corrosion dataset. CorrosionOptiMap attains <i>R</i><sup>2</sup>&gt;0.89 across targets and cuts MAE/RMSE by 50–60% versus raw-parameter models, while SHAP reveals a hierarchy: surface morphology governs <i>C</i><i>R</i>, porosity and spatters control <i>E</i><sub><i>p</i><i>i</i><i>t</i><i>t</i><i>i</i><i>n</i><i>g</i></sub>, and microstructural homogeneity drives <i>E</i><sub><i>c</i><i>o</i><i>r</i><i>r</i></sub>. High-throughput “corrosivity maps” evaluate 6,200 parameter combinations and enable multi-objective Pareto optimization, reducing experimental burden 8̃0 × and guiding practical trade-offs. SEM on 35 prints validates surface area driven kinetics for <i>C</i><i>R</i>. Embedding physics descriptors into ensemble learning yields explainable, scalable predictions that accelerate optimization of corrosion-resistant AM components.</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

CorrosionOptiMap links additive manufacturing process parameters to corrosion of stainless steel

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月20日

查看全文
  • Ayman Karaki,
  • Ahmad Hammoud,
  • Fatima Ghassan Alabtah,
  • Marwa AbdelGawad,
  • Marwan Khraisheh

摘要

Rapid implementation of Additive Manufacturing (AM) for corrosion-resistant components demands understanding complex process-structure-property relationships without costly characterization. Here we introduce CorrosionOptiMap, a physics-informed process-to-property framework that maps laser powder bed fusion (LPBF) parameters directly to corrosion potential (Ecorr), corrosion rate (CR), and pitting potential (Epitting). Engineered physics descriptors feed Pareto-selected stacking ensembles with an ElasticNet meta-learner, trained on potentiodynamic data from 77 SS316L specimens, the largest experimental LPBF corrosion dataset. CorrosionOptiMap attains R2>0.89 across targets and cuts MAE/RMSE by 50–60% versus raw-parameter models, while SHAP reveals a hierarchy: surface morphology governs CR, porosity and spatters control Epitting, and microstructural homogeneity drives Ecorr. High-throughput “corrosivity maps” evaluate 6,200 parameter combinations and enable multi-objective Pareto optimization, reducing experimental burden 8̃0 × and guiding practical trade-offs. SEM on 35 prints validates surface area driven kinetics for CR. Embedding physics descriptors into ensemble learning yields explainable, scalable predictions that accelerate optimization of corrosion-resistant AM components.