Scholarlink

<p>We present a machine learning (ML) framework to characterize the cold spray process of similar and dissimilar metals that include combinations of Cu, Ni, Al, Ag, Au, Pd, and Pt. To this end, data from high-throughput molecular dynamics (MD) simulations of particle–substrate impact as a function of deposition variables (particle/substrate material combinations, particle size, and particle velocity) were integrated and analyzed in conjunction with experimentally determined physical descriptors, resulting in robust ML models for predicting two key process outputs—penetration depth and impact bonding strength. Specifically, to identify the best descriptors for these two key process outputs, our ML framework includes a feature engineering pipeline that critically evaluates 30 potentially relevant properties of input materials spanning physical, mechanical, thermal, and acoustic. The resulting ML models, based on relevant physical descriptors and trained on MD data, accurately predict the key process outputs as a function of deposition variables of cold spray of dissimilar metals. These findings lay the foundation for data-driven selection and optimization of the cold spray process, enabling the development of materials and engineering components with tailored performance.</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

Machine Learning of Impact Behavior in Cold Spray of Similar and Dissimilar Metals

  • 期刊论文

  • 发表日期：

    2025年9月3日

查看全文
  • Y. Nikravesh,
  • K. Muralidharan,
  • G. Frantziskonis,
  • M. I. Latypov

摘要

We present a machine learning (ML) framework to characterize the cold spray process of similar and dissimilar metals that include combinations of Cu, Ni, Al, Ag, Au, Pd, and Pt. To this end, data from high-throughput molecular dynamics (MD) simulations of particle–substrate impact as a function of deposition variables (particle/substrate material combinations, particle size, and particle velocity) were integrated and analyzed in conjunction with experimentally determined physical descriptors, resulting in robust ML models for predicting two key process outputs—penetration depth and impact bonding strength. Specifically, to identify the best descriptors for these two key process outputs, our ML framework includes a feature engineering pipeline that critically evaluates 30 potentially relevant properties of input materials spanning physical, mechanical, thermal, and acoustic. The resulting ML models, based on relevant physical descriptors and trained on MD data, accurately predict the key process outputs as a function of deposition variables of cold spray of dissimilar metals. These findings lay the foundation for data-driven selection and optimization of the cold spray process, enabling the development of materials and engineering components with tailored performance.