Scholarlink

<p>Energy absorption plays a crucial role in laser-based additive manufacturing (AM) through influencing the melting and solidification process. Prediction of energy absorption under varying processing conditions and materials remain a challenging problem. In this paper, we developed a data-driven approach to relate the melt pool dynamics with energy absorption in laser powder bed fusion process. Utilizing the NIST AM&#xa0;bench 2022 Asynchronous challenge problem data, a convolutional neural network (U-Net model) is trained to automatically identify the melt pool and keyhole features which is further related to absorptivity using an attention-enhanced convolutional long short-term memory (ConvLSTM) model. We also incorporated non-dimensional normalized energy density to create a robust, materials-agnostic model to predict absorptivity from the melt pool features. The data-driven model predicts absorptivity accurately and provides insights into how melt pool dynamics affect the energy absorption for both spot and scan lasers. This approach can guide the process design and optimization for target microstructure and properties design in laser-based additive manufacturing.</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

Data-driven insights into melt pool dynamics and energy absorption in the laser powder bed fusion process

  • 期刊论文

  • 发表日期：

    2025年8月22日

查看全文
  • Apurba Sarker,
  • Satyajit Mojumder

摘要

Energy absorption plays a crucial role in laser-based additive manufacturing (AM) through influencing the melting and solidification process. Prediction of energy absorption under varying processing conditions and materials remain a challenging problem. In this paper, we developed a data-driven approach to relate the melt pool dynamics with energy absorption in laser powder bed fusion process. Utilizing the NIST AM bench 2022 Asynchronous challenge problem data, a convolutional neural network (U-Net model) is trained to automatically identify the melt pool and keyhole features which is further related to absorptivity using an attention-enhanced convolutional long short-term memory (ConvLSTM) model. We also incorporated non-dimensional normalized energy density to create a robust, materials-agnostic model to predict absorptivity from the melt pool features. The data-driven model predicts absorptivity accurately and provides insights into how melt pool dynamics affect the energy absorption for both spot and scan lasers. This approach can guide the process design and optimization for target microstructure and properties design in laser-based additive manufacturing.