Scholarlink

<p>The direct metal deposition (DMD) process is a well-established additive manufacturing (AM) technique employed to construct metal alloys through the incremental addition of material. In the DMD process, predicting the temperature gradient is essential for the control of the process and final microstructure. Furthermore, improper cooling rates can exacerbate microstructural defects, impacting the overall part integrity. In this study, the multi-layer DMD of stainless steel 304 is assessed by the Finite Element Model (FEM), and the deep learning data-driven system is employed to predict melting pool maximum temperature and dimension based on the process parameters, including laser power, scanning speed, and laser spot radius. The Birth and death element technique is employed to simulate the addition of deposition with time through the ABAQUS 6.14 commercial software in performing thermal field analyses, and the artificial neural network (ANN) code is developed using the MATLAB platform. FEM provides physics-based insights into DMD, while ANN predicts outcomes like melt pool dimensions. Their integration combines accuracy and efficiency, reducing the need for extensive simulations or experiments, while also minimizing the time and cost associated with long-running FEM simulations and predictive analyses. The results illustrate that due to heat exchange, the maximum melt pool temperature in various layers is different, and the laser power and scanning speed directly affect the temperature distribution and melt pool morphology. Increasing the laser power while reducing the scan speed results in a deeper melt pool, which can lead to keyhole porosity and uneven solidification, significantly contributing to local stress concentration and potential failure<b>.</b> The new combined FEM-ANN model demonstrated remarkable modeling accuracy with an R<sup>2</sup> value of 97.5%, as proven by the following experimental test. Furthermore, the new combined FEM-ANN technique has the potential to generate a high-accuracy, quick, and dedicated model that could suggest the printing parameters based on the expected melt pool morphology.</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

Advanced data-driven insights into melt pool morphology in multi-layer direct metal deposition

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年9月26日

查看全文
  • Mohsen Sarparast,
  • Majid Shafaie,
  • Alireza Behvar,
  • Mahyar Sojoodi,
  • Mohammad Elahinia,
  • Hongyan Zhang

摘要

The direct metal deposition (DMD) process is a well-established additive manufacturing (AM) technique employed to construct metal alloys through the incremental addition of material. In the DMD process, predicting the temperature gradient is essential for the control of the process and final microstructure. Furthermore, improper cooling rates can exacerbate microstructural defects, impacting the overall part integrity. In this study, the multi-layer DMD of stainless steel 304 is assessed by the Finite Element Model (FEM), and the deep learning data-driven system is employed to predict melting pool maximum temperature and dimension based on the process parameters, including laser power, scanning speed, and laser spot radius. The Birth and death element technique is employed to simulate the addition of deposition with time through the ABAQUS 6.14 commercial software in performing thermal field analyses, and the artificial neural network (ANN) code is developed using the MATLAB platform. FEM provides physics-based insights into DMD, while ANN predicts outcomes like melt pool dimensions. Their integration combines accuracy and efficiency, reducing the need for extensive simulations or experiments, while also minimizing the time and cost associated with long-running FEM simulations and predictive analyses. The results illustrate that due to heat exchange, the maximum melt pool temperature in various layers is different, and the laser power and scanning speed directly affect the temperature distribution and melt pool morphology. Increasing the laser power while reducing the scan speed results in a deeper melt pool, which can lead to keyhole porosity and uneven solidification, significantly contributing to local stress concentration and potential failure. The new combined FEM-ANN model demonstrated remarkable modeling accuracy with an R2 value of 97.5%, as proven by the following experimental test. Furthermore, the new combined FEM-ANN technique has the potential to generate a high-accuracy, quick, and dedicated model that could suggest the printing parameters based on the expected melt pool morphology.