Scholarlink

<p>The creation of desired bead shapes is crucial to produce high-quality parts using a directed energy deposition (DED) process. Simulation techniques using data-driven methods can provide an effective method for optimizing deposition parameters to produce high-quality parts. Previous researches have been focused on developing a predictive model for height and width of beads using either regression analysis or machine learning (ML). The goal of this study is to develop a novel method to predict the bead shape in the DED process using ML with limited point data. Coordinates of control points on the bead outline are obtained from image processing. Regression equations are then estimated to predict coordinates between experimental conditions. An optimal basis function to create the bead shape is obtained from a ML model with a multi-layer perceptron. Bead shapes for different conditions are predicted using the ML and limited point data. Finally, the predictive performance and applicability of the proposed method are demonstrated through a comparison of actual and predicted bead shapes.</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

Prediction of the Bead Shape in a Directed Energy Deposition Process Using Machine Learning with Limited Point Data

  • 期刊论文

  • 发表日期：

    2025年10月17日

查看全文
  • Kwang-Kyu Lee,
  • Jungyeon Kim,
  • Seung Ki Moon,
  • Dong-Gyu Ahn

摘要

The creation of desired bead shapes is crucial to produce high-quality parts using a directed energy deposition (DED) process. Simulation techniques using data-driven methods can provide an effective method for optimizing deposition parameters to produce high-quality parts. Previous researches have been focused on developing a predictive model for height and width of beads using either regression analysis or machine learning (ML). The goal of this study is to develop a novel method to predict the bead shape in the DED process using ML with limited point data. Coordinates of control points on the bead outline are obtained from image processing. Regression equations are then estimated to predict coordinates between experimental conditions. An optimal basis function to create the bead shape is obtained from a ML model with a multi-layer perceptron. Bead shapes for different conditions are predicted using the ML and limited point data. Finally, the predictive performance and applicability of the proposed method are demonstrated through a comparison of actual and predicted bead shapes.