Scholarlink

<p>The components built by the Wire Arc Additive Manufacturing (WAAM) process often exhibit dimensional inaccuracies that must be addressed through real-time process monitoring and control. A novel, cost-effective photogrammetry-based methodology is developed to capture layer-by-layer bead profiles during the WAAM of walls. This innovative approach involves converting the captured bead profiles into CAD models using the KIRI engine, which is then sliced with Netfabb to measure dimensional variations for each layer. The measurement from the proposed approach, in comparison with a 3D scanner, shows good accuracy, with an average error of 0.080 ± 0.07&#xa0;mm for bead height and 0.14 ± 0.15&#xa0;mm for bead width along the deposited length. Spatial layer-specific dimensional variations are correlated with key process parameters, including wire feed rate and travel speed. These data are used to develop deep neural network (DNN) models capable of forward and reverse mapping of process parameters to bead dimensions. The forward model achieves an accuracy of 97.80% and 96.59% for bead height and width predictions, respectively. The reverse model achieves a prediction accuracy of 92.47% for wire feed speed and 95.08% for travel speed. The developed forward and reverse models account for instantaneous root mean square (RMS) values of wire feed rate and travel speed during deposition, enabling effective monitoring of local bead dimensions. Thus, the proposed methodology can be implemented for the real-time monitoring of freeform shapes to improve the surface quality of WAAM components and address challenges associated with traditional sensor-based implementations.</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

Novel on-site layer dimensioning and deep learning-enabled predictive modelling for wire arc additive manufacturing

  • 期刊论文

  • 发表日期：

    2025年2月20日

查看全文
  • Srihari Chitral,
  • Bhargav Sri Sai Nama,
  • Sesha Saikrishna Penamakuri,
  • Degala Venkata Kiran

摘要

The components built by the Wire Arc Additive Manufacturing (WAAM) process often exhibit dimensional inaccuracies that must be addressed through real-time process monitoring and control. A novel, cost-effective photogrammetry-based methodology is developed to capture layer-by-layer bead profiles during the WAAM of walls. This innovative approach involves converting the captured bead profiles into CAD models using the KIRI engine, which is then sliced with Netfabb to measure dimensional variations for each layer. The measurement from the proposed approach, in comparison with a 3D scanner, shows good accuracy, with an average error of 0.080 ± 0.07 mm for bead height and 0.14 ± 0.15 mm for bead width along the deposited length. Spatial layer-specific dimensional variations are correlated with key process parameters, including wire feed rate and travel speed. These data are used to develop deep neural network (DNN) models capable of forward and reverse mapping of process parameters to bead dimensions. The forward model achieves an accuracy of 97.80% and 96.59% for bead height and width predictions, respectively. The reverse model achieves a prediction accuracy of 92.47% for wire feed speed and 95.08% for travel speed. The developed forward and reverse models account for instantaneous root mean square (RMS) values of wire feed rate and travel speed during deposition, enabling effective monitoring of local bead dimensions. Thus, the proposed methodology can be implemented for the real-time monitoring of freeform shapes to improve the surface quality of WAAM components and address challenges associated with traditional sensor-based implementations.