Scholarlink

<p>Wire-arc direct energy deposition (DED) is a form of metal additive manufacturing that enables the fabrication of components and engineered materials with high deposition rates and cost-effective processing. However, determining optimal toolpath parameters for multi-bead, multi-layer depositions remains largely heuristic due to complex interdependencies among variables such as heat input, wire feed speed, and travel speed. This work proposes a data-driven approach using a feedforward neural network (FNN) trained on a combination of experimentally and synthetically generated data to optimize printing parameters, specifically, the stepover distance, aiming for flatter, defect-free depositions. An initial design of experiments (DOE) was conducted using 16-pulsed gas metal arc welding (GMAW) trials with ER316L stainless steel filler metal. Overlapping bead geometries were analyzed to quantify dependent variables, and the dataset was augmented using two synthetic data generation techniques: multiple linear regression and statistical mapping. The trained FNN model was integrated into a graphical user interface (GUI) that predicts the optimal stepover based on wire feed and travel speeds, streamlining the toolpath planning process for wire-arc DED. Validation results demonstrate that the proposed method generalizes well across input conditions and provides reliable predictions for smoother, more uniform bead profiles, improving process efficiency for AM and cladding applications.</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

Artificial Intelligence and Statistical Mapping Applied to Additive Manufacturing Toolpath Optimization in Wire-Arc DED

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月15日

查看全文
  • Kaue C. Riffel,
  • Rakhi Bawa,
  • Justin Chan,
  • Rida Adhami,
  • Daniil Gofman,
  • Antonio J. Ramirez

摘要

Wire-arc direct energy deposition (DED) is a form of metal additive manufacturing that enables the fabrication of components and engineered materials with high deposition rates and cost-effective processing. However, determining optimal toolpath parameters for multi-bead, multi-layer depositions remains largely heuristic due to complex interdependencies among variables such as heat input, wire feed speed, and travel speed. This work proposes a data-driven approach using a feedforward neural network (FNN) trained on a combination of experimentally and synthetically generated data to optimize printing parameters, specifically, the stepover distance, aiming for flatter, defect-free depositions. An initial design of experiments (DOE) was conducted using 16-pulsed gas metal arc welding (GMAW) trials with ER316L stainless steel filler metal. Overlapping bead geometries were analyzed to quantify dependent variables, and the dataset was augmented using two synthetic data generation techniques: multiple linear regression and statistical mapping. The trained FNN model was integrated into a graphical user interface (GUI) that predicts the optimal stepover based on wire feed and travel speeds, streamlining the toolpath planning process for wire-arc DED. Validation results demonstrate that the proposed method generalizes well across input conditions and provides reliable predictions for smoother, more uniform bead profiles, improving process efficiency for AM and cladding applications.