Scholarlink

<p>This study introduces an adaptive variable-bead-height (AVBH) slicing strategy for additive manufacturing and demonstrates its implementation in wire arc additive manufacturing (WAAM), utilising a long short-term memory (LSTM)–based travel-speed control. Implemented on a Fanuc ARCMate 100iD six-axis robot, it enables the fabrication of thin-walled structures via WAAM without stair-stepping. Unlike conventional planar slicing, which deposits layers of constant height and produces staircase errors on inclined surfaces, in AVBH, the bead height of each layer is varied to conform to the intended surface profile. Two slicing methods (vertical and normal offset) are proposed and systematically evaluated in combination with corresponding torch orientations to assess geometric fidelity. For a representative sine wave profile, normal-offset slicing with the torch oriented normal to the deposition path performed best, yielding a Pearson shape correlation of &gt; 0.99 and a Spearman’s correlation of &gt; 0.97. This profile, deposited using AVBH, achieved the lowest root-mean-square error (RMSE) of 0.38&#xa0;mm and the lowest mean absolute error of 0.31&#xa0;mm. Furthermore, the capability of AVBH to follow steep surface variation was examined. Even for a representative valley profile with steep slopes approaching ~ 50°, the RMSE error was ~ 1.76 mm. To highlight the elimination of staircase effect, a representative 10 mm step profile deposited using AVBH achieved a root-mean-square error (RMSE) of 0.58 mm, compared with 1.27 mm under conventional planar slicing. To realise variable bead-heights in real time, the LSTM network was trained on multi-layer deposition sequences to predict the travel speed required for target heights. The model delivered an RMSE of 0.84 mm/s in speed prediction. Through incremental correction of deposition deviations, it reduced substrate inclination from 1.49° to 0.10°. AVBH slicing and LSTM-driven speed adjustment complemented each other to enable near-net-shape and smooth curved profiles, substantially improving WAAM part quality and expanding design-for-additive-manufacturing capabilities.</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

Adaptive variable-bead-height slicing with sequence-informed travel-speed control for wire arc additive manufacturing of thin-walled parts

  • 期刊论文

  • 发表日期：

    2026年4月30日

查看全文
  • Anas Ullah Khan,
  • Ajay Venkatramanan,
  • Amber Shrivastava

摘要

This study introduces an adaptive variable-bead-height (AVBH) slicing strategy for additive manufacturing and demonstrates its implementation in wire arc additive manufacturing (WAAM), utilising a long short-term memory (LSTM)–based travel-speed control. Implemented on a Fanuc ARCMate 100iD six-axis robot, it enables the fabrication of thin-walled structures via WAAM without stair-stepping. Unlike conventional planar slicing, which deposits layers of constant height and produces staircase errors on inclined surfaces, in AVBH, the bead height of each layer is varied to conform to the intended surface profile. Two slicing methods (vertical and normal offset) are proposed and systematically evaluated in combination with corresponding torch orientations to assess geometric fidelity. For a representative sine wave profile, normal-offset slicing with the torch oriented normal to the deposition path performed best, yielding a Pearson shape correlation of > 0.99 and a Spearman’s correlation of > 0.97. This profile, deposited using AVBH, achieved the lowest root-mean-square error (RMSE) of 0.38 mm and the lowest mean absolute error of 0.31 mm. Furthermore, the capability of AVBH to follow steep surface variation was examined. Even for a representative valley profile with steep slopes approaching ~ 50°, the RMSE error was ~ 1.76 mm. To highlight the elimination of staircase effect, a representative 10 mm step profile deposited using AVBH achieved a root-mean-square error (RMSE) of 0.58 mm, compared with 1.27 mm under conventional planar slicing. To realise variable bead-heights in real time, the LSTM network was trained on multi-layer deposition sequences to predict the travel speed required for target heights. The model delivered an RMSE of 0.84 mm/s in speed prediction. Through incremental correction of deposition deviations, it reduced substrate inclination from 1.49° to 0.10°. AVBH slicing and LSTM-driven speed adjustment complemented each other to enable near-net-shape and smooth curved profiles, substantially improving WAAM part quality and expanding design-for-additive-manufacturing capabilities.