Scholarlink

<p>This research focuses on optimizing the welding parameters for dissimilar Inconel 625 and Inconel 825 using a novel interlocking joint geometry to improve weld quality. The study examines three joint designs: chamfered edge, circular pin, and a combination of interlocking pin with a chamfered edge. A hybrid approach combining Response Surface Methodology (RSM) and Machine Learning (ML) was applied to optimize the welding parameters. A Box-Behnken design (BBD) with 29 experimental trials was used to analyze the influences of spindle speed, soft load, and frictional load on mechanical properties, including tensile strength, yield strength, and elongation. An Extreme Gradient Boosting (XGBoost) regression model was developed in Python to predict these properties, achieving high accuracy with R<sup>2</sup> values greater than 0.99. The model's predictions were validated by experimental data, with percentage errors of 0.757, − 0.101, and 1.125% for tensile strength, yield strength, and elongation, respectively. Optimal parameters were found to be a spindle speed of 2400 revolutions per minute, soft load of 700&#xa0;Nm, and frictional load of 1400&#xa0;Nm. Under these conditions, Sample 2 (interlocked design) exhibited improved mechanical properties, including a tensile strength of 652.09&#xa0;MPa, yield strength of 583.57&#xa0;MPa, and elongation of 21.22%. X-ray examination showed that all welds were void-free, while microstructural analysis through Scanning Electron Microscopy, Energy Dispersive x-ray Spectroscopy, and Electron Backscatter Diffraction revealed a uniform grain structure, efficient elemental diffusion, and a high density of high-angle and Sigma 3 (Σ3) twin boundaries. These results confirm the effectiveness of the interlocking geometry and the combined RSM and ML optimization for high-performance dissimilar metal joints.</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

Machine Learning-Based Design and Optimization of Rotary Friction Welds in Dissimilar Inconel 625/825 Alloys Using Novel Joint Geometries

  • 期刊论文

  • 发表日期：

    2025年10月7日

查看全文
  • M. Krishna,
  • A. Mathivanan

摘要

This research focuses on optimizing the welding parameters for dissimilar Inconel 625 and Inconel 825 using a novel interlocking joint geometry to improve weld quality. The study examines three joint designs: chamfered edge, circular pin, and a combination of interlocking pin with a chamfered edge. A hybrid approach combining Response Surface Methodology (RSM) and Machine Learning (ML) was applied to optimize the welding parameters. A Box-Behnken design (BBD) with 29 experimental trials was used to analyze the influences of spindle speed, soft load, and frictional load on mechanical properties, including tensile strength, yield strength, and elongation. An Extreme Gradient Boosting (XGBoost) regression model was developed in Python to predict these properties, achieving high accuracy with R2 values greater than 0.99. The model's predictions were validated by experimental data, with percentage errors of 0.757, − 0.101, and 1.125% for tensile strength, yield strength, and elongation, respectively. Optimal parameters were found to be a spindle speed of 2400 revolutions per minute, soft load of 700 Nm, and frictional load of 1400 Nm. Under these conditions, Sample 2 (interlocked design) exhibited improved mechanical properties, including a tensile strength of 652.09 MPa, yield strength of 583.57 MPa, and elongation of 21.22%. X-ray examination showed that all welds were void-free, while microstructural analysis through Scanning Electron Microscopy, Energy Dispersive x-ray Spectroscopy, and Electron Backscatter Diffraction revealed a uniform grain structure, efficient elemental diffusion, and a high density of high-angle and Sigma 3 (Σ3) twin boundaries. These results confirm the effectiveness of the interlocking geometry and the combined RSM and ML optimization for high-performance dissimilar metal joints.