Scholarlink

<p>The present research proposed a machine learning (ML)-based model and meta-model framework for design of friction stir welding (FSW) process applicable to alloys in general. A fourfold model had trained utilizing <i>K</i>-mode clustering, support vector regression, random forest regression, <i>K</i>-nearest neighbors regression and gradient boost algorithms. These meta models build a constitutive framework to progressively identify the required set of parameters when the sheet thickness and alloy(s) are to be welded, which is known. The proposed framework aims to integrate the multiple interconnected predictive ML model to consider all aspects of design of FSW for a wide range of alloys unlike the other reported AI/ML applications, which is limited to specific prediction only. The approach is capable of approximating the possible combinations of tool materials, tool geometry, rotational speed and welding speed of&#xa0;a FSW process. The design of FSW process for a 5-mm-thick AA5083 sheet was investigated with the proposed approach to validate the design approach. Eventually, experimental fabrication of few FSW of AA5083 sheet and their tensile testing were performed with the designed set of parameters and compared their efficiency to validate the model-derived design data. It demonstrated the effectiveness of the proposed design approach with an average error of ± 6%.</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

A Generalized Approach for Comprehensive Design of Friction Stir Welding Process Using Machine Learning Model and Meta-model Framework

  • 期刊论文

  • 发表日期：

    2025年9月5日

查看全文
  • Gaurav Chauhan,
  • Ashutosh Kumar Gupta,
  • Mrinal Sahu,
  • Sudip Kumar Sinha,
  • Subhas Ganguly

摘要

The present research proposed a machine learning (ML)-based model and meta-model framework for design of friction stir welding (FSW) process applicable to alloys in general. A fourfold model had trained utilizing K-mode clustering, support vector regression, random forest regression, K-nearest neighbors regression and gradient boost algorithms. These meta models build a constitutive framework to progressively identify the required set of parameters when the sheet thickness and alloy(s) are to be welded, which is known. The proposed framework aims to integrate the multiple interconnected predictive ML model to consider all aspects of design of FSW for a wide range of alloys unlike the other reported AI/ML applications, which is limited to specific prediction only. The approach is capable of approximating the possible combinations of tool materials, tool geometry, rotational speed and welding speed of a FSW process. The design of FSW process for a 5-mm-thick AA5083 sheet was investigated with the proposed approach to validate the design approach. Eventually, experimental fabrication of few FSW of AA5083 sheet and their tensile testing were performed with the designed set of parameters and compared their efficiency to validate the model-derived design data. It demonstrated the effectiveness of the proposed design approach with an average error of ± 6%.