Scholarlink

<p>A finite element analysis model for embedded substrate chip scale package (CSP) solder joints was established. Finite element analysis under torsional loading was conducted on this model, revealing the stress distribution patterns in the embedded substrate CSP solder joints. A torsional strain measurement platform was constructed, and strain measurement experiments on the embedded substrate were completed, verifying the accuracy of the simulation results. Solder joint diameter, pad diameter, and solder joint height were selected as design variables. Using torsional stress of solder joints as the optimization target, orthogonal experiments combined with grey relational analysis were employed to perform significance analysis and ranking of the influence magnitude of solder joint structural parameters. The genetic algorithm-backpropagation (GA-BP) neural network was utilized to predict torsional stress in embedded substrate CSP solder joints and optimize the structural parameters. Results demonstrate that the GA-BP neural network achieved an average relative error of 0.996% in torsional stress prediction. The optimized solder joint torsional stress was reduced by 24% through optimization, successfully accomplishing the optimization objectives.</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

Stress analysis and optimization of CSP solder joints in embedded substrates under torsional loading using a GA-BP neural network

  • 期刊论文

  • 发表日期：

    2026年1月8日

查看全文
  • Lixiang Huang,
  • Chunyue Huang,
  • Chao Gao,
  • Gui Wang

摘要

A finite element analysis model for embedded substrate chip scale package (CSP) solder joints was established. Finite element analysis under torsional loading was conducted on this model, revealing the stress distribution patterns in the embedded substrate CSP solder joints. A torsional strain measurement platform was constructed, and strain measurement experiments on the embedded substrate were completed, verifying the accuracy of the simulation results. Solder joint diameter, pad diameter, and solder joint height were selected as design variables. Using torsional stress of solder joints as the optimization target, orthogonal experiments combined with grey relational analysis were employed to perform significance analysis and ranking of the influence magnitude of solder joint structural parameters. The genetic algorithm-backpropagation (GA-BP) neural network was utilized to predict torsional stress in embedded substrate CSP solder joints and optimize the structural parameters. Results demonstrate that the GA-BP neural network achieved an average relative error of 0.996% in torsional stress prediction. The optimized solder joint torsional stress was reduced by 24% through optimization, successfully accomplishing the optimization objectives.