<p>The present work aims to investigate the wear behavior of aluminum 5356 alloy samples fabricated by the wire arc additive manufacturing process. An L9 orthogonal array is constructed based on different combinations of parameters, namely load (5&#xa0;N, 15&#xa0;N, 25&#xa0;N), frequency (0.4&#xa0;Hz, 1&#xa0;Hz, 1.6&#xa0;Hz) and testing duration (10&#xa0;min, 15&#xa0;min and 25&#xa0;min) to perform a wear test. A machine learning technique combining Particle Swarm Optimization (PSO) and the limited-memory Broyden-Fletcher Goldfarb-Shanno with Boundaries (L-BFGS-B) algorithm is used to predict material loss and identify the optimum process parameters that minimize material loss during a wear study. Additionally, Taguchi analysis is used to examine the impact of wear test parameters on mass loss and validate the&#xa0;results predicted by the proposed machine learning approach. A minimum mass loss of 0.00177&#xa0;g is observed at a testing load of 25&#xa0;N, frequency of 0.4&#xa0;Hz and duration of 10&#xa0;min. The machine learning-based optimization methodology proposed in the present investigation predicted the mass loss with a minimum absolute error of 7.646 × 10<sup>−5</sup>. The Taguchi analysis reveals that testing frequency has a significant impact on mass loss, followed by testing load and time.</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

Tribological Performance Optimization of Wire Arc Additively Manufactured Al 5356 Alloy Using Particle Swarm Algorithm

  • Jianzhe Jing,
  • Heyan Zhao,
  • P. Sasikumar,
  • Yanhai Cheng,
  • T. Sathies,
  • N. Jeyaprakash

摘要

The present work aims to investigate the wear behavior of aluminum 5356 alloy samples fabricated by the wire arc additive manufacturing process. An L9 orthogonal array is constructed based on different combinations of parameters, namely load (5 N, 15 N, 25 N), frequency (0.4 Hz, 1 Hz, 1.6 Hz) and testing duration (10 min, 15 min and 25 min) to perform a wear test. A machine learning technique combining Particle Swarm Optimization (PSO) and the limited-memory Broyden-Fletcher Goldfarb-Shanno with Boundaries (L-BFGS-B) algorithm is used to predict material loss and identify the optimum process parameters that minimize material loss during a wear study. Additionally, Taguchi analysis is used to examine the impact of wear test parameters on mass loss and validate the results predicted by the proposed machine learning approach. A minimum mass loss of 0.00177 g is observed at a testing load of 25 N, frequency of 0.4 Hz and duration of 10 min. The machine learning-based optimization methodology proposed in the present investigation predicted the mass loss with a minimum absolute error of 7.646 × 10−5. The Taguchi analysis reveals that testing frequency has a significant impact on mass loss, followed by testing load and time.