Scholarlink

<p>The effectiveness of gas tungsten arc welding (GTAW) in fabricating high-quality joints in AISI 1045 steel depends significantly on various pre-welding practices, with preheating being one of the most critical. Recently, AISI 1045 medium carbon steel has become increasingly popular due to its impressive toughness, strength, and resistance to wear. As a result, it is commonly used in a wide range of industries, including automotive, petroleum, and piping, for components such as axles, bolts, connecting rods, spindles, torsion bars, worms, light gears, crankshafts, and more. Preheating primarily reduces hardness in the heat-affected zone, which minimizes the risk of cracking and brittleness often seen in medium carbon steels. By decreasing the temperature differential between the weld area and surrounding material, preheating mitigates thermal shock and potential warping. Furthermore, it enhances fusion between the base metal and the weld filler, leading to improved weld quality and mechanical properties. Therefore, to predict weld quality in terms of strength and hardness under preheated conditions, this study introduces a machine learning (ML)–based framework aimed at optimizing the GTAW process. This research employs various ML models, including artificial neural networks (ANN), Tree models, Gaussian process regression (GPR), and support vector machines (SVM), to forecast ultimate tensile strength (UTS) and hardness in both preheated and non-heated conditions for AISI 1045 steel joints. Key GTAW parameters—welding current (<i>W</i><sub>C</sub>), welding gas flow rate (<i>W</i><sub>GFR</sub>), and welding speed (<i>W</i><sub>S</sub>)—were systematically adjusted. Among the models, ANN demonstrated the best performance, achieving <i>R</i><sup>2</sup> values exceeding 0.99. Sensitivity analysis revealed that <i>W</i><sub>C</sub> had the most significant impact on response measures. Multi-objective optimization using the non-dominated sorting genetic algorithm (NSGA-II) yielded optimal parameters: <i>W</i><sub>C</sub> = 129.95 A, <i>W</i><sub>GFR</sub> = 9.74 lit/min, and <i>W</i><sub>S</sub> = 220&#xa0;mm/min. This approach resulted in improved mechanical properties, confirming that ML-driven optimization enhances GTAW process efficiency and weld quality.</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-driven optimization of mechanical properties in gas tungsten arc welding of preheated AISI 1045 steel

  • 期刊论文

  • 发表日期：

    2025年5月26日

查看全文
  • Muhammad Jawad,
  • Muhammad Sana,
  • Muhammad Asad Ali,
  • Mehdi Tlija,
  • Mirza Jahanzaib,
  • Muhammad Umar Farooq

摘要

The effectiveness of gas tungsten arc welding (GTAW) in fabricating high-quality joints in AISI 1045 steel depends significantly on various pre-welding practices, with preheating being one of the most critical. Recently, AISI 1045 medium carbon steel has become increasingly popular due to its impressive toughness, strength, and resistance to wear. As a result, it is commonly used in a wide range of industries, including automotive, petroleum, and piping, for components such as axles, bolts, connecting rods, spindles, torsion bars, worms, light gears, crankshafts, and more. Preheating primarily reduces hardness in the heat-affected zone, which minimizes the risk of cracking and brittleness often seen in medium carbon steels. By decreasing the temperature differential between the weld area and surrounding material, preheating mitigates thermal shock and potential warping. Furthermore, it enhances fusion between the base metal and the weld filler, leading to improved weld quality and mechanical properties. Therefore, to predict weld quality in terms of strength and hardness under preheated conditions, this study introduces a machine learning (ML)–based framework aimed at optimizing the GTAW process. This research employs various ML models, including artificial neural networks (ANN), Tree models, Gaussian process regression (GPR), and support vector machines (SVM), to forecast ultimate tensile strength (UTS) and hardness in both preheated and non-heated conditions for AISI 1045 steel joints. Key GTAW parameters—welding current (WC), welding gas flow rate (WGFR), and welding speed (WS)—were systematically adjusted. Among the models, ANN demonstrated the best performance, achieving R2 values exceeding 0.99. Sensitivity analysis revealed that WC had the most significant impact on response measures. Multi-objective optimization using the non-dominated sorting genetic algorithm (NSGA-II) yielded optimal parameters: WC = 129.95 A, WGFR = 9.74 lit/min, and WS = 220 mm/min. This approach resulted in improved mechanical properties, confirming that ML-driven optimization enhances GTAW process efficiency and weld quality.