Tungsten inert gas arc welding is not suitable for combining thick portions in one run. Activated flux TIG (A-TIG) welding can enhance weld penetration by 3–4 times in one pass. The base plate was 304 SS flat, and the binary flux used was a combination of SiO2 and TiO2 in the ratio of 4:1, 2.5:1 and 1:1. For creating experimental runs, a three-level, four-factor response surface technique based on Box–Wilson design was implemented. The process parameters were validated using analysis of variance (ANOVA). A reverse mode of automatic differentiation of simulated neural networks was created using MATLAB17 for evaluating width of the weld bead as well as penetration with activated flux TIG welding utilizing the four variables of heat input, flux ratio, gas flow rate, and root gap. The highest depth of penetration of 5 mm (full penetration) and bead width of 8.549 mm were accomplished using the heat input value of 0.85 in kJ/mm, flux ratio value of 4:1, 12 l/min of rate of flow of gas, as well as root gap as 1 mm. The current study additionally investigates activated flux's impact on mechanical properties and morphology such as hardness of metal using A-TIG welding. The microscopic makeup of the area that was welded has been examined employing an optical microscope as well as a scanning electron microscopy. The hardness of the welding zone has also been examined with a maximum value of 65 HRC. The ANN selected has a network structure of 4-10-3. Results indicate that the anticipated values by ANN comply well with that of the experimental ones.

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

Prognosis of the Geometry of Weld Bead Using Artificial Neural Networks Algorithm

  • Samarendra Acharya,
  • Avradip Das,
  • Santanu Das

摘要

Tungsten inert gas arc welding is not suitable for combining thick portions in one run. Activated flux TIG (A-TIG) welding can enhance weld penetration by 3–4 times in one pass. The base plate was 304 SS flat, and the binary flux used was a combination of SiO2 and TiO2 in the ratio of 4:1, 2.5:1 and 1:1. For creating experimental runs, a three-level, four-factor response surface technique based on Box–Wilson design was implemented. The process parameters were validated using analysis of variance (ANOVA). A reverse mode of automatic differentiation of simulated neural networks was created using MATLAB17 for evaluating width of the weld bead as well as penetration with activated flux TIG welding utilizing the four variables of heat input, flux ratio, gas flow rate, and root gap. The highest depth of penetration of 5 mm (full penetration) and bead width of 8.549 mm were accomplished using the heat input value of 0.85 in kJ/mm, flux ratio value of 4:1, 12 l/min of rate of flow of gas, as well as root gap as 1 mm. The current study additionally investigates activated flux's impact on mechanical properties and morphology such as hardness of metal using A-TIG welding. The microscopic makeup of the area that was welded has been examined employing an optical microscope as well as a scanning electron microscopy. The hardness of the welding zone has also been examined with a maximum value of 65 HRC. The ANN selected has a network structure of 4-10-3. Results indicate that the anticipated values by ANN comply well with that of the experimental ones.