Scholarlink

<p>In this study a new approach was taken to designing the mechanical properties of tungsten heavy alloy (THA), based on hybrid artificial neural networks (ANNs) and non-sorting genetic algorithm (NSGA-2) multi-objective optimization. The influence of nickel, cobalt in wt.% and cold working parameters in % on microstructure and mechanical properties such as ultimate tensile strength (UTS), yield strength (YS), percentage elongation (EL) and Vickers hardness (HV) of THA was studied numerically and experimentally. The mechanical properties of the THA were determined by means of a static tensile test and hardness measurements. The microstructure of the THA was characterized using electron microscopy. Firstly, the ANNs were used to modeling of mechanical properties for THA. A multilayer perceptron of ANN architecture with the tree neurons in the input layer, ten neurons in the hidden layer and four neurons in the output layer (3-10-4) with a logarithmic-sigmoidal activation function were selected for the study. High correlation coefficients (<i>R</i>-value) 0.9864, 0.9810, and 0.9925 for training, validation and test sets, respectively, indicated that data predicted by ANN were in good agreement with experimental results. The best validation performance was obtained after 6 epochs, for which the smallest mean square error (MSE = 0.0004268) was obtained. The optimization objectives included simultaneous maximization of UTS, YS, %El and HV for a THA composite. Optimization of chemical composition and cold working parameter was carried out in the following ranges: nickel from 4.0 to 7.2, cobalt from 0.8 to 4.0, and cold working from 0 to 50%. The NSGA-2 was found set of forty-eight <i>Pareto</i> solutions. For example the solution with the highest UTS = 1507 [MPa], YS = 1453 [MPa], HV = 41 and the lowest percentage elongation EL = 2% was found for the following parameters: Ni = 7.2, Co = 2.9 (wt.%) and the cold working 50 [%]. Moreover, the highest percentage elongation EL = 27% and the lowest UTS = 968 [MPa], and YS = 521 [MPa] the genetic algorithm was found for the following solutions: Ni = 6.3 (wt.%), Co = 2.2 (wt.%) and cold working cw = 0 (%).</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

Influence of Post-processing Heat Treatment and Plastic Deformation on the Microstructure and Mechanical Properties of Tungsten Heavy Alloy: Experimental Studies and Artificial Neural Network Modeling

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月8日

查看全文
  • Paweł Skoczylas,
  • Radosław Winiczenko,
  • Ryszard Sitek

摘要

In this study a new approach was taken to designing the mechanical properties of tungsten heavy alloy (THA), based on hybrid artificial neural networks (ANNs) and non-sorting genetic algorithm (NSGA-2) multi-objective optimization. The influence of nickel, cobalt in wt.% and cold working parameters in % on microstructure and mechanical properties such as ultimate tensile strength (UTS), yield strength (YS), percentage elongation (EL) and Vickers hardness (HV) of THA was studied numerically and experimentally. The mechanical properties of the THA were determined by means of a static tensile test and hardness measurements. The microstructure of the THA was characterized using electron microscopy. Firstly, the ANNs were used to modeling of mechanical properties for THA. A multilayer perceptron of ANN architecture with the tree neurons in the input layer, ten neurons in the hidden layer and four neurons in the output layer (3-10-4) with a logarithmic-sigmoidal activation function were selected for the study. High correlation coefficients (R-value) 0.9864, 0.9810, and 0.9925 for training, validation and test sets, respectively, indicated that data predicted by ANN were in good agreement with experimental results. The best validation performance was obtained after 6 epochs, for which the smallest mean square error (MSE = 0.0004268) was obtained. The optimization objectives included simultaneous maximization of UTS, YS, %El and HV for a THA composite. Optimization of chemical composition and cold working parameter was carried out in the following ranges: nickel from 4.0 to 7.2, cobalt from 0.8 to 4.0, and cold working from 0 to 50%. The NSGA-2 was found set of forty-eight Pareto solutions. For example the solution with the highest UTS = 1507 [MPa], YS = 1453 [MPa], HV = 41 and the lowest percentage elongation EL = 2% was found for the following parameters: Ni = 7.2, Co = 2.9 (wt.%) and the cold working 50 [%]. Moreover, the highest percentage elongation EL = 27% and the lowest UTS = 968 [MPa], and YS = 521 [MPa] the genetic algorithm was found for the following solutions: Ni = 6.3 (wt.%), Co = 2.2 (wt.%) and cold working cw = 0 (%).