Abstract

In this work, an attempt is made to identify the optimized parameter combination for improved creep and corrosion properties of AA5083-H111 alloy weldments processed by Spin-Arc gas metal arc welding (SA-GMAW) process. For this, the Artificial Neural Network (ANN) coupled with Genetic Algorithm (GA) was used as a statistical tool. Experiments were conducted by considering the input parameters namely welding current, filler spinning speed and filler spin diameter. The weld quality was assessed by measuring microhardness, corrosion resistance and steady-state creep rate. Initially, ANN was used to establish the relationship between input and output process variables. Different learning algorithms such as quick propagation (QP), back batch propagation (BBP) and incremental batch propagation (IBP) were used for predicting the output parameters. Prediction accuracy of the different learning algorithms was compared, and the best algorithm was used for the GA optimization technique. The optimum parameters were found to be 134 A of welding current, 1050 rpm of filler spinning speed and 1 mm of filler spin diameter. Among the input parameters, the filler spinning speed was identified as the most influential factor (40.11%) that effected the formation and distribution of second phase particles through which improved corrosion and creep properties were achieved.

中文翻译：

---

ANN建模和遗传算法优化在改善自弧弧焊AA5083-H111合金蠕变和腐蚀性能中的应用

摘要

在这项工作中，尝试确定优化的参数组合，以改善通过自旋电弧气体金属电弧焊（SA-GMAW）工艺加工的AA5083-H111合金焊件的蠕变和腐蚀性能。为此，将人工神经网络（ANN）与遗传算法（GA）结合起来用作统计工具。通过考虑输入参数即焊接电流，填料纺丝速度和填料纺丝直径进行实验。通过测量显微硬度，耐腐蚀性和稳态蠕变速率来评估焊接质量。最初，使用ANN建立输入和输出过程变量之间的关系。使用不同的学习算法（例如快速传播（QP），后批传播（BBP）和增量批传播（IBP））来预测输出参数。比较了不同学习算法的预测精度，并将最佳算法用于遗传算法优化技术。最佳参数为焊接电流134 A，填料旋转速度1050 rpm和填料旋转直径1 mm。在输入参数中，填料纺丝速度被认为是影响第二相颗粒形成和分布的最有影响力的因素（40.11％），通过该相变颗粒可以改善腐蚀和蠕变性能。