Bîrsan Dan Cătălin, Susac Florin, Teodor Virgil Gabriel
Department of Manufacturing Engineering, Faculty of Engineering, "Dunărea de Jos" University of Galați, Str. Domnească No. 111, 800201 Galati, Romania.
Materials (Basel). 2024 Sep 18;17(18):4586. doi: 10.3390/ma17184586.
The quality of the refill friction stir spot welding (RFSSW) process is heavily dependent on the selected welding parameters that influence the resultant joint characteristics. Thermomechanical phenomena integral to the process were investigated using finite element (FE) analysis on two dissimilar materials. This FE analysis was subsequently validated through controlled experiments to ensure reliability. An artificial neural network (ANN) was employed to create a neural model based on an experimental setup involving 120 different sets of welding parameters. The parameters adjusted in the experimental plan included pin penetration depth, rotational speed, retention time, and positioning relative to material hardness. To assess the neural model's accuracy, outputs such as maximum temperature and normal stress at the end of the welding process were analyzed and validated by six data sets selected for their uniform distribution across the training domain.
填充式搅拌摩擦点焊（RFSSW）工艺的质量在很大程度上取决于所选的焊接参数，这些参数会影响最终接头的特性。通过对两种不同材料进行有限元（FE）分析，研究了该工艺中不可或缺的热机械现象。随后，通过控制实验对该有限元分析进行了验证，以确保其可靠性。基于一个涉及120组不同焊接参数的实验装置，采用人工神经网络（ANN）创建了一个神经模型。实验方案中调整的参数包括销钉穿透深度、转速、保持时间以及相对于材料硬度的定位。为了评估神经模型的准确性，对焊接过程结束时的最高温度和法向应力等输出进行了分析，并通过从训练域中均匀分布选取的六个数据集进行了验证。
