论文部分内容阅读
采用不同的摩擦头搅拌速度、焊接速度和单位长度上的旋转次数进行AZ91D镁合金和6061铝合金的搅拌摩擦焊,获得表征焊接接头寿命的指标数据;并将其作为计算机预测模型的训练样本和验证样本,构建了两种预测模型进行试验。结果表明,3×12×2三层结构模型的训练误差过大无法满足试验要求,3×12×8×2四层结构预测模型的训练误差和预测误差均较小,预测精度较高,具有较大的实用价值。
Different friction head stirring speed, welding speed and the number of rotations per unit length were used to carry out the friction stir welding of AZ91D magnesium alloy and 6061 aluminum alloy to get the index data which can represent the life of welded joint. And it was used as the training sample of computer prediction model and The samples were validated and two prediction models were constructed for testing. The results show that the training error of 3 × 12 × 2 three-layer structure model is too large to meet the test requirements. The training errors and prediction errors of the 3 × 12 × 8 × 2 four-layer structure prediction model are small and the prediction accuracy is high. Larger practical value.