Comparative experiments are performed in friction stir welding (FSW) of dissimilar Al/Mg alloys with and without assistance of ultrasonic vibration. Metallographic characterization of the welds at transverse cross sections reveals that ultrasonic vibration induces differences in plastic material flow in two conditions. In FSW, the plastic material in the peripheral area of shoulder-affected zone (SAZ) tends to flow downward because of the weakening of the driving force of the shoulder, and a plastic material insulation layer is formed at the SAZ edge. When ultrasonic vibration is exerted, the stirred zone is divided into the inner and outer shear layers, the downward material flow trend of the inner shear layer disappears and tends to flow upward, and the onion-ring structure caused by the swirl motion is avoided in the pin-affected zone. By improving the flow behavior of plastic materials in the stirred zone, ultrasonic vibration reduces the heat generation, accelerates the heat dissipation in nugget zone and changes the thermal cycles, thus inhibiting the formation of intermetallic compound layers.
近年来,深度学习越来越广泛地应用于自然语言处理领域,人们提出了诸如循环神经网络(RNN)等模型来构建文本表达并解决文本分类等任务。长短时记忆(long short term memory,LSTM)是一种具有特别神经元结构的RNN。LSTM的输入是句子的单词序列,模型对单词序列进行扫描并最终得到整个句子的表达。然而,常用的做法是只把LSTM在扫描完整个句子时得到的表达输入到分类器中,而忽略了扫描过程中生成的中间表达。这种做法不能高效地提取一些局部的文本特征,而这些特征往往对决定文档的类别非常重要。为了解决这个问题,该文提出局部化双向LSTM模型,包括MaxBiLSTM和ConvBiLSTM。MaxBiLSTM直接对双向LSTM的中间表达进行max pooling。ConvBiLSTM对双向LSTM的中间表达先卷积再进行max pooling。在两个公开的文本分类数据集上进行了实验。结果表明,局部化双向LSTM尤其是ConvBiLSTM相对于LSTM有明显的效果提升,并取得了目前的最优结果。 相似文献