Computationally, costly FE model is not suitable for large number of repetitive analyses which are often required in an optimization process. Therefore, the FE model for warpage is replaced by a simpler and more efficient predictive model using artificial neural networks(ANNs). The artificial neural network architecture used in this study is shown in Fig.4. It is designed using Matlab Neural Network Toolbox[9] The network model consists of one input,two hidden and one output layers. Hidden layers have 55 neurons each, whereas input and output layers have 5and 1 neurons,respectively. Neurons in input layer correspond to mold temperature (MoT), melt temperature(MeT),packing pressure(PP),packing pressure time(PPT) and cooling time(CT).Out put layer corresponds to warpage.
參考答案:咱们老朋友了啊，哈哈 再帮你一次：：：
计算显示, 昂贵的有限元分析模型不适合在探求最优化过程中反复适应的大量分析。因此，我们使用更为简单高效的基于人工神经元网络(ANNs)的预测模型来取代有限元分析模型以解决翘曲问题。本文采用的人工神经元网络结构图见Fig.4.。它是使用Matlab神经元网络工具箱设计完成的。该网络模型由一个输入层，两个隐藏层及一个输出层组成。每个隐藏层含有50个神经原, 其中输入层含5个神经原，输出层含1个。 输入层中的神经原用于响应模具温度(MoT)，熔化温度(MeT),保压压力(PP),保压压力时间(PPT) 及冷却(CT).输出层用于响应翘曲。
