说明：  基于多层前向网络的诊断模型在设备 故障诊断领域应用比较广泛。 但在多层前向网络的 设计和训练问题上, 单隐层的隐层单元数选取一直非常困难, 一般采用试凑 法, 既费时又 不一定保证收敛。
(Diagnostic model based on multilayer feedforward networks in the field of equipment fault diagnosis used widely. But before multilayer design and training issues on the network, the number of hidden layer unit single hidden layer selection has been very difficult, generally use trial and error method, time-consuming and does not necessarily guarantee convergence.)