神经网络建模及其在食品包装机械中的应用

针对液压伺服系统的强非线性，提出了一种神经网络建模方法。在系统反馈输入信息的基础上，构造一样条函数和常数作为神经网络的输入量，以扩充网络输入量矩阵，使用传统BP算法离线训练网络，得到神经网络模型。应用于食品机械液压伺服系统的建模结果表明，所提出的神经网络模型训练过程收敛速度较快，模型精度高，可有效避免网络陷入局部最优值，而且不需要更多的系统信息。

The neural network modelling and its application to food and packaging machinery

Aiming to strong non-linearity of hydraulic systems, a method of neural network modelling is put forward, in which system backforward data, a spline function and a constant are chosen as the input augment matrix of the network. Meanwhile, the conventional BP algorithm is used to train the network. Application result to the hydraulic system in food and packaging machinery shows that the neural network has high learning speed and high-precision. It also shows that the phenomenon of local minimum can be effectively avoided and no more system data is necessary for training process.