人工神经网络在切削颤振类别诊断中的应用

DIAGNOSIS OF CUTTING CHATTER CATEGORIES WITHARTIFICIAL NEURAL NETWORK

讨论了再生型、模态耦合型和滞后型３种主要切削颤振类别的特征表达参数。使用多层感知器型的前馈神经网络，研究了切削颤振类别的综合诊断方法。通过切削试验获取神经网络的训练和测试样本，使用误差逆向传播算法对神经网络进行训练。训练和测试过的神经网络可对切削颤振的类别作出正确的诊断。

オhe characteristic representation parameters are presented which describe three dominant categories of cutting chatter including the regenerative, modecoupling and lagging types. The feedforward neural network of multilayer perceptron architecture is used. Comprehensive diagnosis method of cutting chatter categories is investigated with the neural network. Training and testing samples of neural network are collected through cutting experiments. Neural network is trained with the error backpropagation learning algorithm. The developed neural network can correctly categorize chatter created in cutting system. The cutting experiments verified the effectiveness of proposed diagnosis method. The machine tool used is CA6140 lathe.