压电陶瓷驱动器的蠕变误差随时间呈现非线性变化,难以实时修正.提出基于BP神经网络的压电陶瓷蠕变预测方法,使用压电陶瓷驱动系统采集数据,对数据进行归一化处理,通过实验设计BP神经网络的隐含层数、隐含层节点数、节点转移函数和训练函数,构建BP神经网络预测模型,建立压电陶瓷蠕变与时间的关系.用BP神经网络模型对压电陶瓷蠕变进行了预测仿真,并将结果与实测数据进行了对比.结果表明,蠕变预测结果与实验数据的最大绝对误差均小于0.1 μm,最大蠕变误差均不超过0.6%,最大均方误差仅为0.0021,可见,BP预测模型具有较高的预测精度,可作为预测压电陶瓷蠕变误差的一种有效手段.%The creep errors of the piezoelectric ceramics have nonlinear change with the time, which is difficult to revise in real time. A creep prediction approach based on back propagation neural network is proposed for the piezoelectric ceramics. The data is collected by the piezoelectric ceramic driving system and normalized for prediction. The parameters of BP neural network including the number of hidden layers, the number of nodes in each hidden layer, the node transfer functions and the training function are designed by experiments.The prediction model of BP neural network is established, and the connection between the creep of the piezoelectric ceramic and the time is built. The creep of piezoelectric ceramics is predicted by the model of BP neural network, compared with the measured data, the results show that, using this prediction model the maximal absolute error is below 0.1 μm, the maximal creep error is below 0.6% and the maximal mean square error is 0.0021. So the BP neural network prediction model has a high prediction accuracy and can be applied to the creep prediction of the piezoelectric ceramics.
展开▼
