Abstract:

MSI, as an active isolation actuator, had advantages including noncontact,high response frequency, high reliability and long lifespan. However, its potential was not fully explored due to the nonlinear and hysteretic behavior in dynamic environments, and there was limited research work in the area. This paper proposed a new ANNbased approach to model the dynamics of MSI. A HA was developed to train the ANN to improve the model accuracy. Results clearly show that the ANN model with the HA approach outperforms the back propagation(BP) approach and modified genetic algorithm(MGA) approach. An experimental platform was developed to test the performance of the magnetic suspension vibration isolation system and the proposed modeling approach. The output force response of the active and passive system under the same excitation are measured. The results show that the active with proposed model control system has much better performance in vibration isolation.

Key words: artificial neural network(ANN), hybrid approach(HA), magnetic suspension isolator(MSI), controller

摘要：

针对磁悬浮隔振器动态电磁力模型存在非线性及磁滞且很难建立其精确模型的问题,提出了基于BP算法、改进遗传（MGA）算法的混合算法的BP神经网络的模型辨识方法，建立了磁悬浮隔振器动态电磁力气隙电流关系的模型。结果表明，基于混合训练算法辨识得到的模型具有更高的精度，能够满足磁悬浮隔振器动态电磁力模型辨识需求。最后，搭建了磁悬浮隔振实验平台，建立控制模型,并验证了辨识模型的有效性。

关键词: 人工神经网络, 混合算法, 磁悬浮隔振器, 改进遗传算法, 控制器