磁流变阻尼器的模糊逼近

The fuzzy approximation of MR damper

由于磁流变液具有非线性特性,所以磁流变阻尼器的输入输出间具有很强的非线性关系。可准确描述其非线性特性的磁流变阻尼器正模型通常非常复杂,难以直接得到逆模型。考虑到某些模糊系统的万能逼近能力,本文提出用模糊系统来逼近磁流变阻尼器逆模型的新思路。根据自适应神经模糊推理系统原理,设计两个模糊系统分别逼近磁流变阻尼器的正模型和逆模型。研究结果表明:无论是正模型还是逆模型,对于训练数据,模糊系统均可以准确逼近,而对于检验数据也可比较准确逼近。正模型的逼近效果稍好,若要提高逆模型ANF IS的逼近精度,将以增加系统复杂性为代价。模糊逼近可以推广到其它的磁流变阻尼器模型中,特别是可对正模型未知的磁流变阻尼器进行建模与控制。

An MR damper has strong nonlinearity between inputs and output owing to the unknown nonlinearity of the MR suspension in it. It is quite difficult to describe the direct model of the MR damper, however, it is much more difficult for the inverse model. The paper presents a novel train of thoughts to approach to the inverse model of the MR damper by using the universal approximation of a kind of fuzzy system. Then two different fuzzy systems are designed to approximate the direct model and the inverse one on the basis of adaptive neuro-fuzzy inference system （ANFIS）. These two ANFIS are similar to their physical counterparts of the MR damper, just considering the number of the inputs and output. The numerical simulation proves that such two fuzzy systems can accurately approximate the direct model and inverse model of the MR damper for the train data, and well approximate for the check data. This idea can be extended to other models of MR dampers. Furthermore, it can be also used to model and control other MR damper with its direct model unknown.