斜拉索基于MR阻尼器的神经网络半主动控制  

SEMI-ACTIVE NEUROCONTROL ON STAY CABLE WITH MR DAMPER

作　　者：肖志荣[1] 孙炳楠[2] 

机构地区：[1]浙江科技学院建工学院,浙江杭州310023 [2]浙江大学建工学院,浙江杭州310027

出　　处：《工程力学》2010年第1期183-187,共5页Engineering Mechanics

基　　金：浙江省教育厅科研项目(Y200804360);浙江省住房和城乡建设厅项目(0903)

摘　　要：该文利用神经网络强大的学习和非线性拟合能力模拟了MR阻尼器的逆动力性能。为了提高神经网络的计算性及泛化性,采用了Levenberg-Marquardt算法与贝叶斯正规化法相结合的方法。与此同时,利用MR阻尼器的神经网络逆模型,提出了一种新的斜拉索神经网络半主动控制策略。为验证所提控制方法的有效性,针对典型算例进行了数值分析,并将其与LQR主动控制方法进行了比较。得出结论:所提神经网络半主动控制方法是有效的,与LQR主动控制效果相比,效果略差,但相差不大。The inverse dynamic behaviour of MR damper is emulated based on neural network theory. In order to improve the calculation and generalization of the neural network model of MR damper, the Levenberg-Marquardt algorithm and Bayes Regularization method are adopted here. With the neural network model of the MR damper, a new control strategy for the stay cables is proposed. A comparison with LQR active control method is also made to check the validity of the proposed control method. Numerical results show that the proposed semi-active neurocontrol method is effective, and its results are comparable to the LQR.

关 键 词：半主动控制:神经网络:Levenberg-Marquardt算法 斜拉索 MR阻尼器 

分 类 号：TU311.3[建筑科学—结构工程] U441.3[建筑科学—桥梁与隧道工程]