基于神经网络的结构振动智能主动容错控制算法研究  

Smart active fault tolerant control algorithm based on neural network for structural vibration control

作　　者：雷永勤[1,2] 杜永峰[1,2] 

机构地区：[1]教育部西部土木工程防灾减震研究中心,兰州730050 [2]兰州理工大学防震减灾研究所,兰州730050

出　　处：《振动与冲击》2014年第13期117-122,160,共7页Journal of Vibration and Shock

基　　金：国家自然科学基金(51178211)

摘　　要：针对重大工程结构振动控制系统传感器失效问题,提出了基于多路RBFNN的控制系统动力特性辨识及传感器故障检测方法,实现了重大工程结构控制系统传感器失效时的智能主动容错控制。通过把传感器反馈信号分解为多路信号进行系统辨识,实现反馈信号的分离,避免正常传感器与故障传感器之间的相互干扰;当传感器正常工作时,控制器按设定的控制算法确定控制力,当多路RBFNN检测到某个传感器失效时,控制系统将自动剔除该传感器信号,切换到考虑此传感器失效时的振动控制算法确定控制力。通过对AMD控制Benchmark模型进行仿真分析,验证了所提出的基于多路RBFNN传感器故障检测技术以及智能主动容错控制策略的有效性和优越性。In allusion to the sensors failure in vibration control system in huge engineering structures,the methods of dynamic characteristic identification and sensor failure detection of control system were proposed based on multi-channel RBF neural network,and a smart active fault tolerant control algorithm was presented for huge engineering structures.The mutual interference between normal sensor and failure sensor in the process of sensor failure detection was solved by decomposing the sensor feedback signals into multi-channel signals in dynamic characteristic identification.When sensors work normally,the controller calculates the control force according to the preset control algorithm.When any sensor fails, the control system will get rid of the failure sensor signal automatically and the controller will switch to the control algorithm which considers this sensor’s failure.The effectiveness and superiority of the sensor failure detection strategy and the presented smart active fault control algorithm based on multi-channel RBF neural network were proved by the simulation of Benchmark model controlled by AMD under earthquake.

关 键 词：结构振动控制 智能容错控制 AMD控制 地震响应 控制算法 

分 类 号：TU352.12[建筑科学—结构工程]