实时子结构试验的时间滞后补偿方法的研究

Investigation on Compensation for the Delay of the Real-time Substructure Experiment

采用神经网络预测的方法,对滞后时间加以补偿。首先建立由磁流变液控制装置、液压伺服作动器、测量系统和数据采集系统所组成的实时子结构试验控制系统,测量该系统滞后时间的具体数值,分析时间滞后对系统稳定性的影响。在此基础上,根据滞后时间确定神经网络预测的实时计算步长,应用训练好的神经网络来补偿滞后时间,使得数值子结构和物理子结构能协调变形。最后,对一栋3层剪切型结构的磁流变液半主动控制系统进行了实时子结构试验,验证了时间滞后补偿方法的有效性。

It suggested that one can make up for time-lag though the method of neural network prediction. Firstly, set up the real-time substructure experiment control system which is made up of magnetic current changing fluid control device, hydraulic pressure servo actuator, measure system and data acquisition system. It is used to measure the specific time-lag data of the system and to analyze the impacts of time-lag on the stability of the system. Based on the above, compute the length of the step according to the real time of the network prediction on the time-lag, train the neural network well enough to make up for the time-lag, which makes the numerical substructure and the physical substructure distort coordinately. Lastly, we have done a real-time substructure experiment on a three-shear structure MR semi-active control system which verified the validity of the method of making up for time-lag.