神经网络多参量地震模拟试验台控制算法

针对液压伺服地震模拟振动台控制系统非线性、构件动力非线性、系统频率响应曲线不平直的情况,在原湖南大学结构实验室振动台FTGT控制器位移内闭环控制的基础上,利用微机开发了一种多参量控制的外闭环控制回路,将力、位移、速度、加速度误差以一定的权值加入控制回路,组建用于加速度波形跟踪的外闭环控制器,以提供更平直的频率响应特性,使输入台面的峰值能量与地震波峰值能量相一致。并且针对实验过程中,结构物理参数的不断变化,将神经网络加入到多参量控制权值的在线调整过程,最终实现加速度波形的高频带高精度跟踪。经过台面模型试验,多参量控制外闭环有较好的自适应效果。

Three Variable Control Algorithm Based on Neural Network for Shaking Table

To solve the non-linear of shaking table and structure, this article has developed a three variables control outerloop base on the displacement innerloop shaking table, used of computer to build up the acceleration controled outerloop, provided an accurate energy peak of shaking table as seismic waves. And forthe physical parameters of structure are constantly changing, a neural network has added to adjust the power of the three variables on-line. Experiments show that TVC outer-loop can provide better shaking table result.