Effect of noises on the active optimal control of partially observed structures for white random ground motion

The effects of measurement noise, number of observations, partial correlation between the measurement noises and the perturbation to the process noise (excitation) on the estimated state and the resulting controlled responses under stochastic excitation are investigated. Linear quadratic Gaussian (LQG) algorithm is used to determine the control force for the partially observed system. Although stochastic control of structures using LQG algorithm is widely reported in the literature, the effects of above parameters on the performance of the algorithm have not been properly investigated. Herein, this investigation is carried out with the help of two example problems, ten story and fifteen story building frames. The results of the study show that the use of Kalman filter for control of the partially observed system using LQG algorithm is only effective within a limited band of measurement noise. If the actual measurement noise deviates from that considered in theory then either the control of response is reduced or the algorithm destabilizes. Further, optimal position of actuator varies for different response quantities of interest but it occurs for the same value of covariance of noise for which maximum reductions in responses are observed. (C) 2016 Institute of Noise Control Engineering.