Improving fragility curves for controlled structures including sensor fault

In this study the effect of sensor fault on damage of active controlled nonlinear structures is investigated. A fault detection neural network and a fault accommodation neural network are proposed to reduce the effect of sensor fault. The fault detection network monitors structural responses and automatically detects faulty sensor that can reduce control performance and effectiveness, while the fault accommodation network accounts for the faulty sensors. Fault is accommodated by using data from the remaining healthy sensors to estimate what the faulty sensors should have been reading. To demonstrate the performance of proposed method, a 3-story full-scale nonlinear benchmark building and several ground motions are selected. The fragility curves are developed for structures using nonlinear dynamic time history analysis through the computer simulation. Fuzzy logic controller (FLC) is employed as a sample of intelligent controller to control the structure using actuators. Here, the fragility curves are represented by lognormal distribution function with two parameters and developed considering three performance levels specified in FEMA 356 includes the Immediate Occupancy (IO), the Life Safety (LS) and the Collapse Prevention (CP). Results show that the sensor fault can reduce the effect of controllers and even can increase the probability of damage compare to the uncontrolled structure. Moreover, results of the proposed method confirm its effectiveness for decreasing the probability of damage of faulty control system.

Keywords:

fault-tolerant controller (FTC) , fault detection , fault accommodation , fuzzy logic controller (FLC) , fragility curves