PID and LQR/LQG controllers have are known to be ineffective for systems suffering from parameter variations and broadband excitations. This paper presents a neural-network design for system identification and vibration suppression in a building structure with an active mass-damper. It is shown both numerically and experimentally that the neural-network controller can reliably identify system dynamics and effectively suppress vibration. For the experimental model, which has a fundamental frequency of about 0.96 Hz, the steady-state vibration amplitude under resonance and random excitation are reduced by 80% and 70%, respectively. In addition, the peak-to-peak displacement under the 7.1 Richer scale Ji-Ji earthquake, Taiwan (Sep. 21, 1999) is effectively reduced by 80%. The controller is also shown to be robust to variations in system parameters.

No abstract

A self-organized, five-layer neuro-fuzzy model is developed to model the dynamics and to forecast air cargo and airline passenger by using the input of previous years’ consumer price index, exchange rate, gross national product, and number of cargo volume/passenger traffic. Simulation results show that the neuro-fuzzy model is more effective than neural network in prediction and accurate in forecasting. The effectiveness in modeling, prediction and forecasting is validated, and the input error from the series-parallel identification method is attenuated by the neuro-fuzzy model to yield better forecasting results.

No abstract

On September 21, 1999, Taiwan was slammed by Taiwan's biggest quake since 1935. The magnitude 7.6 tremor with its epicenter in central Nantou County killed more than 2,300 persons and damaged 82,000 housing units. With the trend toward taller and more flexible building structures, the use of vibration control devices, passive as well as active, as means of structural protection against strong wind and earthquakes have received significant attention in recent years. A mass-damper shaking table system has been considered as means for vibration suppression to external excitation and disturbances.In this paper, the direct experiment method is adopted to determine the control gains for better performance index. No explicitly system identification of the plant dynamics, no membership function and thus no fuzzification-defuzzification operation are required. For effective control performance, a neural classifier controller with genetic algorithm is developed.Compared with the conventional PI controller, neural network and fuzzy controller, the neural classifier controller using genetic algorithm has been presented with the effectiveness of the vibration suppression control. Experimental results show that the neural classifier controller remains effective for building structure vibration suppression under free vibration and forced vibration excitation.