Predictive Active Control of Building Frames with a Solo Neurocontroller

Abstract

We present an efficient generalized control scheme using a solo neural network controller (neuro-controller) for the predictive single mode response control of building frames subjected to earthquake excitation. The control scheme is developed by considering the time delay that exists between the measurement of response and the actual application of the control force. Response measurements are taken from only one point of the structure. The neurocontroller is directly trained to provide the control force signal for a predetermined response reduction, called the target reduction. The neurocontroller is trained such that it can provide control force for target reduction for any building frame ranging between 5 and 15 stories. As an illustrative example, a 10-story shear-building frame is considered whose period of first mode of vibration is about three times that of the second mode of vibration. The efficiency of the control scheme is demonstrated by controlling the first mode response of the frame subjected to El Centro and Treasure Island earthquake records. It is shown that the control scheme provides response reductions close to the target reductions for the unknown problem used for testing the efficiency of the control scheme.