Designing a General Neurocontroller for Water Towers

Abstract

This study deals with the capabilities of artificial neural networks in learning to control water towers of different structural properties that are subjected to earthquakes. To this end, water towers were considered as single-degree-of-freedom systems. First, a number of water towers of different structural properties were controlled by the predictive optimal control method, and then the data collected through this control were used in the training of a general neural network controller, called the general neurocontroller. Capabilities of the general neurocontroller were tested in the control of a number of water towers with structural parameters different from, but in the range of, those used in its training. One of the aims of this study was the introduction of general neurocontrollers as ready-to-use devices that may be used in the design of actively controlled structures, in this case, water towers. Results of this numerical study were promising.