APPLICATION OF ENSEMBLE TRAINING OF A STRUCTURAL CONTROLLER TO THE AMD BENCHMARK PROBLEM

A benchmark structural control problem has been proposed(1) in an atte mpt to evaluate the effectiveness of various control algorithms. The p roblem encompasses the design of an Active Mass Damper (AMD) control s ystem for a Multi-Degree-Of-Freedom (MDOF) building type structure sub jected to earthquake-type excitation. In vibration control for civil s tructures, linear quadratic optimal control is among the most popular techniques. Normally, this approach ignores the external excitation in the time-domain design process. In addition, this technique requires a full-order dynamic observer which is often unattainable. This paper focuses on the development of a new optimal control algorithm which in cludes the earthquake-type exctitation explicitly in the design of con trol systems and the use of prescribed-order, output feedback controll ers. In addition, this approach allows the inclusion of open-loop (fee dforward) as well as closed-loop (feedback) control terms in the contr oller design. The authors have previously designed an algorithm for fu ll state feedback controllers trained on an ensemble of earthquakes.(2 ) A cost functional is minimized on an ensemble of 'known' earthquakes , using analytic gradient information, in order to determine constant control gains. The gradients are obtained in explicit form. The contro l system is then validated by testing on 'unknown' earthquakes. The al gorithm is now modified to develop a prescribed-order, output feedback controller for a specific MDOF system model. (C) 1998 John Wiley and Sons, Ltd.