Files in this item
|(no description provided)|
|Title:||A Study of Neural and Conventional Control Paradigms in Active Digital Control|
|Doctoral Committee Chair(s):||Ghaboussi, J.|
|Department / Program:||Civil Engineering|
|Degree Granting Institution:||University of Illinois at Urbana-Champaign|
|Abstract:||The objective of this study is to investigate/compare some practical issues in structural vibration control associated with conventional and neural approaches. For this purpose two fundamental structural control problems are chosen for investigation. The first problem is a base isolation system which is an example of rigid body structural control. The second problem involves the reduction of excessive vibrations of a nonlinear (geometrically) cantilevered plate structure which is an example of flexible structural control.
For the active isolation example, the external disturbances are assumed to be measurable, while for the plate problem they are assumed to be unmeasurable. In the former case the conventional control is of a feedforward type while for the latter case, it is constituted from an optimal state feedback regulator (LQR).
Results from the conventional control of the base isolation problem indicates that although the tracking characteristics of the controlled plant is satisfactory, the total response is associated with high frequency accelerations which makes the overall behavior of the controlled process unacceptable. Major contributing factors to this behavior are identified.
By replacing the conventional controller with a trained feedforward neuro-controller a vast improvement in the control system's response is observed. Major contributing factors to this behavior are identified.
Results from the conventional control of the flexible plate example indicates that LQR control scheme can produce unacceptable or near failure conditions for cases where the quality of the model is poor or the excitation frequency content continuously excites the unmodeled modes of the system.
Some experiments on the feasibility of applying neural networks to the flexible plate example have been performed. These results indicate that neural networks have the potential and capability of approximating complex and relatively high ordered structural systems. These mappings in the context of the example considered are the forward and inverse dynamics of the plate structure.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1992.
|Date Available in IDEALS:||2014-12-17|
This item appears in the following Collection(s)
Dissertations and Theses - Civil and Environmental Engineering
Graduate Dissertations and Theses at Illinois
Graduate Theses and Dissertations at Illinois