Modeling structure-actuator systems by neural networks



Neural networks are used for identification and solving the coupled equations of motion of a structure and its actuators. First, a neural network is trained to learn the behavior of the structure under study. Second, other neural networks are trained to learn the behavior of the actuators. Third, the trained neural networks of the structure and its actuators are interconnected to form a modular neural network which can be used in the analysis and/or in predicting the future response of the structure-actuators system. Such a modular neural network is called here the "neuro-modeler" or the "neuro-analyzer" of the system. The method is especially advantageous when it is desired to model the structure-actuators system based on laboratory data, and is applicable to both linear and non-linear structure-actuator problems. However, as a first study of the method, the emphasis of this paper is on linear problems. The method is tested on a SDOF frame with one actuator. The robustness of the neuro-analyzer is tested too.