Neural network based optimizing controllers for smart structural systems

Abstract

Neural network based controllers for vibration suppression of smart structural systems have been reported in several recent studies. These studies have shown that in addition to conventional controller design methodologies, neural networks offer an effective basis for design and implementation of controllers. With the introduction of neural network chips like the electronically trainable analog neural network (ETANN) chip i80170NX by Intel and the Ni1000 chip by Nestor Corp., stand-alone hardware implementation of neural network based controllers is possible. In this paper the capabilities of Intel's ETANN chip to implement linear and nonlinear controllers for smart structural systems have been investigated. A neural network based optimizing controller design methodology that integrates the ETANN chip and its capabilities of on-line adaptation has been developed. A priori information on the smart structural systems such as actuator/sensor bandwidth limits and control effort limits can be directly accommodated in this method. Simulation studies of the performance of a closed loop time varying linear and nonlinear system have also been presented with and without on-line adaptation.

Neural Networks in Analog Hardware — Design and Implementation Issues

“…42. In this application, the chip is used to implement linear and nonlinear controllers for smart structural systems.…”

Reduced-order Robust Multivariable Control of a Double-beam Cantilever Structure

“…An iterative procedure was adopted, which was shown to yield a control system that did not exceed the control limits. Damle and Rao (1998) used the sequential quadratic programming (SQP) technique to apply bounded control using LQR control, which, as with Smyser and Chandrashekhara (1997), was applied using an ANN. One major difference was that the system was trained online, which allowed the system to adapt to changes in the plant model, therefore increasing the closed-loop robustness.…”

Adaptive Simultaneous Precision Positioning and Vibration Control of Intelligent Composite Structures

“…Rao and Sana (2001) surveyed the primary feedback control design methods for smart structural systems such as positive position feedback control, independent modal space control, dissipative/ positive real control, robust control, and linear matrix inequalities. Many new control approaches have also been investigated in the vibration control of intelligent structures, such as adaptive control Melcher, 1998, 2003), neural network control (Damle and Rao, 1998;Bosse et al, 2000), fuzzy logic control (Takawa et al, 2000), and their combinations (Mayhan and Washington, 1998;Davis et al, 1999;Shen and Homaifar, 2001). Among these advanced control approaches, adaptive feedforward control is a useful and popular scheme in active vibration and noise control (e.g., see Nelson and Elliott, 1992).…”