Flow Control Using Neural Networks

Flow Control Using Neural Networks

The Wichita State University-Aerospace Engineering

Adaptive neural network controller is obtained by applying a feedback linearlization method. Here, neural network is used to estimate the function of the Wave equation. The dynamic sequential recursive backpropagation algorithm is adopted for on-line training of the neural network.

A feedback linearization method is well known for nonlinear control problems. For a given system,

(1)

, (2)

where

(3)

and

. (4)

are chosen such that the polynomial is Hurwitz.

A neural network can be used to model f(x) and g(x) which can be highly nonlinear and include uncertainty in the system. The controller using the neural network can be defined as

, (5)

where are the estimates from the neural network.

Multilayer perceptron (MLP) is modified for on-line training based on dynamic equential recursive backprogation algorithm. Previous outputs of the system are used for on-line training.

The adaptive neural network controller requires little more calculation for on-line training, but it does not require the off-line training phase. The adaptive neural network controller gives a robust controller for any given conditions than the mapping type neural network controller or other conventional methods. The limit of this approach is that the system should be described in the form of equation (1) to adapt a feedback linearization method.

Research graduate student of aerospace engineering in Wichita State University. The similar work in this field was applied for my master’s thesis titled as Adaptive Control of Unknown Dynamic Systems Using Neural Networks. Paper is accepted for AIAA 4th Shear Flow Control Conference meeting at 29 June-2 July, 1997 in Colorado.

Download this Application Summary in Word97 format.
If you don't have Word97....Download the Word Viewer.