| 202D3 |
|---|
| Identification of Bilinear Systems Using Wavelet Transform |
| AUTHORS |
| A. Tayaoka, T. Furuya, M. Ikemoto, M. Soeda (Kitakyushu National College of Technology) |
| ABSTRACT |
| Some processes in the plants have nonlinear dynamic characteristics called bilinear, which contains products of input signals and state variables. This study proposes some method of identification of the bilinear systems, where the wavelet scaling functions are applied to get the suitable forms for estimation. The scaling functions, having orthognality and two scale properties, approximate the input and output signals. In this method, so called connection coefficients are used to compute derivatives of time functions without any differentiation. So bilinear differential equations, taking inner products with the scaling function, can be represented by time discrete models with the transformed variables and the connection coefficients. The products of input signal and state variables are also computed by the connection coefficients, and the parameters of the bilinear systems are estimated by the least squares method. Simulation is carried out on temperature control of water tank warmed by an electric heater and cooled by a heat exchanger pipe, where a pump circulated water between two tanks. Estimation was performed to obtain the parameters of bilinear differential equation using the 1600 pairs of data excited by saw-teeth shapes of the input signal. The results demonstrate the validity of the method, using connection coefficients for bilinear system identification. |
Back to Program
Time Table
Top of This Site
Annual Conference 2002