Abstract
The aim of this thesis is to show how an industrial plant may be identified under normal operating conditions using pseudo random binary signals. Particular reference is made to a pilot plant scale double effect evaporator.Classical methods for system identification are briefly considered with their advantages and disadvantages.
The properties of a prbs and its use for the evaluation of system dynamics are described and methods are suggested to eliminate or reduce errors and noise.
A simulation package is developed to study the techniques and to reduce the noise by digital filtering. A computer package is also developed for fitting a transfer function to experimental data and for frequency response data analysis.
A package for on-line data acquisition is modified to generate a prbs and to collect data from sections of the double effect evaporator.
The sampled response of sections of the plant to a prbs is transformed by a fast Fourier transform, and the result is divided by the transform of the input signal to obtain an estimate of the frequency response (direct input/output technique).
The ratio of the cross-spectral and auto-spectral densities is also computed to form the transfer function. This correlation method appears to have a good noise rejection.
The method is shown to be suitable for industrial plant where noise is significant and the system may not be disturbed too far from its normal operating conditions.
Finally, possible improvements in experimental methods and the processing
Date of Award | Jul 1978 |
---|---|
Original language | English |
Awarding Institution |
|
Keywords
- On-line computer analysis
- plant dynamics