The application of time-frequency techniques to process control and identification /
Applications of time-frequency techniques, especially the wavelet transform, for system model identification and controller design are studied. An algorithm for simultaneous model predictive control and identification (MPCI) using wavelets is developed. Illustrative examples demonstrate the applic...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1996.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=739667611&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | Applications of time-frequency techniques, especially the wavelet transform, for system model identification and controller design are studied. An algorithm for simultaneous model predictive control and identification (MPCI) using wavelets is developed. Illustrative examples demonstrate the applicability of the cubic spline wavelets and the modulation method. It is shown that MPCI using wavelets identifies true parameters faster than the least-squares method. A multiscale model predictive control (mMPC) algorithm is developed. Unlike the traditional MPC controllers that are only tuned in time domain, the mMPC controller can be tuned in both the time and frequency domains. It turns out that the traditional DMC and the averaging MPC are the two extreme cases of time-frequency domain tuning. Simulations show that the mMPC controller has a stabilizing effect when the high frequency content in the future time in the predicted output is detuned. Application of mMPC to noisy processes shows that 1) by detuning high frequency wavelet components in the future time of the output prediction, the closed-loop output has less strength of high frequency harmonics; 2) for white noise disturbances, nonstationary disturbances described by the integrated moving-average model, and step disturbances, the closed-loop output is closer to be white and gives smaller integral square error (ISE) when the high frequency wavelet components in the future time are detuned. The developed mMPC algorithm is applied to control the conversion of a fixed bed reactor experiencing catalyst deactivation. Simulation studies show that both the MM]PC and DMC controllers regulate the conversion well; however, when a step disturbance enters the process, the output for the DMC controller becomes oscillatory and takes longer time to settle than the mMPC controller. Another application of the mMPC is to a potato chip fryer that is a 3 x 3 system. In this system, one of the measurements is much more noisy than the other two. The mMPC is applied such that the high frequency content of the noisiest output is detuned and the controller performance is improved. |
|---|---|
| Item Description: | Vita. "Major Subject: Chemical Engineering". |
| Physical Description: | xiii, 157 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilms Inc. |
| Bibliography: | Includes bibliographical references: pages 153-156. |