Control loop performance assessment and oscillation detection.

摘要

In this thesis an analysis of process control loop performance based on the Harris's performance index is given. Good and poor performances will be defined and it win be shown that poor performance may be obtained when some loops are cycling. Then a procedure for localizing the oscillating loops in a cascade multiloop system will be introduced. The procedure uses data collected under normal operation and assumes knowledge of the different frequencies of the periodic component of the output signal. An oscillation index is introduced to characterize the oscillating loops.; The estimation of time delay is a major step in assessing control loop performance. To overcome some difficulties of the available methods for its estimation, a new off line extended Least-Squares technique for parameter identification and time delay estimation will be introduced for processes where acting disturbances are white noise (i.e. C( q1) = 1).; The Harris performance index for single-input single-output (SISO) systems is a very useful tool, therefore its extension to multivariable systems becomes necessary. The multi-input multi-output (MIMO) performance indices available in the literature involve use of the MIMO time delay matrix, also called the interactor matrix, to perform the assessment. The method proposed in this thesis does not require knowledge of the interactor matrix, The method uses knowledge of the delays between different input/output pairs of the process and defines an absolute lower bound on the achievable output variance for each output. This bound is then used to define a performance index associated with each output. The sum of these bounds is used to characterize the overall control loop performance. The results are independent of the matrix time delay representation, or interactor. This method is evaluated by application to the loops controlling a lime kiln in a Kraft pulp mill.