| Summary: | Over the years, increased competition and stricter environmental regulations have
placed additional pressure on mills to become more efficient. Chemical additives are
being used to enhance mill performance and improve paper quality. The use of recycle
streams has also reduced the raw material costs. These changes have made control of the
wet end system a challenging problem for mill personnel. The presence of multiple recycle
streams have exacerbated the problems in the wet end. To deal with these problems, the
goal of this work is to provide an understanding of recycle dynamics and control that will
help improve operations in the papermachine wet end.
The first step was to examine recycle dynamics to identify responses that could be
potential control problems. A simple recycle loop system composed of a process model in
the forward path and a process model in the recycle path was used, and the study was
performed using various combinations of parameter values for the process models. The
recycle responses were found to fall broadly into three groups with groups two and three
being identified as having responses that may cause problems for control.
To determine if the wet end exhibited significant recycle behaviour, an IDEAS simulation
model of the paper machine wet end was used. The study looked at the effect of a
filler change in the recycle streams, and in the paper leaving the wire. The effect of tank
sizes on the recycle dynamics was also explored. The process tanks resulted in all of the
responses being first order. It was only when the tanks were removed from the process
that recycle dynamics became apparent, but the extent of the dynamics depended on the
number of recycle streams present.
The problems in the wet end are also attributed to the effect of disturbances being
returned to the process by the recycle streams. Two disturbance types were studied: (a)
Quality disturbances, such as changes in temperature and stream composition, and (b)
Flowrate disturbances which are basically flowrate variations.
Quality disturbances require better control of the recycle process. The common PI
controller was the first controller evaluated on the recycle process. The performance of
the PI controller was adversely affected by the size of the time delays in the forward or
recycle path. Little improvement in controller performance was obtained when dead-time
compensation, in the form of a Smith Predictor, was used.
A new approach using a seasonal model to represent the recycle process for model based
control was presented. The seasonal model approach was compared to the following model
based control solutions found in the literature: (a) the recycle compensator which removes
the recycle effect and (b) the Taylor series approach which approximates the denominator dead time terms of the recycle process. The seasonal model approach provided comparable
performance to the existing methods and was more practical.
Flow disturbances are usually dampened by the surge tanks. However, the degree
of disturbance attentuation is determined by tank level controller. If the controller is
tuned for tight level control, the flow disturbance may pass from the input to the output
with very little attenuation. When the controller is tuned for averaging level control,
disturbances in the flowrate are significantly reduced. In this work, an optimal averaging
level control algorithm developed by Foley et al. (2000) was selected. The algorithm
is based on constrained minimum variance and guarantees the lowest possible flowrate
variation. The algorithm was to handle the case of sampling delays greater than one. The
algorithm was applied to a saveall simulation to reduce the flowrate variability from the
saveall to the blend chest.
|
|---|
