Revelations on Dynamic Process Analysis,
Advanced Control,
and Online Optimization
ProControl’s
Dr. Robert Bartman taught this open-to-all class to Advanced Process Control
Engineers from BP, ExxonMobil, Citgo, and Celanese. Direct quotes follow:
I thought DISCOVER was fantastic. It is easy to use, with great examples and
case studies. I will definitely be using
DISCOVER on a frequent basis when “back at the plant”.
I’m just an inexperienced control engineer,
but this seems to be a very valuable tool.
It showed me a lot of concepts graphically, which helped a lot in
class.
Discover’s multivariable analysis allows very
easy backout of what we already know, without statistics needing to reinvent
anything. A clever
implementation, with great teaching examples.
Having Discover would certainly ease the
analysis of responses with multiple inputs, as well as giving good tuning
constants for levels and pressures, and across the spectrum of PID algorithms –
especially ‘non-Ideal’ algorithms.
I think this tool is incredibly versatile and
useful.
Yes. I
understand the process a lot better. I
am now comfortable enough to go back and make changes on my own.
This has been the best course I’ve ever
taken.
It’s definitely justified – the wealth of
knowledge, the passing-on of experiences, and the reference materials are
invaluable. I will feel much more
comfortable looking for ways to make $$$ now, rather than being purely
reactive. I feel a lot more experienced!
This course was absolutely justified and
worthwhile. This control course was MUCH
more valuable than Emerson’s Modern Loop Tuning Class. This took a much more fundamental
and practical approach to control, rather than tuning just based off of theory.
This course gave great insight into
conditions where PID-based primaries “break down”, and where model-based
controllers offer the simplest effective solution. This was an excellent explanation of
something I’d seen many times in the plant, but couldn’t do anything to
improve.
Almost everything I learned here was new to
me in terms of its application to an actual process. I can envisage this information being
extremely useful for implementing tighter control systems, and for increasing
my own process understanding.
My boss took this course and knew its value
in terms of the amount of control application knowledge it contains. He wanted me to take it to get tuning
controllers down to a science. Most of
the APC engineers in my group have no concept of looking at a loop
and knowing if it’s well tuned or not, whether one mode is dominating or not,
or the precise moves required to improve its tuning. My boss also wanted me to learn how to
convert dynamic process understanding directly into a good set of tuning
parameters, with no guesswork. Now I
have all that!
I learned how to use feedforward control to
minimize impacts of measurable disturbances.
I think most Advanced Process Control engineers at my refinery AVOID
feedforward control whenever possible, because they don’t understand it.
I think this should be a standard course on
the training map for all control engineers.
I learned how to identify important dynamic
process responses, and the constants that describe them, from plant data – both
from ‘opportunistic’ data grabbing, and from formal plant tests.
I thought the topic of constraint control was
great. The custom software displayed
very clearly how a constraint affected min/max of the PV we’re interested in
economically.
I was very interested in constraint control –
it seems to have really good economic potential without the need for
APC, such as DMC.
Constraint control: a good way to make $$ at the DCS
level!
Learning the functions of each mode of a PID, and being able to approach a loop and improve it in a scientific manner, was the most interesting topic for me. You really helped me understand what exactly a loop is doing, and this vastly increased my understanding of process control.
I now understand how P, I, and D really
work, what each mode is responsible for, and when each should /
shouldn’t be applied.
I learned how to tune a loop well – something
that most of the control engineers at my refinery don’t really know how to do,
at least without a LOT of guesswork.
I learned the importance of once-a-second
control intervals, helping me understand whether a variable should, or should
not, be included in an application like DMC (running once a minute) – which our
refinery is always pushing towards.
No one has put level and pressure control in
such simple terms, and frankly I’m not sure if anyone truly understands
levels and pressures this way. Your
examples definitely opened my eyes to a new way of thinking.
The different Level and Pressure control mode
responses were explained to me both in college and in the professional world,
but they have never been explained in such easy to grasp terms. When put in these terms and displayed
graphically as you do, these concepts are no longer a mystery.
I think that the Feedforward control systems
and their implementation you explained will be economically quite
beneficial. We often overfire
furnaces or have sustained periods of poor control during feed changes (both
composition and rate). I think some
straightforward FF control could help overcome some of this and prevent ensuing
effects such as reduced charge rate, and off-spec products.
Now I even know “when to pull the plug” on feedforward control, if disturbance cancellation is impossible past a certain limit. Before this class, the concept of feedforward synchronization itself was a Big Mystery.
The course stressed starting with the
basics: Ask “ Why
do we have this unit?”
Don’t worry about jumping into APC, or any
control really, until you identify where the money is.
Since I work with many 2 product distillation towers, I appreciated knowing how material balance, energy consumption, and quality control “really work” on such towers. Your example gave a great understanding of the nuts and bolts of the tower, and the special conditions under which we’re forced into dual quality (or temperature) control. Great exercise!!!
We still have some smaller units without RMPC
or DMC control, and I think there will be opportunities to implement better
controls and gain economic improvements there.
For the most part I think our units on DMC and RMPC are run too
conservatively, just to ensure the product meets specs.
Two very important facts mentioned every day
become ever more clear:
economics
are the main driver behind most controls, and understanding the process is the
single most important part of control design and tuning.
It was very useful to go through all
the economic considerations necessary for full control scheme evaluation. We often see two different pushes regarding
economics – a push from Operations to minimize energy costs, and a
(conflicting) push from another group to maximize fractionation and side draw
rates.
My boss expects me to analyze what a loop is
doing, so I can set the appropriate tuning parameters without guesswork – and
now I can do that!
I liked your Lubes fractionating tower
example, and the chance we had to analyze its (real) dynamic data. This tower was under APC ‘manipulating
everything’, but no money was being made, and its control of sidestream product viscosity was awful. One of the most important things I learned in
this example was the thought process of understanding tower dynamics, and
carrying this through to a simpler control scheme which got the job done. I thoroughly enjoyed this example, and
thought it was both well explained and portrayed.
It was very helpful to go through the
fundamental differences between feedforward control structures. This level of detail really helped my FF
understanding.
We learned that none of the tuning guidelines
in our control textbooks turn out to be practical, in a real plant. We also learned that Derivative can really be
“Our Friend”, and now I understand why.
Bob is extremely effective at breaking
analysis down into easily understandable fragments. I liked the way that things
were introduced, then returned-to in more detail.
The teaching style was very effective, an
excellent mix of theory, real world data, and simulated data for lab
exercises. Appropriate time was given
each topic. Bob was always willing to
re-visit a topic when needed.
The class was very effective. By making sure we understood each topic, he
ensured that we all take the information home.
We covered some great “real life” examples,
and as we were discussing them I felt very involved, and was forced to think.
He has a great ability to simplify complex
topics, break them down into pieces and explain the subtleties. I also enjoyed his real world examples,
putting ‘theory’ to the test.
Although I don’t always enjoy the work,
writing your nightly summaries helps me solidify the concepts learned each
day. Once again I feel like each concept
has been really learned.
I enjoyed all those computer examples you
used throughout the class. These visual
examples really helped my understanding.
Level looked extremely good – I liked the
idea of throwing your worst real-world disturbance at it, to both select and
tune nonlinear level control algorithms.