Professional Documents
Culture Documents
Carlo Demicheli
Firmware Developer at ZKT Brasil
3 articles Follow
Introduction
When deciding which type of control valve to use, there are many variables involved. Many
other types of valves could be used, that allow an almost linear control. But these valves could
not be the best option due to many aspects, such as space limitation, budget restrictions and so
forth.
In 2009 I was in charge of developing many control loops to automate a gas station for a
customer in Brazil, and one of them presented this type of challenge for the automation
professional.
Blast furnaces produce the BFG, a combustible gas whose main components are (carbon
monoxide), (carbon dioxide) and a small amount of (hydrogen), among other gases. This gas is
produced in the process of iron ore reduction to cast iron. When compared to other gases (natural
gas, for example, presents heat values around 35,000 , we may say that its heat value is low
(typical values around 3,500 ). Because of the amount of gas produced during the iron ore
reduction process, this gas is fully recovered, stored and distributed in the plant, mainly for
heating and power generation purposes. Normally there is a gas holder for buffering, involving
capacities around 100,000 .
When the plant is stable and working efficiently, it is possible to recover and reutilize this gas
flow, that reaches 450,000 of BFG per hour, considering just one blast furnace.
The small BFG consumers, like heat regenerators in the very blast furnace, use more or less
BFG, depending on the number of regenerators simultaneously in the heat cycle. However if a
power plant station suddenly has a turbine trip, there will be an excess of up to 350,000 /h of
BFG. Hence the gas holder control is optimized to try to keep a 50% capacity, being able to
assume or supply BFG as needed. And this surplus of BFG could fill 3 empty gas holders in just
51 minutes.
The only way out of a plant collapse is burning this gas, using flare stacks. Each flare stack has
at least one flare head, capable of burning typical values of 250,000 . Each flare head has a
control valve that works normally closed, being used to burn the BFG surplus and even to
control the BFG pipeline pressure in case of gas holder unavailability.
Many other types of valves could grant us to create a stable and linear control; so why choosing
the Butterfly type? In many cases the answer is cost, because the Butterfly valve is of simple
construction, when compared to other types of control valves. Nevertheless, in this case, it is
more than that: it would be impossible to construct, transport and install a valve of the size
required for this specific case: the piping required for such huge gas flows, reaching 1,000,000 ,
can easily arrive at 160 inches (more than 4 meters) in diameter, depending on the length of the
pipe line, in order to avoid pressure drops. Other types of control valves would be extremely big,
expensive, heavy, and indubitably slow in response time due to its mass and intrinsic moment of
inertia.
The Butterfly element is that kind of bow tie that controls the flow. When it is parallel to the
piping cross sectional area, the valve is closed:
As it is possible to see by just looking at these simple figures, turning the butterfly element (in
figure 2, in the clockwise direction) causes augment in the flow rate, because the space between
the edges of the Butterfly element and the piping walls is bigger.
Let us simplify this figure to 2 simple traces: the Butterfly element and
its projection (shadow) on the plane of the cross sectional area of the piping:
Figure 3 – Butterfly element and its projection over the normal plane
Remember that a control valve element controls flow rate using a modification of area (2
dimensions) and the media transport through this area creates the third dimension that results in
volume transportation (flow). Independently of the special techniques and mechanical blocks
developed and improved by each manufacturer, the first dimension for flow rate definition
depends always on the Butterfly element.
Returning to figure 3 and getting back in time to the elementary school, there is a lesson in
trigonometry that teaches us that the length of the projection of a straight line segment over a
plane is the segment’s length multiplied by the cosine of the angle formed between the segment
and the projection plane:
For the sake of simplification we just define the length of the Butterfly element as 1.0 (one).
When the element is at 0 degrees, its projection has the same length of the element itself. But this
happens when de valve is closed; so we have to invert it. Finally, we can define our “Flow
Factor” Ff as:
Ff = 1.0 - cos(ß)
Where 0.0 implies no flow and 1.0 means total flow, no matter the type of media and bulk of
the flow.
Ff = 1.0 - cos(0)
Ff = 1.0 - 1.0
Ff = 0.0
This information is reliable, because a closed valve should have a flow rate of 0.0. Remember
that the flow factor supplies the first dimension for the flow rate calculation (even if we do not
know the internals of the valve we are using).
The same check can be done for a fully open valve (90 degrees):
Ff = 1.0 - cos(90)
Ff = 1.0 - 0.0
Ff = 1.0
This information is also pertinent.
Now we can use this calculation to build a table and a graphic of the Flow Factor characteristics
of ANY Butterfly valve (that has no hardware or software linearization resources), for control
loop purposes:
Now it is possible to catch a glimpse of the non-linear response of Butterfly valves. For
example, if the valve is moved from 0 to 9 degrees (0 to 10%) the compromise in flow augment
is of only 0.0123 (1%). Going to the other extreme, when the valve is modulated from 81 to 90
degrees (90 to 100%) the interference in flow factor indicates (1.0000 – 0.8436 = 0.1564), that
represents more than 15% in flow rate increase.
That explains why is so difficult to optimize control loops like the BFG flare stack described
earlier in this article:
If you fine tune the control loop to answer quickly and open the control
valve to avoid high pressures in the BFG line, you will end up with a
“waving” control valve in flows bigger than half of the max flow;
If you fine tune the control loop to operate smoothly at higher flows, it
will be lazy and unresponsive for small flows, inclining to poor BFG
header pressure control and angry clients due to pressure peaks and
valleys downstream the BFG pipe rack.
If you still have any expectation in discovering any other solution, you are void of hope. The
only linear control point with simple Butterfly valves happens when the rate of change of Ff is
equal to 1.0. Using mathematics once again, it happens when the derivative of Ff is equal to
1.0:
Ff = 1.0 - cos(ß)
Ff' = sin(ß)
Some manufacturers use proprietary hardware dispositives to achieve linear control and solve
this problem.
Firstly, we will get the equation we formed for Ff, but now calling ß simply as P for position ,
and calling Ff as LP for linear position:
LP = 1.0 - cos(P)
We start changing the position of some elements:
LP = 1.0 - cos(P)
LP + cos(P) = 1.0
cos(P) = 1.0 - LP
I do not intend to get the cosine of P; I need to calculate P. We do get the arccosine of each
side; it does not modify the equality:
P = acos(1.0 - LP)
Now our magic is ready. It is time to prepare another table and graphic, this time for our recently
defined LP to P equation:
Figure 7 – Converting LP to P
Figure 8 – Converting LP to P - Graphic
If you compare this curve with that of figure 6, it is easy to find out that they are complimentary,
i.e., if you subtract the former from the later you will end up with a straight line of coefficient 1:
we have found our solution.
Normally your control program has a setpoint that goes into the Butterfly
valve’s SIC controller:
Figure 9 – Conventional SIC controller for control valves
All that you need to get a linear behavior with simple Butterfly valves is to use a small
calculation box before the valve’s position controller (SIC):
For example, if we need increase a flow from 0 to 10% of the max flow. So we send 10 to the
calc. box. Let’s see the calc. box calculation:
OUT=acos(1-(IN/100))*200/PI
OUT=acos(1-(10/100))*200/PI
OUT=acos(1-0.1)*200/PI
OUT=acos(0.9)*200/PI
OUT=0.4510*200/PI
OUT=28,71
Now look up this value in the first table, at figure 5:
Figure 11 - Figure 5 revisited
28,71% is slightly before 30%, whose Ff is of 0.109 (10.9% of total flow). If we were not using
the calc box, the flow factor would be of only 1%.
Another example, where we need a flow of 80% of the max flow. So we send 80 to the calc.
box. Let’s see the calc. box calculation:
OUT=acos(1-(IN/100))*200/PI
OUT=acos(1-(80/100))*200/PI
OUT=acos(1-0.8)*200/PI
OUT=acos(0.2)*200/PI
OUT=1.3694*200/PI
OUT=87,18
Now look up this value in the table at figure 11: 87,18% is near 90%, and the valve’s
characteristic will produce a flow around 80% of the max flow, as we intended to produce.
NOTES
1. The calc. box equation takes as premise that your PLC/industrial
computer/Soft PLC calculates trigonometric functions based on
radians. If your system uses degrees, change the equation
OUT=acos(1-(IN/100))*200/PI
to
OUT=acos(1-(IN/100))*10/9
Carlo Demicheli
carlo@demicheli.com.br