Eng-Tips is the largest forum for Engineering Professionals on the Internet.

Members share and learn making Eng-Tips Forums the best source of engineering information on the Internet!

  • Congratulations dmapguru on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!

Help With 60" Pipeline Flow 1

Status
Not open for further replies.

mkyauk

Mechanical
Joined
Jul 28, 2006
Messages
5
Location
US
i am looking for a way to come up with a way to detect and pinpoint a crack that may occur in a 60" diameter pipeline that runs for about 5 miles. the flow through the pipe is approx. 120mgd and runs mainly underground. i need to be able to monitor the pipeline constantly and be notified as soon as a crack occurs. i've looked at options from wavealert, flowmetrix, and controlotron but would like a simpler option using pressure gauges or transducers.
 
Here's the "SCADA Leak Detection method.

I will describe the graphic method, you can do the same analytically if you like. With a 5 mile 60", the analytic solution will be better, because otherwise you will be graphically trying to find the intersecting point of two relatively flat lines and its easy to err.

[COLOR=black yellow]At both the beginning and end of the pipeline you must know the flow and the pressure.[/color]

You figure you have a leak, when the two flowrates do not match, now you must estimate where it is.

The accuracy of the solution depends on the accuracy of your flowrate measurements and any transient pressures in the line will give a lot of "noise" to the data that must be filtered out.

Make a profile drawing of the pipeline to scale.

Calculate the total head at the beginning of the pipeline. Here I will assume it equals 300 feet, or around 130 psig if you have water in the line.

Calculate the head loss using a good head loss equation, (Hazen-Williams, Colebrook, or Churchill, etc.) for your pipeline. I assume here that it equals 10 feet/mile of pipeline.


Go to the pipeline profile plot and put a point 300 feet (or whatever your beginning head equals) above the beginning of the pipeline.

From that point draw a line towards the end of the pipeline with a slope equal to your beginning head loss/mile.

Take the outlet head and the flowrate at the end of your pipeline and calculate the head loss/mile there. Here I assume head is 240 feet and head loss/mile is 9 feet/mile.

Go to the plot and put a point 240 feet above the end of the pipeline and draw a line towards the beginning of the pipeline with an upward slope to the left of 9 feet/mile.

Your leak will be located (theoretically) at the intersection of the two head loss lines.

/300 ft
* leak at intersection
| * /
| *O* lne from pipeline end
| | * * <- 9ft/mi
| V * * 240 ft
| | -10ft/mi->* *
| | line from * |
| | beginning of pl * |
+------V-------+-------+-------+--------+
0mi 1mi 2mi 3mi 4mi 5mi
\
leak located at 1 mile



Going the Big Inch! [worm]
 
thanks for the reply ... its seems like this should work ... but i would like to make sure you can use this idea on a pipeline that experiences different internal pressures because this pipe experiences changes in elevation .... there are vaults/blowoff valves as well through out the pipeline ... thanks again
 
It works whenever the conditions of your headloss equation are meet. It obviously makes the same assumptions as whatever headloss equation you use. Typically that iwll be constant flow, incompressible, Newtonian fluid, etc. You must chose a head loss equation appliable for your pipeline. Generally, if your density does not change much with pressure or temperature from one end to the other, i.e. flowrates are usually steady state, its a trivial solution, otherwise it becomes more complex, because you will have to segment the pipe into short lengths over which the density does not change significantly. All volumes leaving from a blowoff or non-metered outlet will have to be accounted for or otherwise be treated as "noise", which will degrade the solution. In that case, you will have to give your solution a margin of error sufficient to tune those out, or you will have to include known volumes removed from the pipeline between segments in a segmented solution, as I described above. Same for vaults. Vaults (I assume are air pressure over water pressure chambers) will have to be modeled at the node in a pipeline segment where they occur, substituting the fluid entry and exit head loss coefficients and pressure-volume relationship inside the vault. If the pressures and volumes change slowly, a "steady state approximation" can be used, if not, you will need a program capable of transient analysis. That is far from a trivial solution. Write me directly if you need some detailed advice. See my webspace, profiles tab, for my e-mail address.

Going the Big Inch! [worm]
 
im not sure if im reading it wrong or interpreting your explanation differently but i was reading your first post about the slopes .... so from you example i begin with a head of 300ft and slope downwards at a rate of 10ft/mi until i hit the end of 5 miles .... this value should be 250ft ....

now i have a leak and the end head value is now 240ft with a head loss of 9ft/mi .... back drawing that slope will get me to 285ft ....with no intersection point ...

sorry about the confusion i was just drawing this out and ran into this and i couldnt figure it out ....
 
Mk,

Right you got me. 240 ft is not correct. This time I did the math. Can I try again.

I used 120 mgd and a Colebrook friction factor gives 84.5 feet drop in 5 miles.

So Line 1 is (0 mi, 300 ft) to (5 mi, 215.5 ft)

Let's say your leak rate is 20 mgd giving a flow of 100 mgd.
If you had 100 mgd and 300 ft head at the beginning of the line, you would have a slope of 12.2 ft/mile and at the end you would have a head of about 240 feet. Except you don't have a head of 240 feet, you actually have less, because in the beginning of the pipeline to the leak point, the flow is 120 and the head loss per foot is more than that for 100 mgd, so 240 is the high value for head at the end of the pipeline.

So your measured head at 5 miles, must be between 215.5 feet and 240 feet. So now that the correct head is bracketed, lets see if I can assume a better value than the 240 ft I picked last time.

I pick 225 ft. Let's say the actual measured head at the end of the line is 225 feet.

OK now draw line 2, from (5 mi, 225 ft) upward to the left with a slope of 12.2 ft/mile and you should get an intersection with line 1 at around 3 miles.

What did I learn? It pays to do the math first.




Going the Big Inch! [worm]
 
The problem you may experience with this type of leak location is sensitivity to the minimum size of the leak you wish to detect. Theoretically the basic method, as I've detailed above, can detect any leak that is measurable by your flowmeters, however there is usually some normal variation in flowmeter readings. Flowmeters for custody transfer of oil and gas typically have a maximum accuracy of +/- 0.5%. In a 120 mgd system, this would be 41 gpm.

The other major factor is variation in the fluid in the pipeline. When pressure increases at the beginning of the pipeline, more fluid is packed into the pipeline until the pipeline stabilizes at a new steady state flowrate. Likewise, when pressures decrease, less fluid is contained in the pipeline. Thus a very accurate means of leak detection must account for these transient flow effects.

A widely accepted method of leak detection that can easily be added to the pressure drop method above is to keep a running total of flows going into the pipeline and flows coming out from the pipeline, then adding the two together. Once line pack has been accounted for, a leak free pipeline should have a sum of flows in - flows out = zero. Theoretically, any deviation to the negative side will indicate a leak has occured. The sensitivity increases with the length of time over which the measurements are taken, as even a very tiny leak that could be "noise" in a daily measurement, when not canceled by + noise the next day, eventually accumulates into a detectable volume at the end of, say 1 week or a month.

A leak detection system using the two methods combined and using statistical techniques, is capable of detecting the smallest leaks, over time.



Going the Big Inch! [worm]
 
This is slightly abstracted, but your problem is a variation analogous to finding electrical faults in the undersea telegraph cable between England and Denmark that was solved by Oliver Heaviside in the late 1800's. I only vaguely remember this as a sidebar story in my eng schooling, but look up Ollie and you might find a nice solution along the lines of what BigInch has posted.
 
thanks for the info i got it to work out correctly this time ... also i was wondering what was the best way if the pipeline was under hydrostatic pressure ... im assuming place pressure gauges throughout the pipeline and if theres a pressure drop somewhere then the leak occurs somewhere within that point ? .... if that works then how do i find where along the two gauges and where should the gauges be placed ? .... thanks again !
 
BigInch,

Help me out here regarding your first post. #1 my brain is on freeze - how do you calculate your total head at the beginning of the pipeline?

Also regarding your graph, won't it only work if you have a straight run of pipe? otherwise you actually have more than 5 mi of "equivalent" feet, and your calculated head loss of 10 ft/mi would have to carry over the entire equivalent distance. In that case, I guess you would just have to corrolate the reading on the graph with the actual pipeline location accounting for elbows, etc.

Am I right about this?
 
Bron,

Its not your head, it was mine. I didn't do the math first and made some bad estimates. See the 6th post where I have corrected the figures and the line coordinates.

The head and flow must be known at the beginning and end of the pipeline, (read off pressure gauges and flowmeters). With that data, bingo you can find the leak.

Actually, this is going to sound strange, but did you know that for long pipelines, you can almost always neglect the minor fitting losses. Fitting losses are only important inside the pump stations or other facilities and only when you are doing a very detailed super hi accuracy model of a pump recirculation line, compressor surge control or some other specific piece of equipment. I also use average wall thicknesses over 100 miles or so of pipeline and get "approximately=exactly" the same answers as when I include every road crossing wall thickness. I've proven it a few thousand times now with actual pressure and flow data.

My rule is, "Never model details that are not important to the objective of the simulation."

In cases of non-flat pipelines, the method still works when all pressures=heads are compensated for actual terrain elevations.

Going the Big Inch! [worm]
 
mkyauk,

I'n not sure what you mean when you say the pipeline is under hydrostatic pressure. Is it a pipeline under water as in "hydrostatic pressure" on the outside of the pipeline?

The method will work between any two points that you have both flow and pressure readings. These will usually be at the beginning and end of a pipeline, but there is nothing that would prohibit you from doing this for any segment of a pipeline where you would want to monitor leaks there only.

For some pipelines it is necessary to model internal segments separately. Consider a long pipeline with a pump station at the beginning and in the middle of the pipeline. Due to the pressure added at the middle pump station, one equation for the entire pipeline will not work, because of the pressure discontinuity at the middle pump station. You have to model the two segments and then write an equation for the pump in the middle, so the change in heads can be accounted for.

You mention that you may want to add pressure indicators at various points along the pipeline. This can increase accuracy, even though you don't have flowmeters there. You can do the calculations based on the points where you have both flow and pressure, then use the internal pressure points as additional "check" points to see how your calculated head loss slope is hitting those points (or not). Every time it misses by an amount greater than the accuracy of your head loss equation, it is a signal that there is a possible leak. It could be "noise", in which case you will eventually learn that it is not a leak and you will be able to adjust your head loss equation to compensate for some fitting losses or reduced or increased roughness factor or some similar adjustment.

There is an optimum number and location for your pressure point indicators which can be determined based on the minimum leak volume that you wish to detect and the accuracy of your instruments and cost of locating them, etc., but its a little too complicated to explain here. For now, just think that the more PIs and or FIs you have, the better you will be able to filter out "noise", increase your ability to detect smaller and smaller leaks and increase your accuracy in pin-pointing the location.

Hope all of this helps, and sorry to everybody about the initial bad choice of data points.


Going the Big Inch! [worm]
 
i meant by hydrostatic as in the water inside the pipeline might not be moving ... ie an earthquake occurs and the pipeline is shutdown due to a possible occurance of a leak and the water is now static .... sorry about the confusion and youve been a great help to my problem ... no need to apologize about ur initial post ... again thank you ...
 
Ah, got it... a static pipeline.

You should be able to work with the data history that the SCADA sys is keeping for you. Just access the data at the time of the earthquake and all subsequent times that the data was collected as the system depressures to the final static condition. You could easily archive flow and head/pressure readings at say, every 5 seconds and calculations done at each time step using each data set should show that all intersection points are more or less the same location. Once it goes completely static (no flow), you will be able to calculate the volume lost. The good thing about it going completely static, is that at least your not leaking any more. You can't model it then, but then there's no need to either.

The more accuracy you build into the model, ie. consider water as a compressible fluid and consider the elasticity and fixity of the pipe and the more interior points you have pressures the more accurate the solution becomes. With the right equations, you can maintain a total simulation, including volume changes due to transient pressures of starting a pump or closing a valve. But cost of the monitoring system (detailed hydraulic model) goes up a lot when you want transient capability. But, all modern petroleum pipelines are using this type of monitoring system running in real time (with hot standby) these days. Nothing miraculous about it or anything. Just costly.




Going the Big Inch! [worm]
 
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top