Big; I'm not sure if they are variable. They seemed to me to be running full-out. I stood next to one that was running. It was in a fire cell. There was a standard black fuel meter/counter mounted on the turbine with big digits on it. Diesel 2 at a third of a gallon a second was clicking by. Twenty gallons a minute... 30,000 gallons/day.
I'm not sure they throttle them, since they have put about half of the pump stations on standby because DRA (Drag Reducing Agent) has allowed them to pump more efficiently. They probably run them full out so they can not, run others at all.
Leaks:
Essentially a hole occurs.
You listen for the shockwave front to come by.
You monitor at several locations.
At each point you detect the wavefront.
You do the time-of-flight calcs.
This gives the origin of the hole.
Detriments:
The accuracy of the location is a function of your quality of time synchronization.
The magnitude of the signals to work with are a function of the hole size.
The discrimination ability of the system is directly related to the signal-to-noise ratio.
The noise is what you need to analyze so you can estimate sensor spacing and minimum detectable leak size.
Also the electrical noise of the transducer is important, some are pretty noisy.
We found that away from pumps the line was generally pretty quiet. We found we could hear leaks pretty well. Even across the hard-to-imagine slack line dropping down to Valdez.
Implementation problems:
Of course there is always the problem of what is a "small enough leak" to
not detect?
How close together do the transducers need to be?
There is a serious communications and synchronism requirement.
The more noise the closer together you need your sensors.
Every line is different and hence you must survey them before any design can take place.
We didn't use AI though there was constant work on the actual algorithm relating to slope, magnitude, and shape of the shockfront.
Do keep in mind this method was not the type used on communications lines where a gas(N) leaking out of a tiny hole, permanently whistles in the ultrasonic region, and can be detected at anytime by just hearing it, and located by tracking down the noise source.
With respect to the Alyeska Pipeline their leak detection was/is basically; Transverse flow meters at each pump station coupled to a detailed flow model of the entire pipeline. If a leak occurs then the model will slowly drift from reality until it becomes obvious. So if a pin hole or more likely a 30-06 hole develops it may take a few hours possibly even 24 to detect the leak. A lot of the leaks are actually detected visually early on. The line is overflown often.. very often, exceptionally often! To be fair, there haven't been many in the line's life time.
They were considering our product in hopes of increasing detection speed. We did the study, they never went very far on it. I don't really know why. I was just the data collector and bear dodger and the guy to put the hardware together. I wasn't in the wheeling dealing, marketing aspect. It was all complicated because we were in bed with a large Japanese corporation that had a foot-in via their extremely sophisticated multimillion dollar pigs.
Keith Cress
Flamin Systems, Inc.-
