NATIONAL TRANSPORTATION SAFETY BOARD
PIPELINE SAFETY HEARING
Leak Detection and Response
Conference Center and Board Room
National Transportation Safety Board
Washington, D.C.
Thursday, November 16, 2000
9:30 a.m.
Day 2 Transcript
Day 1 Transcript
Board Members Present:
JIM HALL, Chairman
CAROL CARMODY
JOHN HAMMERSCHMIDT
GEORGE W. BLACK, JR.
Board Staff Present:
BOB CHIPKEVICH
JOSEPH KRIS
Hearing Officer
ROD DYCK, Associate Director
Pipeline Division
JAMES CASH
DR. ERIC SAGER
ALAN BESHORE
DR. STEVE JENNER
DR. JERRY WEEKS
A G E N D A
AGENDA ITEM: PAGE:
Chairman's Statement 290
Chairman Jim Hall
National Transportation Safety Board
Panels:
Leak Detection Systems 292
Larry Stack, Neles Automaton
Alan Rodecker, Stoner Associates
Dr. Mike Yoon, Simulutions
Dr. Diane J. Hovey, EFA Technologies, Inc.
Kenneth F. McCoy, Tyco International
Independent System Design 401
Daniel W. Nagala, UTSI InternationalCorporation
Al Senftleber, Senftleber and Associates
Byron Coy, PE 439
Office of Pipeline Safety, Eastern Region
Pipeline Operators 460
Ed Nicholas, Alyeska Pipeline Service Company
Valencia Hiebert, Alyeska Pipeline Service Company
Mike Huber, Marathon Ashland Pipe Line LLC
Peter Evans, Marathon Ashland Pipe Line LLC
R.C. (Bob) Darwin, Member of the American Petroleum
Institute Cybernetics Subcommittee
P R O C E E D I N G S
9:33 a.m.
CHAIRMAN HALL: We will reconvene this Public Hearing of the National Transportation Safety Board that is being held on the subject of Pipelines and Pipeline Safety.
I'd like to welcome our guests this morning. For those who weren't with us yesterday, let me make a brief safety announcement.
In the event of an emergency, such as a fire, the building alarm system will activate, and a voice message will instruct persons to vacate the building. You should proceed to the nearest exit. There are emergency exits up front, to the left and to the right of this podium, and at the back of the room.
Also, for your convenience, restrooms and telephones are located in the foyer on the left as you exit the room.
If there is any questions or any assistance that any of the men or women that work here at the National Transportation Safety Board can provide for you during your visit with us, please just ask, and we'll try to see if we can't accommodate you.
I'm very pleased to note this morning that wehave with us the newly-elected congressman from the Bellingham area, Rick Larson. Mr. Larson, I believe, is in our audience in the back. I'd like for you to join me in welcoming the newly-elected congressman.
(Applause)
CHAIRMAN HALL: On a note before we move into this morning's agenda, regrettably, the pig that was located outside the door for observation yesterday will not be here today. The pig had first-class reservations on a USAir flight last night and had to leave. It's a tough crowd to get a laugh out of, I tell you.
We'll begin this morning with a Panel on Leak Detection Systems, and I'd like to ask our hearing officer, Joseph Kris, if he would please introduce the first panel of the morning, or Mr. Chipkevich, whoever.
MR. CHIPKEVICH: Thank you, Mr. Chairman.
On my left is Joe Kris, who's the Hearing Officer. On my immediate right is Alan Beshore, who's a Pipeline Accident Investigator for the Board. Rod Dyck, to his right, who's the Associate Director for the Pipeline Division. Eric Sager, sitting next to him, who's the Human Performance Investigator, and Jim Cash from our Office of Research and Engineering.
Mr. Beshore will be the person who will leadthe Tech Panel.
CHAIRMAN HALL: Thank you.
Please proceed, Alan.
MR. BESHORE: Thank you.
The next panel is the Leak Detection Systems Panel. On this panel is Mr. Larry Stack from Neles Automation, Mr. Alan Rodecker from Stoner Associates, Dr. Mike Yoon from Simulutions, and Dr. Diane Hovey from EFA Technologies, Incorporated, and, finally, Mr. Kenneth McCoy from Tyco International.
Panelists, I would like to remind you that you have 15 minutes to give your presentation. The yellow light in front of you will go on when you have two minutes remaining. The red light indicates that your time is up.
Mr. Chairman, the staff is ready to hear from the panelists.
CHAIRMAN HALL: Let me welcome each one of you, and we appreciate your willingness to come here and make these presentations this morning, and we'll look forward to hearing from each of you, and I guess we'll begin with Mr. Stack.
MR. STACK: Thank you, Mr. Chairman, for the invitation to participate in the panel.
This morning, I would like to take a more general deployment look at some of the technologies that are applied in our industry to the prevention, detection, location and isolation of pipeline leaks.
When we speak of this application and this technology, known as "computational pipeline monitoring", there are several aspects outside of the technology itself that must be considered, and I'd like to discuss those briefly in my presentation.
Including starting of the application --using the application of the technology in the engineering phase of the projects and the pipeline. Some of the pipeline infrastructure considerations that should be taken into account when applying this technology.
Also, the application of the technology itself to the pipeline leak detection prevention location application, and, finally, the role of operations, simulation and training is associated with equipping operators in their daily jobs.
Finally, in conclusion, some effective technology deployment considerations.
First, in the engineering phases, whether you're constructing new facilities or expanding existing facilities, in the engineering phase, thetechnology known as computational pipeline monitoring can be applied in an off-line modeling application to assist in the accurate design and verification of facility expansions.
In this technology, the advantage of using some of these techniques include visualization of the model, using modern drop and drag-type of techniques for construction of the model itself.
The application can contribute significantly to the design phase with variable fidelity, both in the dynamic and steady state calculations associated with the pipeline operations, and certainly linking to the business data that's involved in the economics and optimization of the pipeline.
One of the considerations that today's technology offers, of course, is the fact that one model can be used both in the design and engineering phases, on-line phases, and, finally, in the simulation and training of our operations people and engineering analysts on on-line type of applications, when operating the pipeline.
Traditionally, pipeline design had evolved with data acquisition systems, control systems being designed during the pipeline construction phase, and then, after that, is when the integration of logic,testing and training of the operators took place.
Using off-line technology in the design phase now allows us to do things like construct the actual control logic, begin starting and testing of our designs, and in fact training of the operator during the pipeline construction phase, allowing us to enter the line field phase of the project with well-trained operators who can concisely monitor and detect incidents on the pipeline.
Using the technology today, of course, building these models in a visual and graphic form allow us to better communicate to the engineers and to the operators of the specific design features and functionality of the pipeline system.
Regarding some of the pipeline infrastructure considerations that clearly affect the success of implementation of this technology include some points made here.
First and foremost is the communications infrastructure with regards to gathering data and performing control on the pipeline system. High-performance, reliable and deterministic type of communication is key to the successful and effective use of this technology.
Certainly as well, instrumentation withconsideration for the quality and accuracy of the instrumentation, the quantity of the instrumentation and, most importantly, the proper positioning or location of the instrumentation on the facility, and again during the design phases, this can be considerably improved during the design phases by using off-line technologies.
And an integrated model with the integration of other advanced applications associated with product movements, including monitoring and control, can give a common look and feel to the operator, allowing him to effectively and decisively operate and detect incidents on the pipeline and minimizing operator error.
The first level of application of this technology, and quite honestly most commonly used, is the on-line simple mass balance type of application of the technology. This is relatively inexpensive but appropriate if the infrastructure that we discuss regarding communication and instrumentation is weak.
It has general inaccuracies relative to some of the other technologies that we'll introduce today. It tends to be prone to false alarms during transient conditions and relatively slow than some of the other technologies, in the area of tens of minutes, to detect incidents, with lower location accuracy than can beobtained using advanced techniques.
But it does an effective job by accommodating for line pack and reasonable product tracking capabilities and inventory accounting type of capabilities.
The second and the most common technology looked at today is the on-line transient modeling. It represents a much more responsive type of technology with quite fast detection, in the order of seconds as opposed to tens of minutes, and it allows for these dynamic thresholds to minimize the false alarms which, of course, can contribute to our operations delays in terms of reacting to pipeline incidents, and has the capability to do proper detection through transient conditions on the pipeline.
It, too, provides very accurate product accounting and pig tracking and volumetric accounting applications. An accurate hydraulic profile generated from this technology allows us to monitor an alarm, maximum allowable operating pressures, even between measured points on the pipeline where we have instrumentation.
Some examples of performance attainable in a real application of this technology, the on-line technology, in the test environment are shown here, togive the idea of performance that can be achieved where communications infrastructure and instrumentation infrastructure is applied.
In this case, trucks were used, of course, to draw product from the pipeline. You can see in Test Number 1, 1.7 percent of flow rate or a small leak was introduced, detection times were in the two and three-quarter minute time frame, detecting losses of about less than five barrels.
In Test Number 2, about 50 percent more leak flow was detected in about 33 seconds in this case, limiting the loss during the leak to one and a half barrels.
In the final test, large leak, twice that of Test Number 2, five percent of flow, we were able to obtain with the technology in the order of 28 seconds of detection time and minimizing the leak to 2.3 barrels.
In these cases, this was a crude oil pipeline application, and you can see the performance that can be attained during the test procedures.
One of the key issues associated with applying this technology, of course, is the non-steady state operation, and in this chart, quickly just shows some of the issues that we have to deal with withrespect to heated products, cold products, transitioning in the pipeline, and, of course, the temperature profile shown here indicates injection of products of different characteristics with extremely different temperature issues to deal with as the product moves down the pipeline and cools.
With the simple mass balance type of technology that's generally used, this would be almost impossible. In fact, likely that leak detection system would be turned off during these type of conditions to reduce the false alarm situations.
We also had, just to present to you, some real world type of data. This was a case where we actually lost a gasket on a discharge valve. The real-time transient model was able to pick up the leak in 14 seconds, at about just less than five barrels of leak.
Of course, the operator was able to dispatch field personnel directly to a location within a few miles of the incident. The operator proceeded then under procedures to shut down the line. That took about four and a half minutes.
Proper officials were notified per procedure. Total loss in this incident was a 125 barrels. The good news, the clean-up and restoration was restored in the pipeline within about 11 hours.
Clearly here, the application of the technology during the training and simulation phase gave the operator the confidence to identify the leak situation quickly. False alarms were minimized to ensure that he was confident that he was making the right decision when he dispatched and began to isolate the leak, and quick action by the model, of course, minimized the impact of the incident.
One of the keys that we feel when working with this technology is to use it further in the training simulation area, using the actual pipeline to train the operator through the use of the model, and using the full control system integration with the simulation, so that the operator in fact has a consistent look and feel with his training environment as he does in the real-time operating environment.
This technology can include in-station simulation as opposed to just the pipeline simulation and the station simulation, including equipping the operator to deal with equipment failures in the station which also are common occurrences in his daily routine.
The accurate depiction of sonic events, very rapid communications, allowing resolution of the model every quarter of a second, give us the kind of accuracy that I was able to demonstrate in the real worldapplication.
The full function simulator allows a trainer to work with the operator, introducing malfunctions in equipment beyond just the ordinary scenarios that happen from day to day with the training model in control.
The full scope training simulator takes advantage of using the modeling technology that we know today, coupled with a virtual control system, to ensure that the operator is training on the same man/machine interface, the same look and feel that he is going to work with during the day in his job in the control center.
The trainer is allowed to use this technology to emulate the real field events, prepare scenarios, review procedures and process with the operator. He can be tested and retested to review procedures, and also traditionally these technologies can be used to review events, actual events from the field and to review and more concisely dictate procedures to be followed.
This current technology in respect of using this computational pipeline monitoring technology for real-time training and simulations to better equip our operators allows -- is different than traditionaltrainers, hydraulic trainers, in that the operator trains on a real situation on his real pipeline that he operates each day in his job.
The full control station simulation allows him to perform control. Sonic effects are reflected in the pipeline. So, he's dealing with real life situations.
Key to this is that he is using the same man/machine interface to make him very efficient and accurate in detecting the disturbances on the pipeline, and these technologies can provide up to greater than one percent accuracy for the critical variables that he must monitor and be aware of to detect application problems in the field.
To effectively deploy this technology, some considerations which I mentioned need to be looked at, but one of the key things that we look for the industry to begin to take seriously is to look at using one model in the off-line construction and design phases, in on-line, and in the training simulation application.
Using these techniques and giving the highest fidelity possible, we can have the best-trained operators, concisely able to detect leaks, to identify the incident in the system, to take decisive action without being distracted by false alarms.
The model itself must be self-tuning in order to minimize the false alarms which so often are the results of delays in taking action and isolating these leaks, and by following through with the technology to build effective training simulators, it ensures the operators that will recognize scenarios and pending incidents, practice fault isolation techniques, and make the processes and procedures second nature with respect to quick reaction to pipeline incidents.
With that, I think the technology today -- my final slide here would indicate that we believe the technology today, from the panel that we have here at the hearing today, can deliver. However, we need to employ technologies like advanced CPM modeling to our applications.
We do need integrated control systems and training systems with the on-line operations. We do need very accurate product and pipeline data to properly simulate and to model the pipeline in operation.
One of the problems that we certainly are dealing with is getting this very accurate data configured in the pipeline model.
We need integrated control systems and the advanced applications to allow the operator to be verycomfortable and familiar with the man/machine interface and the look and feel of these systems.
Finally, we need generally better communications. In the case that -- the examples I gave you, the real-life examples, fiber optic communications was being used, which gave us the benefit of being able to look at all pipeline parameters every quarter of a second or 250 milliseconds, giving very accurate representation of what's happening in the pipeline.
And, finally, the best instrumentation and control devices make the best model and the best prevention detection location and isolation of leaks, and we can, given that, provide very effective control in alarming for the operator, not to -- we can minimize false alarms and therefore his reaction time to events, and we can certainly create competent operators which take action immediately and concisely.
So, finally, in conclusion, as you will hear from the rest of the panel, the technology exists today, I believe, to move forward in this area.
Thank you very much.
CHAIRMAN HALL: Thank you.
We'll go to Mr. Rodecker.
MR. RODECKER: Rodecker.
CHAIRMAN HALL: Rodecker. Thank you, sir.
MR. RODECKER: I'd like to begin by thanking Mr. Chairman and the Board for this opportunity to present our technology.
Yesterday, we heard a lot of information about pipe integrity, specifically the pipe wall. Today, I'm going to talk about the technology we offer to the industry, looking inside the pipeline at the fluid itself, because behavior of the fluid in a leaking pipeline is different than the behavior in a sound pipeline.
But let me begin by giving you a short run-down of our current installations. We have more than 25 pipelines licensed with our technology, over 20,000 miles in length, pipe diameters that range from six inches to 48 inches, and the fluids are many crude oils and practically all of the refined products and natural gas.
One that did not make this slide is a CO2 application in South Dakota.
I'm going to run down briefly some leak detection techniques, so that I can give the Board an idea of where our technology fits into the bigger picture.
There are many. There are some physicalmethods, chemical detection, visual and acoustic monitoring. There's statistical methods, and there are those on this panel that can describe those today, and there's algorithmic methods, negative wave, volume, mass balance, and real-time models, and I'm going to focus on real-time models and give the Board an idea on how the technology works.
I think that's important because if you understand how the technology works, you can better understand its application and its limitations.
First, what is a real-time model? Mr. Stack talked about it as a transient model. I use the term "real-time model". It's a transient model. We also use the term "virtual pipeline".
A real-time model is essentially a virtual pipeline on a computer that continuously operates in step with the physical pipeline. It's fed data from the pipeline and keeps pace with actual pipeline operations.
Leak detection using a real-time model. I'll give you a quick overview of how this works. The real-time model continuously compares measured data with the virtual pipeline; that is, measured data coming from the physical pipeline is compared with calculated data from the virtual pipeline continuously.
If a leak develops in the physical pipeline, the measured data from the physical pipeline no longer match the virtual pipeline. The data in the virtual pipeline are then adjusted until the data match, and we look at these adjustments. We call them "hydraulic anomalies", and we look at those adjustments and analyze them.
If they're of sufficient magnitude and pass some mathematical criteria, we alarm a leak to the operator, and that leak alarm will have some information with it. It will estimate the location of the leak. It will estimate the size of the leak, and it will give a confidence level relative to the leak alarm itself.
To better understand how this works, I'll show three views of a pipeline. What we call the pipeline state is a physical pipeline. That's the steel in the ground, the tanks, the pumps, the valves and the fluid inside those facilities. It's continuous, both in time and distance.
That is to say, that all along the pipeline, there is pressure, there is flow, there's compositional analysis, there's temperature.
I've put a graphic up here showing the operator state, and I've got a controller that namedJoe, the operator, that looks at this sample pipeline, and what Joe sees when he controls this pipeline is a whole different view.
His view of the pipeline is discreet in both time and distance. He's looking only at three points on this pipeline, the beginning, the end, and the middle, and that sample data from the pipeline is only communicated to the operations control center periodically, and that could be five seconds, it could be 15 seconds, it could be five minutes, or like the example that Mr. Stack had, it could be fractions of a second.
The third view is the model state. It is continuous. The difference here is that the parameters inside the pipeline are estimated. There's estimated pressures, flows and compositions all up and down the pipeline, and it's continuous in distance, and it's also continuous in time.
I know the real-time model takes calculational time steps. Those time steps can be very small, and therefore for all practical purposes can be considered continuous.
So, the model is driven by data from the physical pipeline. This graphic shows data coming from the sensor devices on the pipeline being presented tothe virtual pipeline, but there's a problem in that the data from the pipeline contain errors or, more accurately, contain uncertainties, and these uncertainties have many sources.
There's some uncertainties relative or sourced in the instruments themselves, repeatability, time stamp, dead band filtering, calibration drift, to name a few.
There's also communication errors or uncertainties, analog-to-digital conversions, pulling or scan rates or even communication failure, and these errors are random. That is, we cannot use mathematic techniques to eliminate them.
So, we use a process that we call "state estimation" to treat these data, so that when they're presented to the real-time model, they allow the real-time model to accurately represent the physical pipeline.
We do this in a number of ways. First of all, the random uncertainties are related, and they're related because they're contained, and they represent a single body of fluid that's continuous all along the pipeline, and that's the behavior of the fluid that I spoke of earlier.
We apply the laws of fluid dynamics toestimate the best fit of every measured data along the pipeline, and they are voluminous.
So, let's look at our pipeline picture again. Here, we have a representation of the physical pipeline communicating raw, unfiltered data to a process, a technology we call "data filtering and state estimation", and from there, that data is presented to the real-time model.
Now, the real-time model must be tuned and calibrated to accurately represent the physical model, and there's many reasons for that, and there's many methods to do that, and I won't get into each one of those today, but I'd be happy to drill down into that technology should the Board desire.
So, what do you do with the well-tuned, real-time model that's running continuously? Well, you can do many, many things with them. One of the things you can do is look for leaks. That's what we're here to talk about.
I have a representation of a physical pipeline with a leak in it. We use two technologies that are working in parallel, working simultaneously to search for leaks in a pipeline.
One we call "leak analysis", and that's driven by pressures, and the other one we call "model-compensated volume balance", and that's volume- or flow-based.
So, let's look first at the technology of leak analysis. On the bottom, I have a representation of a physical pipeline with a leak in the middle. On the top, I have a representation of a virtual pipeline, the real-time model, with no leak.
In the middle, I've got a graph, and that will be recognized by many here as a hydraulic gradient. I'm plotting pressure on the vertical axis against the length of the pipeline on the horizontal axis, and the hydraulic gradient from the measured data, from the physical pipeline, in the white line, is different from the real-time model in the yellow pipeline, and the reason is there's more fluid in the upstream portion than there is in the downstream portion. So, therefore, there's more friction, and therefore, the hydraulic gradient is different.
The technology can look at that difference, and it can adjust the data in the virtual pipeline, so that those hydraulic gradients match, and here we show two flow rates coming out of the virtual pipeline. We call those "diagnostic flow rates", and they're necessary to make those two hydraulic gradients match.
When the two hydraulic gradients match, welook at those adjustments, and if they are of sufficient magnitude, and if they pass certain mathematical criteria, then an alarm is enunciated to the operator.
The second technology is model-compensated volume balance. Again, we look at both a virtual pipeline and the physical pipeline, and in this technology, which Mr. Stack described earlier, looks at the flow into the pipeline, all the flows out of the pipeline, and compensates for changes in inventory inside the pipeline between those two measurements.
If there's a difference, a significant difference between the real-time model, flow in and out, and the inventory, and the physical pipeline, then an alarm is enunciated to the operator.
But that's only half of the equation. The other half of the equation is what does the operator do when an operator gets an alarm? Well, if an operator gets an alarm twice a day every day, he's not going to do much. So, false alarms are very important.
But just as important, the operator has to conform to an approved procedure, and he has to practice response to rare or abnormal or dangerous events, and that can be aided significantly by training in a simulator.
I know there's members of the Board and in this room that are more familiar with the airline industry, which has a very long and successful track record in using simulators for training. So, I'll go through this very briefly.
Training on a simulator will translate knowledge into skill. You can recognize signs of emerging problems and act on those to mitigate their effects before they become dangerous, and, most importantly, you can exercise and practice response to rare events.
I've a comment here relative to operator training and the pipeline industry. The state of the science is such that a simulated environment can be created of such high fidelity that the pipeline operator cannot differentiate between operating the pipeline or operating the simulator.
Mr. Stack went through and explained how that works, but I wanted to point out that the degree of fidelity is such that you can get an operator that literally cannot tell whether they're operating the pipeline or the simulator.
In summary, the real-time model provides functional leak detection. Leaks are inferred from a virtual pipeline. The performance of leak detectionusing a real-time model is unique to each pipeline, and it's relative to the quality of measurements, the quantity of measurements, and the reliability of measurements and communications to the control center.
Operator response is enhanced through simulator-based training.
Lastly, I'd like to thank Mr. Chairman and the Board for giving me this opportunity to present our technology.
CHAIRMAN HALL: Thank you very much. We appreciate that.
Doctor?
DR. YOON: Thank you, Mr. Chairman, for the opportunity to share my experience and my thought in leak detection.
Since technical aspect is already fully discussed by Mr. Stack and Mr. Rodecker, I'd rather emphasize less technical aspect but more non-technical aspect, after I hearing those two individuals' presentation.
Okay. The objective of this hearing is to examine the capability of the existing leak detection and emergency response, and to understand the state of on-going research in this area.
To achieve these objectives, I'd like to putleak detection in proper perspective first, and then discuss what we like to achieve, if we can achieve, and also what kind of status we are in, and what kind of obstacle we have, and, thirdly, what kind of leak detection systems are available in the market.
So, in order to put leak detection into proper perspective, leak detection is one way to manage the risk and reduce the consequence of pipeline failure by ensuring fast and most likely, most importantly, reliable response to leak alarms.
If there are too many leak alarms, then operator tend to ignore that, and we have so many instances that why leak detection system fail in the past.
Okay. Leak detection, however, doesn't prevent and monitor the degradation of the pipe that leads to pipeline failure. In the strict sense, the leak detection has no effect on the technical integrity of a pipeline system. This technical integrity aspect of pipeline has been discussed fully in yesterday's hearing.
Okay. Let me define the scope of leak detection system as I see it. Okay. First, leak detection system includes field instrumentation and data gathering, such as SCADA system, and also leakdetection components, my two predecessors already described this kind of particular CPM method quite in detail level.
Okay. And also, one of the important aspect of leak detection system should include emergency response procedure, such as operator's response and confirmation of a leak, that type things, and also most important factor is, of course, pipeline operator, unless we have closed-loop control ideal system exist, but so far, it -- as far as I know, it doesn't exist.
As a result, pipeline operator is most important system which has to include, in my opinion, in the part of leak detection system.
Then ideally, what we like to achieve as a vendor, as a pipeline operator. Okay. One of the API's literature defines ideal leak detection system should never declare leak incorrectly under any operating conditions, even including select flow conditions.
Secondly, detect any leak always and immediately, and, thirdly, locate and size leak accurately, but I'd like to add a few more. Okay. First, isolation of the leak instantaneously is very important. If it is not isolated and is prolonged, due to whatever the reason, for instance, reliablerecognition, then still the fluid could be leaked.
And also, cost has to be affordable. If it is too expensive, then we're going to have a second thought, and also system is practical and easy to use and maintain. If you need half-dozen -- dozen different kind of rocket scientists to maintain the system, it's not practically usable.
Okay. I'd like to share some of the obstacle as I see. Okay. Pipeline configuration is so diverse, some of the pipeline is short, which is relatively easy to manage, and some of the pipeline system is so complicated and also quite long, and also, as Mr. Stack pointed out, instrumentation is very important, and quite often, sometimes, I won't say quite often, but sometimes, instrumentation is inadequate.
Also, data-gathering and availability is an issue, but these days, data and communication technology is well developed, and this is less an issue these days, and also operational issues, pumps start up and shut down, those kind of things in normal operation, but sometimes this one create will some of the -- create reliability problems, and also certain terrain is so unhospitable, and sometimes it create slack flow conditions. That kind of thing cause uncertainties.
Also, these days, pipeline carries -- single pipeline carries so many different kind of product, and recently, DRAs is quite often used to increase throughput, and the DRA, drag-reducing agent, change the radiological properties. So, simple modeling technique may not be easily applicable in such a situation.
Sometimes pipeline companies have some kind of difficulty to develop a concise leak response procedure and strategy because leak detection system is not 100 percent reliable, and also there seems to be no concise diagnostic tools to distinguish real leak from false alarms.
Also, emergency response is not -- certain company has well-established emergency response, but on the other hand, some companies don't, and also operators' or controllers' confidence levels sometimes is questionable when they are not fully trained.
And here, I'd like to share my experience and also identify some of the issues. Okay. In the past, last 20 years or so that I have been involved in this area, I saw many companies try this new technology in leak detection, but some definitely have a positive experience, but whatever the reason, some companies have poor experience, and either they got rotten systemor the expectation seems too high.
Also, I like to point out, frankly, that vendor tend to exaggerate the capability of leak detection system, and also, at this point, many companies do not have a leak detection system. Of course, I don't know exactly what's the main reason.
My question is that there's no -- they felt that there's no benefits or incentive to have a leak detection system. Particularly if you have unreliable system, of course, you don't have much benefit or incentive either.
So, at this time, some says that leak detection system is sufficient. Okay. I have a bit of reservation on that assurance. Definitely there's technical limitation, and key aspect is reliability, and also sensitivity will rely on instrumentation quite significantly, and also there are some non-technical issues has to be addressed.
For instance, usability, and some of the companies install this sophisticated leak detection system, but virtually they have more than half dozen people who support that. Okay. Those are acceptable or not, that's up to operating companies' discretions.
And here, one of the most encouraging things is API has taken initiative last several years atcoming up with nice guideline to evaluate particularly CPM, computational pipeline monitoring, system.
Okay. So, example include like API 1130, 1149 and 1155. Those are the guideline to follow, to evaluate CPM method, and also I saw that some of my clients have really nice -- they developed a nice response procedure, particularly in line with leak detection technology they have.
And also, one of the aspect we like to emphasize is the training aspect, and already DOT come up with some kind of guideline or some of the directives.
Okay. API 1130 defines several computational pipeline monitoring system. First is line balance. Line balance is applicable to short pipeline with small line pack, and also volume balance method. This one, on top of line balance, it take into account line pack change in the pipeline, and modify the line balance method, which we provide, includes elaborate hydraulic calculations, less demanding than this real-time model.
Of course, we both -- we offer both modified volume balance and real-time model.
Acoustic and negative pressure wave detection method. When there is a leak, then first pressure drops and also line pack is depleted, and the otheraspect is through the hole, it generate a noise, leak noise, and acoustic pressure wave -- acoustic method take advantage of this noise technique, and negative pressure wave takes advantage of this pressure depletion technique, and also pressure and flow monitoring is -- this technique is also used, same phenomena. Pressure drop and flow change.
Statistical analysis. This will be, I understand, discussed by my next speaker.
And also, there are a lot of leak detection techniques available. One is rate of pressure and flow drop devices. This was used to be a popular, but on the other hand, reliability again is a big issue. So, this is no longer that popular, and flame ionization method, this one is basically use the same principle that we have at home. What is it? Smoke detector.
And also, infrared thermography. Basically, it detect change in temperature, and leak detection peaks and hydrostatic pressure test methods were discussed yesterday session in fully.
Okay. As far as I'm concerned, I will say that I know the trend for research activity of every company, but I list through published literature and also through our activities.
I have observed the following trends. Ididn't see, at this point, I didn't see any technology breakthrough, but we continuously improve the existing technology, such as RTM method, and also many companies, I understand, focus on reliability and robustness.
Okay. It can be applicable any operating conditions, and also many companies also start developing emergency response in line with their leak detection system, and also they emphasize operator training.
As a conclusion or summary, definitely there's no ideal system exist, okay, but on the other hand, there are a lot of practical systems exist, and if I have any suggestion to pipeline companies, I rather start with the basic system rather than very sophisticated system.
Rather than run, we got to start walking first, and also emphasize on total solutions rather than just the leak detection technique only, including instrumentation upgrade, if it is necessary, and also leak detection technique, and also other factors, such as response procedures and operator training, and which is not on this slide, but most -- what I'd like to emphasize is a practical system rather than some system so elaborate that only rocket scientists can addressand maintain that system, and that's not workable in my opinion.
Thank you for the opportunity.
CHAIRMAN HALL: Thank you, Doctor, and is it Ms. Hovey?
DR. HOVEY: Hovey.
CHAIRMAN HALL: Hovey. Ms. Hovey. Dr. Hovey.
DR. HOVEY: Okay. Thank you.
Mr. Chairman, I'd like to thank you and the Board for inviting me here to speak today.
CHAIRMAN HALL: The button is out.
DR. HOVEY: The button is --
CHAIRMAN HALL: Not in. It's out.
DR. HOVEY: So, --
CHAIRMAN HALL: Is it working? Is everyone hearing you?
DR. HOVEY: It's working.
CHAIRMAN HALL: Okay. I apologize. Problem with my ears this morning. Proceed.
DR. HOVEY: Thank you.
Okay. The lead-in from the other speakers has been pretty thorough. I think I can be much quicker. Thank you for all of that pre-work.
The system that my company offers containstwo completely independent methods of leak detection. It contains mass balance with dynamic line compensation, which I think has already been talked about to the point of putting everyone to sleep. So, I shan't touch that part.
The part that is definitely unique is our statistically-based leak detection system, and within this package, the user has the option of using both one or the other. Since they're completely independent, it's the user's choice whether or not they want to use a full package or not. It's all in the box, and everything can be removed or added at their choice at any time.
The system can be offered as a stand-alone unit or it can be fully integrated into a SCADA system. Again, that's a user choice kind of issue.
It's presently licensed and monitoring over 300 pipelines around the world, and it's used on lines as short as small hydrant fueling systems through lines as long as a thousand miles that cross the United States, and the products that it covers are crude oil through ethylene, anhydrous ammonia, sulfuric acid, natural gas, methane, a large number of products, and it's the same product that gets put out at all times for monitoring all of those different situations.
Okay. Overview of the preceding methodologies. Mr. Rodecker covered this beautifully. So, I'm not going to touch on it very, very hard, but the computational -- the transient model is doing computations and matching the calculations against what's going on within the pipeline, and when it determines that there is a sufficient difference between those conditions, it declares a leak.
And in order to do this correctly, they need a great deal of information to create the basic hydraulic model, and that includes items like knowing the pipe diameter, and if the pipeline changes diameter, that information needs to be known and where that occurs.
The wall thickness, the pipe material, the pipe length, the insulation, how deep it's buried, its elevations, and if it's a very old pipeline, where we like to say the as-built drawings aren't, it makes it much more difficult to create a good physical database, and this can be extremely difficult when bringing the system on line.
It's simply a fact of life. We used to do a transient model, and I'm eternally grateful that we no longer do that.
The system also requires input from the SCADAsystem. It needs to know the flow rates, the temperatures, the pressures and the liquid properties, in order to be able to make correct calculations for what's going on within the pipeline.
For pressure point analysis, what we're looking at is not so much what's going on but the hydraulic noise and how that's changing. So, it's kind of an inside-out way of handling leak detection.
We're looking at only the most recent information, the most recent readings. In a liquid pipeline, this tends to be about the last five minutes, and if we're dealing with a relatively low-pressure gas pipeline, it might be the last hour, but we really don't care about what happened yesterday or 15 minutes ago.
The system is taking what is occurring now and looking for the most recent readings that are coming in and determining statistically if those are changing in a manner that's characteristic of a leak. So, it's looking for patterns within the noise.
Is there a pointer on this little thing? No. This particular -- is that the pointer? That one. No. Can I go backwards? Okay. I don't see -- okay. We're going to -- there it is. Good.
Okay. This particular plot was done duringthe early development of our pressure point analysis, our statistical leak detection system.
Each of these little bars that's across here is a two-second interval, and this was a leak test that was conducted on an emulsion pipeline. It was about a 20-percent water. It contained particulates, and it contained gas bubbles, about 25 miles long and about 26 inches in diameter, and we initiated a leak at that point, and what the graph is showing you is how that leak affect, that negative pressure wave propagates down the pipeline, and it moves at about the speed of sound in the product constrained by that pipeline.
So, in the case of crude oil, it moves at about 3200 feet per second, and in jet fuel, it moves it about 4000 feet per second, and what we're looking at then is the ends of the pipeline in this case.
We determined that the expansion wave, this trough that you're looking at, travels at about 25 to 30 miles before it begins to attenuate. So, for our method of leak detection, there's absolutely no need to put intervening pressure instruments. Instruments 25 to 30 miles apart see essentially the same effect within the pipeline.
We look at preferably -- where did it go? Flow and pressure at one end, and flow and pressure atthe other end, ideally. What this allows us to do is to monitor what's happening on the pipeline because if it's a leak, we're going to have an increase of flow rate down here at the pump or compressor end because we're essentially now serving a lateral to the pipeline, and down on this end, we're going to see a reduction in speed for the product going through the pipeline because some of it's being siphoned off through the hole in the pipe.
So, if we're monitoring all of those instruments, what we're looking for is the change at a given instrument. We're not comparing values. So, the relational change of the flow meter at this end in the expected direction with the pattern that we're expecting to see, that instrument, that given instrument within the system says I think I see a problem.
The pressure instrument at this end agrees. Then down at this end, if the flow instrument here sees a pattern change that's a decrease in the flow rate, without comparing values, the system will know that it had a leak on the pipeline.
If it had a pump drop at this end, that flow rate's going to decrease instead of increasing. So, the system's logic is able to determine that that was atransient activity and not a leak. So, it can effectively separate transient effects from leaks, and we have a number of installations that do have a 100 percent nuisance alarm-free operation, but it does require proper instrumentation at the ends and good response back to the leak detection system.
We also require a 16-bit resolution because the 12-bit resolution that's fairly common in some PLCs will create a stair-step effect in what the system is looking for, and we will lose leaks in that essence because the small leaks are smaller than the stair-step, and hopefully I haven't lost anyone on this or put them to sleep. Okay. That's our expansion wave.
Essentially, if we're looking at pressure across here, the system is expecting to see it extend out pretty much the same way over time. So, if we have a leak, the statistical filters determine that we have something going on here that is not normal, and, so, our range of detectability is in this window here, and depending on how much noise is normally on the pipeline, because we do have to accept the normal peak-to-peak variation that's caused by bends, valve operation, you know, the usual things that occur within a pipeline, we accept those as normal. They're random.
We're looking for the pattern that's hiddenwithin the random noise. So, we can -- it's gone. We can detect very small leaks out of the noise, providing that they are at the edge or somewhat larger.
So, in this instance, okay, this is a tank facility, and here, there's the arrow, you can see that we have pressure transmitters and flow meters across the pipeline at this location. That separates the portion of the pipeline that's going down to the wharf that we're monitoring. That separates us from the tanks.
So, the system is able to determine when a tank switch or something else has occurred or whether a switch between holes on a ship has occurred, and we have good pipeline leak detection within the facility on lines that are normally considered fairly difficult to monitor.
Flow meters, and this particular facility with these five lines has been nuisance alarm-free since it was installed three years ago.
Okay. What keeps leaks from being seen? That is a big question. Leak detection sensitivity is context-dependent. Okay. The leak detection system cannot respond to anything that the instruments are not reporting back to it. So, those instruments must be capable of responding to the changes caused by a leak.
In our case, we only need the instrument to be repeatable. That means that if it's given the same situation multiple times, the instrument will give us the same answer. It doesn't have to be the right answer, but it's got to be the same answer because we're looking for a relational change, not a specific value change, and communication with the field instruments must be stable, and, yes, it's getting better, but it isn't always great.
You can have a mix of communication systems be radio, microwave and other things coming in, and they aren't always consistent in the data that they're getting to you, and some pipelines that run across Texas, Oklahoma, there are huge sections where there's just no way of getting data back, and, so, those pipelines do need to be monitored with larger spaces than is desirable with the instrumentation.
Instrument noise. I think we've all talked about it in one range or another. It's the instability of the instrument. It's repeatability. Corruption of measurement values, data conversions.
Some PLCs and other equipment as they pass the data in can truncate significant digits that contain information that says that there is a leak on the pipeline, but in order to have fewer visiblechanges on the displays, that information can be removed, saying that maybe it's a one- or two-pound change, so that it stays constant on the display, but for leak detection, that means then you're not getting the information you need.
So, in our case, we need to come in underneath that and take the data before any of the trimming has taken place, and voltage drop and radio interference can also be a problem.
Hydraulic noise. All of the leak detection systems tend to have problems with that. Pumps and compressors. If the instrument's placed too close to a pump, what we're going to be seeing is variations in discharge pressure as opposed to the real pressure on the pipeline.
Control valves that are correcting too quickly. They put a lot of hunt and seek and over-ranging on the pipeline, and, so, models and our system both have to deal with a fluctuation that really doesn't need to be there.
Bends and constrictions on the pipeline create hydraulic noise and obstructions or places and insertion tubes, all of those create fluctuations that have to be dealt with, and how do you know it works? That's always a big question.
When we go out to sell a leak detection system, we actually like to go out and do an on-line test before someone purchases a system, and for our type of technology, that's easy to do.
This particular pipeline runs across the Persian Gulf. This is on the island of Bahrain, and the line is over at Saudi. The leak is -- was generated on the Saudi side. The leak detection system was over on the island of Bahrain. We detected on a loss of three gallons, it's a quarter-inch leak, and that was about a half a percent flow with normal operation on that pipeline, and they routinely test this leak detection system, so that they know that it's operating at the level that it should be.
And that completes my presentation, and thank you.
CHAIRMAN HALL: Thank you very much, and we'll move to our last presenter, Mr. McCoy.
MR. McCOY: Thank you, Mr. Chairman, Members of the Board.
My name is Ken McCoy. I'm the General Manager of the Tracetec Product Group, formerly part of Raychem Corporation, which may be a name recognizable to some of the pipeline people in the audience. Raychem was acquired by Tyco International about 15months ago. So, we're now part of the Flow Control Division of Tyco International.
I'm going to talk about the different approach to leak detection than the four speakers before me. Our system is a direct contact physical or chemical detector. It's a sensor cable that is placed in proximity to the object that we're trying to monitor for leaks, in this case pipeline, and as you'll see, it has quite a different profile of operation than what you've heard about so far this morning.
In many ways, it's a very good complement to the SCADA technology that you've heard so much about.
First of all, I'd like to just dwell a second on the issue of speed. You've heard quite a bit of discussion this morning of how important it is to detect a leak quickly and to shut down the line, and, of course, in the context of major pipeline failures, that's a very important characteristic of the system. You have to be able to stop the pumps, close the valves quickly in the event of a blow-out.
But there are other situations that occur when the leak is essentially a small low-rate leak, but if it goes undetected for a long time, it has very similar consequences to a big leak that lasts a little while.
These are particularly important in environmental issues, contamination of aquifers clean-up. One of the things that's important from a public relations point of view in fact is if you know about a small leak yourself as the operator, you can go out and react to it. You can get your own crew out there before you're notified by public agencies, before a neighbor calls up, before someone smells hydrocarbons or, as in the case of Mr. Johnson from Alyeska yesterday, the last thing you want is that leak detection of last resort, to have a big enough problem that you see it in the weekly fly-by.
So, we're looking at a different range of leaks, but it's an interesting range. This is data that's taken from the Office of Pipeline Safety database on their website. The horizontal axis is spill size in hundreds of barrels, and the vertical columns are just a physical count of the number of incidents that have occurred since 1986. This database goes from '86 to '99, and what it shows you is that there are quite a few small spills.
In fact, 70 percent of their total reported incidents were 500 barrels and less. Now, the big ones, the far right-hand bar, they make the news. That's the blow-outs, the catastrophic failures thatwe've heard about, but as you can see, the actual size of spills, the great quantity, is on the left-hand side, where they're much smaller.
Now, either that's because the SCADA systems are working very well or because a lot of spills are very small, slow leaks that get reported and detected visually.
So, I want to talk just a quick summary. You've heard so much about SCADA, I won't dwell on this. In fact, I'll just jump right to those three indented points in the middle.
What that says is that if you have a large deviation from the predicted flow in the pipeline, the SCADA system can detect it very quickly. That's the kind of percentages we're talking about blow-outs. Let's say, three, four, five percent or higher fraction of the line flow is suddenly missing. SCADA is going to find it. It's going to shut down the line.
The smaller deviations take a longer time. I think Mr. Stack's data, we heard earlier, kind of goes along with that. But the problem really is that very small leaks may go undetected in their entirety.
This curve -- I apologize for the quality. This was taken from the website of one of the SCADA companies. This actually is from a British company. Ichose this particular graph because it's one of the few sites that actually showed this relationship graphically.
The horizontal axis here is percent of the leak or percent of the flow in the line that's leaking, and the vertical axis is the response time in minutes, and this is pretty consistent with the data that we already saw this morning.
Basically, what it says is if the leak is in excess of, let's say, four percent, SCADA is going to find it quickly, shut down the line in a minute or less, but as that leak percentage moves to a smaller and smaller fraction of the flow rate, the response time gets larger, and in fact, if you get down to the flow rates of less than half a percent, certainly down in the tenth of a percent range, the detectability gets very, very long.
Now, if you do a little arithmetic, you can look at the consequences of this. What this slide basically says is that if you take the total flow rate in the line, you multiply it by the fraction that's leaking and multiply that by the response time, you'll have a good estimate of the size of the spill before the system is able to detect it.
So, now we're taking the time parameter,which has been emphasized quite a bit up until now, looking at what that real percentage and time means in terms of total quantity spilled, and what this says is that if you have a high rate of leakage, you detect it quickly, you have a relatively small spill.
If you have a low rate of leakage, you can take longer and still have a small spill, but there are two relatively large or two bad situations. One is that you have a high rate of leakage, and you react slowly. That's going to give you a big blow-out, and likewise if you have a low rate of leakage, but you just failed to detect it, over time, that will grow into a very big problem.
Now, I don't want to try to have you interpret the entire graphic here. This is a little bit too much data, but this is the implications of that curve that I previously showed you, and what I want to draw your attention to is the left-hand lower corner.
What you see in those columns are very small fractional leak rates but allowed to exist over relatively long times, and these are the kinds of spills that are being detected visually today.
This is the fly-over spill or the one where the hiker smells gasoline in the creek bed or calls it in. It's the kind that are not being detected by thesystem, and oftentimes these are being detected by people totally unrelated to the operation of the pipeline.
Okay. Now, I'm going to shift over to the alternative. This is the technology that my company manufactures. This is a leak detection cable. The cable itself is about three-eighth inches in diameter. It reacts to the physical presence of the liquid hydrocarbons. This is not a system for natural gas pipelines. This is limited exclusively to liquid hydrocarbons.
It reacts to the liquid or it reacts to very high vapor concentrations. The time that it takes to react is a physical determined process. It takes anywhere from about 10 minutes for gasoline up to several hours for a heavier, less volatile liquid. A crude oil, for instance.
Well, the implication of that, as I'm sure you can appreciate, is you would not want to use this thing exclusively for a rapid shutdown. You don't want to have a pipeline leaking for two or three hours in a major blow-out condition while you're waiting for this sensor to react.
But the converse is that in slow leaks, this will still react in 10 minutes, as soon as the gasolinegets to this sensor cable, whether the leak is a few gallons per day or a few hundred gallons per minute.
As a bonus, these systems are able to tell us the location of the leak, plus or minus three or four feet, actually a little better than this slide shows, and certainly within backhoe digging accuracy to go out and investigate the problem.
This is a cross-section of the sensor cable that I'm talking about. You can see a bundle of wires in the middle. Those are primarily used for communication within the cable. The two black wires are the actual sensing electrodes. The black band that you see in the middle of this slide, that's the actual sensitive element of the cable.
That's a sponge, if you will, that soaks up jet fuel or diesel or gasoline, and as it soaks up the liquid hydrocarbon, it swells physically. As it swells, it eventually makes contact with those two black wires, and that's the switch that turns on the detection mechanism and helps us locate where the spill has occurred.
For the pipeline application or beneath tanks or buried valves, we put the sensor cable into PVC conduit. This is slightly thicker wall, slightly larger in diameter than what you'd use in your backyardsprinkler system but pretty darn close.
The difference is that there are a series of cuts made into the wall of this tubing. It's essentially an electrical conduit that keeps the dirt and the sand out but allows liquids and vapors to get into the interior, and we use it as a conduit.
We have this installed while heavy construction is going on. We come out to the job site after the heavy physical work is done, and we pull the sensor cable into place, just like an electrician would pull wire into the conduit in your home.
This is a cross-section showing how the sensor is installed in a pipeline application. Typically, we would like to get this conduit and cable system anywhere from the 5:00 position or the 7:00 position relative to the pipeline. It sits in the same trench with the pipe. It sits on the sand pad that the steel pipe rests on.
This shows a new construction example. It's probably the most effective way you can pick up spills a little bit earlier, but it can be trenched in along-side at greater expense obviously, and what we do is we install long runs of this conduit, typically anywhere between 400 to 800 feet between pull points.
This is the same kind of data but for thecable, and the point I want to draw your attention to here is the relatively large numbers on the right-hand columns.
What this says is that if you have a major blow-out, and it's a less volatile fuel, like a jet fuel or a diesel or heating oil, crude oil, you're going to spill quite a bit before you detect it. So, the cable by itself is not the answer either.
But we have a complementary situation here. Where the SCADA is weak, that is, in the small slow leaks, the cable does a very good job at limiting the spill size, and, conversely, where the cable is weak, in that it cannot react quickly enough to a large blow-out situation, the SCADA is quite good.
Graphically, this is what it translates to. If you look at the red line, this is the SCADA curve coming from the upper left and going to the lower right. This is when you translate that earlier graphic I showed into barrels spilled, and I'm using a 5000 barrel per hour flow rate here. This is an 18-inch line. Basically, this is the data that was used for the Longhorn Pipeline being considered in Texas.
If you look at the blue line, that's the cable system by itself, and again you'll see over to the right side the amount spilled at high leak ratesfor cable by itself is just too high.
But the important line is the green line. If you use the two systems together, they complement each other so nicely, we essentially cut the knee off of that SCADA curve, and the two together are a very nice system.
All of the improvements in SCADA that we've been hearing about, the lines of development will essentially push that red line further down and to the left, but they won't change the fundamental shape of that curve.
I won't go over that. Just to show you, this is the same set of numbers but with the two systems combined, and you can see, just like the green curve showed us in the previous slide, the total spill is quite constrained with both systems.
This is the cable reel, and why isn't it being used? Well, it actually was developed in response to some EPA regulations in the mid-1980s. The Underground Storage Tank created quite a market in double-wall pipe and double-wall tank, and this cable has been manufactured since the mid-'80s but primarily installed in double-wall pipe systems. That would be jet fuel distribution at airports within fenced facilities. It has not been used on line pipe to anydegree at all.
We began experimenting with this direct burial technique in the mid-'80s, actually working with the utility company in New York City to monitor oil in high-voltage pipe-type cable, and in those applications, we developed a slotted conduit as the best mechanism for installing the sensor cable.
We began working with single-wall pipe on some pilot facilities in military airfields in the early '90s, and those have been our test beds. Some of the worst case conditions we've seen have been in those brackish water, swampy conditions in South Carolina, monitoring a Marine Corps air station fueling facility there.
More recently, we've had numerous installations of this product using this installation technique beneath aboveground storage tanks in California, under valves, under tanks, under valve manifolds.
The State of Florida has looked at this and decided it's a pretty good technique and has approved it for use under their aboveground storage tanks, but our first true long-line pipeline application is scheduled to go into ground this Spring, and that will be the Longhorn project in Texas.
Instrumentation for this system is we basically look at about a mile-long segment of cable. We look at it with a very low-power small instrumentation node. The instrumentation consumes only about one watt of power. It's actually low enough in power that it could be solar powered, and it can communicate directly with the SCADA communications backbone, although we are looking at some tie-ins to what's known as microburst technology to take advantage of cellular phone systems, essentially to allow this system to operate totally independently and as a redundant system to anything on the SCADA package.
So, if SCADA communications break down for whatever reason, this system theoretically could operate entirely independently of that communications channel.
I apologize. I didn't know that you were not supposed to consider cost. I put this slide in before that, but this is a big significant issue for the pipeline operator. This is at this stage a relatively expensive system, as much as $50,000 a mile for new pipe installation, although we do expect that cost to come down rather quickly as volume increases, and if you're going to ditch it in, I have one anecdotal bit of data that says it costs around $25,000 per mileextra to trench that in along an existing pipeline.
So, in conclusion, SCADA is great. It's good for long pipe runs, but it does have some significant disadvantages, that being that it's relatively insensitive to low leak rates. It is approximate only in its location capability, and as we've heard, there's quite a bit of balancing here between false alarm rate and sensitivity.
The more sensitive you make it, the more likely it is to create false alarms. The more false alarms you create, the less likely an operator is to respond. So, you're in a constant balancing act of sensitivity versus false alarm rate.
The sensor cable. Basic disadvantage there is that it's very expensive, and people just are not going to be willing to put it in the long runs of pipe. The response time is a little slow, so we wouldn't want to operate it by itself. But the two together are a very nice package. They are probably very appropriate for the SCADA covering the long runs, and the cable being used in areas where there is high consequences, extreme environmental sensitivity, or issues of that like.
Thank you very much.
CHAIRMAN HALL: Well, I'd like to thank thepanel, and I'd like to -- there's a whole number of questions in my mind, but I think they'll probably --well, most of them will be addressed by the Technical Panel and my fellow Board Members.
But let me just stress before we go to the panel the importance of your work and the importance of these systems.
I was trying to sit up here, Mr. McCoy, while you were presenting, just thinking about, since I have been the Chairman, the brief period of about six years, the leaks that we have gone to that affected the Tennessee River, which happens to be the drinking water supply for my hometown, Two Mile Creek up in Kentucky, a whole marshland in Grammercy, Louisiana, the Reedy River in South Carolina, the Patuxent River in Maryland, the Rappahannock River here in this area, and another response we made on the water supply -- that leaked and almost reached the water supply for Dallas, Texas.
So, we respond on environmental spills as well as damage to human life, and I just -- you know, this is extremely important, and I appreciate all of you all being here and providing this background and information.
We'll move to the panel. Mr. Beshore.
MR. BESHORE: Thank you, Mr. chairman.
What pipeline-operating conditions and/or physical pipeline system characteristics have the most negative impact on the effectiveness of the CPM leak detection methodologies?
MR. RODECKER: I'll go ahead and take that.
Pipelines are dynamic typically. There are those that operate in stable conditions, but as pumps turn on and turn off, they create pressure waves and transients that bounce around inside the pipeline, and that limits the effectiveness or the performance of the leak detection system.
Also, the instrumentation itself. A well-instrumented pipeline will have higher performance, as Mr. Stack demonstrated with the one that had the fiber running next to the pipeline.
So, it's instrument -- it's got to be instrumented appropriately, and then, when it's dynamic, then the response of the leak detection system, the CPM system, will be degraded during that period of time.
Those that have leak detection thresholds or alarm thresholds that are fixed likely will be turned off during those periods of transient times. Those that are dynamic, the alarm thresholds are dynamic andgo up and down relative to the hydraulic performance of the pipeline, likely will not be turned off. It's a way to minimize false alarms.
MR. BESHORE: Does anybody else have anything to add to that?
DR. HOVEY: I'd like to make one comment, also, that the instruments have to be maintained. If the instruments are not well maintained or if, let's say, a meter isn't functioning properly, it will, you know, create problems for both systems, statistically based or the model based. So, maintenance, I think, is crucial.
MR. BESHORE: Can you describe the impact of data communication losses on these systems, and how quickly they recover once communications are restored?
DR. YOON: These days, communication technology is well advanced. So, many SCADA companies, like Neles Automation, take advantage of that kind of technique fully.
But on the other hand, occasionally there is communication outage, that kind of thing occur. That's unavoidable part of it. So, these days, most CPM method will try to take into account those factors as much as it can, but specifically how do it -- how we do it, each company has different approach.
MR. RODECKER: I might add one comment. In the -- particularly in the natural gas industry, I mentioned real-time models do many things.
One of the things that they can do and do do in application, where there's a communications failure and part of your pipeline goes dark on the SCADA screen, pipelines that have a real-time model can actually operate their pipeline using the model data for a time.
Of course, those estimates get less and less accurate as time goes on.
MR. BESHORE: Anybody else have anything??
(No response)
MR. BESHORE: Now, how about start-up conditions? How quickly after a pipeline starts -- is started up do these types of systems become effective in detecting leaks?
MR. STACK: I would comment that once flow rates have stabilized, and the model can populate them, they can become effective within days of pipeline line fill.
MR. BESHORE: Well, yeah. I wasn't necessarily referring to the initial start-up of the pipeline at the beginning of a new pipeline. I was thinking more in terms of restarting a pipeline, youknow, in regular operational conditions on a day-to-day basis.
MR. RODECKER: Let me respond briefly to that. Pipeline start-up and shut-down is an example of a control move on to a pipeline that creates pipeline transients. It's a significant one and probably an extreme condition, but a pipeline that is shut in, that the pumps are stopped, and the valves are closed, you can still have computational pipeline monitoring.
So, you can go through the process, albeit degraded, with those pressure transients caused by the start-up, and you'd want to be very careful with interpreting alarms during that period because a lot of things happen during pressure transient waves inside the fluid during start-up.
DR. YOON: I'd like to add one comment to that. When pipeline is shut in, then quite often, valve is closed, but if instrumentation is not properly installed, then you are basically cut off because of the valve closure.
So, those kind of effect has to be taken into consideration, and also temperature is another factor. So, temperature can be measured, let's say, in the middle run or something, and it can measure ambient temperature, not the fluid temperature, then it cancause some kind of false alarm, too, if temperature is falsely interpreted.
DR. HOVEY: Okay. We all get to take turns on this one.
With the statistical methodology, when a line is blocked in, and it's static, for us, the sensitivity is considerably tighter, and we can find much, much smaller leaks than we can when the line is flowing, and our system for the State of California has this underground storage tank regulation for -- they don't care what size the pipe is. It must meet certain criteria, which is the three gallon per hour at 10 psi if the line's connected to an underground tank, and our system was able to get third party certification to do that.
But there's no way in the world we're going to have that kind of sensitivity when the line's flowing. When your system then starts back up into operation, how long it takes for the line to get to steady state has a lot to do with how long that pipeline is and what the product happens to be, and for our system, we go to a degraded level of sensitivity, so we can still detect leaks, but prior to start-up, you know that that line was solid because you've been monitoring in a static condition.
Then it transitions into the flowing state and then stabilizes back out, but you will have a degraded period during that transition.
MR. BESHORE: Mr. Rodecker, I think you kind of touched on the next question, but can you describe some of the differences between natural gas transmission and hazardous liquids pipelines in terms of the key monitored variables and the associated modeling equations and leak detection capabilities?
MR. RODECKER: There are two. But, first, two major differences, but first let me tell you that the technology is equally applicable to natural gas as it is to liquid pipelines.
However, the performance is much different, and there are two reasons for that. One is physical. The physics of the natural gas fluid are much different than a liquid fluid in terms of compressibility and wave speed. So, where there's a leak on the pipeline, the fluid properties transmit the signature of that leak up and down the pipeline. That movement is much slower in gas than it is in liquid, and the amplitude of that signature actually dissipates in the compressibility as it goes up and down the pipeline.
So, the first one is the physics. The second one is more the business of gas pipelines. Liquidpipelines tend to measure everything coming in and out and to communicate those measurements to the control center on a very frequent basis.
Legacy Gas Systems, natural gas pipelines and transmission and distribution, measure everything coming in and going out, but the communications of that data to the control center sometimes do not exist. It might be a dial-up information that dials up and reports the flow data once a day, or it could even be paper charts that are collected every seven days, and the technology cannot differentiate between a leak in the vicinity of a seven-day chart that has a big increase in flow. It just cannot differentiate between the two.
So, the one is a physical, having to do with the difference in fluids, and the other one is the way the flows are transmitted to the company.
MR. BESHORE: Dr. Hovey?
DR. HOVEY: Could you give me the question again? I think I was woolgathering.
MR. BESHORE: It was mainly in terms of the differences in approach between natural gas pipelines and hazardous liquid pipeline facilities.
DR. HOVEY: Okay. For statistical leak detection, like Mr. Rodecker said, it's the speed atwhich it occurs in the pipeline, is the main difference. Statistically, it's handled at exactly the same, except that we use a slower update rate, and the window of time that we look at can be as short as the five minutes but frequently it's closer to an hour.
We've got some pipelines that actually operate almost like a fluid line, so that the update rate is the same as a fluid line. We have other gas pipelines that operate at high pressure, going under long segments of water. The instruments are about 240 miles apart, which does limit the lower end of sensitivity, but we're still detecting about one percent of flow leaks, even though our instruments are about 240 miles apart. But again it's because we're looking for differences in the noise that's within the pipeline. So, it makes it easier for our system to be sensitive.
CHAIRMAN HALL: Mr. Beshore, I think that's an appropriate point for us to consider a break. I have been observing the faces in the audience, and I think it is time for a break.
Let me tell you before you move from your seats, please, the plans for the day because we've got a lot of information we still want to get in. We'll take a break now.
Wherever we are at 12:45, we'll stop for lunch. We will only take 45 minutes for lunch, and, so, if you want to slip out a little early or you want to -- whatever you want to do, and then we'll start and reconvene at 1:30.
I don't think the way it's going, we'll be able to not try to break into some questions or something at some point, but that will be -- we'll take a break now, break for lunch at 12:45. The afternoon will start at 1:30.
I'm aware that a number of people have flights and things this afternoon. So, we'd like to try and move it along.
So, we'll stand in recess now for about 15 minutes or come back here about 25 after. Thank you.
(Recess)
CHAIRMAN HALL: I'm going to reconvene this hearing, and the first thing I'm going to do is exercise the prerogative of the Chair, if my three Board Members don't oust me, to ask two questions.
Why don't all companies have leak detection systems? Does anybody want to take that on? I've heard all the technical. You've convinced me that we have many, many aspects, and I'm watching all the --and I'm just wondering why don't all the companies haveleak detection systems.
DR. HOVEY: Okay. I'll stick my neck out.
CHAIRMAN HALL: Dr. Hovey.
DR. HOVEY: We have -- well, first of all, we have a lot of customers that are extremely proactive, that have spared no expense in putting in a good leak detection system.
Then we have other people who come to us who are definitely interested in putting in a leak detection system, but when they start looking at all of the costs that are required for the instrumentation, the leak detection system, and everything else that goes with it, and then they look at their competitor across the street, who has no intentions of putting in a leak detection system, and then they look at the bottom line of what their investors are going to think about the profitability at the end of the year, they elect not to put in the leak detection system, simply because of the bottom line issues, not because they're uninterested or unwilling to do it.
CHAIRMAN HALL: So, the absence of any type of level playing field of minimum regulations in this area leaves it, you know, where it's -- obviously the good soldiers here are bearing an unfair economic hardship many times, if they are being proactive interms of safety.
DR. HOVEY: I believe that's correct.
CHAIRMAN HALL: Mr. McCoy, you mentioned this double pipe. You want to tell us about what a double pipe is, and how extensively that's used by the military or around the country?
MR. McCOY: Well, it's a technology that basically came into being as a result of quite a bit of EPA regulatory efforts in the mid-1980s.
Double pipe is what it sounds like. It's a smaller pipe inside a larger diameter pipe. Many of the companies that produce those type of systems were in the business of making insulated pipe systems for hot water and chill water delivery, and you take the insulation material out, put a set of centralizers in there to keep those two pipes concentric, and you have a double pipe.
One of the problems with it is it's very craft-sensitive to install. It's quite a bit more difficult than welding up single wall steel pipe, and we've sold quite a bit of our product into that marketplace but not without quite a few difficulties.
It's a tough installation. The rewards you get for that extra cost is you can contain a leak when it occurs. You can put it into a very easy-to-detectenvironment. Basically if you put a sensor cable like ours in that space where there is nothing, and suddenly there is a pool of diesel fuel, you can tell that pretty quickly and cleanly and know where it is.
But it's very expensive. It's roughly four to six times the cost of the equivalent single wall pipe. So, it's installed rather limited in sizes, certainly not in the scale of diameters that we're talking about for true pipeline operations. It's primarily installed inside of fenced facilities, military.
Air fields have used quite a bit of it. Commercial establishments might use it. A car factory, for instance, would have underground storage tanks with various liquids, and they would pipe that to the assembly line in double wall pipe. So, it's primarily a smaller scale of installations than what we're talking about here in this hearing.
CHAIRMAN HALL: Are you aware of anyone that has used it for environmental reasons, such as crossing rivers or being near drinking waters or sensitive --
MR. McCOY: Well, there's case pipe crossings frequently used. I can't speak, I'm not that much of a pipeliner to talk about river crossings, but certainly road crossings are -- oftentimes a larger diametershell is used as a conduit to put the line pipe through.
I don't know that that's been done extensively for aquifer protection. It's usually more for protection from mechanical damage at the crossings.
CHAIRMAN HALL: Okay. Well, those were my two. I wanted to get some understanding of -- basic understanding in a non-technical manner.
We'll now go back to the Technical Panel to continue technical questions.
MR. BESHORE: Thank you, Mr. Chairman.
I just had one final question, also, and it was for Mr. McCoy as well.
Nowadays, a lot of the pipeline crossings of waterways and environmentally-sensitive areas are done by directional drilling. Can your product be installed in conjunction with that type of an operation?
MR. McCOY: With difficulty. The problem is that we can't really survive too much mechanical abuse while the product is being installed. If that PVC conduit system or an equivalent, probably a tougher equivalent for a directional drilled pulled-to-crossing, if that can survive the installation process, then the sensor cable can be pulled into place after that.
But we don't have a lot of experience, and I wouldn't want to make any guarantees that that's something -- I don't think it should be a universal recommendation yet because we just don't have the experience and the track record to know that it would work.
MR. BESHORE: Okay. Thank you. Mr. Sager?
MR. SAGER: The emphasis that I have for my questions are a little bit different. I'm representing Human Performance or Human Factors.
CHAIRMAN HALL: Eric, it may be my hearing. Please speak into the microphone.
MR. SAGER: I have several questions that are going to shift the emphasis a little bit. We've alluded to operators. We've alluded to controllers. I'd like to examine them with a little more detail.
Dr. Yoon, you had mentioned that you had felt the controller was the most important element in the system, if I recall correctly.
DR. YOON: Yes.
MR. SAGER: Why did you say that?
DR. YOON: Well, first of all, operation point of view, who is responsible for overall operation, including emergency response? I think it is operator, and, secondly, leak detection system -- I'veheard many good stories about it, but there are so many, if I exaggerate a little bit, horror story, too.
Okay. So, with that kind of past record, if I were the pipeline operator, operating company, I would rely on leak detection system 100 percent. As a result, final say should be on pipeline operator.
MR. SAGER: What kinds of competencies would you associate with a high-performing controller?
DR. YOON: Okay. First of all, I understand DOT. I don't know the detail about DOT directive, but DOT ask each pipeline companies to train pipeline operator, not only in normal pipeline operations but also including procedures, company procedures, and also other abnormal aspect.
When there is a leak, what kind of things you expect from your control center or those kind of nature has to be train so that he knows something unusual things occur, then at least he can respond according to the company procedure.
MR. SAGER: If you were recommending to a company how to select controllers, what would you say that would be valuable?
DR. YOON: Okay. Okay. Selecting controller is difficult for me to say one way or the other, but on the other hand, training aspect can be addressed usingnot only training simulator that those two gentlemen already mentioned, and also a few other means, but at this time, I think a training simulator is one of the best means to address that.
But on the other hand, pipeline operators, if my -- the company that I visit sort of indication, then definitely they are not PTs or even master degree they have. So, okay, that doesn't mean that they are incapable, but on the other hand, it may take long time to train very sophisticated system.
So, as a result, what I said -- I wouldn't say insist, but on the other hand, I firmly believe that we got to walk before we run. So, at least we start with basic leak detection system, whatever the basic leak detection could be. That depends on each pipeline company's size and current system and what kind of upgrade ability, and, you know, to install leak detection, you will shut down the whole pipeline system. So, you've got to gradually build up and get experience with that.
MR. SAGER: Thank you.
For our other four panel members, I'd like to ask two questions. What do you believe should be the issues that we consider in interpreting what we have seen over and over as an inadequate response to alarmsand displays that are actually in front of our controllers? And the outcome of that is delay in responding to leaks and product loss and damage.
Mr. Stack?
MR. STACK: As I alluded to in my presentation, the key here in this situation is an operator is routinely dealing with operations of the pipeline, and these incidents are anomalies or special cases, and certainly training simulators can again help the operator practice to responses, recognize incidents which are not in his normal work day, and his ability to respond is clearly directly related to his confidence in the systems.
So, as vendors and operators, it is our duty to eliminate as best we can these false alarms, present consistent data to the operator, and train him to react to the anomalies as opposed to his normal day-to-day operations.
MR. SAGER: Mr. Rodecker?
MR. RODECKER: I would second those comments, but encapsulate them in one word "confidence".
The pipeline operator, the people that live along the pipeline, those that regulate those operations, and the vendors, such as us, have to have confidence that the alarms are meaningful.
Will there be a world without false alarms? I sincerely doubt it. Can those false alarms be minimized? I believe they can through a number of means.
But the confidence extends not just from the technical system itself and what lies behind the alarm and how to deal with that, but also with the confidence of the controller, the operator, and that confidence can be gained primarily through training and repetition.
Confidence will be instilled in the operator themselves. They'll trust their ability to recognize and react to dangerous conditions, and the management of the company and those that regulate them will have confidence that those operators know how to do the right thing, and back on your last question about what qualities a pipeline operator might have to be a good pipeline operator.
I can't really respond directly with that, but there are pipeline companies that do not allow operators to go on production, live on a control board until they've gone through a number of simulated exercises and have performed well. So, there are means to test for those qualities.
MR. SAGER: Dr. Hovey?
DR. HOVEY: Okay. I think that Mr. Rodecker really summed it up very, very well.
I think there have been some terrible accidents in recent time that are partly the result of, you know, controllers -- how do I want to put this?
I don't want to overlook, I think, the number of times that the control room operators have responded appropriately to unusual activities on the pipeline, which did not result in horrible accidents, like the ones that we've seen recently.
I think that these are somewhat unique, and in the amount of time that I've spent working in oil refineries prior to be coming part of a leak detection vending company and also working in control rooms during start-ups, I have seen operators with a great deal of competence and understanding of what goes on within their systems, and I think that good training drills and other activities that have been referenced are important, but I think that we don't want to lose the focus that many of these people out there are performing extremely well, and that the leak detection system does need to produce fewer nuisance alarms and other activities, so that there's a higher confidence there.
I guess I just -- I'm feeling like, youknow, the focus is so much on these incidences as opposed to the overview of the entire industry.
MR. SAGER: Well, we are the Safety Board and that tends to be our major interest.
Mr. McCoy?
MR. McCOY: Yeah. My comment is a little bit different. My observation of this -- I'm an electrical engineer. So, I look at this as a signal to noise problem, and essentially all of the comments that you're hearing in false alarm issues are the fact that we're trying with the SCADA technologies, whatever model or technique is used, to pull a very, very small signal out of a big noisy background, and you essentially set up a tension between the urgency to make that system as sensitive as possible, knowing that the more sensitive it is made, the more high the probability of false alarms.
So, you put the operators, the physical control room operator in a very precarious position. He doesn't want to call false alarms and shut down the line, but he's under incredible pressure to stop the line in the event of an accident.
Well, what do you do when you have a low signal to noise ratio? You can spend a lot of money. You can optimize the heck out of it, but bottom line iseventually, you get down to just where you can't find much more incremental value.
So, you look for a different signal, and that's my approach obviously. I don't look at what's going on inside the pipeline. I say when you run all you can out of that technology, find something else, and in this case, my case, it's physical presence of liquid where it shouldn't be.
It's got some limitations. It's obviously very costly, and, you know, it's not going to respond fast enough to do things that SCADA can do well, but on the other hand, it's a very clean signal. You know when something is there, and it shouldn't be there, and you can take appropriate action.
MR. SAGER: Thank you. That's all I have.
CHAIRMAN HALL: Let me interject on that again, Ms. Hovey, so you understand on this one item. One thing that is in the statute is that there is a requirement to report where there's a loss of 50 or more barrels of hazardous liquid or carbon dioxide, and while I respect the culture of companies, which is the most important thing in terms of safety, I have personally experienced, while being Chairman of this Board, many people that are in the operating room, it would appear, low-balling initial losses, and I'mconcerned about the training and the proficiency as well as the concern that each of those individuals have for our environment and their legal responsibilities as well.
So, this is -- as far as I'm concerned, is as serious -- is a serious, serious matter, and again I am concerned about the failure for any requirement in this area because while I don't want to put a bad light on those individuals who are attempting to do a good job each and every day, I also don't like to see people and companies that are trying to be proactive be put at a disadvantage in terms of operating their systems that have tremendous potential to cause -- and I guess we saw in the still-under-investigation that we saw in this Chalk Point, Maryland, accident/event, where we -- I won't get into that, but it's -- you know, it was again a situation that the Board's going to fully explore in terms of operations and reporting.
Proceed.
MR. BESHORE: Mr. Cash?
MR. CASH: Good morning. In everybody's presentation and virtually all of the answers, sensors has come up. Sensor reliability, sensor implementation, communication has come up.
In general, the software leak detectionpackages are highly dependent on sensor information from the pipeline, and your systems installed, how do people -- how are the sensors actually implemented? Is it a minimum? Is it people go above and beyond?
Mr. Stack?
MR. STACK: My comment with respect to that is during the initial engineering processes, there are certain standards and levels of instrumentation that are adopted, depending on the product and the environment that the pipeline is installed in.
But, generally, this determination is done at the engineering stages, and again many times with off-line models and simulators, they're able to verify the sensitivity of these applications in a model environment and are able to add additional devices on the line or relocate them to more effective locations for accuracy in this determination.
But, generally, it's determined by standards and practices during the engineering phases.
MR. RODECKER: I believe that's right. The selection of instrumentation is designed at the engineering stage for monitor and control of the pipeline, for moving product through the pipeline.
Unless leak detection is a part of the specification of a new installation, theinstrumentation likely would not be selected with leak detection in mind.
So, we find ourselves in a position of not retrofitting but installing leak detection on existing pipelines. We can look at the instrumentation through what we call leak sensitivity study to evaluate the performance of leak detection in advance of its actual implementation, and, further, we can look at the instruments themselves and do some cost-benefit analysis in terms of changing out existing instrumentation and making it more -- higher resolution in granularity to provide better information, and we can provide that again in advance of implementation.
MR. CASH: Again, on the instrumentation and sensors, I suppose if you put enough sensors on there, leak detection will be very easy. If you had them every several feet, it would be very easy to detect the leak.
Is there any new additions or new technology coming in sensors with new computers, new micritization of chips and sensors? Is there anything on the horizon that will make that easier to install on existing pipelines and new designs?
MR. RODECKER: I'm certainly not an expert in instrumentation technology, but it's not just thesensing of information from the pipeline that's important. It's also the communications of that information to the control center, and Mr. Stack referenced running a fiber optic cable along the pipeline, very, very fast communications of instrumentation.
But along -- I believe it's that same pipeline. My company did a study, back to your earlier comment about putting sensors at very frequent intervals. We did a study on -- I believe it was that pipeline, but a pipeline in California that ran crude oil, and the question that was posed to us is how frequent do we need to measure pressure in order to have good leak detection, and we found that it wasn't -- and this is in a published paper.
We found that it wasn't the frequency of one aspect of the data that was important, but the balance of all the data presented, and in this particular case, having pressure meters every 50 feet would not have made a difference because the fluid properties, that is, defining the behavior of the fluid in the pipeline, was inadequate relative to the sensitivity of the other instrumentation.
So, it's really a balance of all the information that's needed to drive the model.
MR. BESHORE: We have no further questions. Thank you, Mr. Chairman.
CHAIRMAN HALL: Very good. Well, we'll move up here to the non-technical staff. The first non-technical person is Member Carmody.
MS. CARMODY: Thank you. Good morning.
I just have one question. Much of what I was interested in was covered by Dr. Sager. Highly-intelligent questions, I must say, Dr. Sager.
I'm curious about false alarms. We've talked about them a bit, and the fact they're undesirable is obvious.
Do you have a false alarm rate that's acceptable or that's reasonable? Is there a percentage figure that you would accept as a reasonable false alarm rate? This is to anybody. Anybody want to respond?
DR. YOON: Yes, I will respond. When I approach operating companies, different company have different alarm procedure. So, one extreme is they don't want any alarm -- any false alarm at all.
Most of the companies realize the reality, and once or twice a month is acceptable. That's the majority of the companies, and one incidence, they have half dozen alarm, that doesn't bother them at allbecause they are used to it, and they know how to handle that.
MR. RODECKER: I'll comment a little differently. A false alarm is acceptable, so long as you act on the alarm. When you begin the process of disregarding alarms, that's when it becomes unacceptable, and that's different for every company.
I'll tell you about -- briefly about an incident that I'm aware of on one of our client companies, whereby there was an alarm, and a gentleman told me that I shut his pipe down last week, and it was a false alarm.
It was driven -- and I'll tell you why because I think it might give you insight into the sensitivity of the technology. This particular pipeline company uses drag-reducing agent injected into the pipeline to minimize friction, and when they do that, the hydraulic gradient, that is, the pressure decrease along the length of the pipeline changes, and the amount of DRA that they were injecting was different than that that was reported to the real-time model, significantly. I believe it was about a third as much as what they reported.
So, we had a difference. The alarm bells went off. They shut down the pipeline. Thehelicopters flew. He told me later it was a very good exercise that he could not have planned in advance, but on the other side of that coin, he couldn't do that every week either. So, I would say it's a case-by-case.
MS. CARMODY: Reason I was asking is it seems that a number of the accidents I've looked at, when the operator got an indication something was wrong, he didn't believe it, and he took action that was the converse of what he should have done. So, I was just trying to get a sense of what would be reasonable.
But I think, also, when you were answering Dr. Sager, you said that confidence in the system was important as well as company philosophy. I would think the company philosophy of how one responds is probably the key factor.
Is that -- I see nods. I'm not trying to put words in your mouth. It seems to me that would be the critical element, though, how you respond when you get these, what it is you do, and that is, as I understand it, that's a determination made by individual companies, not necessarily as a result of what you might recommend for your system.
MR. STACK: Yes, I agree, and the point that Alan, I think, was alluding to is clear. That companypolicy and procedure of how you deal with false alarms validate or invalidate those alarms as the process and procedure can be clearly beneficial, and some of the best operating companies that I'm aware of are dealing with two to five alarms a week. Many turn out to be false, but they have good process in place to further investigate and take some action accordingly.
MR. McCOY: If I could just add one thought to that. One of the ways that you deal with false alarms is a second source of information, and I think when we look at the records of many of the incidents where the operator was asked to make a judgment call, is this a false alarm, should I override the shut-down, should I start pumps, it's working with one source of information, whatever that SCADA control screen is telling him, and I think specifically of the Bellingham incident.
When there was a period of time when it was indecision in the control room, and the system was restarted, and that amount of time in this case, I think, was in the range of an hour, hour and 15 minutes. It would have been very valuable to that control room to have a totally independent second source of information, and I've thought about my particular product, and, you know, it's really heart-wrenching because that was a situation where if a second alarm bell had gone off and said, hey, not only did your pump shut down, but here's an independent indication that says looks like there's hydrocarbon in the soil at some location.
Chance of them restarting the pump were nil, but they only had one source of information. So, it's a -- if you could afford it, it would be very nice to have two separate independent systems. Unfortunately, it's very expensive and very hard to retrofit, and it's something to think about in the future, but certainly would be nice if we could afford it some day.
MS. CARMODY: That's a good point.
MR. RODECKER: May I clarify one thing? The false alarm in our context needs a little clarification.
The example I mentioned, it was not a false alarm. We did identify a hydraulic anomaly, and that's what the technology does. So, the technology performed just as it should. Did it alarm a leak? No. It alarmed a difference between what was reported to it and what it calculated.
MS. CARMODY: Yeah. Thank you. That's all I have.
CHAIRMAN HALL: All right. Member Black?
MR. BLACK: Thank you, sir.
I guess maybe Mr. Rodecker might be the best, but someone else is more -- is certainly welcome to answer.
The systems that are in existence, I'm sure they're of varying ages, but the large systems, the over-500-mile systems, how old are these SCADA systems now? How long have they been in existence? I see them turning up in accidents as far back as several years ago.
MR. RODECKER: I think I'd like to defer that to Mr. Stack, who represents part of that industry. So, he probably has a better handle.
MR. BLACK: Okay. Whoever.
MR. STACK: Member Black, I would suggest that traditionally the life cycle of this SCADA technology, the data acquisition technology, has been anywhere from seven to 15 years before they turn over.
More recently, with the falling costs of the hardware technology and so on, we're seeing turnover times in these technologies of five years or so.
There are, of course, operating companies there who are operating 10-year old systems, certainly that the -- but I would generally say that the roll-over now is in the five-year range.
MR. BLACK: And this is primarily hardware-based?
MR. STACK: It's primarily due to the cost of the physical technology, the hardware technology particularly has dropped so significantly.
MR. BLACK: Okay. With regard to -- again, this is to anyone who knows the answer. With regard to technology for remotely sensing, I used to have a little experience in that area back when we were using leased telephone lines, and that was not a very reliable system.
What are they using now for remote telemetry?
MR. STACK: You see the entire gamut of still telephone lines in use, UHF-VHF radio, microwave. Fiber optic, of course, is more recent technology that is going into new installations and is much easier to install and cost effective. Even satellite technology.
Of course, some of the concern we have as vendors in this market is deterministic data -- nature of the data, and traditionally satellite systems have not been so effective in helping us do our job with their non-deterministic characteristics.
But you see the whole gamut of installations, including cellular phone as well.
MR. BLACK: That's what I was about to say. There's one railroad that has been experimenting with remote-sensing of the operation of primarily railroad grade crossing signals by this carrier wave -- carrier frequency wave, cell phone technology, and they've been quite successful with it.
I was just wondering if that was being considered.
MR. STACK: It is being considered. Most I've seen have been pilot-type projects at this time. I would say it's not a mainstream technology.
MR. BLACK: This was a pilot project, also, but it was in a large geographical area, and it seemed to work very well based on what they saw.
Okay. Are these systems intended to be operational tools, in addition to just leak detectors? In other words, will they give an operator advice? I was reading through a report here where there was a complicated situation developing pump and pressure wise in a long system, and would this -- and they had a system -- I'm not going to mention the names.
They had a system. They didn't seem to be using the SCADA system to deal with the problem, and then when it detected or it was beginning to detect a leak, they ignored it.
In other words, is this something that sitsthere and waits on a leak, or is this something they would utilize as a system optimization tool? The existing SCADA systems?
MR. RODECKER: I'll go ahead and field that at least first.
Our clients use the real-time model that I described for many purposes. Some of our clients use it only for leak detection. It provides estimates of what's going on in the pipeline network, and they're very good estimates. In some cases, they're better than the data themselves.
So, oftentimes in the control room, where real-time models are -- exist, if you look at the screens that the operators use to monitor and control their pipeline, oftentimes you'll see both SCADA data, that is, data reported from the pipeline, and also underneath it or in a different color, you'll find model data.
That allows you to continue to operate based on those estimates. If there's a communications failure or if there's information that would be very costly to sense out of the pipeline -- in the gas industry, it's very expensive to put in mainline flow meters. So, many of our gas clients and clients from others in the industry use real-time models to get verygood estimates of the flow rates within their pipeline.
So, yes, they're used for many things, more so, I believe, in the gas industry than in the liquid industry.
MR. BLACK: So, the liquid industry, there would be primarily a leak detection or a system -- I guess you also get performance and volume data and that sort of thing out of it, which would be useful in billing or whatever. Okay.
MR. STACK: Yeah. Additionally, in the liquids pipelines, you see them used consistently for volumetric tracking, product tracking, if you're a batching pipeline, and so on as well.
MR. BLACK: Thank you, Mr. Chairman.
CHAIRMAN HALL: Member Hammerschmidt?
MR. HAMMERSCHMIDT: Thank you.
Well, for the record, Carol Carmody and Eric Sager have asked most of my questions jointly, but I was interested -- this is not really a question.
I was interested in Mr. McCoy's presentation when he said that a large rate of leakage, according to this slide, he said, "SCADA will find it quickly and will shut down the line in a minute or less", and it occurred to me that in the world we operate in as indicated by Mr. Sager, we don't always see thatoccurring in terms of the effectiveness of SCADA systems and their dependency on the non-automated systems, their dependency on the human factor to assimilate what is happening and then taking proper action.
MR. McCOY: I concur with your assessment. I was using a curve that was published in -- I'll allude back to one of the comments that was made very early on about maybe some vendor comments on the sales presentations are a little bit more aggressive than what actually happens in reality.
In fact, there was another set of data that I did have access to that I didn't put on there, which was much -- if you think about that curve that I showed and imagine it shifting up and to the right, that's the direction that reality pushes it.
One minute is probably the signal is recognized. The system, if it's in some sort of an automatic mode, probably can close valves and stop pumps, but obvious operational experience has that that doesn't happen too often.
The initial steps may be taken, initial indications may be recognized in that one minute. The line begins to depressurize. There's a whole other issue, and that is, if you have a major rupt