Company & Industry News
Finding the leak
Ken McCoy, TraceTek Group, Tyco Thermal Controls LLC and Julian Waumsley, Aquilar Ltd, explain how TraceTek technology can improve the speed and accuracy of fuel and oil leak detection.
[Click on images for larger versions]
When pipeline operators speak of leak detection are most likely thinking of Supervisory Control Data Acquisition (SCADA) based systems. The most popular form of leak detection for transmission pipelines is based on analysis of pressure and flow data derived from the SCADA system. At its core, leak detection based on the analysis of SCADA data is a simple concept. If the pipeline is operating normally then the measurements should look normal. Similarly, an unexplained imbalance would indicate that there is a possible leak.
SCADA-based leak detection software is an extension of the basic operator responsibility of watching for unusual conditions. Leak detection software attempts to automate the rather boring task of monitoring the read-outs and indicators while simultaneously searching for indications of a problem. While an automated monitoring process seems like a straight forward computer task, there are a host of complexities that make a practical, real world system less than perfect. One problem is sensitivity: if 20% of the flow through a pipeline suddenly goes missing, even a bored operator in the middle of a long night shift would quickly suspect a leak. Likewise, a simple computational algorithm could 'detect' a leak. But what if the loss of flow is a gradual loss of 1% or 2% over several hours? What if the output of a pressure transducer fluctuates for a few seconds only to stabilise again? Was it instrument drift or noise? Was it a momentary telemetry dropout or perhaps a minor operating mode change upstream or downstream that rippled a transient up and down the line?
Figure 1
|
Defining the problem
Any operating pipeline is a dynamic system. Start-ups, shut downs, rate of flow changes, line pack variations, valve changes, number of pumps on line and their output are just a few of the major parameters that give rise to subtle or significant changes in the signals sent from pressure and flow sensors along the pipeline. The challenge to the SCADA leak detection software is to distinguish what is 'normal' fluctuation and what is an unexplainable or 'abnormal' behaviour. To go one step further: is an abnormal measurement caused by a real leak or some other abnormality in the pressure and flow data? Generally speaking, the better the data, the better your chances are of detecting a leak. Therefore, any serious attempt at SCADA-based leak detection usually involves the addition of more, and better quality sensors, compared to what is required for basic pipeline operation. More sensor points require more and faster telemetry. More accurate sensor data requires better calibration and more data requires higher performance computational hardware and software. In the end, there are diminishing returns. Even with the best instrumentation and analysis, there is still a point where the leak is so small compared to the volume flowing through the pipe that it becomes undetectable. The characteristic 'response curve' of all SCADA-based leak detection systems is like that shown in Figure 1.
|
The SCADA response curve quantifies that which is fairly obvious. A fast leak (that is a high percentage of the flow rate) is easy for a computer (or an operator) to detect. As the leak gets to be a smaller percentage of the total flow it becomes increasingly difficult for the system to detect. Eventually, a very small leak is undetectable. More sensors and sensors of a better quality, more computational horse power, and clever algorithms can push the curve down and to the left a bit, but the fundamental shape does not change. Even the most expensive SCADA-based leak detection system reaches a point at which the leak rate is too small to detect. Slow leaks that are below the SCADA detection threshold are what the pipeline industry euphemistically refers to as 'weeps and seeps'. They are a reality in many operating pipelines and a problem that the industry has lived with for many years. Dead vegetation or surface spills spotted during fly-over or drive-by line patrols, reports of water well contamination, or polluted stream beds and other phoned in reports from neighbours, are the consequence of undetected weeps and seeps.
Detecting smaller leaks
During the environmental hearings that were required to obtain construction and operating licenses, the major pipeline project in Texas came up against the limits of the SCADA curve. As a result of well-organised community opposition to the refurbishment and re-licensing of the liquid fuel product pipeline project, the minimum acceptable performance for the required leak detection system was set below the SCADA curve in the lower left corner of the graph. Quite simply, the best SCADA system that money could buy would not meet the performance target that the State of Texas had established. Weeps and seeps could not be tolerated due to the pipeline's close proximity to the Edwards Aquifer, the primary water supply for Austin. Texas. Another approach was required, and the alternative that was selected was the high sensitivity external fuel leak detection cable developed by Tyco Thermal Controls in Redwood City, California. The TraceTek@5000 sensor cable was selected for the project because it uses a technology that is completely independent of the leak rate. The TraceTek 5000 cable detects a leak when a small quantity of liquid hydrocarbon fuel accumulates in the pipeline trench. It may take hours, or days, for a small weep or seep to contaminate the first cubic meter of trench backfill, but the sensor cable responds directly to the presence of the fuel and not to the rate at which it arrived. As a bonus, the leak detection cable can report the spill location to an accuracy of a few meters.
Response time
So which is more important, the fast response of a SCADA based leak detection to a big fast flowing leak, or the direct detection of a slow leak when it totals a cubic meter of contaminated backfill? The answer depends entirely on the frame of reference. Fast response to fast leaks is vital and there are obvious safety concerns. No one would contest that a quick shut-down after a major rupture is critical, but if protection of water resources or prevention of environ- mental damage is important, then detecting a spill while it is still small and easy to remediate is also important. No one wants a big spill and the mistake that many make is to equate a fast leak with a big spill and a slow leak with a small spill. The truth is that a slow leak which goes undetected for weeks or months can become just as big a clean-up problem as the fast leak that was spotted and dealt with promptly.
Cable leak detection The TraceTek@5000 sensor cable system was installed on 12 km of the Texas pipeline where it crosses over, or through, the most environmentally and politically sensitive areas along the right-of-way. It does not compete with the SCADA system; rather it complements the fast leak detection capability of SCADA with the sensitivity to slow leaks that are an equally important part of the leak detection spectrum. In the Texas installation, the TraceTek cable leak detection system was powered and monitored by a series of ten independent sensor interface modules. These modules were connected to the existing landline telemetry sys- tem at a valve station located approximately at the mid point of the section of pipe that required sensor cable and only one central power point was required. In other installations, alternative power and telemetry options have been used but the basic organisation remained the same. Approximately 1 km of sensor cable is assigned to each interface module and each section of sensor cable operates completely independent of other sections of the cable system. Is external leak detection cable appropriate for all liquid fuel pipelines? The answer is probably 'not everywhere', but it is very appropriate in many environmentally sensitive situations or at times or in places where SCADA leak detection is not effective. SCADA performs best in constantly flowing pipelines. During start-ups, shutdowns, or idle periods, the SCADA analysis algorithms are far less sensitive. Most SCADA software offers some form of transient suppression to avoid too many false alarms during periods of unstable operation. Alarms that do occur during these transition phases are sometimes discounted or ignored by the operating staff. Thus, cable leak detection can offer a welcome backup, especially if the pipeline is crossing a sensitive area. Similar 'good' applications for cable leak detection are spur lines that may not be flowing when the main line is in operation. SCADA is virtually blind when the fluid in the line is not flowing. Spur lines to terminals, or lines that operate in a batch mode, are strong candidates for external sensor cable. An excellent example of this scenario occurred in April, 2000. According to the U.S. National Transportation Safety Board report on the incident, approximately 140 400 gallons of fuel oil were released into the surrounding wetlands and Swanson Creek and, subsequently, the Patuxent River, as a result of the accident. The leak occurred during a pigging operation with several operators on each end of the pipeline paying careful attention to the operation. How was this leak detected? Only when operator’s compared tank level readings during a series of phone calls nearly seven hours after the pigging operation began, was it discovered that a leak had occurred. Basic calculator arithmetic told them that they were 'missing' a large amount of oil. The pipeline leak may have been slowly leaking and gone unnoticed for days, or even weeks, before the major rupture occurred. Even with attentive operators, a major spill occurred without any automated indication of a leak.
Conclusion
SCADA is the leak detection tool of choice for most major pipeline operators. Over the past several decades, the technology has grown in sophistication and complexity from a simple valve and pump operating control system to a highly effective monitoring system. However, the real world conditions of mixed fuels, changing flow conditions, instrument drift and noise and telemetry all contribute to the basic characteristic of SCADA response. The smaller the leak as a percentage of the flow, the less likely it will be detected. Too many people make the judgment error that a slow leak is not a big spill. Time is the enemy; even a slow leak can become a big spill and an expensive clean up. SCADA-based leak detection reaches its limits for slow leaks, spur lines or batch flow operations. External cable leak detection like that provided by Tyco Thermal Controls' TraceTek 5000 fills the gap. The systems are not competitors; they are well-matched allies in the quest to operate pipelines in a safe and environmentally friendly manner.
Click here to view the original news article [2.1MB] (First published in the World Pipelines Journal - August 05)
August 2005
