Print this article

Comment

01 June 2003

Oil out of the Amazon Rain Forest...uh oh!

By Carlos Henrique Wildhagen Moura, Judas Tadeu Marques Cavalcante, and Renan Martins Baptista

Leaks in environmental flash point from multiphase-flow pipelines scare all.

Leak detection on pipelines has always been an important issue.

Lately the increasing concerns for environmental preservation have made it even more important. Add to that the Amazon Rain Forest's extreme importance to the world's environmental health.

Now add these two topics to the existence of crude-oil production fields in that forest, and the importance of avoiding oil spillage in such a sensitive area is easy to conclude.

To make things worse, production pipelines do not always behave in a predictable way, presenting a multiphase flow most of the time.

Multiphase flow is one of the most difficult situations for leak detection in pipelines for several reasons: the existence of two or more different and independent phases; variation of each phase's volume along the pipeline; different fluid velocities for each phase; and sometimes a non-Newtonian associated behavior, due to the formation of an oil-water emulsion.

There are two fundamental groupings of leak detection techniques. One group is the models (or computational pipeline monitoring, as stated in API 1130) that monitor the flow in real time (real-time measurement, real-time transient modeling, pressure point analysis, and others) from inside the pipeline. The instrument sensor is actually in physical contact with the fluid.

The second way is to model the flow using a state estimator. Engineers base these systems on dedicated external sensors (such as thermal and mass dispersion) placed on and along the pipeline.

The vast majority of the technologies in the first group rely on volumetric flow rate measurements, which are notoriously inaccurate and ineffective for multiphase flow.

It is also important to mention that the flow pattern in some multiphase pipelines changes randomly and intensively, from tiny bubbles to a severe slug-flow pattern.

This brings unpredictability to those lines, if compared to a regular single-phase line. Consequently, systems based on prediction models tend to be unreliable, inaccurate, and not sensitive.

The acoustic system is an exception to the two groups of technologies previously mentioned. It has, on one hand, a sensor that really touches the fluid (which would suggest it belongs within the first group). On the other hand, there's no flow model behind it, but an acoustic sign-analysis algorithm, acting somewhat like a piece of hardware.

Operate in synchronous time

Acoustic leak detection is based upon the detection and processing of a very low-frequency part of the pressure signals generated by a leak. Any pipeline has its inherent pressure noise pattern, which will be eliminated by filtering techniques within local (or site) stations along the pipeline. It is up to these units to condition the signal for the processing of a master station.

A station is a piece of hardware composed of a processor unit (the main board with TI processor), a global positioning system (GPS) device, and a firmware electronically erasable programmable read-only memory (EEPROM).

The station can communicate with the other stations using common communication media such as radio enlaces and can even link with supervisory control and data acquisition (SCADA), and let SCADA manage the communication task.

As far as localization goes, acoustic technology determines locations of leaks based upon the sound velocity in the process fluid. Thus, due to a random distribution of the two phases within the pipeline, an uncertainty associated with the leak location determination occurs.

This uncertainty is significant if compared to single-phase flow pipelines, where the acoustic velocity is relatively constant. To minimize this loss of accuracy, on-line leak tests are important tools to verify the actual acoustic-wave velocity, and thus mitigate this drawback.

Again, once the leak location is based upon monitoring the time the site stations take to sense it, all nodes, which are parts of the leak detection system, will operate under synchronous time, which is accomplished by external GPS devices.

The system embraces three primary components. The master station runs all the leak location and detection functionality using hardware similar to that used for the site stations (remote units), but with more complex firmware and applications.

Whenever a site station detects an event, the master station waits for a second site station to raise a flag for the same event. If that happens, the location calculation then takes place. Otherwise, the system issues a warning of a nonconfirmed leakage at the human-machine interface (HMI).

The site stations run the time-stamping and signal-filtering functions. As far as filtering goes, different filters operate to wipe out all inherent installation noise and send a clean signal to the master station.

There are special pressure transmitters dedicated to sensing pressure wave signals. They are in contact with the fluid transported by the pipeline and provide 4–20 mA output signals to the site stations.

Algorithms to the rescue

The local stations employ these techniques for reducing false-alarm rates:

• dual-element acoustic sensors at the boundaries of the monitored pipeline segment
• matched filters, which use the signature of a leak from a database as a mask against which one can compare real-time pressure data. This database rises from a comprehensive set of field data, including field leak data as well as experimental leak test data for various pipeline operating conditions, several fluids, and different degrees of dissipation and dispersion. The technique uses a fingerprint-matching active identification on the true leak signal, which attempts to reduce the false-alarm rate, seeking to increase the location accuracy and sensitivity.

Among the filtering techniques, there are:

• digital high-pass and low-pass filters with auto-adjust roll-off frequencies
• moving-average filters with tuned dynamic-adjusted windows to filter out noise
• dynamic threshold logic, using an auto-adjust algorithm to distinguish random noise and other events from true leak signals (It continuously scans, computes, and verifies all incoming data and automatically adjusts the dynamic thresholds.)
• repetitive filters (several types to suppress various unusual noise sources)

The false-alarm rate is a function of frequency of activity (sudden pressure changes on the pipeline associated with operating changes). The use of the previously cited filters is an attempt to reduce false alarms significantly.

Scan stations and calculate

The required leak detection time is the sum of two parcels. First, there is the time required for an acoustic wave to travel to a monitored site, which is calculated by dividing the distance between the leak and the adjacent upstream/downstream monitor sites by the acoustic velocity in the fluid pipeline medium.

The second parcel is the time required for the master station to scan all local stations and calculate the leak location.

When a leak occurs, the GPS times are recorded and archived. The master station uses acoustic parameters that describe the pipeline operating conditions. It also uses the distance between site stations along with GPS arrival time to confirm the occurrence and determine the location of a supposed leak.

The accuracy of a leak location is affected by two principal types of errors: timing errors associated with communications and synchronization, and changes in sound velocity due to changes in temperature, pressure, and other operating parameters.

In the case of loss of communication or other communication faults, the local sensor will still detect the leak signal, and the corresponding local station will register it with a GPS time stamp.

This information will help compute leak location when communication resumes. Tuning acoustic velocity parameters during commissioning will attempt to reduce location errors caused by changes in sound velocity.

This system doesn't quantify the leakage.

Terminology
API	American Petroleum Institute
CPM	computational pipeline monitoring (lead detection calculation)
EEPROM	electronically erasable programmable read-only memory—firmware stores in this microchip. EEPROMs are nonvolatile, which means they do not lose their contents when a computer loses power.
GPS	global positioning system
HMI	human-machine interface
LDS	leak detection system
multiphase flow	any of a complex variety of irregular flow mixed gas, liquid, and solid flow patterns through a pipe
Newtonian fluids	fluids that obey Newton's law of fluid viscosity. Fluids that don't obey are those such as pastes, slurries, polymers, and emulsions.
PPA	pressure point analysis (lead detection calculation)
RTM	real-time measurement (lead detection calculation)
RTTM	real-time transient modeling— simulation of flow within pipelines (lead detection)
SCADA	supervisory control and data acquisition

A complex system disturbance

The pipeline chosen for this pilot test is located in the state of Amazonas in Brazil.

Its service is to carry oil production from an oil and gas production field named Leste do Rio Urucu to a gathering and treating station named Polo Arara Station.

From beginning to end, the pipeline is almost 36,000 meters across the forest. This 14-inch pipeline collects production from 40 wells scattered along the way. These wells present average gas-to-oil ratios of 800 and produce light crude.

With this significant amount of gas together with such light oil, it is necessary to install six inhabited separation stations along the pipeline to remove the incoming gas. These inlets are also additional sources of disturbance to the pipeline flow behavior.

Due to the high pressures in the separators, there is still a great amount of dissolved gas in the oil leaving the separators. This gas gradually releases along the pipeline on its way to the station, forming a multiphase flow.

Still, pressure along this pipeline substantially decreases along its route; flow arrives at the end gathering station with one-fifth the pressure it had at the beginning. The combination favors the phase change from liquid to gas as pressure and temperature experience changes along the pipeline length.

The complexity of this system was one of the prime reasons for choosing it to pilot test the new technology.

Leak detection system description

Detectors were installed in the pipeline after a detailed study, which tailored the acoustic technology to this particular application.

Defining the adequate quantity and location of sensors, considering the existence of secondary branches, process vessels, and equipment such as pumps and valves, as well as compensating for the effect of extraordinarily long sections of pipe were all addressed.

This study led to a final number of 13 sensors monitored by five site stations: two working with four sensors each, two more with two sensors each, and one station with only one sensor.

These field-located site stations report to a master station installed in the Polo Arara Central Control Room. The adopted HMI was a PC, running the Windows NT-based Intellution Dynamics' iFIX supervisory software, which provides system information and operational and leakage detection alarms to the station operators.

To implement the data transmission among site stations and the master station, we evaluated several options ranging from optic fibers to radio links. Considering the involved distances, installation time, cost, and previous local experience, the natural choice was the use of radio link for this service.

Half-duplex radio modems were our choice. They modulate in frequency and transmit in the band from 500 to 542 megahertz, with output transmitting power of up to 5 watts and a communication rate of 9600 bits per second. This system had already proven itself in this same environment where attenuation by the forest is a major consideration.

To eliminate the application's inherent noise, it was necessary to develop specific filters. This happened by setting acoustic detection thresholds higher than the noise level detected in each of the site sensors' installations. These filters were based on acoustic records of the existing noise and then downloaded to the site stations' EEPROMs.

All sensors were implemented and tested by producing an actual leak, by opening valves previously installed along the length of the pipeline for this purpose, and by bleeding the content of the pipeline to accumulation tanks.

Downstream from these valves, workers installed calibrated orifice plates and pairs of rupture disks. The orifice plates made it possible to test a variety of hole sizes. The rupture disks simulated the typical noise generated by the abrupt rupture of a pipeline wall.

These disks also provided a means of controlling the start of the leak. By depressurizing the chamber between the disks, they would burst.

During the sequence of tests the plates' orifice sizes varied such that we could ascertain the sensitivity limit of the system.

This sort of system has limitations in detecting leaks that result from the gradual erosion of pipe walls. This type of leak does not involve a fast and sudden rupture.

Gradual leak detection is a limitation because a gradual leak does not generate an acoustic transient big enough to surpass the preset system-response threshold. This technology cannot detect that kind of fracture.

However, considering all possible leak occurrences in a pipeline and the existence of other methods to detect the gradual degradation of wall thickness, this method still presents advantages, such as its fast response and ability to accurately locate leaks.

Generate no false alarms

Because it was the first time this technology was tested in a multiphase scenario, and to guarantee the tests would be conclusive regarding the method's effectiveness, reliability, and stability, we conducted two rounds of tests.

By spreading the testing over three months, we could also more certainly evaluate the system's immunity to false alarms. After the first test round, the system remained in operation to evaluate its stability to spurious alarms.

Since then, the system generated no false alarms. This result proved its stability. After the testing, analysis of the obtained data showed the system had the capability of detecting and locating holes ranging from 0.2 to 0.5 inches.

The average error on the leak location was +/– 200 meters, which when considering that the length of the pipeline is 35,900 meters, was within the range of expected and acceptable results.

Regarding time response, the average time delay between a leak simulation and its confirmation by the system was less than fifteen seconds. ST

Behind the byline

Carlos Henrique Wildhagen Moura is an electronic engineer. He designs oil and gas onshore and offshore production plants for Petrobras in Brazil. Judas Tadeu Marques Cavalcante is an electronic engineer. He works on automation projects and project coordination, also at Petrobras. Renan Martins Baptista is a chemical engineer, and he works on pipeline flow assurance, simulation, and leak detection for both single and multiphase flow at Petrobras.