Key takeaways
- Internal (CPM) methods use the flow, pressure and temperature data a pipeline already produces. External methods sense the pipeline’s surroundings directly.
- Each internal method has a blind spot, which is why mass balance, RTTM, statistical analysis and pressure-wave detection are usually combined.
- Instrument quality and communications set the ceiling on sensitivity, whatever the software does.
- On high-consequence lines, pairing internal and external detection adds redundancy and faster confirmation.
What a pipeline leak detection system has to do
A leak detection system has four jobs. It has to notice a loss of containment quickly, tell the operator where along the line it happened, stay quiet when nothing is wrong, and keep working through the normal upsets of pipeline life: start-ups, pump trips, valve moves and batch changes. A system that alarms too often gets ignored, and one that alarms too late is no use. Most design decisions come back to balancing those two failures.
For liquid pipelines, API RP 1130 sets out how computational pipeline monitoring (CPM) systems are designed, tested and operated, and API RP 1175 covers managing a leak detection programme as a whole. In the United States, 49 CFR 195.444 sets leak detection requirements for hazardous liquid pipelines. Other regions apply their own rules. Our Karbala Refinery project, described below, was delivered against API 1130 and TRFL guidelines. Our article on API RP 1130, API RP 1175 and 49 CFR 195.444 explains how the three documents fit together.
Internal leak detection: computational pipeline monitoring
Internal methods run on data the pipeline already produces: flow, pressure, temperature and density at the inlet, the outlet and intermediate stations. Software compares what the data says should be happening with what is happening. There is no new hardware along the right of way, but the result is only as good as the instruments and communications underneath it.
Mass and volume balance
The simplest idea is that what goes in must come out. Compensated mass or volume balance compares inlet and outlet flow, corrected for line pack, which is the change in the amount of fluid stored in the pipe as pressure and temperature move. It is robust and easy to explain, and it is good at confirming a sustained loss. Its weakness is speed on small leaks: separating a genuine imbalance from line-pack noise takes averaging over time, and a balance on its own says little about where the fluid is escaping.
Real-time transient modelling (RTTM)
An RTTM runs a hydraulic model of the pipeline in real time, fed with the measured boundary conditions and the properties of the fluid. If measured pressures and flows drift away from what the model predicts, something has changed. Extended RTTM (E-RTTM) adds statistical processing of those residuals. Modelling copes well with transients and multi-product lines, and it can estimate leak location, but it needs accurate pipeline data and careful calibration for each line.
Statistical analysis
Statistical methods look at the pattern of deviations over time rather than a single threshold. By learning what normal noise looks like, they can judge whether a deviation is consistent with a leak, which cuts false alarms. They need a learning period and clean data, and they work best alongside a balance or model-based method.
Negative pressure wave detection
A sudden leak sends a pressure wave in both directions along the pipe at close to the speed of sound in the fluid. Sensors at known points record when the wave arrives, and the difference in arrival times gives the leak position. It is fast, and good at locating abrupt leaks and ruptures. It is far less sensitive to slow, gradual leaks, and it depends on accurate time synchronisation.
| Method | Strong at | Watch out for |
|---|---|---|
| Mass / volume balance | Confirming sustained losses; easy to understand | Slow on small leaks; limited location information |
| RTTM / E-RTTM | Transients, multi-product lines, leak localisation | Needs accurate pipeline data and per-line calibration |
| Statistical analysis | Reducing false alarms in noisy data | Needs a learning period and good data quality |
| Negative pressure wave | Fast detection and location of sudden leaks | Poor on slow leaks; needs precise time synchronisation |
External leak detection: sensing along the route
External methods put sensors where a leak would show itself. The best known option is distributed fibre optic sensing. Distributed temperature sensing (DTS) picks up the local temperature change a leak causes, and distributed acoustic sensing (DAS) listens for the sound of fluid escaping and for digging near the line. Hydrocarbon-sensing cables and point sensors are used in some designs.
Because they measure the leak itself, external methods keep working when the line is shut in and static, when flow-based methods have little to compare. They also give a precise location. The trade-off is that the sensor has to be installed along the route, so coverage depends on where the cable is. For a closer look, see our guide to fibre optic sensing for pipelines.
Why dual-layer detection is the sensible default
Internal and external methods fail in different ways, which is the point of using both. Internal detection covers the whole line from instruments you already need for operations. External sensing gives independent confirmation and a precise location, and it keeps watching when the pipeline is static or when instrumentation is degraded. When both layers alarm on the same event, the operator can act with more confidence, and a single failed layer does not leave the line blind.
This matters most on high-consequence sections, such as crossings near water, settlements or sensitive land. That is where an external layer is most often worth the added installation effort. For a side-by-side comparison, see RTTM vs fibre optic leak detection.
In practice: leak detection for eight pipelines at Karbala Refinery
At Karbala Refinery in Iraq, we designed, integrated and commissioned a turnkey computational pipeline monitoring and leak detection system, both internal and external, for eight pipelines carrying refined products and LPG. The scope included leak detection hardware and software, telecommunications, RTUs, field instrumentation, SCADA integration and operator training. Four decisions shaped the result:
- One model per pipeline. Each line carries different products with different fluid properties, so each received its own real-time transient model, calibrated against its product and operating profile.
- Instruments built for the site. Ambient temperatures above 50°C and blowing sand called for ruggedised RTUs and field instruments in climate-controlled, surge-protected, sand-resistant enclosures.
- Communications as part of leak detection. With no existing backbone along the routes, we installed UHF/VHF radio links and microwave backhaul so every measurement point reached the control room with low latency.
- Operators in the design. A unified SCADA platform brought alarm management, hydraulic profile visualisation, automatic leak localisation and decision-support dashboards into one interface.
The result was continuous monitoring across all eight pipelines, meeting API 1130 and TRFL guidelines, with leak response reduced from hours to minutes. You can read the full case study on our projects page.
How to choose a leak detection approach
There is no universal answer, but a handful of questions narrow the options quickly:
- What is in the line? A single product or batches, liquid, gas or LPG. Fluid properties drive how complex the model has to be.
- How steady is operation? Frequent transients favour model-based methods over a simple balance.
- What instrumentation exists, and how accurate is it? Sensitivity cannot beat the quality of the measurements, so upgrading a few key meters can matter more than new software.
- What communications are available? Slow or unreliable links limit what the software can see and how fast it can alarm.
- What do the regulations and the risk assessment require? Standards and local rules often set minimum performance, and consequence analysis shows where a second layer is justified.
- Who responds, and how? Define alarm handling and how operators will use leak localisation before commissioning, then test with real procedures.
If you are weighing these options for a specific line, our team can help from feasibility through commissioning and support. See our leak detection systems and pipeline management systems, or get in touch.
Frequently asked questions
What is computational pipeline monitoring (CPM)?
Computational pipeline monitoring is a family of software methods that detect leaks by analysing pipeline data such as flow, pressure and temperature, rather than by sensing the leak directly. Mass balance, real-time transient modelling, statistical analysis and pressure-wave detection are all CPM methods.
What does API RP 1130 cover?
API RP 1130 is the recommended practice for computational pipeline monitoring on liquid pipelines. It addresses how CPM systems are designed, implemented, tested and operated. API RP 1175 is its companion for managing a pipeline leak detection programme.
How small a leak can a system detect?
It depends on the method, the quality of the instruments, the fluid and how long the system waits before it raises an alarm. Internal methods trade sensitivity against false alarms, and external sensing can respond to small local changes along its route. Detection performance should be tested on the actual pipeline rather than assumed from a datasheet.
Is internal or external leak detection better?
Neither is better in every case. Internal methods cover the entire line using existing instruments, while external methods give direct detection and a precise location wherever sensors are installed. Pairing them gives redundancy, which is why we usually recommend both for high-consequence pipelines.
