How to Evaluate a CPM Leak Detection System per API 1130
API RP 1130 characterizes leak-detection performance along four attributes rather than a single accuracy number, and engineers who only quote one of them get caught out in audits. This page walks each attribute in the order you would test it, so an integrity engineer can put structure around a vendor's claims and a controller can understand what the system will and will not catch. It describes the evaluation logic, not a substitute for your operator's formal test plan.
CPM Evaluation Attributes (RP 1130) in one line: To evaluate a CPM leak detection system per API 1130, characterize it against four attributes: sensitivity (how small and how fast a leak it can flag), reliability (how often it alarms correctly versus falsely), accuracy (how well it estimates leak size and location), and robustness (whether it keeps working during transients, instrument dropouts, and off-design conditions). No single number captures a leak system - trade one attribute up and another usually moves down.
Establish the Baseline and Test Conditions
Start by fixing what you are testing against. Document the pipeline hydraulics, the instrument suite feeding the model, and the operating states the line actually runs in - steady flow, ramping, shut-in, slack line, and product interfaces. A leak system tuned for steady flow may fall apart during a transient, so the test conditions must span the real duty, not just the easy case.
Decide up front how a leak will be simulated - drawoff tests, withdrawal at a valve, or a modeled leak injected into the data. The method matters because it bounds what the results mean. Capture the instrument accuracy for pressure, flow, temperature, and density going in, because the leak system can never be better than the flow computer and transmitter data it consumes.
Characterize Sensitivity
Sensitivity is the smallest leak the system detects and how quickly. It is properly stated as a curve, not a point: a large rupture is caught in seconds, while a small seep may take hours or slip under the threshold entirely. Push the test across leak rates to map that trade-off between minimum detectable leak size and detection time.
Be honest about the floor. Below some fraction of throughput, a mass-balance method cannot separate a real leak from measurement noise and normal line-pack imbalance and swing. Fast, large leaks are where a pressure-wave method such as negative pressure wave detection earns its place alongside the balance model.
Characterize Reliability
Reliability is the alarm-integrity attribute: how often the system produces a valid alarm for a real event versus a false alarm for a non-event. False alarms are not a nuisance to be tuned away casually - each one trains controllers to distrust the system, which is the real failure mode. Test both directions: injected leaks that should alarm, and known-clean operating swings that should not.
Look specifically at the conditions that generate false positives on your line - pump starts, valve moves, batch interfaces, and instrument glitches. A reliable system rides through those without crying wolf. Track the alarm history the way you would for any critical alarm, since the same discipline that governs SCADA alarm rationalization applies here.
Characterize Accuracy and Robustness
Accuracy is how close the system's estimate of leak volume and location is to the truth. Location accuracy in particular drives response - a leak flagged to within a valve segment is actionable; one flagged to 'somewhere on the line' is not. Test estimated versus actual for both, across the leak-rate range.
Robustness is survival under stress: does the system keep detecting when a transmitter drops out, telemetry is late, or the line is in an unusual state such as slack flow. A robust design degrades gracefully and tells the controller it is degraded, rather than going quiet. Document how the system behaves when an input it depends on fails, because that is exactly when a leak is most likely to be missed.
Common Mistakes
The biggest mistake is quoting sensitivity alone and calling the system evaluated. A very sensitive system with poor reliability floods controllers with false alarms; a very reliable one with poor sensitivity misses the small leaks that matter environmentally. The four attributes are a set.
The second is testing only in steady state. Leaks do not wait for steady flow, and the transient behavior is where most systems fail. If your test plan never exercises a ramp or a shut-in, it has not evaluated robustness - it has evaluated the easy 20 percent of the duty cycle.
Frequently Asked Questions
What are the four API 1130 performance attributes?
Sensitivity, reliability, accuracy, and robustness. Sensitivity is the smallest and fastest leak detectable; reliability is valid-versus-false alarm behavior; accuracy is leak size and location estimation; robustness is performance under transients and instrument faults.
Can one attribute be improved without hurting another?
Rarely. Tightening a threshold to catch smaller leaks (sensitivity) usually raises false alarms (reliability). The evaluation exists to make that trade-off explicit and set it deliberately for your line, not to chase a single perfect number.
Do I need physical leak tests to evaluate the system?
Some form of controlled test is expected, whether a physical drawoff or a validated modeled-leak injection. The method must be defensible. A system that has never been challenged with a leak-like signal has claims, not evaluation results.
Automation services
Need help turning this into a working system?
Merobix integrates SCADA, programs Allen-Bradley and Siemens PLCs, and designs and fabricates industrial control panels.
Meeting requests are reviewed before confirmation.