interview with leakgeek creator

How would you explain LeakGeek™ in one sentence?

LeakGeek is an AI/ML driven small leak detection algorithm that reduces the risk of false positives in detecting small leaks that are <5% of nominal flow in liquid pipelines.

What was the catalyst to develop this product/service?

A major pipeline operator approached us with a clear problem – they needed to detect small leaks with a massive reduction of false alarms in the leak detection. Current CPM (computer pipeline monitoring) methods are reliable in detecting larger leaks, but struggle in smaller leaks due to the high risk of providing false positives.

Pipeline operations usually shut down when any leak alarm is raised in the control room. Therefore, false positives pose a threat to pipeline operations. The goal was to create an algorithm that can detect these small leaks, while at the same time reducing the risk of pipeline shutdowns because of false positives in its leak predictions.

How is the product different from a traditional CPM?

Most CPM systems are not designed for high-noise environments like gathering lines.  This algorithm works in tanks, metering stations, gathering, and transmission lines. Through its workflows, it only does leak detection when data is properly validated. Meaning, it does not do leak detection when it does not believe the data are candidates for leaks, this is the first step in false positive reduction. It then uses pattern recognition, and machine learning to infer on small leaks that are usually missed by traditional
CPM.  The algorithm can adjust its sensitivity based on pipeline geometry,  and operational cycles, making it more versatile than static threshold-based systems. With planned automation features, the algorithm can be integrated into enterprise systems for real-time, autonomous monitoring. 

How does the product work?

LeakGeek is comprised of 3 components or stages:

1. Data validation: Using UTSI’s DQM, we use AI/ML, and advanced analytics to validate all metered signals and we also assign quality scores to the measured signals to make sure we are making leak predictions with the best possible data. SCADA in pipelines are complex architectures with many moving parts, it is not uncommon for a data acquisition system to have measurement errors, communication errors, or problem with specific tag data quality.
2. Data Labeling (account for dynamic states): using UTSI’s AI/ML driven data labeling tool, we map all dynamic states of the pipeline. High quality data is data with context, which is why this step is very important. Once we know the operational state of the pipeline we can analyze for leaks in each of the pipeline states. Data looks different in different operations – this allows us to make better predictions during each identified pipeline state
3. Domain Leak detection: Using our AI driven algorithm, we use domain leak detection knowledge for a final leak inference. The algorithm is designed to detect leaks  <5% of nominal flow. Results have shown a 90%+ risk reduction of false positives, and has detected leaks 5 hours faster than operators and current CPM systems on average. It has also detected leaks that current CPM systems have missed.

Is there a hardware component to LeakGeek? 

There is no hardware component.  However, deployment into different hardware platforms can be done in different ways, by the use of docker containers, or even embedded code when possible.

Why do you think LeakGeek is a game-changer? 

The focus on false positive reduction is the key. Additionally, it has been tested with success in both gathering and transmission lines.

What are the anticipated market and cost savings impact?

1. Early Leak Detection = Massive Cost Avoidance
   • Small leaks, if undetected, can escalate into major spills. Early detection prevents:
     • Pipeline shutdowns (which can cost millions per day)
     • Environmental cleanup costs
     • Regulatory fines and penalties
     • Legal liabilities from landowners and communities

2. Reduction in False Positives
   • Current CPM (Computational Pipeline Monitoring) systems often trigger false alarms, leading to:
     • Unnecessary operational disruptions
     • Dispatch of field crews for non-events
     • Erosion of trust in monitoring systems
   • A smarter algorithm reduces these costs by filtering out noise, improving operational efficiency.
3. Asset Protection and Lifecycle Extension
   • Detecting small leaks early helps maintain pipeline integrity, reducing long-term maintenance and replacement costs.

Where do you see this product in 5 years?

I see the product deployed in over 600 segments of liquid pipelines.

View the LeakGeek case study results HERE.

For more information on LeakGeek product, view our product page HERE.

View Data Quality Management case study results HERE.