Flowstate co-founder and CEO Jerad Stack recently returned to the Pipeliners Podcast to join host Russel Treat for Episode 460, “Leak Detection in Compressible Fluids.” The conversation explores one of the more challenging problems in computational pipeline monitoring (CPM): how to reliably detect leaks in pipelines transporting compressible fluids such as natural gas, CO₂, NGLs, and ethylene.
Historically, computational leak detection has been much more accessible for traditional liquid pipelines. Compressible fluids introduce an additional layer of complexity because changes in pressure and temperature can significantly affect the amount of product stored within a pipeline. What goes in doesn’t necessarily come out at the same rate, making it difficult to distinguish normal operating conditions from an actual leak.
In the episode, Jerad and Russel discuss how advances in machine learning, combined with traditional physics-based modeling, are opening new possibilities for addressing this challenge.
What Makes Compressible Fluids Different?
Unlike traditional liquid pipelines, where volume balance methods can provide a relatively straightforward indication of product loss, compressible pipelines require operators to account for changes in internal line inventory.
Pressure, temperature, fluid properties, and operating conditions all influence how much product is contained within the pipeline at any given time. Without accounting for these variables, normal pipeline behavior can look like a leak, creating unnecessary alarms or making smaller leaks difficult to identify.
Jerad explains how Flowstate combines physics-based calculations with machine learning to better understand these complex hydraulic behaviors. By accounting for expected inventory changes and learning how a pipeline operates across different conditions, the technology can help distinguish normal operational transients from potential leak events.
Bringing a New Approach to Compressible Pipeline Monitoring
A key theme throughout the conversation is the role of modern machine learning in solving problems that have historically required extensive engineering, complex hydraulic models, and significant manual tuning.
Flowstate’s new compressible-fluid solution builds on its existing leak detection platform, combining statistical volume balance methods with physics-informed machine learning to estimate pipeline inventory and identify unexplained deviations.
The approach is designed to leverage data operators are already collecting through SCADA, rather than requiring an entirely new instrumentation or data infrastructure investment.
Jerad and Russel also discuss the importance of data quality, pressure and temperature measurements, time synchronization, and understanding a pipeline’s full operating envelope. These considerations are critical to successfully implementing leak detection on compressible systems.
Expanding What’s Possible for Pipeline Leak Detection
The conversation highlights an exciting opportunity for the pipeline industry. Advances in computing and machine learning are making it possible to tackle increasingly complex pipeline monitoring challenges with greater sophistication.
For operators transporting compressible fluids, effective computational leak detection no longer has to be out of reach.
Flowstate is excited to bring this new capability to the market and help operators improve pipeline monitoring, reduce false alarms, and strengthen their overall leak detection programs.
Listen to Episode 460: “Leahttps://pipelinepodcastnetwork.com/leak-detection-in-compressible-fluids/k Detection in Compressible Fluids with Jerad Stack” on the Pipeliners Podcast.
Learn more about Flowstate’s Leak Detection Solution for Compressible Fluids.



