Every water distribution system produces a continuous stream of pressure data. In most utilities, that stream flows into a historian, accumulates for months, and is consulted only after a main break — to reconstruct what happened, not to prevent it. That habit is worth reconsidering.
Pressure transients are not random noise. They carry structural information about the condition of buried pipe, and that information is readable weeks or months before a segment reaches visible failure. The challenge is that extracting it requires treating the pressure signal as a diagnostic channel, not just an operational parameter.
What a transient actually tells you
A pressure transient is a rapid, temporary deviation from the hydraulic grade line — typically lasting milliseconds to a few seconds. They originate from pump starts and stops, valve actuations, hydrant flushes, and demand surges at service connections. Every distribution system generates dozens to hundreds of them per day; in a well-instrumented network, they're visible as sharp spikes in the SCADA pressure record.
On an intact, structurally sound main, a transient propagates through the pipe wall with predictable wave speed. The wave speed in a buried water main is a function of the elastic modulus of the pipe material, wall thickness, diameter, and the bulk modulus of water — the classic Joukowsky relationship. For a 12-inch ductile iron main in good condition, you'd expect wave speeds in the range of 1,200–1,400 m/s.
When that same main has developed internal corrosion pitting, wall thinning from tuberculation, or a developing circumferential stress fracture, the effective stiffness of the pipe wall decreases. The transient wave speed drops — sometimes by 5–15% in moderately deteriorated sections, and by significantly more in severely compromised segments. A continuously logged pressure record, analyzed with appropriate signal processing, can detect that change.
This is not a new observation. Transient-based condition assessment (sometimes called TBCA) has been an active research area since at least the mid-1990s, with foundational work by researchers at the University of Adelaide and elsewhere. The constraint has always been practical: the sensing infrastructure required to capture high-resolution transients across a large distribution network was expensive, and the analytical burden was substantial.
The data most utilities already have — and aren't using
Here's the inconvenient reality: a meaningful number of SCADA pressure loggers already sample at 1–4 Hz, fast enough to capture the shape of many transient events. Utilities that have deployed AMI meter infrastructure often have pressure recording capabilities at endpoint nodes that were never configured for high-resolution logging. The data gap is smaller than it appears.
The bigger gap is in how the data is being processed. A standard SCADA historian aggregates pressure readings into 5- or 15-minute averages for operational display. Transient peaks are invisible at that resolution — they're compressed out. The raw, sub-minute samples are retained in most systems for a configurable rolling window, typically 24–72 hours, before being downsampled. Almost no utility has a workflow that captures and indexes those raw readings for condition-assessment analysis.
Consider a 2024 scenario we see repeatedly in the Southwest: a utility with a 1970s-era 16-inch unlined cast iron trunk main running through a commercial corridor. The main carries high daily demand fluctuation because it serves a mixed residential-commercial zone — restaurants and irrigation accounts cycle pressure aggressively. The SCADA data shows normal operating pressures, 72–85 PSI, and the 5-minute averages look perfectly stable. But if you pull the raw 1-Hz samples from a nearby pressure logger during a pump station start, you see transient peaks that are 8–12 PSI above the steady-state pressure, and the waveform decay pattern has become progressively more irregular over the prior 18 months. The main broke catastrophically during a demand peak in June. None of that degradation signal was visible in the data the operations team was actually looking at.
The signal-to-prediction problem
We should be clear about what transient analysis can and cannot do on its own. A single transient reading, or even a month's worth of readings from one logger, is insufficient to confidently identify a specific failing segment. The wave speed inversion problem — deducing where along a pipe run the condition has changed — requires multiple measurement points and careful interpretation of wave reflections, which gets complicated in networked systems with branches and tee intersections. Any vendor claiming a simple "transient signal = imminent break" model is oversimplifying.
What the transient signal does well is serve as a probabilistic early-warning layer within a multi-variable risk model. When a 200-meter segment has elevated transient deviation compared to its historical baseline AND that same segment has unlined cast iron installed in 1965 AND it sits in a soil profile with high shrink-swell index AND the adjacent segment broke twice in the past eight years — the combination of those signals produces a risk score that is genuinely predictive, not just a heuristic.
This is the architecture behind how Watsynq uses pressure data: not as a standalone diagnostic, but as one input channel — weighted according to its statistical relationship with historical break events in the training set — fused with GIS pipe attributes, soil classification data, break records, and operational history.
Practical implementation: what the operations team needs to change
Getting value from transient data does not require replacing your SCADA infrastructure. The practical requirements are more modest: configure your existing pressure loggers to retain raw 1-Hz or higher resolution samples for a minimum 7-day rolling window rather than 24-hour. Identify the 15–20 loggers in your network that are positioned closest to your highest-risk pipe segments (the AWWA M36 asset management framework provides a methodology for identifying those segments if you don't have an existing risk ranking). Ensure those loggers are transmitting to a historian that can accept and store the raw samples without downsampling.
The analytical layer — detecting anomalous transient signatures, correlating wave speed changes over time, flagging segments for elevated attention — is where software matters. The pattern recognition problem across a large dataset of irregular transient events is not tractable with manual review. But the data collection side is largely a configuration change, not a capital project.
A note on what transients won't catch
Pressure transient analysis is most effective for metallic pipe with distributed corrosion deterioration — unlined cast iron, older ductile iron, and steel mains. It is considerably less useful for asbestos cement (AC) pipe, which deteriorates through a different mechanism (delamination and chemical attack rather than wall-thinning), and is largely not applicable to PVC or HDPE, which are effectively elastic enough that their transient profiles don't change meaningfully until near the point of failure. For AC pipe specifically — and a substantial portion of Southwest water networks installed between 1955 and 1975 include AC distribution mains — acoustic leak detection methods and direct inspection remain the more reliable condition signals.
Transient analysis also doesn't help you identify failing service connections, meter manifolds, or valve seats — those failure modes don't produce the pipe-wall stiffness change that the transient signal encodes. Point your transient monitoring effort at trunk mains and large-diameter distribution mains where the structural information content is highest and the consequence of failure is greatest.
The data is there. Most of it is being thrown away every 48 hours. That is worth fixing before the next emergency crew shows up at 2 a.m. looking for the break.
Ethan Morales is CEO and Co-Founder of Watsynq. He spent several years in water infrastructure consulting before founding the company in Phoenix in 2023.