The most dangerous fault in a SCADA system is often the one that looks fine. A frozen value is a tag that still reads Good quality and shows a plausible number - but that number never moves. Behind it may be a failed sensor, a plugged impulse line, or a communication path quietly serving the same cached last-good value over and over. Because the quality flag is Good, nothing complains, and an operator can trust a reading that stopped being real hours ago. This page explains what a frozen value is, how to tell a stuck transmitter from a genuinely steady process, and why a frozen Good value is more dangerous than an honest Bad flag.
Frozen value / stuck transmitter in one line: A frozen value is a SCADA tag whose reading stays fixed at one number and never updates, even though the system still reports it as Good quality. It typically indicates a stuck or failed transmitter, a plugged impulse line, or a comm path repeatedly returning a cached last-good value. It is dangerous precisely because nothing flags it - the value looks live and trustworthy while the real measurement has stopped changing.
A frozen value is deceptive because it passes every obvious check. The quality flag says Good, the number on the screen is within a believable range, and if an operator glances at it once, nothing seems wrong. The tell is only visible over time: the value simply does not move when the process should be moving it. A tank that is filling should show a rising level; a flowing line should show flow that varies; a temperature under load should wander. When such a value sits at exactly the same figure for far longer than the process would ever hold it, the reading has frozen even though the system still trusts it.
Several physical and data-path faults produce this. A transmitter can fail in a way that latches its output at the last value instead of dropping to a fault state, so it keeps reporting a fixed signal. A plugged or frozen impulse line on a pressure or level transmitter isolates the sensor from the real process, so it faithfully measures a pressure that is no longer connected to anything and never changes. And a communication or caching path can serve a stale last-good value repeatedly - the underlying poll may be failing while the layer above keeps handing out the last number it successfully read, with the quality still marked Good. In every case the value is frozen while the system has no idea anything is wrong.
The hard part is that a perfectly flat reading can be completely legitimate. A tank at a stable level, a pressure held tight by a good controller, or a temperature in steady state can all sit at nearly the same value for a long time. So you cannot call a value frozen just because it is not moving - you have to distinguish a value that should be moving but is not from one that is legitimately still. The practical distinction is expectation: does the process context imply this value ought to be changing right now, and does the reading show even the small natural variation that a live measurement almost always carries?
That is where detection techniques come in. A change-of-state or no-update watchdog checks whether a value has changed at all within an expected window and flags it if it has been perfectly static too long. Timestamp aging looks at whether the value is actually being refreshed - a live measurement gets new updates even if the number happens to be similar, whereas a truly frozen value may stop getting fresh timestamps altogether. Real sensors also produce a little noise or dither; a reading pinned to an identical value with zero fluctuation is more suspicious than one that jitters slightly around a level. Cross-checking against a related measurement helps too: if flow is clearly running but the downstream level never moves, one of them is lying. Combining these - expected behavior, update freshness, natural noise, and correlation - is how you separate a stuck transmitter from an honestly steady process.
A Bad-quality tag announces its own problem. It grays out, it alarms, and an operator immediately knows not to trust it and to go investigate. A frozen value does none of that - it hides in plain sight wearing a Good flag, which is exactly what makes it more dangerous. Decisions get made on a number that stopped being real, control can drift because a feedback signal is stuck, and a genuine excursion can go completely unseen because the very sensor meant to catch it is reporting a frozen value that never crosses an alarm limit. The failure is silent, and silent failures are the ones that turn into incidents.
This is why active frozen-value detection is worth building in rather than relying on quality flags alone, and where a cloud platform's history and watchdogs help. It matters to operations when the system can notice that a Good-quality tag has not changed within its expected window and raise a no-update or frozen-value alarm, because that converts an invisible fault into a visible one. A cloud SCADA platform such as Merobix continuously historizes every tag, so a value stuck at an identical figure while its neighbors move is detectable, and stale-data and change-of-state checks can flag a transmitter that has quietly stopped reporting a real measurement. That turns the most dangerous kind of fault - the one that looks healthy - into an alarm someone can act on before it hides a real event.
Because quality flags describe whether the system can obtain and vouch for a value, not whether the value is changing. A transmitter can fail latched at its last output, a plugged impulse line can leave a sensor measuring a fixed isolated pressure, or a caching path can keep serving the last successfully read number - all while the quality stays Good. The value is frozen but nothing in the quality flag reveals it.
Use context and freshness rather than the number alone. Ask whether the process should be making this value change right now, whether the value is still getting new updated timestamps, and whether it shows the small natural noise a live measurement usually carries. Cross-check against a related tag too - if flow is running but the level never moves, one of them is stuck. A no-update watchdog and timestamp aging automate these checks.
A Bad-quality tag announces itself - it grays out and alarms, so operators know to distrust it. A frozen value wears a Good flag and shows a plausible number, so it is trusted while the real measurement has stopped. Decisions get made on stale data and a real excursion can go unseen because the stuck sensor never crosses an alarm limit. The danger is that the failure is silent.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.