Short answer: a frozen temperature channel is a value that stops updating while still looking plausible. Detection works best by combining timestamp freshness, repeated-value duration, process movement and comparison with relevant neighbouring channels.
An open circuit may produce an obvious invalid reading. A frozen value is harder because 742 °C can look reasonable on an HMI for a long time. The fault may be in the sensor, wiring, input module, communication path, application cache or display—not necessarily at the thermocouple tip.
Signals that support a frozen-value rule
- Freshness: did a new source timestamp arrive, or is the screen repainting an old value?
- Exact repetition: has the raw value remained bit-for-bit identical beyond a configured duration?
- Expected movement: is the furnace ramping or cooling while the point stays flat?
- Group behaviour: are comparable channels moving together while one remains still?
- Independent system health: did a communication, module or database event occur at the same time?
Exact repetition alone can create false alarms during a genuine stable soak, especially when displayed resolution is coarse. Conversely, noise may make a stuck source appear to move if the fault occurs after filtering or if the application reuses a computed value.
A credible decision pattern
| Observed condition | Possible interpretation | Useful next check |
|---|---|---|
| Timestamp old and value unchanged | Data transport or acquisition stopped | Check module heartbeat and last successful write |
| Timestamp fresh, raw code unchanged during ramp | Input or upstream source may be frozen | Compare raw input and an independent reference |
| One channel flat, neighbours moving | Condition consistent with channel-specific trouble | Inspect wiring, terminal and sensor path |
| All channels flat together | Real soak or common system issue | Compare process state, controller and acquisition health |
Preserve evidence before resetting
Record the last changing value, source and server timestamps, furnace state, neighbouring channels, alarm transitions and any module restart. A quick power cycle may restore the display while erasing the best clue to root cause.
Pyrolog logic can be configured to identify conditions consistent with a frozen reading where the topology and process provide enough context. It should not be advertised as universal AI fault diagnosis.