This case covers a 120,000 m² commercial complex with a 5-story high-end mall, a 20-story Grade A office tower, and a 2-level underground garage. Daily foot traffic is around 30,000 and doubles on holidays. In 2024, the plaza passed the emergency management department’s acceptance review for a fire protection monitoring demonstration project. Seventeen smart monitoring terminals run online 24/7, and data from seven categories of fire protection equipment (smoke detectors, temperature sensors, water pressure, water level, sprinklers, fire doors, and emergency lighting) is aggregated to the fire control center over LoRa.
Data is not the shortage; judgment is. A smoke alarm triggers, but why is it 8 minutes later than it should be? The end-line water pressure drops below the safety line. Is the fire pump the problem, or did the sprinklers discharge too much water? The standby pump interlock has activated. What hidden risk sits behind that signal?
In a traditional fire protection system, every detector reports independently, and there is no chain linking the alarms. Faced with a pile of alarms, operations staff know where an alarm triggered, but explaining why it triggered and what may happen next means pulling reports from three systems and relying on their own experience.
Why fire protection monitoring needs connected alarm context
Fire protection is different from industrial production. Many industrial processes let faults propagate from upstream to downstream over windows measured in hours. Fire protection runs a short chain: smoke detector → temperature sensor → sprinkler → water pressure, and the whole sequence, from anomaly to completed interlock, can take under 3 minutes. Those 3 minutes are exactly where causal coupling matters most:
- Every 1 ppm rise in smoke concentration raises the ambient temperature by 5 to 7 °C after a 12-minute lag. That is the physical delay of hot smoke diffusion.
- Every 10 m³/h increase in sprinkler flow drops the end-line water pressure by 0.04 MPa after a 2-minute lag. That is the pipe network behavior of hydraulic conduction.
- When the main pipe water pressure stays below 0.5 MPa for 10 minutes, the standby pump interlock activates after a 30-minute lag. That is the protection logic of the automatic interlock.
None of these couplings shows up in any single sensor. When the smoke detector triggers, you know there is smoke, but not that the temperature is climbing behind it. When the pressure drops, you know the network has lost pressure, but not whether the sprinklers discharged or the pump failed.
Two typical problems, two analysis paths
Problem 1: the smoke detector triggered, but why was it 8 minutes late? cross-sensor linkage root-cause tracing
Across inspection batches F-003 to F-007, smoke detector SD-M3 in the third-floor food court of the mall was the first to act up. Its smoke concentration rose gradually from a normal 0.05 ppm to 0.18 ppm but did not trigger an alarm, because it had not reached the 0.5 ppm threshold. Grease deposition had caused the detector’s sensitivity to drift. That is the root cause, but you cannot see it at this point.
During the day shift of batch F-004, a wok caught fire and the smoke concentration jumped from 0.18 ppm to 4.2 ppm. The alarm should have triggered at 11:20, but the drift slowed the detector’s response, and it did not trigger until 11:28, 8 minutes late. In those 8 minutes, the fire kept spreading.
Temperature sensor TD-M3 then picked up the relay. The ambient temperature rose from 29 °C to 78 °C, crossing the 57 °C threshold, and the temperature rise rate hit 12 °C/min. Sprinkler control valve SP-M3 opened 6 minutes after the temperature alarm, and the sprinkler flow surged to 85 m³/h. The heavy discharge dropped the end-line water pressure from 0.65 MPa to 0.28 MPa, below the 0.35 MPa safety line, so the most remote sprinkler head failed. The fire water tank level fell from 4.8 m to 2.6 m.
The full chain runs: smoke detector drift → delayed alarm → temperature sensor relay → sprinkler activation → end-line pressure loss → water tank drop, six links, each with a clear physical lag to the next. A traditional system scatters these six alarms across six pages, and the operations engineer has to click through them one by one.
TDengine’s approach is to chain these alarms into one causal chain:
Step 1: detect the anomaly from the smoke concentration trend. SD-M3’s concentration curve starts rising from a flat 0.05 ppm line, clearly deviating from the normal baseline. More importantly, the alarm trigger time (11:28) lags the threshold crossing time (11:20) by 8 minutes. That time gap is the signal of sensitivity drift.
Step 2: trace the interlock process along the propagation chain. After SD-M3 alarmed, TD-M3 ambient temperature reached 78 °C at 11:50, SP-M3 sprinkler flow reached 85 m³/h at 11:58, WP-End end-line water pressure fell to 0.28 MPa at 12:03, and WL-Tank water level fell to 3.2 m at 12:08. Overlaying the five parameters on one trend chart makes the causal propagation obvious.
Step 3: confirm the root cause. Comparing SD-M3 with the baselines of the other four smoke detectors in the plaza, SD-M1, SD-M2, SD-O1 and SD-G1 all hold steady at 0.03 to 0.05 ppm, and only SD-M3 has drifted to 0.18 ppm. Grease deposition in the food court shifted the sensitivity. That is the only explanation.
Problem 2: fire pump discharge pressure keeps falling. Is the seal the root cause? multi-parameter diagnosis of mechanical degradation
Across inspection batches F-010 to F-014, the discharge pressure of fire pump FP-G1 in the underground parking garage started to slide. But a falling pressure by itself says little. It could be a pipe network leak, a mechanical problem with the pump, or normal fluctuation.
TDengine’s real-time analysis triggered an alarm when FP-G1’s discharge pressure fell below 0.5 MPa, but the real judgment comes from overlaying three parameters: the pump discharge pressure is falling, the pump body temperature is rising, and the vibration value is rising. The combination of directions, pressure down + temperature up + vibration up, is a typical signature of mechanical seal aging. It is not a pipe network problem, because a network leak would only drop the pressure, not raise the temperature and vibration. And it is not normal fluctuation, because in normal fluctuation the three parameters would not all deviate in the same direction.
Tracing down the propagation chain: FP-G1 discharge pressure falls → WP-Main main pipe pressure drops from 0.80 MPa to 0.48 MPa → WP-End end-line pressure drops from 0.65 MPa to 0.38 MPa. The causal chain of hydraulic conduction is confirmed. Then FP-G2, the standby pump, interlocks and starts after the main pipe pressure stays below 0.5 MPa for 10 minutes, and a cascading response occurs.
The root cause is finally confirmed by comparing FP-G1’s seal leakage trend (0.1 → 2.8 L/min, far above the 0.5 L/min alarm line) with FP-G2’s seal leakage (steady at 0.1 L/min). FP-G1’s mechanical seal has aged. Further tracing shows that the visual inspection during the F-010 monthly deep maintenance failed to spot the micro leak, and the last intervention window was missed.
From alarm location to root cause and downstream risk
Smart fire protection linkage does not depend on how many sensors are connected or how many big screens are built. It depends on whether the data can form an actionable causal judgment:
- Cross-sensor root-cause tracing: from “the smoke detector triggered” to “the smoke concentration trend climbs, the alarm lags the threshold crossing by 8 minutes, comparison with the other detectors’ baselines confirms the drift, and the root cause is grease deposition”. A single point alarm becomes a complete root-cause story.
- Multi-parameter coupled diagnosis: from “low pump discharge pressure” to “three parameters deviate in the same direction, pressure down, temperature up, vibration up, the typical signature of mechanical seal aging, confirmed by the seal leakage trend, and the missed maintenance window”. Experience-based judgment becomes data evidence.
- Cascading propagation tracing: from “end-line water pressure fell below the safety line” to “tracing backwards along the pump, main pipe, end-line hydraulic conduction chain, locating FP-G1’s seal degradation, and recognizing that the standby pump interlock is the effect, not the cause”. The causality between alarms is no longer guesswork.
TDengine organizes the time-series data scattered across 17 fire protection terminals into business objects with physical coupling relationships, so an alarm is no longer an isolated signal but a node on a causal chain. When the fire control center receives an alarm, the operations engineer no longer sees only “a detector triggered” but the whole propagation chain: where the root cause is, how the event propagates, where it ends, and what may happen next.
TDengine comes with a high-performance, distributed time-series database, Industrial Ontology modeling and an Industrial Agent Runtime, providing a full-stack solution for industrial data streams from collection and storage to real-time analytics, visualization, event management, and root-cause analysis. To learn more about TDengine, visit www.tdengine.com and download it for free.
Try it yourself
Install and deploy TDengine Visit the TDengine Download Center, select TDengine All-in-One, choose the deployment platform and architecture that matches your environment, and follow the guided steps to complete the installation.
Load the sample data
On first activation, on the sample data loading screen, select Smart Fire Monitoring and Warning and wait for loading to complete.
If you have already activated the product, click your avatar in the top-right corner, select the Management Console, choose Sample Data on the left, then select Smart Fire Monitoring and Warning to load it. Wait a few minutes for loading to finish, and you are ready to explore.
Smart Fire Monitoring and Warning
