Why it matters
Absolute limits cannot see the abnormal-but-legal reading. A value that is unremarkable against a fixed limit may be extraordinary for that room, and that gap is where a developing problem hides.
Prediction & analysis
Baseline deviation
Baseline deviation measures how far a current reading has moved from what is normal for that specific sensor in that specific location.
How it works
A baseline is captured per device, typically at boot or over an initial learning window. Current readings are expressed as a difference from it, usually scaled by the channel's own variability.
Example
A solvent area normally reads 180 on its VOC channel. A reading of 260 breaches no limit and is a large deviation.
In Alfa HazardX
Baselines are captured per device at boot and used as the floor for percent-over-baseline metrics. This assumes the air is clean at that moment - a node powered up in a contaminated room learns the wrong zero.
Read further
Related terms
- Z-scoreA z-score expresses how many standard deviations a value sits from the mean of its distribution, making deviations comparable across channels with different units and scales.
- Time to thresholdTime to threshold is an estimate of how long a monitored channel will take to reach its alarm limit if its current trend continues.
- Sensor driftSensor drift is the gradual change in a sensor's output over time for an unchanged input, caused by ageing, contamination or environmental exposure.