Sensors and condition alerts

What is being measured continuously, what has breached a limit, and which sensors actually raise maintenance.

What the instruments on your machines are reading, what has crossed a limit, and which of those readings turn into work.

A trace on its own is decoration. 4.2 mm/s is neither high nor low until you know the limit, and the limit is not interesting until you know whether crossing it raises a work order or just draws a red line. This screen is built around those two questions.

The header

Figure	Meaning
Sensors	How many are registered
Open alerts	Breaches nobody has looked at
Feeding maintenance	Sensors bound to a meter

A banner appears when any sensor sits on a device that has stopped reporting. This is worth saying out loud: a silent sensor raises no alerts, which reads exactly like a healthy one.

The sensor list

Each row carries what it measures, its latest reading, when it was last heard from, how many limits watch it, and how many alerts are open against it.

The Feeds column is the one that matters most. A sensor bound to a meter writes its readings through to that meter, so the rounds already scheduled against the meter — every 2,000 run hours, every 500,000 cycles — come due on their own. A sensor feeding nothing is a chart. See meters and readings and the plan builder.

Click a row to open its trace.

The trace

Readings over the last 24 hours, 7 days or 30 days, drawn against the limits the sensor is judged by.

• The line is the average across each bucket of time.
• The band is the highest and lowest reading inside that bucket — not error bars. It is drawn because a spike that lasted one sample is exactly what an average hides.
• The dashed lines are the limits. Every limit is forced into the scale, so a limit is never drawn off the top of the plot where nobody can see they are under it.
• A gap in the line is a gap in the readings. A line drawn straight across an outage would be a picture of a sensor that was working.

Beneath the chart, each limit is listed with what it watches for — above, below, outside or inside a value — how many consecutive samples it takes to trip, what it takes to clear, and whether tripping it raises a work order.

Alerts

One alert per breach, not one per reading. An alert opens when a limit is crossed for as many samples in a row as the limit requires, and closes on its own when the reading comes back past the clearing band.

Each row shows what tripped, the reading that opened it, the worst reading since, the severity, and a link to the work order it raised if it raised one. An alert that opened at 7.1 and peaked at 22 is a different event from one that hovered, which is why both numbers are there.

Acknowledge says somebody has seen it. It does not clear it — the reading clears it. The row stays until the machine says otherwise.

Acknowledging needs permission to edit assets. Reading the screen does not.

Predictions

A tab that is normally empty, and says so rather than saying “no data”. Nothing in the product produces a prediction yet. The tab exists so that when something does, what it expected and the day it said so are on record — which is what makes it possible to grade a prediction afterwards rather than admire it on the day it was made.

Registering a sensor

Sensors, the devices they sit on, and the limits that watch them are set up in Administration, under the IoT lists. See administered lists.

A sensor becomes useful the moment it is bound to a meter. Until then it is a number on a screen.