TwinEdge / Platform · Connecting the sensorsThe model of the plant
Connecting the sensors
Give every signal its place.
Bind tags to equipment, components, and measurements. Reuse those mappings so twins and analytics start with the right physical context.
What a mapping records
| Where it comes from | The device, and the address or node on it |
| What it is | The machine, and which input on its model |
| In what units | Converted once, at the point it arrives |
| How to read it | Any scaling or offset the instrument needs |
| When to doubt it | The conditions that make this reading untrustworthy |
§ 01
What it handles.
- 01It finds what is there
Browse a device and see what it is willing to tell you, rather than starting from a documentation file that was last accurate in 2019.
- 02One reading, one input
Each sensor is attached to a specific input on a specific machine model. That connection is the thing that makes a number mean something.
- 03Units sorted out once
Bar to kilopascals, gallons to litres, Fahrenheit to Celsius. Set it where the reading arrives, and every calculation after it is in the units you work in.
- 04Scaling and offsets, written down
The raw count that has to be multiplied by 0.1, or the transmitter reading 4 mA at zero. These live in the mapping instead of in somebody’s head.
- 05A reading knows when not to be trusted
Stale, out of range, or impossible values are marked as such and are not fed to models as though they were fresh. A dead sensor reports a beautifully steady number.
§ 02
Doing it once for a kind of machine.
The saving is not in mapping one pump. It is that the second pump of the same model, at another site, wired by a different contractor, is nearly the same job — so it becomes a matter of confirming rather than starting again.
The mapping for a kind of equipment is written down once and applied wherever that equipment appears. What differs is the handful of things that genuinely differ: the addresses, and the one channel somebody wired differently.
This is why the tenth site is quick and the first one is not. Anyone promising that the first site is quick has not done this before.
“Eighteen months in, nobody can say for certain which tag is which pump.”
Where analytics projects actually die
Send the tag list nobody wants to touch.
Thousands of rows, half of them abbreviated by somebody who left, units in a column that was never filled in. That is the realistic starting point, and it is the one worth testing against.