Data Quality
A value is only useful when you can trust it. When a sensor breaks, a PLC answers a read with an error or a device drops off the network, UpBlue stores that too: every datapoint carries a quality next to its value.
| Quality | Meaning |
|---|---|
| Good | A normal reading. |
| Uncertain | The device delivered a value but doubts it, for example an OPC UA server that reports the last usable value of a sensor it can no longer read. |
| Bad | The device reported a fault, or the reading failed. A failed reading has no value. |
Where quality comes from
- OPC UA: the status the server sends with each value. Good, uncertain and bad map one to one.
- Modbus: a read that fails (an exception from the device, a timeout, a lost connection) is stored once as a bad datapoint without a value. The next successful read makes the tag good again.
- UpBlue devices: the same for the device’s own Modbus reads. When the collector loses the connection to a device, the device’s tags are marked bad until it delivers values again.
So a machine that stops answering on Friday evening shows as Bad from that moment, instead of keeping its last value as if nothing happened.
In charts
A chart shows where the readings failed, without any setting:
- A failed reading has no value, so the line breaks there, and the period gets a light hatched band. Gap fill stops at a failed read: it never draws a line through a period without readings.
- An uncertain period gets a subtle marker along the bottom of the chart.
- Hover over a marker to see the tag, the times and the reason, for example the status the OPC UA server sent or a Modbus timeout.
Numbers do not change: a failed reading never counts in an average, a sum or a minimum, and uncertain values stay in.
An event whose end UpBlue set because the Collector or device went silent is drawn dashed, with “estimated: collector offline” (or “device offline”).
On the tag page
Next to the latest value, the tag page shows its status: Valid, Uncertain or No reading, since when, and the reason. The raw datapoints list shows the quality of every datapoint; a failed reading appears as a row without a value and quality Bad.
Attention items
A tag whose source answers but gives no valid reading for longer than the organisation’s offline time (15 minutes by default, set in Edit organisation) becomes an attention item: “No reading for more than 15 minutes” or “Only uncertain readings for more than 15 minutes”. It links to the Flow card of the tag, which shows the source and rule it comes from. This covers tags read through a Collector. A tag that just stops sending raises no item: use a No data alert rule for that.
Alerts
A No data alert rule also fires when the latest datapoint in its window is a failed read: the source answers, the sensor does not. You need no separate rule for broken sensors.
Existing No data rules on sensors that report failed reads can start firing.
Using quality in dashboards
You rarely need this: charts already show quality. Under Advanced in a data query of a dashboard panel are two settings:
- Field: Value (default) charts the values; Quality charts the quality per datapoint instead (0 good, 1 uncertain, 2 bad). With the aggregate Max you see the worst quality of every bucket, for example as a status bar under a trend.
- Quality: only use datapoints of the selected qualities. Select Good to keep doubtful values out of an average or a total.
Leave the Quality filter empty to use all datapoints. A failed reading has no value, so it never counts in an average, a sum or a minimum.