Troubleshooting

Common symptoms, what usually causes them, and what to check — grouped by where they show up. Most issues are resolved by the checks below; when they aren't, the last resort in each case is a support ticket with the specifics named (device serial, gateway, instance name, and a timestamp).

Two habits that prevent most surprises

Don't judge a device by its status icon alone — also check its last communication time (see Devices). A device can read Good while it has been silent for weeks.
Make sure the right people are subscribed to notifications — don't rely on logging in to catch an alert (see Notification settings).

Your invitation email contains a link to set your own password — it does not contain a password. Follow the link, create a password, then sign in with your email and that password at your instance URL. Invitation and password-reset links are valid for 72 hours; an expired link stops working but costs nothing — your account still exists, and a fresh link works the same way. If the link expired or you never received it, an administrator can Resend Invite or Reset Password from Access.

I can't see assets my colleague created

Views only show content in the groups currently selected in the group selector. Selecting a parent group does not automatically select its sub-groups — assets in a sub-group outside your selection are invisible. Open the group selector and select the sub-groups, or use the search box, which searches your whole permission scope. If a whole area is missing, your role may not grant access to that group — ask an administrator.

"Couldn't create asset. Reason: Bad request"

Asset names are unique across the whole instance. An asset with that name already exists — possibly in a group outside your current selection. Search all groups for the name, or choose a site-qualified name.

Devices and connectivity

A gateway shows offline (and its sensors went quiet)

A gateway is the uplink for every sensor that reports through it, so a gateway outage takes all of its sensors with it. Check, in order:

Power-cycle the gateway. This alone restores it in a large share of cases.
Power and network cabling on site — a pulled Ethernet cable or lost power is the most common cause.
Cellular/SIM gateways — confirm the data plan hasn't hit its monthly cap; a blocked SIM looks exactly like an outage.
Check the gateway's status LED on the unit.

If it doesn't return after a power-cycle, raise a support ticket with the gateway identifier and roughly when it dropped.

A sensor reads "healthy" but hasn't sent data for a long time

Don't trust the health icon on its own — a sensor that has stopped reporting can still show its last-known (good) status. Check each device's last-communication time and treat a device that has been silent for days or weeks as needing attention, even if it looks green. A long-silent sensor usually needs a power-cycle / battery reseat and possibly a firmware update. It's worth periodically reviewing your devices sorted by oldest last-communication.

A gateway's "last communication" looks stale, but its sensors show fresh data

These are two different signals: a gateway's own heartbeat to the cloud can lag behind the sensor payloads it is relaying. Before assuming a gateway is down, confirm whether sensor data is still arriving through it — if it is, the gateway is working.

A sensor isn't on the network map

A node missing from the network map usually hasn't joined the network yet, or hasn't reported since the last refresh. Check it's powered and in range of a gateway or a neighbouring node, confirm the group filter isn't hiding it, use the history slider to see when it last appeared, then Refresh. Treon Industrial Node X reaches the gateway over the Wirepas mesh (it can hop via neighbours); Treon Industrial Node C needs to be within Bluetooth range of its gateway.

"Get Settings" takes a long time, or a config change / "Measure Now" doesn't take effect

On-demand actions travel through the gateway to the device, which may be asleep, so allow from seconds up to a few minutes. Crucially, the device must be reachable when the action is issued — a Bluetooth Treon Industrial Node C buffers nothing, so it has to be connected at that moment. If an action seems stuck: confirm the device is online and recently communicating, retry, and follow the operation in Operational Telemetry. A device that's asleep picks up a saved configuration change at its next wake-up.

A new or long-idle Treon Industrial Node C won't come online

A Bluetooth Treon Industrial Node C that shipped from stock or sat unused can be in deep sleep and may not wake on its own. Tap it with an NFC-enabled phone running the mobile app to wake it — a dormant node revives reliably this way — then confirm it's within Bluetooth range of its gateway.

A device's configuration shows blank, or "Failed to load device configuration"

Live measurements can keep arriving while the configuration reads empty or all-disabled and Get Settings / Measure Now return nothing. A node reports its configuration structure at start-up (and periodically after); if the platform hasn't received it, config read/write fails even though data ingestion is fine. Reboot the node so it re-reports its configuration, make sure it's online, and check it's on current firmware.

Installation and commissioning

NFC provisioning seems to freeze

Provisioning a sensor by NFC takes several seconds — noticeably longer for Treon Industrial Node X than Treon Industrial Node C — with limited on-screen progress, so it can look frozen when it isn't. Hold the phone flat against the sensor's NFC spot and keep it still for the whole duration, and wait for the completion confirmation before moving it away. Pulling the phone away early is the most common reason provisioning fails.

Mobile commissioning stalls partway through

If commissioning a new node via the mobile app halts mid-progress and never finishes, this is usually an instance/backend issue rather than a faulty sensor or phone. Retry; if it still stalls, note the instance name and raise a support ticket — mention that commissioning stops part-way.

Data is plotted at the wrong time just after onboarding

If a freshly-onboarded node's measurements appear at implausible times — and on-demand measurements fail with a time error — the node's clock didn't sync at onboarding. Current platform versions deliver the clock as part of onboarding; if you've just upgraded, reboot the node to force a re-sync.

Measurements and data

Many KPIs show "No data found" or "Failed to load data"

A device only produces the KPIs its sensor type and measurement profile are configured to compute — not every metric in the picker. Velocity RMS, Acceleration RMS and Temperature are the broadly-available ones; advanced metrics (kurtosis, crest factor, peak-to-peak, zero-to-peak, per-axis and frequency-domain values) need a Treon Industrial Node X (Treon Make) sensor with those features enabled in its measurement profile. Pick KPIs the device actually outputs, and enable the ones you want in its configuration.

A measurement point has no Spectrum or Waveform

Spectra and waveforms are a Treon Make (Treon Industrial Node X) capability — a Treon Flow (Treon Industrial Node C) point is KPI-only and has none (see platform and solutions). On a Treon Industrial Node X point that shows none, the feature isn't enabled in its calculation settings, or none has been captured yet — use Measure Now to trigger one.

A waveform arrives truncated, or not at all (Treon Industrial Node X)

On some node firmware, requesting an FFT and a waveform in the same measurement, or setting the spectrum/waveform resolution too high, can cut a capture short or prevent it arriving. Update the node to current firmware, keep the acceleration spectrum/waveform resolution within limits (lower the Fmax / apply a high cut-off), avoid combining FFT and waveform in one measurement on affected firmware, and re-trigger the capture.

A report shows "No Data" for a KPI

The report lists a KPI the selected sensor doesn't produce (see "Many KPIs show No data found" above). Choose KPIs the device is configured to output. Spectrum/FFT report elements apply only to Treon Industrial Node X points.

A temperature alert is late, or a reading looks impossible

Sensors poll temperature on a periodic cycle (roughly every 15 minutes) to save battery, so a temperature threshold is only checked at the next poll — expect an alert to lag and slightly overshoot a fast-changing temperature. Treat temperature thresholds as coarse alerts rather than precise capture. Separately, disregard isolated, physically-impossible values (for example −256 °C) among otherwise-sensible readings; those are invalid samples, not real measurements.

Events and alerts

An event won't go away / keeps coming back

Resolving or rejecting an event moves it to History. If a new event for the same measurement point keeps appearing, the underlying condition is still occurring — the value is still over its threshold, or the AI model still sees the pattern. Investigate via the measurement trends, or adjust the threshold / ISO class if the limit itself is wrong for this machine.

Too many alerts / false alarms

A few common causes and what helps:

A single incidental spike (a passing load, a one-off knock) trips a simple threshold. Tune the threshold to the machine, and prefer AI anomaly detection, which only raises an event once a deviation persists rather than on one sample.
A warning and an alert for the same exceedance — crossing a higher limit can also satisfy the lower one. Set limits with that in mind.
A large batch of temperature alarms across many machines at once is almost always ambient (e.g. a heatwave), not a fleet of simultaneous faults — check the weather before the machines.

I missed an alert

Alerts reach you two ways beyond the dashboard: email and mobile push. Make sure every responsible person is subscribed (per group and severity) in Notification settings — it's easy to finish onboarding without subscribing anyone. Install the mobile app and enable notifications so alerts reach you in the field.

"Devices Trained" shows fewer than my device count

The Devices Trained number counts devices whose AI model has finished training. Training needs enough measurement history, so newly commissioned devices show as untrained until they've collected sufficient data. This is expected and resolves itself; a device that never trains despite weeks of running data is worth a support ticket.

Events fire while a machine is stopped or idling

A machine that's switched off or idling vibrates very little, and without a floor that near-zero state can be misread as abnormal. Each device has an off-state cutoff — a minimum vibration level (around 0.5 mm/s velocity RMS by default) below which readings are treated as stopped, not faulty and left out of AI scoring. If a duty-cycled machine raises events when idle, check that cutoff suits the asset's real idle level, and clear the false events with the correct cause ("no failure detected" or an operating-condition change) so the AI model learns that state as normal.

After a firmware or platform update

Advanced measurements stopped, or a device's configuration can't be edited

After an update, if magnetometer/RPM, waveform, or acceleration FFT go missing, only one axis arrives, or you can't edit a node's profile, the node firmware and the platform version may be out of alignment (the node is on older firmware than the platform expects, or vice-versa). Confirm the device is online, check its firmware in Devices → Firmware updates, and keep spectrum/ waveform resolution within limits. If it persists, raise a ticket with the node identifier and a sample of what's missing.

Shorter battery life than expected

Higher measurement cadence, wake-on-motion, and accelerated AI-training modes all trade battery life for more data — use the recommended default cadence unless you need more. A specific firmware build can also affect consumption, so keep firmware current. Treon Industrial Node X has a replaceable battery module; a Treon Industrial Node C battery is non-replaceable, so the sensor is replaced at end of life.

Known limitations

A few current UI limitations that are expected rather than faults:

Trend/comparison legend can hide when many series (roughly more than four) are plotted at once — plot fewer series to keep the legend.
Some view choices (column selection, filters, chart style) don't persist between sessions.
Devices-list exports may not include the asset name/identifier for each device.
A grey battery indicator is informational — it shows the raw battery voltage, not a colour-coded health status. Grey does not mean a fault; read the numeric voltage.
Some advanced configuration options are intentionally hidden on production instances, so two instances on the same version can expose different Treon Industrial Node X options. If a feature you're licensed for is genuinely missing, contact support — it's an instance setting, not a version mismatch.