Predictive maintenance on NCD vibration sensors without a vibration analyst

General
1

Most people who buy industrial vibration sensors are not vibration analysts. They run a plant, they know a motor is going bad when it starts sounding wrong, and they want to know that three weeks earlier instead of on the day it stops.

This tutorial provides a step-by-step guide to integrate NCD vibration sensors with the TagoIO platform using the native TagoIO app for the Atrium gateway. By following these instructions, you will stream sensor telemetry into TagoIO and configure those sensors remotely from your TagoIO dashboard, with no code and no payload parser to write.

The sensors already solve the signal processing half of the problem. NCD, a US sensor manufacturer, computes RMS velocity, peak acceleration, displacement and the top peak frequencies from an on-sensor FFT, then transmits around twenty numbers per report instead of a raw waveform. What is left is storing that data, learning what normal looks like for each individual machine, spotting the slow drift that precedes a bearing failure, and getting an alert to a human before the machine trips. That is a memory and judgment problem, and it does not belong on a battery powered sensor.

Why remote configuration matters here

A vibration sensor reporting every thirty minutes is fine for a machine behaving normally. The moment one starts trending badly you want tighter sampling to see what is developing, and today that usually means someone drives to the site and opens the gateway.

The Atrium app closes that loop. The gateway publishes each sensor’s configuration schema to TagoIO, TagoIO builds a settings dashboard for that sensor automatically, and any change you make there is delivered to the sensor in the field and confirmed back. Both halves of the integration are covered below.

Prerequisites

Before you begin, ensure you have the following:

  • TagoIO account: if you don’t have one, register here.
  • Atrium gateway on version 2.3.2 or later: earlier versions do not include the app center.
  • Sensors provisioned on the Atrium: add your vibration sensors to the gateway before starting.

Step 1: Install the TagoIO app on Atrium

  1. Open the app center

    • On your Atrium gateway, click Browse App Center.
    • Find the TagoIO app.

  2. Install and open it

    • Select the app and install it from the app center.
    • Open the app. It shows as not connected. The next two steps authenticate it against your TagoIO account.

Step 2: Create the Profile Token in TagoIO

  1. Access your organization tokens

    • Click your name in the bottom left corner of the TagoIO Admin.
    • Go to My Account, select your profile, then go to the tab Tokens.

  2. Create a new token

    • Click to create a token and give it a meaningful name (e.g., Atrium app).
    • Set it to full permission and choose an expiration.
    • Generate the token and confirm with your password.

  3. Copy it into the Atrium app

    • Copy the token, paste it into the Profile Token field in the Atrium app.
    • Select the region of your account: US or Europe.

:information_source: Note: this token is what allows the Atrium app to create devices in your account. If it expires you have to load a new one into the gateway, so never avoids maintenance. If your security policy requires rotation, set one month or one year and put a reminder somewhere you will see it.

:warning: A wrong region is the most common reason the connection fails.

Step 3: Create the Service Authorization in TagoIO

This second token is what allows the gateway to create devices and publish their telemetry.

  1. Access Authorizations

    • Go to Devices in the TagoIO Admin and click Authorization.
    • In the Service Authorization page, click on Add Authorizatino on the right.

  2. Create the authorization

    • Click Add Authorization and give it a name (e.g., Atrium app).
    • Set the token format to TagoIO TiP.
    • Create it. The authorization appears in the list with a purple TagoIO TiP label.

  3. Copy it into the Atrium app

    • Copy the token and paste it into the Authorization field in the Atrium app.
    • Click Connect.

The app now shows your account name, region and profile. Verify they look correct.

:warning: The TagoIO TiP token format is the step most people miss. A standard authorization will not work here.

Step 4: Select the sensors that report to TagoIO

  1. Choose your devices

    • Open Devices in the Atrium app’s left menu.
    • Select the vibration sensors you want reporting to TagoIO.

  2. Check their state

    • Each selected sensor gets a green icon and a state of valid, meaning the device now exists in your TagoIO account.
    • Last seen is the last wireless transmission the Atrium heard from that sensor.
    • Last push is the last time the Atrium forwarded that sensor’s telemetry to TagoIO.

  3. Trigger the first transmission

    • Last push reads never until the sensor transmits again. Sensors report on their own schedule.
    • To force it, press the sensor’s reset button or trigger the reed switch with a magnet.

  4. Verify data reception

    • In TagoIO, refresh Devices. Your sensors are listed.
    • Open the sensor you woke and check the Data tab: RMS velocity per axis, peak acceleration, peak frequencies, temperature, RPM and battery.
    • Devices that have not transmitted yet show their variables with no data. That is expected until their next report.
    • You can also use the Live Inspector to watch incoming data in real time.

Step 5: Configure sensors from the TagoIO dashboard

When the Atrium sees a sensor, it publishes that sensor’s configuration schema to TagoIO: the sensor type, the settings it exposes, and what a valid value looks like for each one.

  1. Confirm the schema published

    • Open the Configuration page in the Atrium app and check that the schema published for your devices.

  2. Find the dashboard

    • In TagoIO, a dashboard for that sensor now exists. Nothing was built by hand.
    • Because it comes from the schema, a new sensor model works the same way without anyone writing a form for it.

  3. Change a setting

    • Change the sampling interval, for example from five minutes to ten.
    • The field flags as unsaved. Click Apply changes.
    • The status becomes waiting for the sensor to wake up.

  4. Understand the timing

    • The Atrium holds the pending change and pushes settings to sensors roughly once an hour.
    • The sensor wakes on its own schedule, sends a data packet, then sends a sync packet asking whether any settings are waiting.
    • If a change is pending, it is applied and confirmed back up to TagoIO.

  5. Force it, or wait

    • To see it immediately, trigger a sensor reset.
    • Within seconds the setting applies, the sensor confirms it, and the dashboard clears the pending flag and shows 10 minutes as the value the sensor itself reported. Not the value you requested: the value the hardware confirmed.

:information_source: Note: after setup you never need the Atrium interface open for this. Change the setting in TagoIO, click apply, and the confirmation arrives on its own.

What to turn on once the data has history

Raw vibration numbers on a chart are monitoring, not prediction. Three things convert them, and none require code or a data scientist.

Severity zones on the chart. ISO 20816 defines RMS velocity bands per machine class, from good through to danger. Draw them on the velocity chart and anyone reading the dashboard sees green, yellow or red without knowing what a spectrum is.

Anomaly detection against a learned baseline. Fixed thresholds are noisy: a machine at full load vibrates differently from the same machine idling, so static limits either miss real problems or cry wolf until people stop reading the alerts. Models trained on each machine’s own stored history score new readings against how that specific machine normally behaves, learn its operating modes so a load change does not fire an alert, and catch the slow weeks-long drift that precedes most bearing and imbalance failures. You can also compare identical machines against each other and find the one pump out of twelve behaving differently, without setting a single threshold.

Time to threshold. Point a forecast at RMS velocity, set the limit at the ISO zone boundary for that machine class, and the output is a countdown in days plus a severity tier. That turns spectrum analysis into a planning horizon, which is the language a maintenance planner works in.

All three arrive as ordinary TagoIO variables, so they feed Dashboards, Analyses and Actions, and a branded portal your own customers can log into.

:information_source: The honest limit: this tells you which machine needs attention, how urgent it is, and roughly how much time you have. It does not tell you that the outer race on bearing two is cracked. Diagnosing a specific failure mode is spectrum work for an analyst, on the few machines that get flagged. For most plants that split is the point: you do not need an analyst watching every motor, you need to know which three motors deserve one.

On building this yourself

You can get vibration data into a database in an afternoon with an AI assistant. For one gateway on one machine, it works.

What tends to surface later, in roughly this order: provisioning sensors without hand-editing config, keeping tokens and transport secure, handling the configuration schema so remote settings changes are possible at all, alert routing that reaches a human who is not looking at a screen, several sites in one view instead of one gateway at a time, and enough retained history for a baseline model to mean anything. Each one is solvable. Together they are a product, and maintaining that product is not the business most people are in.

The reasonable position is not to avoid AI, it is to put it on top of something that already handles the parts above. Build your application layer with an assistant, use the MCP server to query and build against your own data, and let the platform carry ingestion, storage, permissions and security underneath.

Troubleshooting

The Atrium app will not connect

  • Wrong region: confirm the region selected in the app matches your TagoIO account region, US or Europe.
  • Wrong token format: the Service Authorization must be created with the TagoIO TiP format. Check for the purple TagoIO TiP label in the authorization list.
  • Expired profile token: if the profile token has passed its expiration, generate a new one and load it into the gateway again.

Devices created but no data

  • Sensor has not transmitted: Last push reads never until the next wireless transmission. Press the reset button or use a magnet on the reed switch.
  • Device not selected: confirm the sensor is selected on the app’s Devices page and shows a valid state with a green icon.
  • Check the Live Inspector: open the Live Inspector while the sensor publishes to see the requests reaching TagoIO.

Settings change stays pending

  • Waiting on the sensor: the Atrium pushes settings roughly once an hour and the sensor has to wake up and request them. Trigger a sensor reset to force it.
  • Schema not published: check the Configuration page in the Atrium app. Without a published schema there is no settings dashboard to write to.
  • Value out of range: the schema defines valid values per setting. A rejected value stays pending rather than applying.

The TagoIO app for Atrium was built by the team at National Control Devices.