BRENNEN

Industries / Industrial equipment and services / Industrial services / Vibration analysis

Wireless monitoring

Wireless monitoring leaves sensors on the machine that send their readings by radio or over a network, so that nobody has to walk to the machine to measure it.

4

Milestones achieved, from 2007 to 2017.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 2007

    Vibration over mesh radio

    Emerson Process Management introduced the CSI 9420 Wireless Vibration Transmitter at its 2007 Emerson Exchange with a 900 MHz radio, and in 2008 added a version for WirelessHART, the 2.4 GHz mesh network standard (see IoT and industrial automation). In a mesh network each device relays the others' messages to a gateway. Emerson's 2013 data sheet calls the CSI 9420 the first device to send full vibration data over a self-organizing wireless network, including overall levels and readings from PeakVue, Emerson's own measure of impacts.

    automationworld.comemerson.com

    Credit

    Emerson Process Management

  2. 2012

    Certified to an insurer's rules

    An online system records a machine's vibration continuously and sends it to analysts far away. In 2012 Germanischer Lloyd, a certification body, certified WindSafe, a monitoring system for wind turbines made by DMT GmbH & Co KG of Essen, Germany, against specifications set by the insurer Allianz, according to DMT's announcement as printed by the trade site Offshore Wind. DMT says an earlier version of the system was in 2003 the first online system of its type to gain AZT approval, and that the concept has since been accredited by all the major German insurance carriers.

    offshorewind.biz

    Credit

    DMT GmbH & Co KG; Germanischer Lloyd; Allianz

  3. 2013

    Trains power their own sensors

    Perpetuum, spun out of the University of Southampton in 2004, makes wireless sensors powered by a harvester in which the train's vibration moves a magnet across a fixed coil. In June 2013 the Institution of Mechanical Engineers reported that Southeastern had become the first UK rail company to buy such self-powered sensors for its fleet, to watch bearing and wheel wear on all 618 of its cars after a year of trials. Southampton's 2021 impact case study says Perpetuum's systems run on fleets in nine countries, and that Hitachi Rail bought Perpetuum in August 2020.

    imeche.orgresults2021.ref.ac.uk

    Credit

    Perpetuum; University of Southampton; Southeastern

  4. 2017

    Radio measured inside plants

    Metal machines and walls reflect radio waves, so a signal reaches its receiver by many paths at slightly different times. The US National Institute of Standards and Technology (NIST) measured this in a car transmission factory, a machine shop and a steam plant, and published the results, raw data and code in January 2017. It found average delay spreads, the smear in arrival times, below 500 nanoseconds, which it said may not matter to a 256 kbps protocol but could severely limit much faster ones, and wrote that existing technologies must be tested in such places.

    LessonEchoes inside a plant do not doom wireless sensors, but each radio has to be tested where it will work.

    nvlpubs.nist.gov

    Credit

    Richard Candell, Kate A. Remley and colleagues, NIST

Watch for

  1. Ten years on one battery

    TDK says its SensEI edgeRX Pro sensor node, launched on 19 August 2026, combines vibration, sound, magnetic, temperature and rotation sensing with AI processing on the device, and gives up to 10 years of battery life for long-term monitoring; on wired USB power it samples faster and runs more robust AI models.

    Achieved whena user or an outside laboratory publishes the battery life and the faults detected by installed units, with the settings they ran at.

    automation.com

    Claimed by

    TDK, August 2026

Open

  1. 01

    What snapshot sensors miss

    A replay of public run-to-failure bearing records at the settings that battery-powered sensors use: one short capture every hour, every 12 hours or once a day, with the frequency band cut to 1, 4 or 10 kHz. For each setting, measure how much later the fault shows than in the full recording.

    The brief
    Credit · unclaimed
  2. 02

    Raw waveforms in one format

    An open exporter that pulls the raw recordings and their settings (sample rate, axis, units and time) from the cloud services of several brands of wireless vibration sensor, through their published interfaces, and writes them into one documented open file format, so that a plant can analyse its own data or change supplier without losing its history.

    The brief
    Credit · unclaimed

Open 01 · the brief

What snapshot sensors miss

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A replay of public run-to-failure bearing records at the settings that battery-powered sensors use: one short capture every hour, every 12 hours or once a day, with the frequency band cut to 1, 4 or 10 kHz. For each setting, measure how much later the fault shows than in the full recording.

Why it is still open

Murata, a maker of wireless vibration sensors, writes that a short sampling window may miss the impacts of a fault on a slow machine, and that stretching the window while lowering the sampling rate may filter out the high frequencies that early bearing damage produces. The nearest study, by Hyung Jun Park, Nam Ho Kim and Joo-Ho Choi (Sensors, 2022), kept only every second, fourth or eighth measurement of three bearing run-to-failure records and scored how well the remaining life was then predicted; it did not shorten the captures, cut the frequency band or report warning time lost. A Crossref search in November 2026 found no study that does.

murata.comweb.mae.ufl.edu

Built with what exists

NASA's Prognostics Data Repository, which hosts the University of Cincinnati IMS bearing data and the FEMTO-ST bearing data; Python with NumPy and SciPy; a simple detector based on the RMS (root mean square) level and the envelope of the signal as the baseline.

nasa.gov

  1. First step

    Download the IMS bearing data, run the baseline detector on the full record, then rerun it on the record thinned to one capture per hour, per 12 hours and per day.

  2. Done looks like

    A table and code giving, for at least 15 bearings run to failure, the warning time lost at each setting compared with the full recording, at a fixed false-alarm rate.

Open 02 · the brief

Raw waveforms in one format

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open exporter that pulls the raw recordings and their settings (sample rate, axis, units and time) from the cloud services of several brands of wireless vibration sensor, through their published interfaces, and writes them into one documented open file format, so that a plant can analyse its own data or change supplier without losing its history.

Why it is still open

When Amazon Web Services closed its Monitron sensor service to new customers on 31 October 2024, its notice pointed users to partners' alternatives and did not describe how to move recorded data to them. Each maker offers its own route to the raw signal: ERBESSD Instruments, for one, says the full time waveform and the spectrum "are not available over OPC UA", the common industrial data standard, through its gateway, and points users to its own MQTT feed. GitHub searches in November 2026 for tools that export waveforms from wireless vibration sensors returned nothing.

aws.amazon.comerbessd-instruments.com

Built with what exists

The sensor makers' published REST interfaces; Python; HDF5 files written with h5py; a small shaker or test rig that gives every sensor the same known signal so that exports can be compared.

  1. First step

    Buy one sensor from each of two brands that publish an interface, record the same test signal on both, and write the format description from the fields each returns.

  2. Done looks like

    Exports from at least three brands, each checked by recomputing the maker's own overall vibration value from the exported waveform to within 2%, with the format description and code under an open license.