vibration sensor with fieldbus or PC connection (no vendor cloud lock-in)

I have motors and pumps we'd like to start gathering general vibration data on for use in predictive maintenance. in the ~ 1000rpm, 10kW class.

I'm looking for a digital-output vibration sensor that I can read from in "some documented way from" a ~ linux PC:
  • Must:
    • Fannout/connection: no proprietary edge "gateway" device required, eg a full PC or PLC in it's own right that would require a cloud connection for vendor updates or data access. Implies data format must be documented (any of open/standard, publicly documented, under NDA, etc)
    • data: read at reasonable latency (1sec~10sec): the standard "point values" of velocity & acceleration peak & RMS in order to provide a "live view" dashboad
  • Strongly preferred:
    • fannout: no "bus masters" required beyond standard PC interfaces such as rs485, TCP/IP, Bluetooth, etc.
    • data: ability to dump on-demand / infrequently the full waveform (pre FFT) or spectrum (either works) for more detailed analysis. This may interrupt sensor operation to perform the download; may take 10's of seconds to download.

As an example, take the IFM VVB001. This meets all those requirements EXCEPT requires an IOLink master, adding about $200 to each sensor all-in and complicating the wiring/install and failure points. But it does provide access to the point values at ~100ms cadence or faster, as well as supporting dumping a couple seconds of waveform for deeper analysis over the same bus.

Our edge ~gateways are linux-based PC's, used to talking "open" protocols to devices over protocols that don't require "vendor gateways" to gather and interpret the results. On the simple end that's just modbus/rs485/various protocols over TCP/IP. On the slightly more complex end that's SAE protocols over CAN, or custom (but documented) protocols over Bluetooth.

I am not finding anything better than the IFM IOLink device above. Most seem to require a full proprietary path to the cloud, which doesn't fit for how we need to use & access the data in the future.

Thank you!
the list price of the VVB001 is ~$450 btw, which once you include an IOLink master (amortized over the number of ports it supports), you're at ~~$600+ per sensor.

So anything in the $1000/sensor "up to the standard interface" is fair game - eg up to the point I can read from it with rs485, TCP/IP etc.
<Sorry for the multiple posted, i keep missing the edit window>

I should add about the wireless solutions such as Bently Ranger Pro - i have a small amount of knowledge of whats out there (still pretty limited) but it seems that such solutions are not that conducive to continuous readings (stated battery life implies readings on minute(s) cadence, not seconds).

A wireless network like wireless HART would (I think, I haven't deployed one) work just fine - except I would really want the sensors powered - it's just too expensive to have to replace the batteries once, let alone multiple, times per year. But the wireless comms is just fine, and assuming a gateway can talk to 10+ wireless devices (and download raw waveform as in the OP), it would probably justify the modem.
Wired HART is WAY too slow: 2-3 updates per second.

I don’t know what the wireless HART update rate is, but
- it requires a proprietary gateway just to get data into Modbus server registers.
- battery life/replacement is a serious issue

Take HART off the table.
Ironically, I got an email offering “Master the mystery of vibration application engineering“ today.

I have no idea about the Link’s content, but its teaser matches your interests.
Well, i entered fake info and downloaded the pdf. It's a few pages, mentions the difference between raw waveform and the basic point values (RMS accel, vel, etc). It's a trivial whitepaper sales article for a company called IMI; incidentally they seem focused on 4-20ma outputs. Not exactly informative, or relevant to my question, but thanks for the thought.