The Squeaky Wheel Gets the Repair

In our chat session last month, one of the topics that came up really caught my attention — probably because of my work in the industrial environment and as a railroad maintenance-of-way worker. It concerned preventative maintenance on rotating equipment.

This generally means detecting problems in motors. The problem areas are primarily the bearings and, secondarily, motor winding overheating. If you monitor the motor temperature — and probably motor winding current, voltage, and phase angle while you're at it — you can detect problems well before they become catastrophic. If you listen to the bearings with a MEMS (micro-electro-mechanical system) sensor, you will be able to detect vibration that portends bearing failure.

In consideration of just these few sensor and data acquisition functions, it's obvious there is a lot of functionality here. It should be combined and integrated. An integrated analog approach is needed that will accept these inputs:

  • Voltage, phase A
  • Voltage, phase B
  • Voltage, phase C
  • Current, phase A
  • Current, phase B
  • Current, phase C
  • Multiple temperature sensors — assume a minimum of 3
  • Multiple MEMS vibration sensors — assume a minimum of 2

But there is another important function we've left out. What do you do with the information from all these sensors? Certainly you will digitize and multiplex it — but then what? How do you get it to the PLC (programmable logic controller) that is running the motor? You could run a data cable next to the three-phase power cables that are already routed between the motor and the PLC. Even though the wire ways or conduit are present, clearly that method is fraught with peril.

In the case of monitoring bearings in a gearbox or on a conveyer belt, there probably is no easy method for data cable routing. Running such cable, materials and labor become a significant cost in factory installations. And conveyer belts are notorious for static electrical charge accumulation (think of a Van de Graaff generator), which can damage sensitive circuitry.

A sensible approach to work around these problems would be to use a low-power RF data transfer link such as Bluetooth. Not only will this approach help with the problems described above, but it should be possible to integrate the functionality with the sensor circuitry.

Besides use in the industrial environment, this same integrated analog functionality can be used to monitor railroad rolling stock. In some versions of this, MEMS and temperature sensors could be installed on the trucks of all rolling stock to send information to the engineer or to wayside monitoring equipment. In other versions, MEMS sensors could be installed on the rail web to listen for wheels with flat spots. Flat spots develop when emergency braking occurs. The flat spots pound away at the railhead, damaging it and further degrading the wheels.

The railroads are already using some technology to detect problems:

  • BNSF has provided some information on technologies they are pursuing.
  • LBFoster/Salient Systems is building some interesting trackside products that can detect bad bearings, flat spots, and hunting trucks.
  • Another member of the UBM family, Design News, published this article last November that describes in more detail how bad bearings are detected.

I may build up a strain gauge bridge circuit + amplifier and glue the strain gauge to the web on a section of rail on some nearby tracks and listen to the amplifier output with headphones. It should be interesting to see what it sounds like.

The task now is to move the existing technology onto an IC so that power consumption and cost can be reduced. All of the preventative maintenance methods would benefit. And there should be plenty of opportunity to sell the technology — the market looks to be huge.

Have you had any experience with this sort of preventive maintenance technology?

Related posts:

15 comments on “The Squeaky Wheel Gets the Repair

  1. CarlWH
    June 27, 2013

    @Brad – nice article on a massive field.


    Condition monitoring is very prevalent in the oil and gas markets, where you have the additional complexity of ATEX requirements. But if you look at the cost of being able to do preventative maintenance, rather than waiting for a failure (potentailly catostrophic), then it becomes a no brainer.

    Other Techniques include, thermal monitoring using thermopiles, (not a detailed image so more private) or CMOS sensor arrays. As well as machines these can be applied to the well being/assisted living field also. So the idea is sensors around the house detect behaviour, (e.g. get up at a certain time) and if you get events outside this then an alarm is raised. The algorithm will be adaptive and learning. This may be considered a big brother approach, but what it allows is for people to reside in their own homes rather than care homes, and it gives relatives peace of mind.

    It will be interesting to see the privacy debate, although recent events suggest big brother is already watching you, (not a debate for here).

    As we move to the IoT you will see more and more sensor arrays being deployed. How you power these, (energy harvesting or low power design) and how they communicate is one thing, but what you do with the data is another.

    All these applications require analogue skills

  2. Steve Taranovich
    June 27, 2013

    I agree with CarlWH—

    Wireless sensor networks can take advantage of the mechanical vibration energy as well as thermal energy from heat in the motor and convert this energy to drive the wireless sensor networks in some cases without a bat­tery (energy harvesting). 868-MHz wireless is a band that is a possibility here since it is used in thermostats, burglar alarm systems and fire systems.

  3. Scott Elder
    June 27, 2013

    @Brad, I would think that a Wifi connection would be valuable as well to report the status over the internet.  The manufacturer could maintain a huge profile database for failures and map it against the realtime data coming off the local unit.  Statistical prediction is probably more valuable than a determinstic design prediction if the worldwide failure logs had a reasonable amount of data collected for a particular production model.

    I think this is an IoT device.  

  4. Bob.Groh
    June 28, 2013

    Lots has been done and is being done in this field over the last 30 or 40 years (probably more).  Monitoring motor bearings with vibration devices is done routinely in addition to temperature.  Many motor control devices (e.g. variable speed drives) already provide data on motor currents etc.

    WRT railroad applications, in my prior life at Harmon Industries (a RR supplier – which later became a part of GE Transportation Systems), we developed and produced trackside bearing thermal detectors (HBD or Hot Box Detectors) which examined the trains as they rolled by (at speed) and analyzed the data to detect (and announce via voice msgs on the railroad's radio system) and alarm actual or pending bad bearings ('hot boxes' in the RR venacular).

    The Harmon engineering staff also played around with acoutiscal 'bad bearing' detection by 'listening' to the cars as they went into the hump yard.  Experienced people could detect a bad bearing by the sound but trying to figure out a way for a computer to do it?  Difficult.

    Back to the sensor question. Lots of work on 'self powered' sensor packages today which in combination with some form of RF to transfer the data is a very interesting idea (well, I'm sure several firms on working it right now). 

    As you might guess, I love the design part!  Gets me going…again.



  5. PCR
    June 30, 2013

    It's a great thought Steve, so it will be more useful to use energy from the wastage for the sensors. 

  6. Netcrawl
    June 30, 2013

    @Scott you're probably right, we live in a world filled with sensors, from workplace to public facilities, or even our home, sensors are getting smaller but more powerful and much sophisticated. Unfortunately, the cost would be a different story, the cost of installing sensors has not kept pace, the cost of placing those wires to carry power and data would dwarfs the cost of the sensor. 



  7. Steve Taranovich
    June 30, 2013

    Right Ranasinghe, thermoelectric generators (TEGs) and piezoelectric generators are ubiquitous and low cost nowadays and are provided by a multitude of suppliers. All you need in addition are the power management path to efficiently convert sensor output to usable voltage/current (many of these devices have this as built-in modules too), maybe a battery for storage (optional) and the low power wireless node device.

  8. Brad Albing
    June 30, 2013

    Thanks Bob – the RR technology is slowly creeping upward. Interesting to see how it's evloving. I remember when it was all just batteries and relays for track occupency circuits, then track coding for in-cab signals….

    Just a ton more sophisticated stuff now.

  9. Brad Albing
    June 30, 2013

    @Scott – I have to assume that there are lots of companies working on this now. Freight traffic is increasing every year, so there is money to be made by developing fancier methods to track freight shipments and keep the freight safe via monitors for the rolling stock and the track itself.

  10. BillWM
    July 1, 2013

    Just about every place that does large motors does this —  The algorithyms are starting to get fairly sophisticated for failure prediction in some cases.

  11. RedDerek
    July 3, 2013

    Thing about all the pieces that have been discussed on this forum.

    1. Energy harvesting methods – TEGs, piezeo, etc

    2. Low VT MOSFETs

    3. Low power SDRs (software defined radios)

    4. Various instrumentation devices

    5. Super capacitors

    This all adds up to a device that can use temperature and/vibration for energy generation to store charege in a super cap. Then grab a reading and broadcast it to a local receiver. The local receiver collects data from 100 to 1000's transmitters and bundles it to broadcast data at a higher power to the next level device that then transmits multiple local receiver data to a central processor for the final computation and data manipulation. Call this web data integration via wireless and wired data transmissions.

    Technology is there. Now to just fund the development of everything to be merged into a system. I am sure this is being worked on by someone out there with money for big development projects.

  12. WKetel
    July 4, 2013

    Instrument amplifier ICs have been around for years, and some are quite cheap. And at least one model will also provide the regulated bridge excitation. So that part is easy. Getting the signal away from the tracks is a bit more of a challenge, but wires are simple and have the advantage that you can also run power to the sensor amplifier. What you do with the signal is another thing. Many early packages simply had a meter with some arbitrary units on the scale. But a set of narrow band sensors can be very useful if you already know the frequencies of interest.

    But for the trai tracks a better choice might be to stick a piezo-electric sensor to the side of the rail, and a similar sensor to the bottom, and then compare the signals.

  13. Brad Albing
    July 4, 2013

    I know it's being worked on. I'll try to get some more specific commentary from some of the design engineers from some of the companies that I know are working on this stuff. So stand by for that.

  14. Brad Albing
    July 4, 2013

    That dual piezo approach is good. That should provide sufficient info from the rail in 2 different axes.

    But I still think any design that can do away with the wires altogether is good – especially in railroad apps. The wire is expensive. The labor to install the wire is expensive. And due to the environment, damage is quite likely (weather, pounding from the rolling stock, vandalism). So I like wireless for this along with suitable energy harvesting techniques.

  15. Brad Albing
    July 6, 2013

    I'll see if I can get a comment from Rockwell.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.