David Clark, Condition Monitoring Expert, www.bachmann.info
Wind-farm operations and maintenance crews often fixate on a condition monitoring system’s hardware and software specifications, mostly because the specs are somewhat easy to quantify. But these people should be equally interested in who will run and configure that system. Without a proper configuration, you really do not have a system – it’s just hardware. Although hardware is usually what is on a request-for-proposal, there are two other equally important but difficult elements to put on a specification sheet. Car races, like the Indy 500, provide a useful analogy. In racing terms:
The car ≡ CMS (Condition Monitoring System) hardware
Crew chief ≡ Person who configures the system installation and measurements
Driver ≡ Analyst or data interpreter
Sponsor ≡ End user or customer
For this analogy, the car is the CMS system. It is an assembly of components, the sensors, cables, an acquisition device, software, and other items that together make a CMS system. These parts are usually found on an RFQ or proposal. A few boxes of parts can meet some specifications. But that part collection is no guarantee a CMS will effectively function, just as a car will not run without a driver.
It is surprising there is such a huge variance in CMS parts for wind. This is not true for other vertical markets and applications. CMS produces standard vibration measurements regardless of manufacturer. Other applications do not have the variability that wind does in the CMS offerings. Some are much better than others due to parts selection and application knowledge. You do not see that in standard industrial applications.
However, the observation of a large variance in hardware does affect the data acquired. Or rather, the lack of quality parts affects the system’s effectiveness.
A few signs you have a bad “car” include a vibration sensor that reports nothing while a tech can tell something is wrong, or a system that provides no alarms, or an analyst that can’t interpret the data.
To fix the problem, specify the best parts from the beginning. Use the do-it-once, do-it-right approach. An insurance specification is helpful, but it is no guarantee that you will have a working, effective CMS. For example, I still see systems that do not use sensor cables that are lubrication-rated. Oil can soften some insulation so they peel away from the conductors. But if you like paying for and changing sensor cables in every tower in five years, disregard this suggestion.
The crew chief
The crew chief adjusts and manipulates the CMS system for results. Adjusting it for results and a required performance comes with experience. A CMS system incorrectly tuned will never produce required results without a crew chief to properly tune the system. By tuning, I mean selecting measurement parameters, setting alarms, providing communications and data storage, correct sensor placement, and more. This role is inclusive of application knowledge as well as vibration analysis. Communication knowledge and wind-farm operation also play roles.
How to know if you have a bad crew chief? When the CMS system is constantly in alarm, it sounds false alarms, misses detections, and generally does not give the required results. Time and again I hear that a CMS system missed a failure, has been in various states of alarm for months, or needs nine months to determine where to set the alarm. These events indicate a bad crew chief.
How do you find a good crew chief? To start, look at the total number of installs. This is similar to looking at championships for a racing team. Few in the wind industry have this ability.
The driver for this analogy is the vibration analyst. This person is responsible for driving the CMS system to get results from the data coming off of the system. Without a good “car” or crew chief, a driver will never be successful. So without good hardware and proper alarming, an analyst is hamstrung to getting good results.
Over the last 15 years, I have used six different CMS systems outside the wind industry, and they are all similar in function, measurements produced, and results. I wish the same held true in wind, but not all CMS systems are the same.
I consistently still see the wrong sensors, cables, measurements, sensor locations, and 10 other basics absolutely missed. It’s frustrating to hear CMS stereotyped as not being effective because of these rookie mistakes.
Part of the driver’s responsibility is knowing the track. That is, the analyst must understand the application. I recall the first time inside a hydro dam taking vibration data and seeing a pump that rattled every bearing on the impeller shaft. All vibrations were exceedingly elevated across the spectrum. I thought the pump must be nearly destroyed…and then the cavitation stopped. All the vibration disappeared. That was not an optimal operating condition for the pump, but it illustrates that not all vibration is indicative of a failing component. Not understanding that particular application could have led an analyst astray.
Two examples on wind farms recently came to light. One site had all sensor locations on all 75 towers in alarm. Wow. Not only were they false alarms (it should seem grossly improbable that every condition would be in alarm at every location and on every tower) it illustrates my point. In this case, the driver clearly did not understand the track.
In a second example, someone was explaining that CMS is known for missing detections or detecting issues too early, or both, that is, non-alarming and false alarming. The comment is similar to saying that a particular car is known for running into things and going over curbs. Who’s driving the car?
How to know you have a bad driver. Ask: Are you getting useful results? If not then you likely have a bad driver. It’s that simple.
How to find a good driver?
One with a vibration certification is a good start but that is not an assurance of one’s ability for this application. I know of a few level 2 and 3 certified vibration analysts who have zero wind application knowledge and some have never been up a tower. The problem is that less than 40% of most certification courses focus on wind-industry issues because most certifications focus on portable analysis which is not used in wind, for safety reasons. For instance, insurance usually precludes anyone from being in a running nacelle. Secondly, a portable device means being in a tower when the wind blows in comparable manner to the last measurement. Still, the certification is a good start.
Application knowledge is an indicator of a good driver. Sadly, there is little resource here in the U.S. for wind specific analysts. So a certified person with application knowledge puts you a step closer to getting good data. Again, it is no guarantee.
I see end users as sponsors because they pay for the results. Sponsors would like to see their car win. They want to see a CMS provide accurate assessments of health – reports they can read and understand.
So how do you determine if a problem is with the car, driver, or crew chief? If there are missed calls, false alarms, or general ineffectiveness of CMS you know then that the issue is within the driver (analyst) or crew chief (measurement and software set-up). A large assumption here is that the system is installed correctly with application knowledge. There is also an assumption that the hardware is capable of acquiring good data. Once that is determined, it falls on the shoulders of the software. I had a recent conversation with a person who had a certification, application knowledge, but could not figure out why they were missing a basic measurement that turned catastrophic. The answer was simple: The sensor was in the wrong location and would never give good data to anyone, (crew chief) or analyst (driver). Time and again rookie mistakes come from the crew chief or driver, but the mistakes are made to sound as if they are inherent shortcomings of condition monitoring.
The mistakes fall into several categories:
• False alarms for months on end
• Incorrect alarms for months. False alarms are generated when there is no component failure. Incorrect alarms are the absence of alarms when there is component failure. Both are guesses based upon initial guesses. Neither are acceptable and both completely avoidable.
• Cannot detect issues in certain areas of the drive train
• Short lead times to prediction
• The feeling that someone is learning at the sponsor’s expense
• Reports that make no sense
A quality and effective CMS program runs on the backs of people. Making sure you have the right people and the right tools will help garner the right results. Remember, a CMS program is more than the parts. It takes someone to tune the system correctly, drive the system correctly, and get results one would expect from a predictive maintenance tool. WPE