Monitoring Slow-Speed Bearings With Ultrasound

An ultrasound program associated with the critical coal-handling conveyor system at Dakota Gasification’s Great Plains Synfuels Plant is proving that catastrophic slow-speed bearing failures can be avoided.

This maintenance professional’s account of his site’s experience with ultrasound technology lays out details that can help others.

By Ron Tangen, CMRP, Dakota Gasification Co.

Dakota Gasification Co. (DGC), a for-profit subsidiary of Basin Electric Power Cooperative, Bismarck, ND, owns and operates the Great Plains Synfuels Plant, a coal gasification complex near Beulah, ND. The plant produces pipeline-quality synthetic natural gas and related products.

It’s a given that coal gasification requires a conveyor system to transport coal. As others may have, Dakota Gasification has experienced its share of frustration with conveyor-bearing performance. An evaluation of the problem suggested that it was impossible to eliminate failures. With this in mind, we set a goal to simply minimize the occurrence of catastrophic slow-speed bearing (SSB) failures. This meant finding a technology that allowed us to accurately monitor bearing condition and maximize service life, as well as alert us when SSBs reached the end of their life, so we could remove them from the system.

DGC’s Coal Handling Operations group recognized the SSB issue several years earlier and, as a result, established a policy of weekly walk-downs of the conveyor system. During these walk-downs, technicians would perform a physical-senses evaluation of each of the main pulley bearings. At some point, a handheld infrared pyrometer was added to their toolbox to help identify failing bearings. Even though it operated close to the bottom of the P-F (potential failure) Curve, this strategy was better than no program, and some successes were realized. Still, we experienced two to four failures per year in a 400-bearing system.

Predictive maintenance (PdM) strategies operating this close to equipment failure inherently have some shortcomings—even when looking at SSBs. With our coal conveyors, the short window for failure detection is a key disadvantage. Bearings often don’t enter the “visual senses” portion of the failure curve until just days or hours prior to catastrophic failure. Operations personnel might perform inspections and report no problems, only to have a failure in the same week. Another disadvantage is the increased cost of catastrophic failure over planned replacement: safety aspects, manpower, production loss, and collateral damage costs far exceed bearing cost. Such issues helped justify transition to an ultrasound-based program.

Leveraging ultrasound for SSB monitoring, we’ve been able to move from one end of the P-F curve to the other—from reactive/failure to predictive maintenance. We would have considered ultrasound a success if it had provided even a few days or weeks of early detection. With this technology, however, we can now produce a report that identifies bearings at risk of failure as much as 12 months out.

Every five weeks, technicians on nine different routes collect data from the main pulley bearings in the coal-handling conveyor and then download it into ultrasound software for archiving, trending, and analysis.

Background

DGC had been using ultrasound for about 15 years in other applications. As we considered our SSB failure problem, the structure-borne nature of bearing applications led us to look into acoustical ultrasound. After trial testing the technology on our conveyor bearings, we conducted a hands-on field demonstration with our operations superintendent. This, we hoped, would be the key to securing high-level management buy-in that was so crucial to the program’s success.

Once the operations superintendent donned the ultrasound equipment and was able to “hear” a couple of bearings for himself, he began working his way down the conveyor galley to listen to the rest. Upon reaching the end of the galley, he directed us to implement the technology at DGC as “soon as possible.”

The Great Plains Synfuels Plant, a commercial-scale coal gasification complex near Beulah, ND, produces pipeline-quality synthetic natural gas and related products.

Program details

Since our coal-handling system includes nine major buildings, we created nine ultrasound routes. Technicians from the operations group perform data collection every five weeks. There’s no magic to this interval: We simply wanted a short-enough frequency that would provide two or three readings within a failure cycle. Also, as operations personnel were accustomed to weekly conveyor-system walk-downs, a five-week interval would help gain program buy-in through reduced manpower and avoidance of the “monthly route” syndrome.

Data from routes (collected only on main pulley bearings, not idler bearings) are downloaded into the ultrasound software for archiving, trending, and analysis. Two key elements of the data are the recorded ultrasound wave file and bearing decibel (dB) value. The wave file contains several seconds of digitally recorded sound; the dB value represents its intensity. Without these elements, analysis wouldn’t be possible.

Inner-race spalling in a failed bearing.

Analysis process

Several features of our ultrasound tools have been particularly helpful.

Our ultraprobe data collectors allow us to digitally record several seconds of wave-file (sound) data. In addition, a high level of ultraprobe sensitivity allows us to detect small failure modes, i.e., we can track a bearing’s health through its entire life. For example, many bearings are classified as having a “zero-dB fault”—meaning a cyclical fault inside them can clearly be heard (and seen), while the ultrasound equipment is registering a 0-dB sound level (Fig. 1) Although the bearing reflected in Fig. 1 has a fault, it may be years, or even decades, from failure. Typically, we don’t consider replacing a bearing until the sound level is in the 25-to-30-dB range.

Fig. 1. Zero dB fault.

The ultrasound software lets us replay wave files to interpret sound signatures and to analyze dB-value changes over time. This helps establish trend history and future risk of failure. The software also helps us see the amplitude and pattern of a sound—which is a valuable capability when analyzing bearing health.

One challenge that comes with high sensitivity is the presence of competing (structure-borne) ultrasound sources that might be heard along with the bearing sound signature. Competing sources can come from coal (product) falling onto belts or through metal chutes, gearbox noise, or a nearby bad idler bearing. Listening to the sound of bearings, what you actually hear are impacting and/or white-noise frictional forces.

Impacting is short-duration frictional forces caused by failure modes such as particle contamination, pitting, spalling, fretting, or broken parts, i.e., a cracked race.

White noise is caused by constant frictional forces. Even “good” bearings have some white noise, i.e., all have some level of constant frictional force acting on them as they rotate. Elevated levels could be a result of new/tight bearings, high dynamic loading, inadequate lubrication, or misalignment.

Fig. 2. Classification is ‘OK.’

Fig. 3. Classification is ‘Moderate Impacting.’

Fig. 4. Classification is ‘Moderate White Noise.’

The recorded wave file is transferred to the software for analysis. This is where the bearing’s signature is established. Each wave file has a unique signature relating to the bearing condition. To simplify things, we evaluate this signature in terms of its impacting and white noise. While signatures may clearly be dominant in one way, a bearing typically reflects a mix of impacting and white noise. Determining how much and what level of each helps establish the bearing’s overall health (Figs. 2, 3, 4). This signature and the dB value of the bearing provide insight into the level of deterioration.

Since our bearings operate at speeds between 70 and 80 rpm, FFT (Fast Fourier Transform) analysis isn’t feasible. As a result, all analysis on these SSBs is done through our ultrasound equipment’s time series analysis software and interpretation of trended decibel values.

Analyzing SSBs with ultrasound allows us to accurately monitor bearing health over the component’s entire life. This, in turn, allows time to recognize a bearing that’s nearing the end of its useful life or catch one that’s moving into catastrophic failure.

Decibel values charted within the ultrasound software provide important information regarding a bearing’s historical and current conditions—and can even be used to gain insight into the future failure risk. Since individual data points seldom plot in a straight line, “normalizing” the information by drawing a straight line through it provides a more linear perspective in establishing a bearing’s historical and projected trends. The line’s slope represents the component’s rate of failure:

• A slowly rising slope indicates a bearing that’s deteriorating/failing slowly.
• A rapidly rising slope indicates a bearing that’s deteriorating/failing rapidly.

Extending the historical trend line into the future allows users to anticipate future bearing health and risk of failure (Fig. 5). The intersection of this line with the established target decibel level for bearing replacements (30 dB in Fig. 5) can provide good clues as to remaining operating life and the timeframe available for a planned replacement.

Decibel values charted within the ultrasound software provide important information regarding a bearing’s historical and current conditions. Extending the historical trend line into the future allows users to anticipate future bearing health and risk of failure. The line’s slope represents the component’s rate of failure: The intersection of this line with the established target decibel level for bearing replacements (30 dB) can provide good clues as to remaining operating life and the timeframe available for a planned replacement. Typically, DGC doesn’t consider replacing a bearing until the sound level is in the 25-to-30-dB range.

Based on analysis of the sound signature and recorded dB level, the bearing can be classified, or graded, regarding its condition. To maintain consistency in our grading process, we developed a failure-classification chart. This chart has been periodically updated over the years to align with the actual deterioration in bearings we’ve removed from the field. Moreover, it only applies to DGC’s conveyor bearings—mostly 3- to 4-in.-dia. spherical roller designs from one manufacturer. The bearing-classification information is also documented in the ultrasound software for purposes of reference and reporting.

Rolling-element spalling.

Value-added insight

Our ultrasound program has provided some value-added insights beyond what we had expected from the technology. Examples include helping track new bearing “wear-in.” We originally anticipated that when a “bad” bearing was replaced, the decibel reading on the “good” bearing would be significantly lower, i.e., close to a normal operational baseline. In fact, after pulling 25- to 30-dB impacting bearings from the conveyor system, we’ve discovered 25- to 30-dB white-noise signatures on their replacements.

Although manufacturers suggest a bearing will wear in over a few days or weeks, ultrasound has shown us a few months to years may be more likely. This isn’t to say OEMs are wrong, but rather to point out ultrasound’s high degree of sensitivity to frictional activity within bearings.

This graph represents what the author has come to identify as a “typical” ultrasound bearing life cycle. It shows an upward slope as the bearing goes into “failure,” and a downward sloping bearing wear-in period as it returns to baseline. In this example, he manually inserted two ‘markers’ into the trend data: one to represent when the bearing was replaced, and another to establish the anticipated new baseline

DGC’s ultrasound experience also suggests correctly lubricated SSBs can survive failure modes that quickly fail bearings in high-speed applications. Some of our SSBs have operated with a cracked race for several years. One that we removed and analyzed had significant spalling on the inner race and rolling elements; its outer race had a chip, crack, and break.

A failed bearing with an outer-race crack, chip, and break.

Ultrasound has also provided insight on the ability of SSBs to recover from certain levels of failure modes. In one case, we saw a bearing decibel trend move, over a 3-yr. period, into accelerated failure and recovery four times. At first glance, the random and inconsistent trend seemed to point to a data-collection problem. A closer look, however, indicated the bearing actually went through insipient failure and recovery. In the end, as we normalized the data, we were still able to establish an overall failure rate.

Despite some challenges since its implementation, no one has said the slow-speed bearing ultrasound program isn’t worth the effort or is not delivering value for the site’s owner/operator Dakota Gasification Co.

Where we are

Although we at DGC still have much to learn about ultrasound, I’m pleased with the progress of our SSB monitoring program. Despite some challenging situations, no one has said the program isn’t worth the effort or not delivering value to our company. Just as important is the fact that managers are now requesting quarterly predictions of high-risk bearings rather than a single annual report.

To date, we don’t have a calculated statistical level of improvement for the program, but, the number of visibly damaged components in our bearing showcase continues to grow every year. Each of these bearings represents a catastrophic failure that was avoided—and dollars added to DGC’s bottom line. MT

Ron Tangen is a maintenance-engineering specialist for Dakota Gasification Co. (dakotagas.com), a for-profit subsidiary of Basin Electric Power Cooperative (basinelectric.com). A Certified Maintenance and Reliability Professional (CMRP), Tangen is based at the company’s Great Plains Synfuels Plant, a commercial-scale coal gasification complex near Beulah, ND. For more information on this article, email him at rtangen@bepc.com.