Predictive Maintenance Integration

James Sullivan, The New York Times

The newspaper industry is unique not only in its demanding production schedules but in its limited availability for planned maintenance. With the need for equipment reliability, it is more important than ever to have an effective predictive maintenance program that can incorporate several of the advanced maintenance technologies.

For more than 100 years, the newspaper industry has been using the maintenance theory of run-to-failure (RTF). Within the last couple of decades, preventive maintenance has become a large part of the maintenance thinking.

It is only in the last decade that newspapers have begun to embrace the whole concept of predictive maintenance (PdM). By incorporating several of the advanced maintenance technologies such as vibration analysis and monitoring, thermography, tribology and ultrasound, a facility can reap the full benefits of a successful PdM program.

The New York Times Edison Plant is enjoying the benefits of the successful PdM program and has been able to justify the program’s cost.

Establishing the Program

There are several different PdM technologies that can be incorporated into a program. When choosing technologies, it is important to consider who will be using it as well. An operator or a maintenance person on the floor can use a simple technology such as ultrasound. Ultrasound is quick, easy to use and requires minimal training, but it is not as accurate for true diagnosis and severity.

Whereas vibration sampling can be looked at as having multiple skill levels, a person can collect vibration data with minimal training. However, basic analysis and database manipulation require more rigorous formal training. For true diagnostic ability and complete database set-up, a combination of formal training and years of experience is necessary but can reap tremendous benefit.

Thermography also requires formal training and has a rather large initial start-up cost. Tribology, or oil analysis, can reside at either end of the cost spectrum depending upon the level at which it is deployed. At the high end of the spectrum, it may not be cost-effective to set up your own oil analysis lab and train your personnel for this type of analysis.

When establishing a program, it is important to consider available resources and how to best deploy them. If creating a program from scratch, an estimated time at which solid results will start to be seen must be established. If hiring from within, a great deal of time will be required to train the people not only in the PdM technologies, but also in the true maintenance diagnostic discipline.

Personnel must rely on data, not merely upon what they are told. Personnel must also have an understanding of machinery and know when to use a little of their own diagnostic instinct. There will always be a learning curve, so it is important to be prepared for novice-related mistakes. On the other hand, if immediate results are needed, hiring someone with experience should be considered.

This will bring predictive maintenance ability to the program early on, but that person must be educated about the company’s particular industry and practices. People with this type of experience come with a price. The bottom line is that a timeframe and budget must be established up front.

At The New York Times, a satisfactory level of PdM integration has been found. The general maintenance personnel will use ultrasound and stroboscopes for suspected problems during a production run. The general maintenance personnel will also collect vibration and oil samples on regularly scheduled PdM tours. Predictive maintenance technicians collect diagnostic vibration data and thermographic images.

They then analyze and categorize all findings with a priority rating of one to five; one being an immediate shutdown and five being a distant early warning. All report findings are submitted to maintenance planning for the scheduling of required repairs.

Multidisciplinary Program

The folder is the heart of a printing press. All the pages are married together and are cut and folded together by this unit. The press runs an average of 65,000 copies per hour or approximately 18 complete newspapers per second.

Precision and timing are critical for the tucking blade assembly of the folder. Newspapers must be in the customers’ hands by 6:30 a.m. or else the newspaper is considered no good. With such a demanding print schedule, downtime is forbidden.

Vibration data are sampled on a weekly basis from the press folders during live production runs. Each product is unique in its vibration signature, so all paging is noted and a library of vibration data is filed in accordance with its paging.

For example, a 40-page product on a certain press will have a different signature than a 42-page product on the same press; and different presses with the same exact product will have different vibration signatures from press to press. Similar products are compared whenever possible on a weekly basis. Empty folder readings are gathered every six months to trend “no product” and compare with less variables.

Because the folder tucking blade assembly is a rotating planetary gearbox, data are collected from the center point of rotation from within the folder via internally mounted sensors connected to an external junction box. A monthly oil sample from the tucking blade assembly is taken and analyzed for viscosity metal content and overall lubricant performance.

Dual Technologies Save More Than $2.4 Million

In April 1998, a rise in the spectrum at 139 Hz and 153 Hz alerted The New York Times maintenance staff to a possible problem (Figure 1).

The monthly oil samples also exhibited a problem with a rise in the iron and chromium (Figure 3). This was monitored closely until November 1998 when the vibration and oil data indicated the problem was worsening at an accelerated rate (Figure 2).

Figure 1. Press No. 27
in April 1998
Figure 2. Press No. 27
in November 1998

The folder was then put on a “daily vibration sampling” status. When the oil sample came back, a high level of iron was present. The oil was immediately flushed between production shifts and a vibration sampling was then taken on a “per shift” basis. Two days, or four shifts later, the vibration signature began to clearly exhibit a bearing problem.

Another oil sample was taken to indicate the amount of wear in two days of running. It had been decided that there was a problem, but the next full maintenance window for a complete change out was three days away. Oil change outs were then done on a “per shift” basis and vibration readings were done on a once- if not twice-per-shift basis.

The decision was then made to pull the tucking blade assembly out of the folder. A new spare tucking blade assembly was installed into the press on a scheduled maintenance window. The printing press did not have one minute of unscheduled production downtime.


Figure 3. Oil Trend Chart for Press No. 27 Folder

Upon inspection of the pulled assembly, it was found that a bearing retainer had worn away and an entire shaft was floating within the unit. Had a critical failure occurred, there would have been an estimated $275,000 in damage to the printing press. Not counting availability of some parts, there would have been a minimum of five days lost production time worth an estimated $2.4 million.

The cost of time and material to install a new tucking blade assembly was approximately $62,000. The cost to repair the damaged tucking blade assembly at a leisurely pace was approximately $18,000. The integration of oil analysis and vibration analysis not only alerted the staff to a problem, but it also gave the staff more confidence in running the unit for as long as we did.

Multiple technologies were again used in pinpointing a gear failure in a press unit drive roller train (Figure 4).


Figure 4. Section of a Unit Gear Train

Due to the nature of many different components rotating at similar speeds, it is rather difficult to pinpoint which gear is the actual problem. An audible sound was heard coming from the press. A press color tower is approximately 25 feet high and has an estimated 120 gears. Due to the oscillation of some roller trains and the fact that many of the rollers are hollow drums, it is difficult to pinpoint the precise location from which the noise is being generated.

Ultrasound was the first line of PdM technology to be used. The machinist on the floor was able to narrow the noise to unit No. 6 on the B level. An oil sample was taken. The lab’s analysis yielded a dramatic increase in the lead and iron content.

With this information, an in-depth vibration analysis began. Due to the location of the gears, it was safer and more feasible to collect data from the bearing caps inside the roller section of the press, which mount on the back wall of the gearhouse. All roller speeds and gear tooth counts that were accessible were gathered. Calculations were done for the missing information so that all gears had a tooth count and a speed in hertz.

The readings were collected and analyzed. The source of the primary noise appeared to be the No. 18 cylinder gear in the oscillator drive (Figure 5).


Figure 5. No. 18 Oscillator Gear

There were some high amplitude readings on the No. 17 and No. 19 gears as well but at first, it was suspected to be transmitting over from the No. 18 through the oscillator pinion.

Due to the involvement of the repair, as a precaution, parts were ordered to repair all three gear assemblies and the oscillator pinion. Once the oscillating unit had been disassembled, an inspection of the gears revealed that the No. 18 gear (Figure 6) and the pinion were the most severely damaged, and No. 17 and the No. 19 gears were slightly damaged. All four assemblies were changed out and the unit was once again operating as normal (Figure 7).


Figure 6.
Actual Bad No. 18 Gear


Figure 7. No. 18 Oscillator Gear

Another example of PdM integration was a line clutch on a press lineshaft. A thermographic scan of the entire lineshaft was conducted during a live production run. Scans and reporting for a 10-unit press were completed in two hours. One of the problems discovered was the Press 25-07 Line Clutch with a temperature of 168ºF, approximately 60 percent above normal (Figure 8).


Figure 8. Spectrum, Time Waveform and
Thermographic Scan of Press 25-07 Line Clutch (Before)

Vibration data were collected for only the questionable component, which took two hours for collecting and reporting. The typical time needed to collect and analyze vibration data for an entire lineshaft is approximately 12 hours.

Both the vibration data and the thermographic data complemented one another in confirming a definite problem with the line clutch. The thermographic scan indicated a high operating temperature and the vibration data indicated excessive looseness. A priority 2 was assigned and a repair/replace job was requested for the next available maintenance window in accordance with the priority rating.

The maintenance crew found a worn out bearing housing and worn shifter pins on the line clutch. They replaced the whole assembly with a rebuilt unit. The following day, another thermographic scan was conducted and vibration data were collected to verify the repair was done properly and to establish a new baseline in the database.


Figure 9. Spectrum, Time Waveform and
Thermographic Scan of Press 25-07 Line Clutch (After)

As shown in Figure 9, the repair for the most part was a success with an operating temperature of 105°F and vibration readings that were reduced approximately 90 percent. Upon a closer look at the vibration data, there is an indication of a slight parallel misalignment in the horizontal direction.

Although corrected better than before, further work will be performed with our laser alignment equipment. The PdM department was able to evaluate the entire lineshaft and diagnose a problem area in a total of four hours with two different technologies that agreed with one another.

Vibe and Infrared Partner to Detect Overgreased Bearing

After the prior example, it might seem that thermography is more valuable as a stand-alone technology. This is not always the case. Take for example another problem found on another press. Press 22-10 Line Clutch was found to be running at 140°F. Vibration data were collected for the questionable component. Even though the thermographic scans indicated a possible major problem with the line clutch, the vibration data indicated no problem with the clutch.

What was observed when collecting the vibration data was a large amount of grease being flung all over the press. A quick search in the maintenance records revealed that an untrained employee had overgreased that unit one week prior, resulting in too much grease in the line clutch bearing. This is a good example of how the technologies did not agree and could have resulted in this line clutch being replaced due to a misdiagnosed problem.

Thermography in the Mail Room

Predictive maintenance is not limited to just the pressroom. The mailroom is another department in the newspaper where all the sections are compiled into a final product. In this process, an inserter is used to marry all the preprinted advertisements, magazines and coupons with the main sheet of the newspaper. The eight inserters in this plant contain 104 hopper units.

These hoppers are a subsection of the inserter with a series of cams and many pivoting mechanical components that do a great deal of banging and knocking. A major maintenance project was being planned for the following year. An evaluation was needed to summarize the overall condition of each of the hoppers. Two forms of PdM technology, vibration analysis and thermographic scans, were recommended.

Two axial readings per hopper, one at each end of the main camshaft bearing, were taken. The magnitude value of each was then plotted. Thermographic scans of the same main camshaft bearings were made and the results were plotted in a similar manner. The vibration data and the thermographic data yielded similar results.


Figure 10. Inserter Hopper Evaluation Curve

Figure 10 shows all the hoppers plotted with the condition rating on the X-axis and the number of hoppers with those ratings on the Y-axis. The maintenance department began to systematically rebuild the hoppers beginning with the four hoppers that were three standard deviations out of the curve.

They found a tremendous amount of backlash in these units and parts that had excessive wear. The maintenance department then disassembled the two units that were only two standard deviations off the curve. Upon inspection, those units showed some wear but were still in good running condition.

Instead of proceeding with overhauling of all the units, only the few bad ones were rebuilt. Now there is a standard to which future evaluations can be compared. This integration of vibration analysis, thermography and statistics minimized the need for a major maintenance project.

With the use of PdM, The New York Times has also been able to streamline our other maintenance tasks. A good example of this is when the maintenance department checks the pipe rollers on a press. The old method was to have a maintenance person spin the roller by hand, tap it and listen for some kind of odd rattle sounds.

Not only was this a tiring and lengthy process, but also it was inaccurate. With the use of thermography, a whole series of pipe rollers on a press can be checked while they are running. The pipe rollers with the hotter ends, such as the one shown in Figure 11, tend to have deteriorating bearings. Because these are not driven rollers, it is not possible to use vibration analysis.


Figure 11. Bad Pipe Roller Found During Production

Another example is on the folder nipping rollers. Even when set properly, they will sometimes generate a rattling sound. Although considered normal, this can mask the sound of a bad nipping gear getting ready to fail. With ultrasound, any maintenance professional with minimal training can spot-check the area while the press is running production.

A chart for these nipping gears has been created. Anything with an intensity reading of less than 40 is normal. If the reading is between 40 and 100, a vibration reading is scheduled for sometime that week. Anything more than 100 on the ultrasound intensity meter must be reported immediately and a vibration reading is scheduled as soon as possible. This is shown in Figures 12 and 13.


Figure 12. Press 27 Nipping Rollers at 40


Figure 13. Press 26 Nipping Rollers at 110

To have a successful predictive maintenance program, personnel need to understand the tools available and have the practical knowledge of when to apply each technology. It may not be possible or even recommended to begin using multiple technologies from the start. Until some expertise in each technology is established, there is a strong possibility that data could be misinterpreted.

The advantage of using multiple technologies is that a problem can be cross-diagnosed and decisions can be made with more confidence. That is, if two or more technologies can point to a fault, the chances of misdiagnosing a problem become less of a threat. By making more accurate diagnoses, personnel can order parts more appropriately and schedule production run time and maintenance downtime with greater confidence.

This is successfully working at The New York Times and will hopefully be implemented in more newspaper printing facilities around the world.

Editor’s Note
Article provided courtesy of FLIR Infrared Training Center, www.infraredtraining.com

Subscribe to Machinery Lubrication

About the Author