If you’re going to implement a predictive maintenance program, be prepared to change the way you do business. Two years ago, Guardian Industries’ DeWitt facility started its own in-house predictive maintenance program. From the first day, the program changed the way we operated.
In retrospect, our greatest strength was the willingness to learn from the data and accept the fact that some of our prior beliefs were wrong. This has accelerated our progress and has helped increase our predictive maintenance technicians’ ability to analyze equipment problems.
Bill Jacobyansky (olive green shirt),
Paul Thorngren (blue shirt) and
Wes Ehlers (light blue shirt)
analyze pieces of equipment.
The Guardian Industries plant manufactures glass, using the float glass process where the raw materials are melted, mixed (homogenized), and then floated onto a bed of liquid tin to begin the cooling and annealing process.
The annealing is continued on a long run of conveyor sections where heating elements and air fans are used to cool the glass in a controlled manner. Glass is easy to manufacture. The difficulty comes in making glass perfectly every second of every day during the year. Consistency is mandatory, and this was the driving force that prompted us to implement the predictive maintenance program.
For us, predictive maintenance was a new way of thinking and was seldom understood beforehand by all interested parties. Our customers, those working in the production areas of the plant, needed cost-effective results to be convinced of the program’s validity.
These individuals were worried that some of the best employees would be transferred from wrenching to monitoring. An expensive equipment outlay at the same time could have been quite scary to them. Fortunately, our plant operates in an environment where new ideas are supported even when staff are apprehensive.
It was fortunate when the plant was built eight years ago, that vibration equipment was purchased as part of the initial tooling. An Entek DATAPAC 1500 was found in the tool room and was like new! To minimize start-up costs, the predictive maintenance program was started with vibration analysis and lube oil analysis.
There were two expectations at the start of the predictive maintenance program. We thought it would take two years for our predictive maintenance technicians to become proficient at their jobs. We also expected that the implementation of equipment monitoring would increase our work load for at least one year as our standards changed from “is the equipment functional?” to “how well is the equipment running?”
The strategy was to let the program show us what we needed to do and when to do it. The last sentence is not in regard to repairing plant equipment. It was an acknowledgment that we were entering a new area and we didn’t know how to staff it, how to monitor equipment, what equipment to monitor, and when or what direction to expand the program. We had read many articles and knew their recommendations, so we had a starting point.
Before predictive maintenance, our bearing greasing was calendar-based. In addition, we also had the destructive habit of applying new grease until it was seen coming out the sides of the bearing housing.
Somewhere in the early years of the plant, this had become the recommended practice and it was ingrained in many of our maintenance craftsmen. Though we vehemently preached against overgreasing, it was difficult to eliminate a bad habit.
It is difficult to get craftsmen to change their practices until it is proved that the new way is better. It was obvious that the greasing practices had to improve, but we needed a way to determine when we were doing it correctly. Equipment monitoring and predictive maintenance enabled us to do this.
Fans were chosen as an initial focus for predictive maintenance. They are critical to the combustion and annealing process, and they are a good application for vibration data analysis. As the vibration technician started collecting data and learning his craft, he quickly learned how to spot problems of inadequate grease in the fan bearings.
This knowledge came through trial and error as every symptom of high vibration was immediately treated with the application of additional grease. It was similar to giving children Tylenol every time they develop a fever.
We soon refined our efforts using vibration analysis to determine when a bearing actually needed more grease. (Hint: using vibration analysis, bearings that needed more grease would show an increase in spike energy with no indications at any of the bearing failure frequencies and some increase in high-frequency acceleration.)
Vibration data analysis is great for determining greasing needs, but it is also cumbersome. Because of the gap between data collection and data analysis, vibration data was impractical for determining just how much grease needed to be added to a bearing. This caused us to add grease cautiously at first, take a reading, add a little more grease, then repeat the process.
As the technicians became more experienced, the number of iterations was reduced to a manageable, though still inefficient level.
Relatively inexpensive acoustic monitoring equipment was added to the predictive maintenance arsenal. An Ultra-Lube acoustic attachment for the grease gun enabled staff to listen to a bearing while injecting grease.
We expected that there would be some synergies between our vibration equipment and acoustic monitoring equipment, but we were stunned by how strong the synergies were. The two pieces of equipment work together to enable one to find and determine equipment problems.
The acoustic grease gun enabled us to inject the correct amount of grease on the first try. This meant that we were able to apply the correct amount of grease that the bearing needed at that time. Here, our knowledge of bearing greasing took another large step forward.
We found that a new bearing needed to be greased several times before it could be put on a standard monitoring schedule. As part of the predictive nature of the program, vibration data was collected on all critical fans once a month. If a new fan bearing was installed, the next month’s vibration data would show that the bearing needed additional grease.
The same thing would occur the following month. Fortunately, new bearings were monitored on a much greater frequency. Daily monitoring and grease addition were often required during the first week.
The frequency of monitoring and greasing was extended based on vibration or acoustic monitoring results. It took at least a month to get the proper amount of grease in most bearings to allow them to run well for an extended period. This initial greasing process could take as long as two months if there were additional mechanical circumstances.
Figure 1. Horizontal Acceleration Readings
from Inboard Fan Bearing
Figure 1 shows the horizontal acceleration readings taken on an inboard fan bearing - one of Guardian Industries’ earliest experiences with the greasing of a new fan bearing using predictive maintenance tools.
The bearing is on a horizontal fan with a 2 3/8-inch shaft that rotates at approximately 1,938 rpm. Every point indicates that vibration data was taken. After the installation of the new bearing, we reacted to all of the high points by adding grease.
As expected, the predictive maintenance program did increase the workload and much of it came from raising the bar on equipment performance.
What we did not realize at the beginning was how much of the additional work load would come from finding unsuspected problems such as: incorrect belt tensions, poor alignment, worn vibration dampers, cracked welds, and incorrect equipment applications with bearings and sheaves. Some of these problems were self-induced; others were created when the equipment was designed. Predictive maintenance enabled us to learn that there was a problem.
The determination of the root cause often required additional effort and hours of conversations with equipment vendors and peers in other plants. Predictive maintenance made us look deeper into equipment problems than we ever did before.
Our primary vibration technician became accomplished at data analysis well ahead of the two-year schedule, due to the conscientiousness of the individual and the commitment to have him work full-time as a data collector and analyst. While our primary vibration technician is competent, his knowledge continues to increase.
We still have much to learn as we continue down the predictive path. The goal is to eliminate the majority of our time-based PMs and replace them with periodic equipment monitoring. Our predictive maintenance program results have been impressive and we are still expanding our capabilities.
We recently added a Raytech TI30 infrared camera and we are looking at acoustic monitoring equipment later this year. We are also increasing our knowledge of how the equipment in our plant runs and how to properly maintain it.