Preventative maintenance through balancing and vibration analysis
By Jerry Justice
Have you ever had a dead battery in your car? Did you discover it right when you really needed to get somewhere? Do you wish you could prevent things like this from happening? The key word here is prevent.
Most of us take preventative measures in many areas of our day to day lives. We change the oil in our cars to prevent excessive engine wear and check the battery for its electrolyte level to prevent it from going dry and dying on us at inopportune times.
Structures and components on aircraft are inspected to prevent unexpected failures. An unexpected failure is bad. A catastrophic failure is...well, catastrophic.
There are limits set on how much wear, crack, leak, and a host of other common faults or malfunctions is acceptable. All these limits were established to prevent failures. We measure these inspection points with the tools we have available and base our repair, replace, or condemn decisions on the known factors, past behaviors, or experience by which those limits were set.
What if you had an automated means of detecting a failure before it happened? Would it be beneficial if you could say with reasonable accuracy that a component would probably fail within the next hundred hours of operation? Could you save money by removing and replacing a component for overhaul or exchange before it experienced a structural failure?
All these things can be accomplished. They have been routinely used with industrial machines for over 25 years! It began with vibration trending and predictive analysis. Industrial-based predictive maintenance now uses even more advanced techniques, which provide even greater odds of successful prediction and thus greater economy of operation.
In aviation today, vibration analysis and balancing have, to a great degree, been relegated to the task of reducing discomfort to passengers and crew caused by noise or to certify the engine as airworthy following overhaul or, in some cases, STC completion. While noise reduction is one great use of modern balancing and analysis tools, it falls well short of achieving the long-term benefits of vibration trend analysis and predictive based maintenance these tools are capable of. Balancing is still one of the best methods of reducing component stress and premature failure. Predictive maintenance can play an irreplaceable role in life extension and damage control. Leading edge technology available today can collect, analyze, and even predict useful component life based on various factors including vibration.
The U.S. military is currently attempting to perfect prognostic hardware and software that will predict an impending failure with greater accuracy. If adequate warning is provided, logistics and planning requirements can be completed in a timely manner, thereby reducing spares inventory, preventing unexpected failures, and providing a basis for improving design and engineering techniques for components and processes. Trending may also be expanded to include oil analysis, ultrasonic, and thermographic, methods.
Trending engine performance factors
Engine performance factors may also be trended for tracking thermodynamic characteristics and other performance factors. Trending isn't a new science, but one that is just beginning to mature in the aviation arena. The following is a very basic synopsis of how a trending program works.
Where to start
To begin a trending program you must decide what you want to trend. You may collect an overall survey that includes all the operating frequencies for the engine or you may wish to trend discrete spectrum data. While the overall data is good for establishing initial thresholds or the gross overall condition, many outside forces including aerodynamic loads on the airframe can affect the data quality. For this example, we will trend discrete spectrum data on three components of an engine system. The drive speed is 10,000 RPM, which is N1 at 100 percent. For the sake of simplicity, the three fictitious components are driven at whole multiples of the drive speed.
1. A hydraulic pump operating at 2X drive speed.
2. A low-pressure fuel pump operating at 3X drive speed.
3. A generator operating at 4X drive speed.
Remember that these speeds are only used as examples for this article and not indicative of a normal operating speed for the actual components.
Next, you must decide on an interval for collecting data. For this example we will collect a vibration spectrum every 50 hours. Intervals may be extended or shortened according to specific requirements. Accelerated wear in components may require shorter intervals. If no significant change is noted in any of the components, a longer collection interval may be called for. Try to collect data on or as near as possible to the scheduled interval you decide on.
Standardization is key
In order to ensure you are comparing like data, standardize the method of collection. All parameters should be the same or as near the same as possible for each collection. The engine speed, operating temperature, and ambient conditions, especially wind, should also remain the same whenever possible. Be certain the same type sensors, cables, charge converters, and analyzers are used to collect all data. You must also ensure the data is collected using the same engineering units and modifiers (IPS, Mils, Gs and Peak, Peak-to-Peak or RMS). Remember that differences introduced into the collection process will affect the data and make meaningful analysis difficult or impossible.
Establishing a baseline
As a starting point, you must establish a baseline. This will be the measuring stick by which you compare and evaluate a change in the condition of the individual components being trended. The baseline should be collected on new or very low-time components. Ideally, this will be the best condition of the components you can expect to see during their life cycle. You should ensure all, if any, break-in times and requirements have been met before collecting the baseline data.
In the X, Y, Plot illustration (See Figure 1, pg. 63), you can see the frequency, in RPM, along the horizontal (X) axis. This scale may also be in Hertz (Hz) or cycles per second. By moving from left to right along this axis you can easily locate the four components for our example at 10,000 RPM, 20,000 RPM, 30,000 RPM, and 40,000 RPM. By comparing the top of each of the peaks horizontally to a relative position on the vertical (Y) axis you can read the amplitude of the vibration for the component operating at that frequency. The units of measure in this case are Inches Per Second Peak (IPS Peak). For example, the N1, turning at 10,000 RPM is generating vibration with a value of approximately 0.32 IPS Peak.
If we use this data as our baseline, we now have a basis of comparison for future changes in the condition of the four components. Remember that this is a very simplistic view of how trending works. With that in mind, you should also remember that a rotating component might display various vibration characteristics according to a specific type of wear, misalignment, or imbalance condition. These characteristics include harmonics of the fundamental frequency, which would, in this case, conflict with the frequencies of one or more of the other components in our example.
Now that the baseline has been established, collect data on time, on schedule, and under the conditions as noted above. In this example, and again for the sake of simplicity, we will collect data on a schedule of every 50 hours of operation. When the individual spectra are plotted on a comparison plot, as shown in Figure 2 on page 64, changes can be measured visually with the assistance of various software tools. Analysis and evaluation can also be accomplished automatically by setting limits or flags for the target frequencies and amplitudes associated with those frequencies. A report generated when those limits are met or exceeded will then provide the information critical to the corrective action decision and the success of the trending program.
In Figure 2 above, the amplitude of the generator turning at 4X drive speed or 40,000 RPM, shows an increase (depicted in red) with each subsequent data collection. The other components remain at the same amplitude for the sake of comparison in this graph.
The amplitude of the generator baseline was 0.18 IPS. The progressive increase of that amplitude is shown in the graph above where it is shown in comparison to all the other monitored components and below as compared to time only. Above, the last three collection points at 400, 450, and 500 hours show an increase up to 0.4, 0.48, and 0.55 IPS respectively. The change in the vibration condition of the generator is progressing in a linear fashion as shown in Figure 3. While this data does not provide an indication of an impending failure, it does establish a trend that can be directly related to time and a corresponding change in vibration.
If the generator is replaced because of other non-vibration-associated defects, the baseline for the generator may remain the same and the data for the life of the generator used as a comparison in the life of the new generator. However, an exception to the rule arises when the initial data indicates a significant difference, either above or below, the established baseline. If this occurs, you must investigate the situation to determine which generator is closer to the norm. If the vibration data from several generators of the same operating time are available for comparison, it becomes a matter of elimination as to which generator falls outside the norm.
It is easy to see that the more data you gather, the more accurate your trending program becomes. In Figure 4, page 65; four generators, A, B, C, and D are plotted to the 500 hour point. Notice that generator C failed prior to the 500-hour point. The sharp increase in the vibration amplitude just prior to failure is typical. With this knowledge in hand, you may choose to remove and replace the generator when a similar trend is noted in the future. While you cannot accurately predict the failure down to an hour of operation, you can heed the warning signs that were produced by other generators in your database prior to their failure.
You must also realize that some failures will produce little, if any, warning prior to a failure. If this proves to be the rule rather than the exception you should reduce the time between data collection intervals. If, in the case of Generator C, a failure mode is clearly in its advanced stages, you should set the limit for the generator at a point prior to the anticipated failure. For example, in this case you should set the limit at 1.2 IPS. Remember that these vibration levels are only used for this example and do not reflect actual values or recommended limits for any component.
When a sufficient quantity of data is collected to establish a limit, you can install a limit line in your software and/or analyzer if it is capable of this function. Some analyzers are capable of storing a database of frequency ratios associated with the components you wish to monitor. These ratios are synchronized with a tachometer input of the driving component. The database is then queried for the turning ratio of the requested component. A key press can then identify individual components. The database may also contain additional display information such as vibration limits of the component set by manufacturer standards or by the user from established trending limits. (See the example below). This analyzer also has the ability of allowing the user to install a limit line in a Setup, which then shows the established limits of the displayed frequencies on screen in the form of an encapsulating line as you collect and view the spectrum. (See Figure 5). If a limit is exceeded, you are aware of it instantly.
The idea behind any trending program is to take corrective action for an impending failure before it gets expensive. Although still in its infancy, trending in aviation applications is already saving untold thousands of dollars in overhaul cost for operators of corporate fleets and regional airlines. Having been in the vibration analysis and balancing business for a few years I still think the largest challenge to establishing vibration analysis as a viable aviation tool is the education of A&P technicians in its use. Reluctance to use something you don't understand is human nature. Several vibration equipment manufacturers offer on site training as well as scheduled courses in understanding and using vibration analysis. With the current advancements being made, those who are ready to meet the coming challenge of their use will be the recognized leaders in this little understood but invaluable area of aircraft maintenance.