Level Up Your Reliability Skills: Get Certified! Boost your career now!

Elevate your industry profile at The RELIABILITY Conference.

Sign Up

Please use your business email address if applicable

It seems that the majority of analysts today are only comfortable working with linear graphs of the FFT spectrum. This graph displays frequency or orders in the X-axis and amplitude in the Y-axis. The rationale is that the highest peaks in the graph are the ones you need to watch. Those very small or almost invisible peaks probably aren't important. Unfortunately, there are many cases where this is not true, and data that indicates other problems is overlooked.

The problem with a linear FFT graph is that the amplitude scale is usually set automatically according to the highest amplitude in the graph (autoscaling). If the same machine also has other, more subtle problems, the fault frequency peaks that you can be much smaller amplitude and barely visible - but they are there. Today's high dynamic range instruments will measure even very low amplitude signals and preserve them in the data set, but if you're not very careful, you may not see them when you review the data. You could very well not see an impending problem simply because you're looking at the data in a linear display.

Let's look at an example of what could happen. Let's say you have a machine with rolling element bearings that has some unbalance. The unbalance has been diagnosed quite easily. At the same time, there's a problem developing in the outer race in the load zone of one of the bearings. As you know, unbalance creates additional forces on the bearings, especially in the load zone. Because of the mass of the rotor and stiffness of the machine, you are going to have a high amplitude peak at the running speed of the machine. In the linear graph, you will see a high 1X peak, and pretty much little else. Because of increased loading due to the unbalance condition, instead of a normal wear-out timeline for a bearing, this bearing is probably going to go from mildly damaged to complete failure in much less time than usual.

In the linear graph view of the FFT spectra for this machine, you might not even see the advent of bearing tones as the rollers pass through the defect in the load zone. They are going to be very small initially compared to the unbalance amplitude at the running speed. On an analyzer/data collector screen this could be just one pixel, practically part of the noise floor. However, if you toggle your display over to a log-linear display in units of dB, you would see bearing tones up in the higher, non-synchronous frequencies. Because the lower amplitudes are expanded in scale and the higher amplitudes are condensed in scale, you can see the bearing tones as soon as they start to occur, while still seeing the full range of amplitudes. Without even knowing the fault frequencies as calculated by the bearing geometry, you would probably recognize a bearing fault by its classic signature. Knowing that you have an unbalance situation in the machine, this would be the ideal time to go ahead and release a work order, get the machine balanced and replace the bearing.

But don't take my word for it. Try it yourself: The next time you're looking at a machine that you feel comfortable with, one that has a known fault with high amplitudes, and some other fault frequencies, try toggling between a graph of the fault frequencies in linear mode and in log-linear mode. Do other fault frequencies suddenly seem to jump off the screen? Maybe there's something to this log-linear graph thing after all.

Tip Provided by: Bill Slonaker, Marketing Manager
Mobius Institute
Tel: +1 877.550.3400

Mobius Logo

“R.A.I.” the Reliability.aiTMChatbot

You can ask "R.A.I." anything about maintenance, reliability, and asset management.