How to Diagnose Machine Failures

When a machine fails, evidence of the failure is often destroyed along with the component. The initial evidence is so mangled by the actual failure itself, it becomes indistinguishable from other elements.

The destruction of incipient failure evidence often leads one to blame the lubricant for a failure that was caused by another mechanism. Conversely, failures that are caused by the lubricant or the lubrication are often attributed to something else.

There are many possible cause-and-effect sequences where the evidence about the cause is lost in the evidence of the effect. The absence of good evidence is what leads to costly poor decisions to act - or not to act. Moreover, you cannot even determine if the event is addressable, or if it is random and should be accepted as such.

Have you experienced recurring failure because you never seem to get a handle on why a machine is failing? How many unnecessary preventive maintenance (PM) procedures are in your system as a result of a failure, the cause for which you could not pinpoint?

Implementing a bunch of PM procedures is a common response when nobody can really explain why a machine failed. Once unnecessary PMs get into the system, they are like a rash - they just won’t go away. And when nobody remembers why the PM procedure was instituted, the process of rationalizing maintenance activities and removing it can be long and difficult.

A lubrication failure occurs when the lubricant cannot effectively separate in relative motion. However, there are numerous mechanisms by which this can occur. In my article “FMEA Process for Lubrication Failures” I identified 13 common lubrication failure mechanisms, and I only scratched the surface. I bet I could come up with a list of 50.

When you start looking at submechanisms, the list increases geometrically. How can one equate a scenario where vibration causes the drain-plug to come off the sump - which causes the machine to drop its lubricant - to one where high levels of particle contamination cause excessive abrasion? Yes, both events are a lubrication failure, but from a maintenance and prevention perspective, they are incomparable.

So how can you improve the quality of evidence and your ability to examine it? Often, the lubricant or the lubrication system contains evidence even though the evidence on the component itself has been destroyed. Don’t throw away the evidence - seize it! I have outlined several items below to think about when considering the evidence:

1. Oil Analysis

Despite the mangled appearance of the component where evidence is destroyed, the lubricant still contains evidence about the time leading up to and including the catastrophic event. Sometimes, sump or tank bottom sampling can help in the process. The debris accumulates in the bottom of the tank, creating a veritable history book of the machine’s operation since the last oil change or tank cleaning.

Many of the particles contained will be incipient wear particles. Others will be catastrophic. They can help you piece together the story. A word of caution - this approach to sampling does not enable you to categorize the particle with respect to which component created it, nor the point in time when it was created.

2. Filter Analysis

The filter also serves as a history book. It captures the particles generated since the last filter change and leading up to the failure event. By opening up the filter, liberating the particles with an ultrasonic bath and depositing them onto a slide or filter patch, the evidence of the incipient event leading up to the catastrophic failure can be evaluated. Like tank bottom sampling, this method does not segregate the particles with respect to component or time of production.

3. Thoroughly Inspect Components In Response to Condition Monitoring

Frequently, condition monitoring detects a failure in its early stages, but when the mechanics pull the component and inspect it, everything lofoks fine. They reassemble the unit and a few weeks later the machine fails. Naturally, the assumption is that the intrusive act of inspecting the machine caused the failure - and the condition-monitoring technicians look like fools.

The typical inspection of a bearing is visual, often with the lubricant still present; or the mechanic may spin the bearing to see if it turns freely. Five-micron to 50-micron particles can be analyzed with wear debris analysis, which means that the machine has pits or grooves that are roughly the same size. It takes a microscope or other special instruments to evaluate the particles. How will a mechanic see those little pits on the machine’s surface with the naked eye?

Once a component starts to fail, the progression usually accelerates. Inspect pulled components with a microscope after cleaning the surfaces to see what is really going on.

4. Train the staff

Training programs for millwrights, mechanics and operators focus upon repair or restoration programs. They also require knowledge about failure mechanisms to understand how the machines reach a failed state, methods and techniques for reducing failure (proactive maintenance), including proper lubrication, machine condition-monitoring and failure analysis techniques.

This will better enable them to find and interpret evidence about machinery failure, and to employ their knowledge in pursuit of improvement.

We are living in a new world. Many organizations have thrown out their old book on equipment maintenance, and are rewriting a revolutionary new one. You can’t create new practices that eliminate problems at the root without evidence that accurately describes the problem. The lubricant and the lubrication system offer clues even though the evidence on the machine has been destroyed. Learn how to access and use this evidence - it is the basis for sound decisions.