Root Cause Failure Analysis

Reliability Center, Inc. Root Cause Failure Analysis - www.Reliability.com Interpretation of Fatigue Failures 804-458-0645 Neville Sachs, President of Sachs, Salvaterra & Associates [email protected] Reliability Magazine, August 1999

In the article, Understanding Mechanical Failures that appeared in the June 1999 issue of Reliability® Magazine, we looked at a series of failures and discussed the basics of mechanical failure analysis. The valuable part about conducting these analyses is that the parts always tell you why they failed. Certainly, there are times when for one reason or another clues are difficult to interpret, but most of the time the fracture face is far more reliable then an eyewitness.

By far, the majority of mechanical failures happen from fatigue. (If someone asks you why something failed, just tell them fatigue. You’ll be right 90% of the time and that’s a great average in any activity.) But saying that fatigue caused a failure is like saying the hill caused a car’s brakes to fail. The value of failure analysis is that you can use it to look at the broken parts, determine the type and magnitude of the forces involved, and then do something to prevent their recurrence. The appearance of the fracture face, the shape of the progression marks, the location, shape and size of the instantaneous zone (fast failure zone), and the direction of the failure propagation really tell how the fracture occurred. In some situations high-powered microscopy is required, but most of the time careful observation with good lighting and low power (5 to 25X) magnification is all that is needed.

We will build on the last article’s basics to help in your failure analysis.

PROGRESSION MARKS

These are the clam shell-like marks on the surface of a fatigue face. Many people call them beachmarks, because they look like the marks waves leave on a sandy beach. We prefer the term progression marks because they show the path the crack takes as it grows across the part.

Progression marks show gross load changes on a part, for example the change in a pump load from 70 to 90 amps. This change in motor load and the corresponding change in shaft stress cause variations in the rate of crack growth and visible irregularities in the surface of the crack face. A common misconception many people have is that all fatigue failures have progression marks. In reality, only those components that see varying fatigue stresses have progression marks, so the shaft on a pump that always runs at a steady load will not show these marks.

The sketch below shows two views of a shaft that failed from plain bending. On the left you can see the progression marks, but many times there are invisible fatigue striations between the progression marks. These fatigue striations are the result of each stress cycle and can only be viewed using very high magnification, typically 1000X and up. Fatigue striations are relatively easy to see in aluminum alloys and some steels, but almost impossible to see in many others. Much of the time they are of no importance, but there are failures when it is critical to try to determine exactly how long it took from detectable size to catastrophic failure. In those cases metallurgists use an electron microscope to try to count the striations and estimate the crack life.

Sometimes progression marks are easy to see and at other times it is very difficult. When inspecting a failure face, we almost always look at it with a bright light set at a very shallow angle. The shadows make even minor surface variations relatively obvious.

Below we compare a plain bending failure with a rotating bending failure. In the rotating bending failure you see that the origin is not aligned with the bisector of the instantaneous zone and the growth of the progression marks is skewed. From this we know that rotating bending is involved. If only rotating bending is involved, the deviation between the IZ bisector and the origin is usually at least 15 degrees. Therefore, a fracture face where the difference is less than 15 degrees tells us that the forces involved a combination of rotating bending and plain bending. (Note that the shaft to the right below was turning clockwise during the crack progression.)

A good example of these differences can be seen in the breast roll failure below. The paper machine recently had several problems and we were asked to look at the failure. This is a shaft that is normally subjected to a substantial rotating bending load from the machine wire, i.e., the machine operating loads. But in this case, you can look at the fracture face and see that the deviation from the bisector is only a few degrees. The fracture face says the load is almost all from plain bending. Based on this, further investigation was initiated, revealing that in an effort to reduce the time needed to change the rolls, plant personnel had developed an installation procedure that inadvertently bent the shaft as it was being installed.

The shaft is from a gear driven machine used to cut a fabric, and an end view of the machine is shown below. The lower shaft is driven by a timing belt and the two shafts are interconnected by a set of split gears, adjustable to maintain a minimum backlash. The plant had suffered at least six failures and had made major revisions in the two rolls to reduce the load on the shaft before they finally decided to do a failure analysis on the shaft. The analysis not only found there were metallurgical problems but it also showed the stress that rotating bending was largely causing the failure. If the primary source of the loading had been the cutters, the failure would show a plain bending load. Looking at the failures face you can see that the initial failure growth was skewed. In this application the only source of the rotating bending stress was the zero backlash gears.

Again referring to the failure, it is easy to see that the crack growth in the middle of the shaft was straight across the shaft from plain bending because of the cutter loads. But the important question was "What initiated the failure?" The progression marks tell us the initiating force was the result of the gear mesh forces.

One last comment about progression marks is that they can be used to tell the direction of shaft rotation. (In the sketch above, the shaft is turning counterclockwise.) this is a valuable tool because most of our equipment is driven by three phase motors and the chance of connecting them with the correct rotation is 50%. Centrifugal pumps will pump and squirrel cage fans will blow air whether they are running in the correct direction or not. However when they run in the wrong direction the result is poor performance and unusual loads. I can clearly remember looking at an 8" (203mm) pump shaft and saying "Look at that fracture face" You can see that pump was running clockwise". Then the client said "Oh, no. That pump runs counterclockwise!" Oops! Case closed, physical failure cause found.

FAST FRACTURE ZONE (INSTANTANEOUS ZONE)

The size of the IZ, instantaneous zone, tells us the load on the shaft when the final failure occurs. Using the fracture mechanics, it can actually be used to calculate the load on the shaft at the time of failure. However, it can also be deceiving if the loads change with time. Looking at this failure you see a shaft with a tiny IZ, less that 5% of the total shaft area. This tells us that the load at the time of failure was low, but does not tell what the load was like when the crack initiated. Meanwhile, the progression marks are stating that the load varies greatly with time. With a small IZ care must be taken to investigate the loading that initiated the failure.

The same effect applies to rotating shafts except that stress concentrations are more frequently involved. As mentioned in the previous article, ratchet marks are an indication of stress concentrations, but the location of the IZ is also a significant clue. In many applications the size of the IZ tells the magnitude of the final stress on the shaft and the location is an important indication of the relative significance of load and stress concentration. Above we see two motor shafts from the same application. They were generated about 24 hours apart. The complete lack of progression marks tells us they ran at constant loads. The size of the fast fracture zones tells us that the load on the right motor was a little more than that on the left, while the eccentric location of the left failure tells us the stress concentrations were less than on the right example. The reason for this is that the centered IZ reveals that the failure began at many locations around the perimeter of the shaft and then grew inward. If the shaft was lightly overloaded with a low stress concentration, the cracking would have begun at one point and the final fracture would have looked like the rotating bending failure above illustration. Between these we have the example below where there was a moderate overload and a moderate stress concentration.

Neville Sachs, P.E., is President of Sachs, Salvaterra & Associates, Inc., which was founded in 1986. The consulting firm specializes in improved plant and equipment reliability and technical support services. Among the firm’s capabilities are vibration monitoring, mechanical failure anlaysis, corrosion and materials engineering, design reliability analysis and a wide variety of nondestructive examination methods. Previously, Neville was Supervisor, Reliability Engineering for Allied Signal Corporation where he was instrumental in developing one of the first large predictive maintenance inspection programs in the nation. Mr. Sachs received a Bachelor of Engineering Degrees in both Mechanical and Chemical Engineering from Stevens Institute of Technology. Visit his web site at http://www.sachssalvaterra.com.