During endurance testing of a new product, some screws were found to loosen and break. When I examined the broken screws, I found that they had laps or folds under the head. Laps or folds are defects, which can be introduced during the heading process. The fatigue cracks in the screws had started from these defects.
As soon as it was discovered that the screws were defective, the vendor went to work finding a new source and providing replacements. They were able to get us new, fully conforming screws in record time.
Meanwhile, I continued to investigate the failure. It's one thing to identify a non-conformance, and another thing to identify the root cause of a failure. Yes, the screws were defective, but this still didn't fully explain why they failed. Screws don't tend to experience bending fatigue unless they are loose. Why were the screws coming loose in the first place?
While I was pondering this question, an assembly line worker came to me with a new problem. When installing the new screws, the internal threads on the mating part were stripping out. What now?
The aluminum from the mating part lodged in the threads of the screw showed me that there were only a few threads of engagement. These threads had to support the entire clamp load of the joint. This load, over the surface area of these threads, exceeded the shear strength of the aluminum, causing the threads to strip out.
Why didn't this happen with the old screws? As it turns out, the old screws were about half a thread longer than the new screws. (Both were within the acceptable length tolerance.) This extra bit of engagement was just enough to ensure that the internal threads wouldn't be ripped out of the aluminum when the screw was installed. However, the stress was still high enough for the internal threads to yield. This caused a reduction in clamp load, which allowed the screws to loosen and fatigue.
The root cause for both problems was the same: a joint design which didn't allow for enough thread engagement. How much thread engagement is enough?