Our Beloved Bathtub Curve Might Not Commonly Occur

Picture of Adam Bahret
Adam Bahret

What if our beloved bathtub curve doesn’t exist in the form we always have thought of it. It is always presented as a “theoretical” model of course, “Your results may vary”! But it does dictate that we will find three distinct product life phases, infant mortality, useful life, and wear-out. Infant mortality being driven by quality defects, useful life being driven by use stresses, and wear-out being driven by …well…wear-out. So we end up with a curve that looks like this.
bathtub curve
So here is the question. Does that center section being driven by “use stresses” make sense? If the product is designed to handle it’s use stresses and expected variabilities, it has margin, would there be a noticeable failure rate associated to them?
What does the curve look like if we say that failures associated to use stresses are very rare and when they do occur it’s because of some extreme case and we don’t care about what happened in regard to modeling reliability.

If an angry spouse throws a TV out of a second floor window, the manufacturer isn’t going to do a thorough root cause investigation into why it broke and change their modeling and predicted reliability test methods. That data point is not going to be in the product status update at the weekly executive steering meeting. I guarantee it! It’s already hard enough to try and explain why there is a 2% no fault found in the failure Pareto because people won’t even read the “quick start” card that came with the darn thing. So why account for these failures in the original model? So if we only include quality and wear-out influences in the life curve what would it look like? If manufacturing is well controlled and the product has good margin on wearout then it may look like this.
failure free bathtub
There would be a group of quality defects that would be screened out. The  curve may have a quality defect tail extending into the warranty period, a failure free period, and then a point where the wear-out curve pops on the radar. If quality defects are less controlled and wearout starts earlier the curve may look like this.
bathtb U

Do either of these curves match the actual field data you have seen? I can think of examples that fit these two curves very well.

I want to thank Fred Schenkelberg of FMS Reliability for his input on this article. Visit FMS at www.fmsreliability.com/accendo/

Share this post