Comparing accelerated ageing with "real" ageing

Question

In the electronics industry (and many others fields, but I'm interested in electronics here), it is common to submit electronic components and boards to accelerated ageing processes. The aim is to detect early-life failures and avoid putting the concerned product on the market in order to improve overall reliability. And also, it can be used to assess long-term reliability of the product.

My question is : on which scientific basis can we affirm that accelerated ageing is "similar" or "comparable" to "real" ageing ? How is an electronic board that spent 100 hours in an oven at 90°C "similarly aged" than the same board that spent 10 years under operation ?

Answer 1

"How is an electronic board that spent 100 hours in an oven at 90°C "similarly aged" than the same board that spent 10 years under operation ?"

It is not 'similarly aged'. There are two types of failure at work here, one is use-related and one is and underlying weakness that needs to be forced to fail. The purpose of 'burn-in' is to force any weak spot to fail and not to 'age' the electronics. Most electronics do not follow an age degradation profile (with the known exception of some components such as electrolytic capacitors) and therefore do not have a wear-out life. In fact according to Nowlan and Heap only 2% of equipments failures (mechanical and electrical) follow the tradition 'bathub' curve and a total of 14% of equipment failures conform to a wear-out pattern. The concept of an increase in probability of instantaneous failure with time has largely been discounted. These curves have been investigated time and again and form the basis for the RCM 'industry'. It is dealt with in great detail in Mowbray's RCM II and MIL-STD-3034 (US) and DS 02-45 (UK).

Most of the burn-in of boards and equipment is a calculated mix of thermal cycling, electrical load cycling and shock & vibration cycling. This is designed to cause manufacturing weaknesses to fail such as poor quality solder joints that flex under thermal and vibration cycling and then fatigue and crack. It will also cause poorer quality electronic components to fail. In theory these boards, once the failed components have been replaced, will be more reliable in the field.

Answer 2

While it is deceptively simple, after all you place your DUT (Device under test) in a chamber and crank up the heat, there is a lot more going on to these tests than most people suspect.

Ideally you want to come up with a result that allows you to have a certain comfort level that the product will last past it's design lifetime. And the best tool we have is the Arrhenius equation, FIT's (Failures In Time) and other techniques. And these techniques do work and have been verified (by running very long tests). However, as the design complexity increases and you have components made from very different materials and processes and then this technique becomes harder to apply. It is a matter of complexity and not veracity.

The key to a proper long term estimate is to determine the activation energies of the various failure modes within the DUT. Once these are understood then the application of the time and temperature etc. is straight forward. However, determination of activation energies is never straight forward and can involve a whole series of tests at different forcing temperatures and then subsequent inspection and failure analysis.

The fundamental limits are, as always, how much time do you have? and how much money do you have?

All of the above is assuming that the test for long term failure. There are also short term stress tests and burn-in tests that are to ensure that the infant mortality failures do not enter into your sales channel. These tend to be very different failures that long term failures.