Acceleration models are easier than you think

Picture of Adam Bahret
Adam Bahret
life stress curves

I often find customers seem to feel there is some magic to Accelerated Life Test (ALT) modeling.  It’s a little cool to be seen as a magician or some genius guru but the truth is that they are based on very simple principles that can be communicated easily.

My motivation for this article is that I don’t want the people who care most (product developers) about these life projections to not understand how they were created.  If they do they can add a great deal more to the process by contributing their knowledge about the product to the method.   It can be very helpful if a team member highlights that their is a condition where an item might temperature cycle even though the higher level assembly doesn’t temperature cycle.  They may only think to mention this if they understand the modeling and critical factors.

To assist with getting over the hurdle of understanding and feeling engaged in ALT, I am going to create a multiple part series in this blog just on ALT.  But first an overview.

Some of the most common ALT models are the Arrhenius, Coffin Manson, and Inverse Power Law.  Each of these are selected based on what is believed to be driving stress behind the primary wearout failure mode.  Those last four words, “primary wearout failure mode”, are the most critical to a successful ALT program.

Primary:  There will be many wearout failure modes for a product.  Each will be driven by different stresses and conditions.  An ALT can not study them all simultaneously.  There may be more than one in a program but it is more likely that separate ALT programs will be created to fully study the product wearout. The ALT results that will initially be of greatest interest is the wearout failure mode that will take the item out of service indefinitely.

Wearout:  There are three main classifications for failure modes, Infant Mortality, Uselife, and Wearout.  it is important that failures that are classified as quality defect driven “Infant Mortality” and use stress or use conditions driven “Uselife” are not included in the ALT test data set. We don’t know if or how these failures are being accelerated by the applied ALT stresses. We can’t attempt to make a correlation to when they would happen under normal usage with the derived ALT model.

Failure Mode: We have to clearly define what a failure is before starting the program.  Are we looking at soft fails (recoverable), hard fails (not recoverable), performance variability within spec or out of spec but operation continues?  With out clearly documented definitions of performance stats the data set used in the ALT model will be contaminated and can significantly skew the results.

Here is a quick overview of the three common models I mentioned above.  There will be separate articles that go into each in more depth.  If you read the Arrhenius article there is an added bonus of learning how to caramelize onions which is great because it’s such a versatile side dish.


Arrhenius Model:  The Arrhenius model is based on the principle that the primary failure mode is driven by a chemical reaction that creates material property change.  This may be oxidation, oil grease lubrication/viscocity changes, plastic arheius blockbrittleness, or enamel wire coating dielectric breakdown. The Arrhenius model uses elevated temperature to accelerate these failure modes.   If the model is correct it will demonstrate that a failure occurring while the unit operates at a higher than normal temperature can be correlated to time to fail while operating under nominal conditions.  An example would be that a product ran to failure after  4 weeks in a 70C environment. Using the derived model it can be said that this failure mode would occur under 6 years of similar operation in a 25C environment.

Coffin Manson:  The Coffin Manson model is based on the principle that mechanical stress driven by temperature change is driving the primary wearout failure mode.  This may be solder fracture from expansion and contraction cycles, delamination of materials, or material fatigue failures. The model increases the range of cycling temperature during ALT testing and then correlates it cofin manson graphicto life when temperature cycling matches that of normal field usage. An example would be; a PC board is cycled between -20C to 60C at a ramp rate of 10C/hour and a dwell time of 1 hour at each extreme.  Solder begins to break/detach after 250 cycles, the primary wearout failure mode.  Using the Coffin Manson model it can be stated that this same failure mode is expected to occur after 5,000 cycles when operating in the normal field temperature cycling range of 10C to 30C.

Inverse Power Law:  The Inverse Power law is the most dynamic of the three models.  It can be derived and manipulated to fit many applications and stress to failure mode relationships.  The law states that the time to fail under a high stress  can be correlated to a nominal stress condition by taking the ratio of the stresses and raising it to a defined power.  Now this is a a simple three variable epower law graphicquation to start but can become very complex once all conditions and peripheral factors are included.  But there is an elegance to it because it allows for what can be a very complex relationship mathematically to be trimmed down by inputting experimentally discovered constants.  These constants are found by creating a preliminary set of experiments that run the device at multiple stress levels.  By plotting the failures and their associated stresses on a log-log scale a straight line should emerge.  The characteristics of this line, like slope and Y intercept, will yield the needed constants to create an accurate Inverse power Law equation.  This is done without having to incorporate what could be very complex kinematic, structural, thermo dynamic equations to define the relationships.

Look for the detailed descriptions of each of these equations in other articles in the blog.

Share this post