The essential difference
A test under real conditions exposes the product to the use and environment it will see in service, for the length of life of interest. An accelerated test increases a stress, such as temperature, voltage, humidity or cycling, to cause the same wear in less time. The former is faithful but takes years; the latter is fast, but its conclusions rest on an assumption.
The assumption that decides everything
Acceleration is valid only if the failure mechanism is the same at the accelerated level and in normal use. If the temperature is raised until the material changes phase or a different reaction appears, you end up studying a failure that will never occur in service and missing the one that would. A physical basis for the acceleration model is therefore needed, not just a data fit.
When each one is appropriate
- Accelerated: in development, when designs or materials need to be compared quickly, and when a recognized model exists for the suspected mechanism.
- Real conditions: when the environment is complex and combines many factors, when safety demands direct evidence, or to validate the accelerated model.
How they relate
The usual approach is to combine them. Accelerated tests are used to identify failure modes and estimate trends, and these are confirmed with a smaller sample under real conditions or with field data. That confirmation is what gives confidence in the acceleration factors used.
What the engineer sees in practice
A very aggressive accelerated test produces rapid failures, but the subsequent failure analysis must show the same mode observed in the field. If the failure mode is different, the result is of no use for predicting life. The data are interpreted with life statistics, for example with Weibull distributions, and with the uncertainty that entails when there are few samples.
A practical criterion
Always ask: which mechanism do I want to accelerate, and why do I believe the chosen stress accelerates only that mechanism? If the answer is not clear, a less aggressive test or an additional verification under real conditions is advisable.