TABLE OF CONTENTS Many of the devices produced today for complex technical systems have very high reliability under normal use conditions. The time involved in a life test such as those described in Section 9.1 would therefore be prohibitive. Furthermore, the device is likely to be obsolete by the time the test is completed. The questions then arise of how to make the optimal choice between several types or designs of a device and how to collect information about the corresponding life distributions under normal use conditions.
A common way of tackling these problems is to expose the device to sufficient overstress to bring the mean time to failure down to an acceptable level. Thereafter, one tries to "extrapolate" from the information obtained under over stress to normal use conditions. This approach is called Accelerated Life Testing (ALT) or overstress testing [2 8, 11].
Depending on the kind of device in question, the accelerated testing conditions may involve a higher level of temperature, pressure, voltage, load, vibration, and so on, than the corresponding levels occurring in normal use conditions. These variables are called stressors (stress variables or covariates). In a specific situation, there may be one or several ( m) stressors s 1, s 2, , s m acting simultaneously. The vector s = ( s 1, s 2, , s m) is called the stress vector.
In simple situations, there is only one stressor
