Convex optimization of durability testing profiles for autonomous vehicles

Durability life of an automotive product depends not just on structural characteristics of the product but also on time and frequency domain characteristics of the road load it is exposed to over its life. Accelerated durability life testing is performed in a proving ground with a well defined set of damaging road events or `stressors' performed at a set of predefined speeds. A testing profile consisting of a mix of proving ground features is built. The objective of this profile is to closely match the total structural damage accumulated in test to the total damage due to `road load' accumulated by the automotive product over one design life, while also minimizing the total test time. In this paper we show that the problem of selecting a mix of proving ground features given calculated relative damage and time taken, is convex and hence solvable. Also, by fitting a continuous spline function to our field calculated damage values for 5 pre set velocities, we expand our convex optimization problem to include all possible velocities based on the sampling of the continuous function generated above. In the test profile developed, we were able to reach 9 out of 12 damage targets precisely over 2 design lives of the vehicle. This new test profile also takes approximately 175th the time of a similar testing profile derived manually. We then also develop three different testing profiles each with its own unique advantages.