Optimal stopping in a burn-in model

Abstract

Burn-in is a widely used procedure for screening out defective items in a production before it is delivered to the customers. Given some cost and reward structure, one problem related to burn-in is to determine a burn-in time ζ which maximizes the expected net reward. This optimization problem can be viewed as an optimal stopping problem which we solve by means of a semimartingale approach. Furthermore, we investigate how the use of estimators instead of the true parameters of the model affects the obtained burn-in time.

Keywords:

• burn-in
• optimal stopping
• reliability