Abstract
Burn-in is an engineering method extensively used to screen out infant mortality failure defects. Previous studies have attempted to determine the optimum burn-in time and cost for a device or a system. However, for the mathematical model, many assumptions are inappropriate due to practical concerns, and for the cost model, the required costs are difficult to find. How to effectively determine the optimal burn-in time and cost has perplexed manufacturers for quite some time. In the actual manufacturing process, a new electronic product is always extended from an old product, called the base product. By adopting the relationship between new product and base product, this study presents a neural network-based approach to determine the optimal burn-in time and cost without any assumptions. A case study of the production of a switch mode rectifier demonstrates the effectiveness of the proposed approach. © MCB University Press.