Similarity based reliability qualification of electronic components

Qualification analysis and reliability testing of electronic components represent a major activity in the process of development of electronic equipment. Electronics manufacturers have to adopt often costly test programmes to ensure qualification standards and reliability requirements are met. This paper presents a similarity-based reliability qualification approach for assessing electronic parts, and in general electronic products, as an alternative to physical testing. This approach relies on the formulation and use of product attributes that impact the reliability risks with respect anticipated failure mechanisms. These attributes define and underpin the similarity-based clustering of previously qualified products. Using an unsupervised neural network known as Self-organising Map, new products can be mapped through similarity to respective clusters of previously tested parts with known qualification status. A demonstration of the proposed approach in qualifying electronic parts to the thermal loads induced by hot solder dip refinishing process is presented and discussed.