An approach to the dynamic detection of outliers in electronics production

Abstract

There are many cases, especially in the electronics industry, of manufacturing activities involving very high volumes of components that, whilst individually costing very little, can cause very expensive failures or recalls upon unexpected failure. The identification of such outliers in component supply is therefore of great interest. These items are often within design technical specification but their deviation from the characteristics of the rest of the population may be indicative of some underlying cause that can lead to their premature failure in service. Identification of outliers using conventional statistical process control (SPC) and Control Charting techniques is often uncertain and most other published methods are inappropriate for dynamic monitoring. This paper proposes a simple and practical technique that has proved successful in identifying such components whilst minimizing ‘false alarms’. In many cases these items have subsequently been found to pass short-term inspection tests but fail well short of their design life.

Keywords:

• Outliers
• Quality
• SPC
• Statistical modelling
• Moving difference