Using big data to detect hardware manufacturing defects

Dale Humby bio photo By Dale Humby

Modern hardware devices have hundreds of discrete components, and finished products cost thousands of Rand each. A defect rate of even a few percent multiplied across thousands (or millions) of devices leads to huge losses and unhappy consumers who receive devices that are ‘dead on arrival.’

In this talk I am going to show how Nomanini automatically tests printed circuit board assemblies (PCBA’s) in the factory and uses Google’s BigQuery analytics database to mine test output in real time to detect anomalies, flag defects and alert us to changes in the underlying manufacturing process before they impact our customers.