Self-adapting test-plans in production line: An application to vision control stations

Automatic Quality Control (AQC) in production line needs a constant design and update of test plans, in order to face with incoming new models produced, variations in sub assembly components and modifications of environmental conditions. This time consuming and fatiguing activity is usually carried out by human operators, otherwise the AQC system could drift away from optimal operating conditions, thus providing unreliable result. In the present work, an innovative approach for self-creation and self-updating of test plans using only the knowledge gained from the production data (through a Multi Agent System architecture) is presented, along with its application to an AQC system based on machine vision. Real images from a washing machines production line are used to validate the algorithms. This work is being developed within the FP7 European Project GRACE (inteGration of pRocess and quAlity Control using multi-agEnt technology).