Data Mining in Product Traceability Data: Knowledge Discovery applied to the process measurement data,
Résumé
Data mining derives its name from the similarities between searching for valuable business information in a large database and mining a mountain for a vein of valuable ore. This evolution began, forty years ago, when business data were stored on computers. It continued with the databases techniques and more recently, since 1990, with advanced algorithms for processing large masses of data. Nowadays, data mining technology are mainly applied for automated discovering of previously unknown patterns, especially in business and financial services. During the last twenty years, the combined progress of computers technologies and digital transmissions techniques has allowed to constitute important databases in various sectors (medicine, meteorology or the industrial production). The industrial production has widely benefited of these progresses, especially at the level of the production machines that are automated and computerized. A computerized process records large masses of data during its functioning for giving a visibility on the good functioning of the machine and for insuring the product traceability. A Product Traceability system records measurement data and/or processing data. Therefore, it could be interesting to apply data mining techniques to these databases in order to find ‘hidden knowledge'. But, nowadays, one of the crucial issues facing many manufacturing facilities is not the gathering of data, but how to efficiently use the acquired data. The chapter describes the evolution of Datamining and Knowledge Discovery and its use today in the information-intensive industries. Secondly, we discuss about the idea of using datamining techniques to extract knowledge from data coming from an industrial process: what type of knowledge, for which use? In the last section, we develop an example of knowledge discovery in engineering with KDD method. A French trucks manufacturer has decided to use data recorded to trace the processed product (a diesel engine) for optimizing the engine test process itself. The initial goal was to reduce significantly the time of process (about 25%). The goal has been reached: the process time has been reduced of 28%.