Documents modeling and similarities searching - Proposition of a framework for the management of defect documents to optimize corrective and preventive maintenances

The daily practice of an activity generates a set of knowledge that results in a know-how, a mastery, a skill a person gains over time. In order to take advantage of this experience, capitalization of knowledge has become an essential activity for companies. Our research work aims to model and implement such a system that extracts and formalizes knowledge from defects that occur in the context of industrial production, and to integrate it into a framework in order to facilitate corrective and preventive maintenance. This framework organizes the knowledge in the form of defects' groups. These groups can be compared to patterns: they represent a problem to which one or more solutions are related. They are not defined a priori; the analysis of past defects generates relevant groups, which may change with the addition of new defects. To identify these patterns, a complete process of knowledge extraction and formalization is adopted, Knowledge Discovery in Databases, well known in the domain of knowledge management. This process has been applied in very diversified fields. In this work, we give a new dimension to this process, the processing of defects, especially those that occur during industrial production processes. The generic steps that compose it, from the simple data selection to the interpretation of patterns that support knowledge, are considered. A specific processing, relevant to our applicative context, is assigned to each of these steps.