Extracting a Unified Database from Heterogeneous Bills of Materials via Data Integration Approaches

Since the beginning of the information era, manufacturing companies have collected a huge amount of data and acknowledged the importance of proper collection and curation policies. Nevertheless, an open problem is the exploitation of historical and legacy data. Such data is heterogeneous: it has often been accumulated throughout decades by different curators with different degrees of attention toward the data. In this work, we focus on the historical data collected by Bedeschi S.p.A., an international company dealing with the design and production of turn-key solutions for Bulk Material Handling, Container Logistics and Bricks. Our task is to extract knowledge, available under the form of partially structured Bills of Materials (BoMs), to distil a unified database over a mediated schema that can be used by designers and sales department to rapidly have a global overview of the products developed over the years. We present a feasibility study investigating the performance of several traditional approaches of data integration applied to our specific task. In particular, we treat BoMs as sources that need to be aggregated in the unified mediated schema. We achieved satisfactory results indicating the direction that the company should follow to properly extract knowledge from their historical data.