Abstract
Data-driven models, and maturating data-treatment tools, now allow for automation in a wide range of domains. The data-rich domain that is NC machining presents a strong potential for automation, in this regard, and the smart factory calls for knowledge capitalisation in manufacturing.
The field of machining saw numerous data formats, and is still led by heterogeneous sources. To exploit this massive amount of data, capitalisation of information from those sources is compulsory. Such system should allow aggregation of different sources and formats, and accommodate for information about diverse elements regarding manufacturing – Parts, tools, machines…
In this paper, automated process planning generation methods are reviewed. The localisation and formalisation of data in manufacturing documents is considered, and emphasis is made on linking trajectories and features, to define strategies. Attempts to use machine learning techniques in manufacturing are reviewed, and the lack thereof in manufacturing knowledge retrieval is underlined.
Then, a framework is proposed to realise knowledge retrieval from legacy programs, and several locks and improvement possibilities are identified.