고장 분석 문서에서 자동 고장 지식 추출

Most engineers in manufacturing industries routinely handle large volumes of engineering documents that are stored in PLM system. The challenge for extracting knowledge from the engineering documents is to quickly gather the right information from unstructured text in engineering documents into a structured knowledge. Since context is very important information to understand knowledge in unstructured text, it should be extracted and reflect to organize a structured knowledge. Several studies to extract knowledge from unstructured text focus on extracting concepts and their relationships using matching between terminologies in text and concepts in a predefined ontology. They did not extract context information from unstructured texts. This paper undertakes the in-depth analysis of failure analysis documents collected from a manufacturing company and provides a method for failure knowledge extraction based on failure context, which enables to prevent potential failure in a design process by securing the failure knowledge. Two-layered model, failure-context and product-component layers, is also developed to exploit both failure knowledge and relevant design information through a series of proposed linguistic operations.