(A) study on design knowledge extraction and reuse based on text data using semantic processing

Abstract

During the product design stage, engineers should make various design decisions considering product quality, cost, and delivery. In order to make effective and efficient design decisions, reuse of existing design knowledge is necessary. Explicit design knowledge in the manufacturing companies is generated in various forms such as documents, product specifications, and CAD models. In particular, a large amount of design knowledge is generated in the form of text data such as design documents, technical reports, and failure analysis documents. Although most manufacturing companies store a vast amount of text data, design knowledge extraction and reuse from the text data depend on relevant document search using keywords. Therefore, it is very urgent to develop methods for extracting semantically structured design information and knowledge automatically, and methods for design knowledge reuse. This study proposes an approach for automatic design information and knowledge extraction and reuse from text data based on semantic processing. First, a domain ontology is constructed to extract semantic information from domain-specific text data of manufacturing companies. A domain ontology is a semantic reference model in which concepts and their relationships that are used during product development are defined. Second, design information is extracted from text data. Design information in text data of the manufacturing companies is mainly described using nouns. Thus, component information of product, property information of the component, and value of the property are extracted and structured using noun-based approach. Extracted design information is used for design document based CAD model retrieval, in which similarity between design information of an input design document and design information of CAD models is calculated. Next, design knowledge is semantically structured from unstructured design knowledge in text data based on combining context information and design information in sentences. Since a sentence contain various context, same design information can be interpreted in different meaning according to context. Context information is extracted using linguistic patterns that are generated based on information needs and analyzation of text in documents. This study focus on extracting failure knowledge that is one of important design knowledge using failure context from failure analysis documents. Lastly, Extracted design (failure) knowledge is reused in the design stage. Interrelations between components and failures are extracted using quantitative analysis. Rules for failure prediction are automatically generated. Potential failures are predicted by rule reasoning and input design information, and design recommendations are provided to engineers. In order to evaluate performance and feasibility of the proposed approach, tire design documents, tire CAD models, and tire analysis documents are used as examples.