27 駐在監督報告書からの不具合知識獲得手法に関する研究

概要

論文の詳細を見る
A method to retrieve knowledge of trouble in the workshops is proposed in this paper. This method can collect similar troubles from a collection of surveyor's reports composed of natural language using text-processing and ontology. This method has two steps. First, text-processing for the surveyor's reports is performed. By analyzing the result of text-processing, the information on trouble in the fabrication is normalized to a pair of component and trouble names. Second, the words which have similar meanings with the trouble information are listed by ontology. The trouble information and the listed words are then stored in a database as one record. This database is called failure information database. Classification of similar troubles can then be performed by searching the failure information database based on the surveyor's reports. The presented method is verified by using actual reports in a shipyard. The result shows that the method can obtain knowledge. For example, it was found that machine parts are often located in wrong or inappropriate positions. This kind of knowledge is useful in the design phase to avoid committing the same mistakes.