Level: Intermediate
【Presenter】
Akira Hososhima, Representative of Tobayashi Consulting
[Presentation summary]
There are notable use cases of text mining methods in the field of engineering, such as quality trend analysis and reports that optimize maintenance by text mining maintenance records. This report introduces a case study in which quality and productivity problems were discovered and the cause investigated through text mining of production performance data.
Production performance data includes dozens of variables such as machine and operator numbers, processing conditions, material conditions, and special notes. Special notes are text data that was manually entered by the operator when an irregular event occurred, and the content includes words unrelated to the abnormal event, so they are registered as stop words and excluded from analysis. Then, synonyms such as “mistake, mistake, mistake” are replaced with a single word by recoding. Through this preprocessing, the presence or absence of frequently occurring abnormal words is stored in a column, and the presence or absence is analyzed in relation to production conditions and two variables to identify abnormalities such as machine, operator, processing conditions, material conditions, etc. The correlation between the two conditions was revealed, and the cause of the abnormality could be investigated.
