[2H1-J-2] Machine learning: modeling in businesses

[2H1-J-2-04] Productivity Improvement through Yield Prediction and Machine Combination Optimization

Yoshiaki Suzuki¹, 〇Manabu Kano¹, Akira Soga², Takeshi Yanagimachi², Ryo Murao², Masaya Takaki² (1. Kyoto University, 2. Toshiba)

Keywords:Machine Learning, Production Scheduling

In a multi-process production system, the yield rate of the final products depends not only on the goodness of each machine but also on that of machine combinations at different stages. To maximize the productivity, it is crucial to use good machine combinations by priority. In the present work, we proposed a modeling method that can estimate the yield rates of unused machine combinations and a production scheduling method that can optimize machine combinations by taking account of the yield rates. The proposed modeling method uses field-aware factorization machines (FFM). A case study demonstrated that FFM can estimate the yield rates with great accuracy even when yield rates are available in only 20% of all machine combinations and also that the proposed scheduling method improved the productivity by more than 10%.