Job cycle time estimation in a wafer fabrication factory with a bi-directional classifying fuzzy-neural approach

Estimating the cycle time for every job in a factory is a critical task. It was recently reported that job classification noticeably enhanced the accuracy of job cycle time estimation. In pre-classifying approaches, whether the pre-classification approach combined with the subsequent estimation approach is suitable for the data is questionable. Conversely, the difficulty in classifying a job according to only the estimation error not the various attributes is a problem to post-classifying approaches. To tackle these problems, a bi-directional classifying fuzzy-neural approach is proposed in this study. In the proposed methodology, jobs are not only pre-classified but also post-classified. The results of pre-classification and post-classification are aggregated into a suitability index for each job. A job is then assigned to the category to which its suitability index is the highest. A radial basis function network is also constructed to predict the suitability index of a job according to the various attributes. To evaluate the effectiveness of the proposed methodology, a practical example was used in this study. According to experimental results, the estimation accuracy of the proposed methodology was significantly better than those of many existing approaches.