Embedding a back propagation network into fuzzy c-means for estimating job cycle time: wafer fabrication as an example

Abstract

Cycle time is the time required for a job to be processed by a factory including the time required for processing, transportation, and waiting. Estimating the cycle time of each job is a critical concern in managing a factory. To address this concern, classifying approaches in which jobs are classified before or after forecasting their cycle times have recently been proposed. However, none of these approaches can guarantee the compatibility of the job classifier with the forecasting mechanism. To overcome this difficulty, a new classifying approach is proposed in this study. In the proposed methodology, the training of the forecasting mechanism is embedded into the iterations of the job classifier. Consequently, the classification and forecasting stages interweave with each other, improving their compatibility. The proposed methodology was tested using data on 120 jobs. According to the results, the proposed methodology surpassed five existing methods in forecasting accuracy. Compared with two existing classifying approaches, the proposed methodology statistically significantly reduced the mean absolute percentage error by 56 and 38 % for testing and unlearned data, respectively.