برآورد زمان پردازش در سیستم های تولید زمینه آگاه

Due to high volatility and dynamics in today's markets, manufacturers are required to react more quickly (e.g. frequent changeovers of products) to changing environments, but still ensure efficient production plans and accurate throughput estimates. Since, generally, processing times of machines are not fixed and depend on several variables, such as product type or material quality, accurate estimations of these times need to be available as basis for implementing production plans. Therefore, calculating high-accuracy estimates of processing times of all activities involved in manufacturing is an important task. The employment of statistical learning models (e.g. regression) for processing time estimations is not straightforward, since there are many situational dependencies to take into account. We refer to such dependencies as manufacturing context and propose a framework that integrates context information into a flexible production planning and scheduling scenario. We show that for frequent product changeovers, it is crucial to continuously adapt processing time estimators to the corresponding situation. Our approach shows increasing stability and accuracy of processing time estimates compared to state-of-the-art adaptive learning models. These estimates ensure more reliable production plans, and furthermore, context also reveals variables that influence estimation accuracy.