Data analytics for time constraint adherence prediction in a semiconductor manufacturing use-case

摘要

Semiconductor manufacturing represents a challenging industrial environments, where products require more than several hundred operations, each representing the technical state-of-the-art. Products vary greatly in volume, design and required production processes and, additionally, product portfolios and technologies change rapidly. Thus, technologically restricted rapid product development, stringent quality related clean room requirements and high precision manufacturing equipment application enforce operational excellence, in particular time constraints adherence. Product specific time constraints between two or more successive process operations are an industry-specific challenge, as violations lead to additional scrapping or reworking costs. Time constraint adherence is linked to dispatching and currently manually assessed. To overcome this error-prone manual task, this article presents a data-based decision process to predict time constraint adherence in semiconductor manufacturing. Real-world historical data is analyzed and appropriate statistical models and scoring functions derived. Compared to other relevant literature regarding time constraint violations, the central contribution of this article is the design, generation and validation of a model for product quality-related time constraint adherence based on a real-world semiconductor plant.