Abstract
Predictive maintenance (PdM) is a key application of data analytics in semiconductor manufacturing. The optimization of equipment performance has been found to deliver significant revenue benefits, especially in the wafer fabrication process. This chapter addresses two main research objectives: first, to investigate the particular challenges and opportunities of implementing PdM for wafer fabrication equipment and, second, to identify the implications of PdM on key performance indicators in the wafer fabrication process. The research methodology is based on a detailed case study of a wafer fabrication facility and expert interviews. The findings indicate the potential benefits of PdM beyond improving equipment maintenance operations, and the chapter concludes that the quality of analytics models for PdM in wafer fabrication is critical, but this depends on challenging data preparation processes, per machine type. Without valid predictions, decision-making ability and benefits delivery will be limited.
TopIntroduction
The semiconductor industry (SI) is one of the most capital-intensive industries, with significant capital investment in equipment, and optimization of equipment performance has thus received noteworthy attention. SI manufacturing processes generate large amounts of metrology data that can be used to analyse and understand failure patterns and to improve the yield of high quality products (Munirathinam & Ramadoss, 2016). Speaking at the January 2000 Industry Strategy Symposium, the former Intel Senior Vice President, Michael Splinter stated that one hour of downtime for a critical unit of process equipment could be translated into $100,000 of lost revenue. Hence, in a generic wafer fabrication facility, a downtime reduction of only 1% on the 50 most critical tools can lead to revenue opportunities and cost savings of around $100 million annually. By improving response times and repair times, and by predicting the point in time when problems may occur, a reduction of unscheduled downtimes can be achieved (Munirathinam & Ramadoss, 2014).
Rødseth & Schjølberg (2016), Chiu et al. (2017) and Motaghare et al. (2018) are amongst a number of authors who see predictive maintenance (PdM) as a data-driven approach to address these goals. At its core is the use of advanced analytics capabilities to support the development of maintenance strategies and activities. This chapter assesses the potential impacts of a PdM strategy on the equipment maintenance (EM) operations and manufacturing performance in the SI, and is based on primary research undertaken in the period 2016-2020 (Preis, 2021). Following this introduction, relevant background information is set out. The research methodology, based on a case study at a wafer fabrication facility is then explained, and in the following section the manufacturing process at the case study company is outlined. The findings from the case study are then set out and analysed. Finally, the research questions are addressed, the main themes emerging from the study are reviewed, and future areas for research in this field are discussed.
TopRelevant Literature
Data analytics has become a critical capability for modern businesses. For instance, big data analytics have been utilized in the context of the financial stability of Asian banks (Zhu & Yang, 2021), and the performance of preventive medicine in the healthcare industry (Ivan & Velicanu, 2015). Gronwald (2015) classified data analytics into five categories: (1) descriptive analytics, (2) predictive analytics (PA), (3) prescriptive analytics, (4) sentiment analysis (SA) and (5) text mining. He defined each category with different orientations, techniques and goals, indicating the necessity of defining the goal before selecting an appropriate approach. For instance, prescriptive analytics focuses on the underlying causes as well as the predicted result, while PA focuses only on the predicted result. In addition, this classification provides a clear overview of the various capabilities of data analytics. Among others, important technologies that enable data analytics are NoSQL databases, knowledge discovery tools, in-memory data fabric, and data integration and data quality tools (Maruti Techlabs, 2017). According to the recent Gartner Magic Quadrant for Business Intelligence and Analytics tools, some of the most important tools in this area are Microsoft Power BI, Tableau and Qlik (King, 2021).
Key Terms in this Chapter
Flow Factor (FF): A multiplier that states how much the CT exceeds the raw process time.
Wafer Fabrication: The frontend part of the semiconductor value chain that is concentrated on the actual chip technology.
Predictive Analytics: A fuzzy concept that methodically overlaps with Data Mining and focusses on the prediction of future events.
Equipment Maintenance (EM): A discipline in the manufacturing business which aims to avoid or respond quickly to equipment failures under consideration of maintenance costs.
Meantime to Repair (MTTR): The average duration to repair a failure at a particular workcenter within a defined timeframe.
Predictive Maintenance (PdM): A data-based approach for reducing the number of unscheduled downtimes by fostering the prediction and prevention of machine failures.
Yield (Wafer Fabrication): Refers to a) the percentage of wafers that successfully leave the production versus wafers that entered the production or b) t he percentage of chips that pass the functional tests.
Enterprise Resource Planning: A mainstream business software package that integrates multiple business processes based on a single consistent database that stores enterprise resource data.
Work in Progress (WIP): The non-finished goods in a factory which require further processing before they are delivered to the customers.
Cycle Time (CT): The duration required to produce a good, considering all process steps and wait times.
Meantime Between Failure (MTBF): The average duration between failure events at a particular workcenter within a defined timeframe.