What Is the Research About?
At its heart, this research tackles one of the most pressing challenges in the semiconductor industry: How can we effectively manage and schedule operations in Research and Development (R&D) facilities?
R&D semiconductor manufacturing systems differ significantly from conventional manufacturing fabs. These differences stem from the limited characterization and control of factory processes, the scales of production control, and the range of automatable decision processes.
Considering these complexity factors, this study focuses on enhancing scheduling operations in R&D fabs. It specifically challenges, evaluates, and improves an existing dispatching rule-based heuristic applied in real-life settings.
What’s New About This Research?
Managing R&D semiconductor fabs is no easy feat. Let’s break down the core differences:
- Scale of Production
- R&D fabs juggle a wide variety of products, produced in small quantities,
- Specific to R&D systems, the concept of a campaign can be used to launch production events.
- Quantitative operations management systems
- No bill of material is available to manage material procurement and usage operations.
- The concept of a product line is generally used to express and manage the processes’ requirements.
- Production Control
- R&D experiments often take unpredictable paths, requiring frequent reworks, adjustments, and inspections,
- Limited historical data and incomplete process routes add to complexity.
- Automation Challenges
- The deployment of automation is not so straightforward as in production fabs, where operating conditions are known beforehand, and manufacturing schemes are predictable,
- R&D fabs operate in the realm of “unknown unknowns,” demanding custom automation solutions to support decision-making.
In this study, an existing dispatching rule-based heuristic, running in R&D settings, is challenged, investigated, and improved.
Practical Applications of the Research
Numerical experiments are conducted based on real-world data from CEA-Leti, a leading European R&D center for semiconductor manufacturing. Several levers of improvement are considered in the study, including monitoring Key Performance Indicators (KPIs) including monitoring cycle time, throughput, and on-time delivery, analyzing processing time variability, and assessing the quality of applied dispatching rules.
What’s Next?
This study lays the groundwork for further research to:
- Better Decision-Making
- Integrating detailed knowledge of product routes to enhance both toolset-level and fab-wide scheduling strategies.
- Savvy KPIs
- Rethinking conventional performance indicators to better align with the high-level of uncertainty, low-control environment of R&D fabs.
- Production Efficiency
- Measuring the production efficiency in R&D environments characterized by a low level of production control conditioned by technological reasons.
Looking ahead, the adoption of simulation and optimization techniques will play a pivotal role in developing smarter, more adaptive scheduling solutions tailored to the dynamic nature of R&D fabs.
Why It Matters for ACCURATE
By tackling the complexities of scheduling operations in semiconductor R&D fabs, this research supports the broader vision of smart manufacturing. Easy-to-use, real-time data exchange and performance monitoring, as well as automation of decision processes within manufacturing systems are among the main specifications of Manufacturing as a Service.
