Semiconductor: Incorporating Nonlinear Phenomena in Semiconductor Supply Chain Planning Models

Purdue University, West Lafayette IN

Investigators

Abstract

The grant supports scientific and mathematical aspects of a joint project co-sponsored by NSF and the Semiconductor Research Corporation (SRC), with data-gathering, validation and testing supported separately by SRC. The combined effort will investigate a series of refinements and extensions to the investigators prior work on production planning models using nonlinear Clearing Functions (CFs), which express the expected throughput of a resource as a concave function of the expected work in process inventory (WIP). Extensive computational experiments have shown that this model is capable of capturing the interaction between resource utilization and lead times in manufacturing environments while retaining the tractability for large-scale use. New research will address how CFs can be fitted to various workcenters without the burden of simulations and enhance the models to encompass three new elements impacting the dynamic of complexity production operations: lot-sizing/setups, safety stock, and dynamic/promotional pricing. Taken separately, each of these has been shown to significantly impact congestion, throughput, and lead times. The new research will attempt to exploit the flexibility of CF-formulations to derive an integrated multiperiod planning model incorporating all three. In high-technology, capital-intensive industries such as semiconductor manufacturing, global supply chains, short product cycles and rapidly changing market conditions render effective supply chain coordination a requirement for success. Applications of company-wide production planning models that support this coordination have been repeatedly shown to yield significant benefits. Still, most methods tractable enough for large-scale application are unable to model the congestion and lead-time dynamics of real operations. Success in this research will make the already promising CF approaches more ready to model large and complex operations, as well as incorporating other issues now handled externally. The result could be a significant improvement in productivity and service levels of critical manufacturing industries.