Integrating data mining and simulation optimization for decision making in manufacturing

Abstract

Manufacturers are facing an ever-increasing demand for customized products on the one hand and environmentally friendly products on the other. This situation affects both the product and the process life cycles. To guide decision-making across these life cycles, the performance of today's manufacturing systems is monitored by collecting and analyzing large volumes of data, primarily from the shop floor. A new research field, Data Mining, can uncover insights hidden in that data. However, insights alone may not always result in actionable recommendations. Simulation models are frequently used to test and evaluate the performance impacts of various decisions under different operating conditions. As the number of possible operating conditions increases, so does the complexity and difficulty to understand and assess those impacts. This chapter describes a decision-making methodology that combines data mining and simulation. Data mining develops associations between system and performance to derive scenarios for simulation inputs. Thereafter, simulation is used in conjunction with optimization is to produce actionable recommendations. We demonstrate the methodology with an example of a machine shop where the concern is to optimize energy consumption and production time. Implementing this methodology requires interface standards. As such, this chapter also discusses candidate standards and gaps in those standards for information representation, model composition, and system integration.