Üretim sistemlerinin gözlemlenen olaylar arası sürelere dayalı modellenmesi ve kontrolü: Analitik ve ampirik bir araştırma

Technological advances allow manufacturers to collect and access data from a production system effectively. The objective of data collection is deploying the collected data in developing decision support systems for performance evaluation, problem identification, and production control. The goal of this dissertation is to investigate how the collected data can be used to evaluate performance and optimize manufacturing systems, analytically and empirically. In the first part of the thesis, I investigate the question: \textit{How can the collected data from the shop-floor be used in efficient control and design of manufacturing systems?} In order to investigate the impact of possible dependency in the inter-event times on the optimal control and performance measures of the system, first, a manufacturing system that is controlled by using a single-threshold production control policy is analyzed. It is shown that ignoring autocorrelation in interarrival or service times can lead to overestimation of the optimal threshold level for negatively correlated processes, and underestimation of the optimal threshold level for the positively correlated processes. Then the optimal control problem of a production/inventory control problem system with correlated inter-arrival and service times modeled as Markovian Arrival Processes is considered. It is shown that the optimal control policy that minimizes the expected average cost of the system in the steady-state is a state-dependent threshold policy. In the second part of the thesis, an exploratory data analysis is conducted by using a large industrial data set that includes 17 million rows of data related to flow of 17000 unique products that are processed in 500 different machines at a semiconductor manufacturing plant. The product flow dynamics that include inter-arrival, service, and inter-departure distributions and autocorrelations, and work-in-process and cycle time dynamics are investigated at different levels of detail. Then, the data-driven cycle time prediction problem is considered. In order to develop effective prediction methods, a methodology is developed to determine the most important product and system-state related features. The performance of different prediction algorithms is compared by using the selected features. The analytical and empirical results presented in this dissertation show that the effective use of the collected data from a manufacturing system enables controlling the manufacturing system effectively and predicting its main performance measures accurately.