| Discrete event sequences record different types of events with corresponding time stamps of their occurrences in the system. They are very common in practice. For example, in medical diagnostic imaging systems, the events related to various machine activities and behaviors, critical system failures, operator/user actions, task status, etc. are collected automatically when the system is in use. As another example, in service or production systems, activities such as "customer arrival", "service start", "customer departure" for each customer can also be recorded and form a sequence of discrete events. Despite the differences in context, these discrete event sequences share the same characteristics: (i) different events are usually inter-correlated, and their occurrence times are often randomly distributed; (ii) the mechanisms generating these events are not readily available. However, it is generally believed that these data provide rich information regarding the system working conditions and could be used for condition monitoring, diagnosis, and optimal maintenance. The rapid development of computing, communication, and sensing technologies has provided us numerous opportunities to utilize discrete event data for the improvement of quality and productivity.;In this research, we try to develop a systematic methodology to integrate engineering domain knowledge and advanced statistical methods into a unified framework for better modeling and analysis of the discrete event data with stochastic evolutions. Our research focuses on the distributions of time between events (TBE), with context varying in different applications. Numerical results and case studies from real industry applications demonstrate the effectiveness of our methods in modeling and analyzing the event data. The proposed methodologies can achieve systematic utilization of discrete event data and possess wide applicability to various engineering systems. |
