Variability is a major factor associated with the costs of poor quality performance. Improving process stability and manufacturing capability by reducing variability is the key to competitiveness for manufacturing and service enterprises. Another key is identifying and meeting customer needs, which can be accomplished through extracting implicit, previously unknown and potentially useful information from data. Due to advances in measurement and data collection technology, abundant in-process measurement data is available.
This data has huge potential for monitoring, diagnosing, and controlling variation in manufacturing and business processes. However, current data processing methods have limitations and cannot take full advantage of the complex measurement data: (1) the spatially and/or temporally dense structure of the data often violates assumptions required for existing methods, (2) the various types and high dimensionality of the data outstrip the capabilities of traditional statistical, graphical, and data visualization methods, and (3) the current focus of statistical process control (SPC) on monitoring, as opposed to root cause identification and elimination, overlooks diagnostic information buried in the massive quantity of data. Six-sigma methods are becoming ubiquitous in manufacturing and other industries, in efforts to reduce variability. However, current six-sigma tools are not designed to fully utilize large quantities of in-process measurement data.
The research objective of the QUAD laboratory is to develop a general framework for effectively and efficiently extracting potential information and monitoring, diagnosing, and controlling variation in data-rich environments such as manufacturing, computer and network systems and the healthcare and service industries. We believe that this will provide a basis for achieving enterprise goals in modern data-rich manufacturing and business processes.