Process Monitoring and Improvement in a Realistic Scenario

With the technological advancements and popularization of Six Sigma tools, the number of defects/faults in products and services observed today is no more than a few per thousands. But with time and increasing competition, the cost per faulty unit/service has also increased many-folds. This is because dissatisfied customers are more likely to communicate grievances, commonly known as ‘negative word of mouth’. To sustain and thrive in today’s competitive markets, it is therefore essential to continuously improve the quality levels by reducing faults. Current developments in quality control and data mining literatures are insufficient in adequately addressing together the objectives of fault detection and diagnosis, when dealing with high dimensional, highly imbalanced (1:1000 faulty-to-normal ratio) nature of process monitoring data. This research develops a knowledge base for adequately addressing these issues by building on the quality control literature and recent developments in data mining.