Kim Phuc Tran - Control Charts and Machine Learning for Anomaly Detection in Manufacturing

Todays modern systems have become increasingly complex to design and build, while the demand for reliability and cost effective development continues. Reliability is one of the most important attributes in all these systems, including aerospace applications, real-time control, medical applications, defense systems, human decision-making, and home-security products. Growing international competition has increased the need for all designers, managers, practitioners, scientists and engineers to ensure a level of reliability of their product before release at the lowest cost. The interest in reliability has been growing in recent years and this trend will continue during the next decade and beyond.

The Springer Series in Reliability Engineering publishes books, monographs and edited volumes in important areas of current theoretical research development in reliability and in areas that attempt to bridge the gap between theory and application in areas of interest to practitioners in industry, laboratories, business, and government.

**Indexed in Scopus and EI Compendex**

Interested authors should contact the series editor, Hoang Pham, Department of Industrial and Systems Engineering, Rutgers University, Piscataway, NJ 08854, USA. Email:hopham@rci.rutgers.edu, or Anthony Doyle, Executive Editor, Springer, London. Email:anthony.doyle@springer.com.

More information about this series at http://www.springer.com/series/6917

This Springer imprint is published by the registered company Springer Nature Switzerland AG

The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Anomaly Detection is a set of major techniques with an aim to detect rare events or observations that deviate from normal behavior. Process monitoring and Anomaly Detection are becoming increasingly important to enhance reliability and productivity in manufacturing by detecting abnormalities early. For example, a vibration level in an electric motor exceeding the permissible threshold can be considered as an anomaly, it might not be considered as a fault. However, if the vibration level continues to rise and leads to motor destruction, it can be considered as faulty. Therefore, Anomaly Detection can provide advantages to manufacturing companies by reducing their downtime due to machine breakdowns by detecting a failure before this results in a catastrophic event that may cause degradation of the process and product (Lindemann et al. []). There have been various data-driven and model-based approaches to detect anomalies occurring in manufacturing systems. The most common approach to Anomaly Detection includes Control Charts and Machine Learning methods. In manufacturing, Control Charts are effective tools of Statistical Process Control (SPC) for continuously monitoring a process as well as detecting process abnormalities to improve and optimize the process. There are many different types of Control Charts that have been developed for this purpose. In addition, Machine Learning methods have been used a lot in detecting anomalies with different applications in manufacturing such as, detecting network attacks, detecting abnormal states in machines. Finally, the interference of these two techniques is also found in literature such as the design of Control Charts, anomaly signal interpretation, and pattern recognition in Control Charts using Machine Learning techniques.

In recent years, the rapid development and wide application of advanced technologies have profoundly impacted industrial manufacturing. The recent development of information and communication technologies such as smart sensor networks and the Internet of Things (IoT) has engendered the concept of Smart Manufacturing (SM) that adds intelligence into the manufacturing process to drive continuous improvement, knowledge transfer, and data-based decision-making. In this context, the increasing volume and quality of data from production facilitate the extraction of meaningful information, predicting future states of the manufacturing system that would be impossible to obtain even by human experts. Due to recent advances in the field of SPC, there are a lot of advanced Control Charts that have been developed, thus SPC can become a powerful tool for handling many Big Data applications that are beyond the production line monitoring in the context of SM Qiu [] categorized anomalies in machines, controllers, and networks along with their detection mechanisms, and unify them under a common framework to allows the identification of gaps in Anomaly Detection in SM systems that should be addressed in future studies solutions.