Quality control and anomaly detection in production

Quality control and anomaly detection in
production is the content of nugget 11, developed by the Czech Technical
University of Prague. Learning outcomes of this nugget are: drawbacks of classical
statistical process control methods, concept of one class
classification and application of one class
classifier to statistical process control.

One of the problems
with classical control charts is that nowadays, hundreds of variables are
commonly measured, and the usage of hundreds of individual uni-variate control
charts is unacceptable. Further, data is often not normally distributed and
structural patterns change by the underlying distributions. Machine learning
based methods are more flexible and robust to these mentioned problems, however
they need enough training data to lead to correct results. Anomaly detection is
explained by several charts. A practice with Python and Excel is part of the
learning.<br>