Detecting anomalies with a control chart
A control chart defines boundaries for a normal signal based on the statistics of the training set, which, in this case, refers to values from a correctly working system. During deployment, an alarm system is built to warn when the signal is exceeding the boundaries.
We will introduce and build a control chart based on the IoT data using the following steps:
- Introducing a control chart
- Implementing a control chart
- Deploying a control chart
In the first subsection, we will show you how to define and visualize a control chart.
Introducing a control chart
A control chart defines the normal functioning of a process by a statistical range. We define this range by calculating the following statistics over the training window:
- Upper limit: Mean+12*Standard Deviation
- Lower limit: Mean-12*Standard Deviation
Once these boundaries have been set, we can visualize the control chart in a line plot. As an...
