Developments in Monitoring Dynamic Data
Major technological advances in sensors and mobile devices have led to the availability of high volume – high frequency data originating from technical and social networks. These data offer opportunities to perform real-time monitoring with respect to both physical quantities and structural information. Different application areas for real-time monitoring include manufacturing processes, public health, finance, and telecommunication and communication networks.
Techniques for monitoring dynamic data (i.e., data changing over time) are known under different names depending on perspective or scientific community: statistical process control, surveillance, changepoint detection. Specific changes in such data are indicated with words such as anomalies, trends and drift.
In this talk I will first introduce the statistical concepts behind monitoring dynamic data. Then I will discuss which concepts and methodologies need to be adapted in the case of data streams that have higher levels of volume, high velocity and/or variety than the traditional levels for which these concepts and methodologies were developed.
I will illustrate some of the concepts with a maintenance case study on wind turbines performed within the Dutch national project DAISY4OFFSHORE.