Graph of a model

Example of graph design using g2sys-generated model of a hydraulic rig.

In this section, a movie is proposed that shows how to build the graph of a system using the following steps:

  1. Use g2sys to find the parsimonious static relationships between sensors. This step is shown in a accelerated manner as this functionality has been shown elsewhere in this site1.

  2. Each time such a static relationship is found, the sensor used as label is removed from the next steps. This is because this sensors can be expressed as a static relationship involving the other sensors and hence it should be useless in discovering the remaining relationships.

  3. One no more good-enough static relationships can be found and all the sensors involved (as labels) in the previously found static relationships have been removed. Some of the remaining sensors might represent the state components of the system and their dynamic relationships can be looked for using the g2sys parameter recursive set to True.

These steps are shown in the following video that uses the publicly available Zema dataset that is available at UC Irvine Machine Learning Repository.

Works also with pwpol

Notice that the implicit relationships provided by the piece-wise polynomial modelling tool can also be visualized as a graph of connection. This uses mainly the same prinicple with the two differences:

  1. There is currently no static vs dynamic relationships.
  2. The graph cannot be used to design digital twins since the relaionships are not explicit.

Footnotes

  1. See for instance the movies of the g2sys used for the hydraulic rig system, the industrial valves and the compressor system.↩︎