MizoPol

A set of tools for identification, normality characterization and anomaly detection in industrial time-series.

---

MizoPol is a set of python modules¹ that addresses the problem of identifying (piece-wise) polynomial relationships between time series. This identification is detrimental in solving the following problems:

• Anomaly detection by normality characterization
• Understanding the coupling between sensors beyond linear correlations
• Identifying dynamics governing nonlinear unmodelled systems
• Creating virtual sensors
• Generating parametric subtil defaults to auto-train models
• help designing digital twins

Domains of application

Achieving the above listed tasks paves the way towards addressing many important domains of interest in industry such as:

• Predictive maintenance.
• Nonlinear control design.
• State / parameter estimation of nonlinear systems.
• Introducing relevant parametric anomalies beyond bias or relative error
• Context detection.

We’ll come back to these applications as we explain the different modules and their illustrative use-cases.

Reading-frienly introductory slides

A reading-firendly set of slides introducting the MizoPol suite of tools, their advantages, raison d’être compared to existing options can be downloded here

Footnotes

1. MizoPol is not rigorously speaking a python package. This is the reason why it is referred to as a python suite or a set of modules. Nevertheless, it is possible that the work package is used here and there by mistake.