Data Science for industrial time series
Discover Data Science for industrial Time Series
7 steps to understand our Data Science methodology for blind fault detection and sensor data exploration.
Our Data-Scientists show you how to make the most of your data, and how to perform in-depth exploration with DiagFit.
Through our Chalk & Talk concept, we aim to introduce simple Data Science concepts through short tech videos, hosted by our team. Our goal is to explain our know-how and innovations within the DiagFit software in a highly popularized way.
1. The “blind” approach
Characterize nominal equipment operation to create the normality space.
2. Characteristic generation
Extract more information from the time series to generate an accurate and more robust normality space.
3. Equipment health score
Define the boundary of the normality space to obtain the best compromise to meet the challenges of the use case.
4. Labeling anomalies
Improve the model with new data and label unhealthy behaviors.
5. Time series analysis and processing
Clean a sensor signal to extract relevant information.
6. Signal decomposition and virtual sensors
Filter signals from industrial time series and create virtual sensors.
7. From sensor subgroups to the overall health indicator
Model the healthy behavior of equipment by breaking down the overall behavior into simpler sub-elements.
From the analysis of raw signals to the generation of one or more anomaly detectors in blind mode, all these Data Science steps are present in DiagFit. Thanks to these semi-automated functions, business experts can make the most of their data.
