The vf-OS Data Analytics module, this task covers the creation of the building blocks for the analytical processing of sensor inputs. This includes machine learning algorithms supporting supervised and unsupervised scenarios. A broad and diverse range of sensors provides a massive amount of data as inputs for the Data Analytics component. The data from the sensors will be published in real-time as a data stream and can also be stored in databases as Historical data. The Historical data is used by the Machine Learning (ML) algorithms to build models which will be able to make Anomaly detection and Predictions in real-time on a data stream. The result of the processing is provided in a form that is useful for decision-makers and for software systems able to act on the provided analytics.
The tasks required for processing streaming data include Anomalies detection, Prediction of the future evolution of the data, Prediction for systems control, and Interpretation of data and visualisations.
The below figure shows the relationship between the Data analytics components (and its sub-components) and the functionalities provided as one of the vf-OS components.
This vf-OS module is subdivided in lower-level modules:
Historic Data Analytics refers to the techniques and processes that are related to the investigation of large sets of data. Within the arena of manufacturing, data analytics can be used to examine historical process data, and then identify patterns and relationships within that data, so as to optimise the factors that are seen to have the greatest impact on production. By adding data analytics processes, a company may significantly improve product quality and yield. Furthermore, the use of Predictive Analysis can also reduce the amount of downtime that is created by malfunctions, that may not be noticed otherwise.
Streaming Data Analytics subcomponent receives data from the PubSub Module (probably coming from some sensors) and performs calculations using user-defined rules to decide if the data received can be classified as an anomaly. Rules are defined using variables, mathematical functions and relational operators. In order to simplify and organize the definition of rules in a company, the user can define several kinds of entities (datasets, modules, sentences, thresholds). A REST API is provided to create, edit, query and delete every entity. The definition of the entities is stored in a relational database, using the vf-OS relational storage component.
Additionally, this page includes a Tickets entry to handle specific issues regarding this module as a whole (which may include management, integration, new features, etc.).