Introduction
The Data Historian Accelerator captures real-time telemetry from data historians like OPC UA and OSI PI. A custom HTML5 web interface provides the user the ability to visualize the object hierarchy of the historian, and create subscriptions on the nodes or tags of interest. The accelerator receives this data in real-time and assembles the points into logical data sets that can then be passed to business rule modules that implement decision tables or data science models. The data and model output are streamed into Live Datamart for visualization in Spotfire.
The Intelligent Equipment Accelerator is a similar offering, but is a generic platform for any data provider, whereas the Data Historian Accelerator focuses specifically on systems like OPC UA and OSI PI.
And here's a video showing how the Accelerator works with OSI PI.
Download
The Data Historian Accelerator can be downloaded from the Exchange.
Business Scenario
Data Historians are applications that retrieve production and process data from manufacturing and other process-oriented systems. They store data in an efficient database reducing the requirement for large amoutns of disk space. They also provide quick access to the data through API-based queries.
Historians are a mature technology. OSI PI is over 40 years old, for example. They have large existing install-bases and are well-integrated with manufacturing and process-based technologies through DCS and PLC control systems. However they do not contain native advanced analytics capabilities and the ability to execute machine learning models.
Concepts
The Event Manager implementation handles the connection to the historian systems and full processing of the data. The following concepts and terms apply to this component.
Data Source is an implementation of a connection to a historian system. It includes some standard web services that allow the user to browse the hierarchy on the historian and setup Subscriptions to data points of interest. The Accelerator refers to these data points as Nodes but this will map on to different concepts depending on the historian being connected to. In OPC UA these are also referred to as Nodes but in OSI PI these refer to Tags.
Once a Subscription is setup to one or more Nodes, we can assemble these into a logical set of data points called a Feature Set, with each data point mapping to a single Feature. Each Feature Set may have one to many Features, which typically will all come from the same Subscription or Data Source, but may span different Subscriptions and Data Sources if necessary.
A Feature Set can then be directed to call one to many Indicators. An Indicator is an implementation of a business rule that requires certain data to operator, which is modelled as the Feature Set. Indicators are called in an pre-defined order, with the output of earlier Indicators passed into the subsequent Indicators, along with all Features from the initial Feature Set, unless these have updated by preceeding Indicators.
An Indicator can implement any kind of arbitrary business logic as required. The Accelerator provides examples of Indicators that compute the mean of all a set of Features using TERRTM and Python. There is also an example that computes a cluster using PMML. Indicators may implement any other standard EventFlow logic, including Decision Tables. They may output a Feature Set of Features which can include new Feature values or updates to Features passed in to the Indicator.
All raw data from a Data Source are passed to Data Sinks for storage. The Accelerator implements a single Data Sink for Spotfire Live Datamart, but this could be extended to other sinks such as CSV files. All Features in a Feature Set are also sent to the same Data Sinks for storage. This includes both original Feature Sets as well as new and updated Features generated by Indicators.
Benefits and Business Value
By integrating Spotfire® Data Science, Spotfire® and Spotfire® Streaming with these Data Historians, process data can be captured and and analysed to detect patterns. Models can be developed using languages like R or Python, or developed using more advanced tools like StatisticaTM which can then be deployed to a running SpotfireStreaming engine for real-time model execution. These models could be used to detect anomalies, detect production quality issues, or do predictive or condition-based maintenance.
Technical Scenario
The Accelerator shows how to integrate Spotfire, Statistica, and Spotfire Streaming to capture historian data and execute against some simple models.
Two datasets are provided:
- Electric Submersible Pump (ESP) telemetry for intake pressure and current draw
- Power plant telemetry for a gas-fired turbine plant which generates electricity
Simple model implementations in R and Python are provided to show how to integrate with inbound historian data. These models simply compute a mean of all provided features.
A more sophisticated K-Means Clustering model is provided using PMML that can be used to detect anomalies in the power plant telemetry data.
Components
Related Accelerators
This Accelerator is one of several used for industrial telemetry monitoring. These other Accelerators may also be of interest:
Recommended Comments
There are no comments to display.