Michael O’Connell, TIBCO’s chief data scientist, and Hayden Schultz, a TIBCO architect, discussed and demonstrated an event-handling example using remote sensor data with Spotfire and Streambase. One oil company may have thousands of submersible pumps moving oil up from well, and these modern pumps include sensors and telemetry to allow them to be monitored and controlled remotely. One of their oil and gas customers said that through active monitoring and control such as this, they are avoiding downtime worth $1000/day/well, meaning an additional $100M in additional revenue each year. In addition to production monitoring, they can also use remote monitoring in drilling operations to detect conditions that might be a physical risk. They use standards for sensor data format, and a variety of data sources including SAP HANA.
For the production monitoring, the submersible pumps emit a lot of data about their current state: monitoring for changes to temperature, pressure and current shows patterns that can be correlated with specific pre-failure conditions. By developing models of these pre-failure patterns using Spotfire’s data discovery capabilities on historical failure data, data pushed into Streambase can be monitored for the patterns, then Spotfire used to trigger a notification and allow visualization and analytics by someone monitoring the pumps.
We saw a demonstration of how the pre-failure patterns are modeled in Spotfire, then how the rules are implemented in Streambase for real-time monitoring and response using visual modeling and some XML snippets generated by Spotfire. We saw the result in Streambase LiveView, which provides visualization of streaming data and highlights those data points that are exhibiting the pre-failure condition. The engineers monitoring the pumps can change some of the configuration of the failure conditions, allowing them to fine-tune to reduce false positives without missing actual failure events. Events can kick off notification emails, generate Spotfire root cause analysis reports, or invoke other applications such as instantiating a BPM process.
There are a number of similar industrial applications, such as in mining: wherever there are a large number of remote devices that require monitoring and control.