Stanford’s particle accelerator using Azure Machine Learning to communicate with scientists

 |   Microsoft IoT

As part of a pilot program at the SLAC National Accelerator Laboratory in Palo Alto, Calif., technicians are testing the potential of the Internet of Things (IoT) to harness the massive volumes of data generated at the facility — using the open source project and Azure Event Hubs to link tens of thousands of sensors from a klystron cooling system to Azure Machine Learning in the cloud.

Built in the 1960s, SLAC is a two-mile-long particle accelerator capable of beaming electrons at nearly the speed of light. It is used across a range of scientific disciplines to help understand processes at the subatomic level. Over its five decades, six scientists have been awarded Nobel prizes for research conducted at SLAC.

The pilot IoT program focuses on the klystron cooling systems that run the length of the facility to maintain constant temperatures. The current, manual process involves operators monitoring hundreds of trends, fed by thousands of data sources, to spot anomalous conditions. The initial goal of the project is to automate much of that legwork so technicians can focus on resolving any issues that arise

Only two months in, the system is already showing the potential to take the pressure off the operators, by instantly flagging conditions that are out of established thresholds. With technicians continually feeding information back into the system about the root cause of detected anomalies, the expectation is that SLAC will eventually begin diagnosing itself — before a problem occurs.

According to officials there, the building management systems are just the beginning: Ultimately millions of sensors at the world-famous lab could become connected, improving the quality of scientific data that researchers need to unlock clues to the very nature of matter, energy, time and space.

Get the full story on the Microsoft News Center, and go deeper into the technology on Microsoft’s Open Technologies blog.


Tags: , , , ,