Advanced controls framework enabled through data science

To accommodate advanced data science applications to particle accelerators and their user experiments and other applications, there is a need for advanced diagnostic sensors as well as for advanced controls. In particular, there is need to create a modern runtime environment conducive with facility control systems such as EPICS while enabling advanced data science methods, such as passing sensor data to advanced optimization-based and/or AI-based algorithms to predict an action for tuning or for “dialing in” the machine configuration to achieve a certain beam parameter (e.g. luminosity, emittance, energy spread) at a particular location. To achieve this, we need to create a seamless link between facility control systems, such as EPICS, and their protocols, the new data science techniques (to analyze and “write” the machine action back through these protocols), a method to display the data and analyses for the operator, and a way to pull from the systems-based engineering models (e.g. surrogate models). The need for a flexible runtime environment to enable data science applications will only become more important as future particle accelerators become more complex. We should address this here and now. With our team consisting of members from Element Aero, the Facility for Rare Isotope Beams at Michigan State University, and the Thomas Jefferson National Accelerator Facility, we plan to architect, design, simulate, and test an extendable framework to process and display real time accelerator data in Phase I of the grant at two nuclear physics accelerator facilities. In Phase II, it is the intent of refining the system and delivering a prototype device. Then we would move into Phase III – the commercialization phase. What we can design and develop here will be of utility not only to many accelerators, but industries with complex systems also having the same goals. Imagine for example that these frameworks could help “dial in” quickly for the high-throughput sterilization of a variety (different machine configurations) of medical devices coming off the production line requiring packaging or could help hospitals remotely treat cancerous tumors in rural areas with fewer highly-skilled medical professionals through data being displayed real time to help tune the machine for a specific patient.