The Innovation Platform delves into Verena Kain, project leader of CERN’s Efficient Particle Accelerator (EPA) project, and innovative strategies used to achieve the goals of the project, the rationale behind its inception, and the ambitious goals.
The Efficient Particle Accelerator (EPA) project is a groundbreaking initiative aimed at revolutionizing the operation of particle accelerators at CERN.
The EPA project aims to increase overall efficiency of not only LHC but also future accelerators in the future with a difference in the shift of large hadron criders to the high luminosity phase. By focusing on improving performance, flexibility, repeatability and sustainability, the project occupy a key blueprint for next-generation particle physics research.
As CERN navigates the challenges posed by energy consumption and complex operational demands, integration of innovative technologies such as automation, artificial intelligence (AI), and machine learning (ML) becomes most important.
To learn more, the Innovation Platform spoke with EPA project leader Verena Kain to explore the goals of the project, the rationale behind its founding, and the innovative strategies being used to achieve the ambitious goals.
Can I provide an overview of the EPA project? What is its main goal and purpose?
The Efficient Particle Accelerator (EPA) project aims to increase the efficiency of particle accelerators, especially CERNs. The hope is that this project will serve as a blueprint for future accelerators aimed at improving performance, flexibility, repeatability and sustainability.
How and why did this project happen?
The main catalyst was another project, LHC injector upgrade (LIU). The LHC is scheduled to end in 2026, and will then undergo an upgrade to become a High-Light LHC (HL-LHC), marking the next and final stages of the LHC accelerator.
It took place between 2019-2021 and involves six particle accelerators, so an upgrade of the LHC injector was required to increase the brightness of the injection beam. The Liu project led to a series of workshops that revisited actual performance metrics from other facilities for the first time since LHC was installed. It has become clear that while the machine is functioning properly, there is a significant potential for improvement.
In 2022, a workshop was held to assess the energy consumption of machines amid the energy crisis. It has been revealed that hysteresis mitigation measures in iron-dominated magnets have a major impact on energy consumption and do not fully compensate for it. Hysteresis means that the field within the magnet used for saturation depends not only on the current provided by the power source, but also essentially on which field the magnet is played previously.
One mitigation measure that is typically used is to severely limit the magnet cycles or programs that can be played one after another. As a result, there is a significant decrease in flexibility. Another energy expenditure measure is to fully regenerate a particular cycle, even if the beam cannot be injected to ensure a reproducible field.
Why is automation so important in improving particle acceleration efficiency? Can you explain the role of AI and ML in EPA projects?
The focus is to keep hardware as long as possible, for example, maintaining magnets and RF cavity, and leveraging software to automate processes.
Current equipment operations are designed for human decision-making, except for real-time scenarios that require split 2-second decisions. Real-time feedback systems and electronics have been automated for some time, but automating slower controls and loop-in processes has challenges.
Existing systems will work, but may not be suitable for the size of your future machine. Operator training can be very long as it requires acculturation to handle the vast array of components within the accelerator. Despite efforts to simplify the process with layers of abstraction, the task remains complex.
Additionally, operators have different skill sets and experiences, with different repair times and results. The possibility of managing a particular task simultaneously autonomously can significantly improve efficiency, and operator attention may be directed towards more abstract decisions.
AI and ML are becoming increasingly important in a variety of areas, including CERN. For example, LHC, the largest and most complex machine in CERN, requires minimal human intervention for advanced automation. However, it is not completely autonomous and requires human surveillance.
Traditional automation requires knowledge of systems and the ability to represent symbolically. These technologies allow for analyzing large amounts of data and identifying patterns in different data structures without additional input into parametric relationships of data, which eliminates this limitation.
For example, it is very challenging to describe what appears in an image symbolically. AI can accomplish this task, computer vision without defining rules. This feature offers great advantages for learning patterns or mappings that have been difficult to accurately express in written form or spoken language.
What other innovative approaches are being investigated in the EPA project to increase the efficiency of particle accelerators? Could you please explain in detail about the nine work packages?
AI is used for tasks that lack traditional solutions. To support this, we have established infrastructure requirements and blueprints for smart equipment related to science labs to promote AI efforts. The goal is to develop an AI-Ready accelerator by providing hierarchical building blocks within a cohesive system. For example, scientists use Python primarily for programming, but don’t want to get used to all the complexities of the underlying control infrastructure.
To address this, systems are designed to hide the complexity of the control infrastructure, while allowing for the integration of advanced AI and other algorithms. Additionally, a shared GPU system is implemented to allow easy access to computing resources without the need for individual purchases. Next-generation devices will become (more) autonomous. And the idea is that the final accelerator will become an ensemble of hierarchical autonomous systems. The distribution of information will also be revisited.
What are the key challenges faced in increasing the efficiency of particle accelerators?
Many of the tasks performed here at CERN have never been attempted before. There are technical hurdles, but they are not overcome.
For quite some time, the biggest challenge has been persuading people the need to improve efficiency. These accelerators have been running for quite some time, and were skeptical of the potential for improvement and the benefits of investing in such projects, and perhaps still.
However, when we presented the project and demonstrated the minimum initial investment required for the first step, these arguments were no longer standing. In fact, people are now predicting results. According to our timeline, the key results of some work packages should be clear by the end of next year.
Additionally, the entire project is time-limited, and only takes 5 years to complete. This adds an extra layer of challenges and interest, so you have to keep focused and not falling aside.
Your team comes from a variety of CERN backgrounds. How important is project collaboration?
The nine working packages may look different, but they are all very interconnected. For one of them to succeed, they all need to flourish. Collaboration is essential to defining requirements from the start, and everyone must contribute to defining the infrastructure needed and expected outcomes. This level of collaboration may not be typical, but in our case it is important. This project includes infrastructure projects and work packages that implement solutions on accelerators to address specific issues, but act as one team.
How do you envision EPA projects that will affect the field of particle physics? What is the next step?
The idea for this project came in preparation for the analysis of existing accelerators and for the future of these accelerators. The goal is to develop a blueprint for future accelerator exploitation models. The team works closely with the team working on the Future Circular Collider (FCC) research at CERN.
The current business model as a business is not affordable for the FCC. So there is no way to surround innovation. Our project aims to establish new operation and maintenance models for future accelerators, and set new standards for efficiency and sustainability, as well as a set of guidelines for self-supporting systems.
This article will also be featured in the 20th edition of Quarterly Publishing.
Source link
