Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userReal-time Artificial Intelligence for Accelerator Control: A Study at the Fermilab Booster
73
0
0.0
(
0
)
Authors
Jason St. John
Ask ChatGPT about the research
No Arabic abstract
We describe a method for precisely regulating the gradient magnet power supply at the Fermilab Booster accelerator complex using a neural network trained via reinforcement learning. We demonstrate preliminary results by training a surrogate machine-learning model on real accelerator data to emulate the Booster environment, and using this surrogate model in turn to train the neural network for its regulation task. We additionally show how the neural networks to be deployed for control purposes may be compiled to execute on field-programmable gate arrays. This capability is important for operational stability in complicated environments such as an accelerator facility.
rate research
Read More
The Fermilab Booster is a fast ramping (15Hz) synchrotron which accelerates protons from 400MeV to 8GeV. During commissioning of a transverse digital damper system, it was shown that the damper could provide a measurement of the machine tune throughout the cycle by exciting just 1 of the 84 bunches with minimal impact on the machine operation. The algorithms used to make the measurement have been incorporated into the damper FPGA firmware allowing for real-time tune monitoring of all Booster cycles.
The Fermilab booster has an intensity upgrade plan called the Proton Improvement plan (PIP). The flux throughput goal is 2E17 protons/hour which is almost double the current operation at 1.1E17 protons/hour. The beam loss in the machine is going to be an issue. The booster accelerates beam from 400 MeV to 8GeV and extracts to The Main Injector (MI). Cogging is the process that synchronizes the extraction kicker gap to the MI by changing radial position of the beam during the cycle. The gap creation occurs at about 700MeV which is 6msec into the cycle. The variation of the revolution frequency from cycle to cycle is larger at lower energy and it is hard to control by changing the radial position because of aperture limitations. Momentum cogging is able to move the gap creation earlier by using dipole correctors and radial position feedback, and controlling the revolution frequency and radial position at the same time. The new cogging is going to save energy loss and aperture. The progress of the momentum cogging system development is going to be discussed in this paper.
The development of magnetic cogging is part of the Fermilab Booster upgrade within the Proton Improvement Plan (PIP). The Booster is going to send 2.25E17 protons/hour which is almost double the present flux, 1.4E17 protons/hour to the Main Injector (MI) and Recycler (RR). The extraction kicker gap has to synchronize to the MI and RR injection bucket in order to avoid a beam loss at the rising edge of the extraction and injection kickers. Magnetic cogging is able to control the revolution frequency and the position of the gap using the magnetic field from dipole correctors while radial position feedback keeps the beam at the central orbit. The new cogging is expected to reduce beam loss due to the orbit changes and reduce beam energy loss when the gap is created. The progress of the magnetic cogging system development is going to be discussed in this paper.
The Fermilab booster has an intensity upgrade plan called the Proton Improvement plan (PIP). The flux throughput goal is 2E17 protons/hour, which is almost double the current operation at 1.1E17 protons/hour. The beam loss in the machine is going to be the source of issues. The booster accelerates beam from 400 MeV to 8 GeV and extracts to the Main Injector. Several percent of the beam is lost within 3 msec after the injection. The aperture at injection energy was measured and compared with the survey data. The magnets are going to be realigned in March 2012 in order to increase the aperture. The beam studies, analysis of the scan and alignment data, and the result of the magnet moves will be discussed in this paper.
A new beam injection scheme is proposed for the Fermilab Booster to increase beam brightness. The beam is injected on the deceleration part of the sinusoidal magnetic ramp and capture is started immediately after the injection. During the entire capture process we impose Pdot=0 in a changing B field. Beam dynamics simulations clearly show that this method is very efficient with no longitudinal beam emittance dilution and no beam loss. As a consequence of preserved emittance, the required RF power on a typical Booster cycle can be reduced by ~30% as compared with the scheme in current operation. Further, we also propose snap bunch rotation at extraction to reduce dP/P of the beam to improve the slip-stacking efficiency in MI/RR.
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
