ANL, Lemont, Illinois, USA
We explore the possibility of using a Machine Learning (ML) algorithm to identify the source of occasional poor performance of the Particle Accumulator Ring (PAR) and the Linac-To-PAR (LTP) transport line, which are parts of the injector complex of the Advanced Photon Source (APS) at Argonne National Lab. The cause of reduced injection or extraction efficiencies may be as simple as one parameter being out of range. Still, it may take an expert considerable time to notice it, whereas a well-trained ML model can point at it instantly. In addition, a machine expert might not be immediately available when a problem occurs. Therefore, we began by focusing on such single-parameter anomalies. The training data were generated by creating controlled perturbations of several parameters of PAR and LTP one-by-one, while continuously logging all available process variables. Then, several ML classifiers were trained to recognize certain signatures in the logged data and link them to the sources of poor machine performance. Possible applications of autoencoders and variational autoencoders for unsupervised anomaly detection and for anomaly clustering were considered as well.
ANL, Lemont, Illinois, USA
Ion trapping occurs when a negatively charged beam ionizes residual gas inside an accelerator vacuum chamber, and the resulting ions become trapped in the beam potential. Trapped ions can cause a variety of undesirable effects, including coherent instability and incoherent emittance growth. Because of the challenging emittance and stability requirements of next generation light sources, ion trapping is a serious concern. To study this effect at the present APS, a gas injection system was designed and installed at two different locations in the ring. The system creates a controlled and localized pressure bump of nitrogen gas, so the resulting ion instability can be studied. Measurements were taken under a wide variety of beam conditions, using a spectrum analyzer, pinhole camera, and bunch-by-bunch feedback system. The feedback system was also used to perform grow-damp measurements, allowing us to measure the growth rate of individual unstable modes. This paper will present some of the results of these experiments. Simulations using the IONEFFECTS element in the particle tracking code elegant will also be presented.
ANL, Lemont, Illinois, USA
The Advanced Photon Source (APS) is upgrading the storage ring to a design that requires on-axis injection. Matching between the incoming beam and the ring is important to ensure high injection efficiency. Toward this end, we have developed and tested a method for measuring all σ matrix elements except those related to the time coordinate. We report on challenges inherent in this technique, based on simulation and real-world trials.
ANL, Lemont, Illinois, USA
The Advanced Photon Source (APS) is upgrading the storage ring to a hybrid seven-bend-achromat design with reverse bends, providing a natural emittance of 41 pm at 6 GeV. The small dynamic acceptance entails operation in on-axis swap-out mode. Careful consideration is required of the safety implications of injection with shutters open. Tracking studies require simulation of multiple simultaneous magnet errors, some combinations of which may introduce potentially dangerous conditions. A naive grid scan of possible errors, while potentially very complete, would be prohibitively time-consuming. We describe a different approach using biased sampling of particle distributions from successive scans. We also describe other aspects of the simulations, such as use of 3D field maps and a highly detailed aperture model.
ANL, Lemont, Illinois, USA
RadiaSoft LLC, Boulder, Colorado, USA
UCSC, Santa Cruz, California, USA
The Advanced Photon Source Upgrade will use a multi-bend achromatic lattice to reduce vertical and horizontal beam emittances by one- and two-orders of magnitude respectively; in addition operating current will double. The resulting electron beam will be capable of depositing more than 150 MGy on machine protection collimators creating high-energy-density conditions. Work is underway to couple the beam dynamics code Elegant with the particle-matter interaction program MARS and the magnetohydrodynamics code FLASH to model the effects of whole beam dumps on the collimators. Loss distributions from Elegant are input to MARS which provide dose maps to FLASH. We also examine the propagation of downstream shower components after the beam interacts with the collimator. Electrons and positrons are tracked to determine locations of beam loss. Beam dump experiments conducted in the APS storage-ring, generated dose levels as high as 30 MGy resulting in severe damage to the collimator surfaces with melting in the bulk. The deformed collimator surface may lead to beam deposition in unexpected locations. A fan-out kicker is planned to mitigate the effects of whole beam dumps on the collimators.
ANL, Lemont, Illinois, USA
Fringe fields effects in dipoles can give rise to important linear and nonlinear contributions. This paper describes how to extend the classic results of Brown [1] and the more recent calculations of Hwang and Lee [2] to Cartesian dipoles with transverse and/or longitudinal gradients. We do this by 1) introducing a more general definition of the fringe field that can be applied to longitudinal gradient dipoles, 2) allowing for quadrupole and/or sextupole content in the magnet body, and 3) showing how to employ the resulting fringe field maps as a symplectic transformation of the coordinates. We compare our calculation results with tracking for longitudinal and transverse gradient dipoles planned for the APS-U.
[1] K.L. Brown, Report SLAC-75, 1982.
[2] K. Hwang and S.Y. Lee, Phys. Rev. Accel. Beams, vol. 18, p. 122401 2015.
ANL, Lemont, Illinois, USA
RadiaSoft LLC, Boulder, Colorado, USA
In the Advanced Photon Source (APS) storage ring at Argonne National Lab, trips in the magnet power supplies (PSs) lead to a complete electron beam loss a few times a year. This results in unexpected interruptions of the users’ experiments. In this contribution, we investigate the historical data for the last two decades to find precursors for the PS trips that could provide an advance notice for future trips and allow some preventive action by the ring operator or by the PS maintenance team. Various unsupervised anomaly detection models can be trained on the vast amounts of available reference data from the beamtime periods that ended with an intentional beam dump. We find that such models can sometimes detect trip precursors in PS currents, voltages, and in the temperatures of magnets, capacitors and transistors (components of PSs).
ANL, Lemont, Illinois, USA
We have developed a set of tools to simulate particle dynamics in the full magnetic field using the generalized gradients representation. Generalized gradients provide accurate and analytic representations of the magnetic field that allow for symplectic tracking [1]. We describe the tools that convert magnetic field data into generalized gradients representations suitable for tracking in Elegant, and discuss recent results based upon tracking with the full field representations for all magnets in the APS-U storage ring.
[1] A. Dragt. Lie Methods for Nonlinear Dynamics with Applications to Accelerator Physics. University of Maryland (2019).
ANL, Lemont, Illinois, USA
Particle accelerators require continuous adjustment to maintain beam quality. At the Advanced Photon Source (APS) this is accomplished using a mix of operator-controlled and automated tools. To improve the latter, we explored the use of machine learning (ML) at the APS injector complex. The core approach we chose was Bayesian optimization (BO), which is well suited for sparse data tasks. To enable long-term online use, we modified BO into adaptive Bayesian optimization (ABO) though auxiliary models of device drift, physics-informed quality and constraint weights, time-biased data subsampling, digital twin retraining, and other approaches. ABO allowed for compensation of changes in inputs and objectives without discarding previous data. Benchmarks showed better ABO performance in several simulated and experimental cases. To integrate ABO into the operational workflow, we developed a Python command line utility, pysddsoptimize, that is compatible with existing Tcl/Tk tools and the SDDS data format. This allowed for fast implementation, debugging, and benchmarking. Our results are an encouraging step for the wider adoption of ML at APS.