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
The Advanced Photon Source (APS) particle accumulator ring (PAR) was designed to accumulate linac pulses into a single bunch using a fundamental radio frequency (rf) system, and longitudinally compress the beam using a harmonic rf system prior to injection into the booster. The APS Upgrade injectors will need to supply full-current bunch replacement with high single-bunch charge for swap-out injection in the new storage ring. Significant bunch lengthening is observed in the PAR at high charge, which negatively affects beam capture in the booster. Predictions showed that the bunch length could be compressed to better match the booster acceptance using a combination of higher beam energy and higher harmonic gap voltage. A new 10-kW harmonic rf solid-state amplifier (SSA) was installed in 2021 to raise the gap voltage and improve bunch compression. The SSA has been operating reliably. Initial results show that the charge-dependent bunch lengthening in PAR with higher gap voltage agrees qualitatively with predictions. A tool was written to automate bunch length data acquisition. Future plans to increase the beam energy, which makes the SSA more effective, will also be summarized.
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
RadiaSoft LLC, Boulder, Colorado, USA
The Linac Extension Area at the Advanced Photon Source is a flexible beamline area for testing accelerator components and techniques. Driven by the Advanced Photon Source electron linac equipped with a photocathode RF electron gun, the Linac Extension Area houses a 12 m long beamline. The beamline is furnished with YAG screens, BPMs and a magnetic spectrometer to assist with characterization of beam emittance and energy spread. A 1.4 m long insertion in the middle of the beamline is provided for the installation of a device under test. The beamline is expected to be available soon for testing accelerator components and techniques using round and flat electron beams over an energy range 150-450 MeV. In the present work, we describe this beamline and summarise the main beam parameters.
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.