Oak Ridge Spallation Neutron Source Research Articles

Distance Preserving Machine Learning for Uncertainty Aware Accelerator Capacitance Predictions

Abstract Accurate uncertainty estimations are essential for producing reliable machine learning models, especially in safety-critical applications such as accelerator systems. Gaussian process models are generally regarded as the gold standard for this task; however, they can struggle with large, high-dimensional datasets. Combining deep neural networks with Gaussian process approximation techniques has shown promising results, but dimensionality reduction through standard deep neural network layers is not guaranteed to maintain the distance information necessary for Gaussian process models. We build on previous work by comparing the use of the singular value decomposition against a spectral-normalized dense layer as a feature extractor for a deep neural Gaussian process approximation model and apply it to a capacitance prediction problem for the High Voltage Converter Modulators in the Oak Ridge Spallation Neutron Source. Our model shows improved distance preservation and predicts in-distribution capacitance values with less than 1% error.

Uncertainty aware anomaly detection to predict errant beam pulses in the Oak Ridge Spallation Neutron Source accelerator

High-power particle accelerators are complex machines with thousands of pieces of equipment that are frequently running at the cutting edge of technology. In order to improve the day-to-day operations and maximize the delivery of the science, new analytical techniques are being explored for anomaly detection, classification, and prognostications. As such, we describe the application of an uncertainty aware Machine Learning method using the Siamese neural network model to predict upcoming errant beam pulses using the data from a single monitoring device. By predicting the upcoming failure, we can stop the accelerator before damage occurs. We describe the accelerator operation, related Machine Learning research, the prediction performance required to abort the beam while maintaining operations, the monitoring device and its data, and the uncertainty aware Siamese method and its results. These results show that the researched method can be applied to improve accelerator operations.

Oak Ridge Spallation Neutron Source superconducting rf linac availability performance and demonstration of operation restoration with superconducting rf cavity off

Recent advances in superconducting rf (SRF) technology allow for increasingly more reliable and higher power hadron linacs. They hold enormous potential for economic, scientific, medical, environmental, and national security fields. Moving this technology from one of a kind laboratory demonstration devices to industrial scale applications imposes strict requirements on the linac performance in terms of its stability, reliability, and availability. The Spallation Neutron Source (SNS) is currently the highest average power superconducting proton linac in the world that has been routinely operated since 2006. We present statistical data on the SNS SRF linac reliability obtained from its practical operational experience during 2020. We analyze the frequency and duration of SRF cavity trips and identify their causes. These data will show that SRF cavity trips are the most common source of single-point failure within the linac. In an industrial environment, where redundancy is a necessity to prevent interruption of facility operation, the linac beam can be quickly restored by redistributing the failed cavity function to downstream energy reserve SRF cavities. We present a practical demonstration of this approach. We intentionally turn off one of the cavities and recover the beam by readjusting the phases of downstream cavities to maintain the same linac beam energy output. We discuss limitations of this approach at SNS and how they can be overcome.

Predicting particle accelerator failures using binary classifiers

Particle accelerator failures lead to unscheduled downtime and lower reliability. Although simple to mitigate while they are actually happening such failures are difficult to predict or identify beforehand. In this work we propose using machine learning approaches to predict machine failures via beam current measurements before they actual occur. To demonstrate this technique in this paper we examine beam pulses from the Oakridge Spallation Neutron Source (SNS). By evaluating a pulse against a set of common classification techniques we show that accelerator failure can be identified prior to actually failing with almost 80% accuracy. We also show that tuning classifier parameters and using pulse properties for refining datasets can further lead to almost 92% accuracy in classification of bad pulses. Most importantly, in the paper we establish there is information about the failure encoded in the pulses prior to it, so we also present a list of feasible next steps for increasing pulse classification accuracy.

Background studies for the COHERENT experiment at the Spallation Neutron Source

The COHERENT experiment is designed for a first measurement and N2 dependence study of coherent elastic neutrino-nucleus scattering (CEvNS) at the Oak Ridge Spallation Neutron Source (SNS). CEvNS is a standard model process that is important in understanding supernova neutrinos, the structure of the weak interaction and backgrounds for dark matter searches. The COHERENT collaboration is placing a suite of detector technologies in a basement location at the SNS: 14 kg CsI[Na] crystals; 185 kg NaI crystals; 35 kg single-phase LAr; 10kg high-purity, point contact Ge. Previous attempts to measure the CEvNS process have grappled with very high background rates. One class of troublesome backgrounds is accelerator-correlated neutrons because a simple accelerator on/off background subtraction procedure fails to remove these events. The collaboration has completed a comprehensive background study in the basement region where the experiments are located. We conclude from these studies that the neutron background is sufficiently low for successful CEvNS measurement in the SNS basement region. Neutron measurements from the Indiana University SciBath detector and the Sandia Neutron Scatter Camera are presented here.

Noninterceptive method to measure longitudinal Twiss parameters of a beam in a hadron linear accelerator using beam position monitors

A new method of measuring of the rms longitudinal Twiss parameters of a beam in linear accelerators is presented. It is based on using sum signals from beam position monitors sensitive to the longitudinal charge distribution in the bunch. The applicability of the method is demonstrated on the superconducting section of the Oak Ridge Spallation Neutron Source linear accelerator. The results are compared to a direct measurement of the bunch longitudinal profiles using an interceptive bunch shape monitor in the linac warm section of the same accelerator. Limitations of the method are discussed. The method is fast and simple, and can be used to obtain the initial parameters for the longitudinal matching in linear accelerators where interceptive diagnostics are not desirable.

Field distributions and particle optics in main bending dipoles of Oak Ridge Spallation Neutron Source accumulator ring

The SNS accumulator ring employs 32 electro-magnetic dipoles to bend proton beams. The dipoles are typical sector magnets with relatively large aperture and short length. Thus, how to correctly treat magnetic fringe fields in the devices remains as a question. We have performed 3D computer simulations to study magnetic field distributions in the dipoles. Further, we have analyzed particle optics based on the space-dependent curvature and focusing functions in the magnets. The effect of magnetic fringe fields on the particle motion, especially the focusing/defocusing and dispersion, is investigated. The lens parameters, including the second-order aberrations, are derived and compared with the design hard-edge parameters used in the ring lattice calculations.

Dynamics of uncaught foil-stripped electrons in the Oak Ridge Spallation Neutron Source accumulator ring

The Spallation Neutron Source accelerator utilizes ${\mathrm{H}}^{\ensuremath{-}}$ charge exchange injection to merge a 1 GeV, 1 MW beam from a linac into an accumulator ring. The design calls for the 545 keV foil-stripped electrons to be absorbed by an electron catcher located below the foil. However, multiple observations of hardware damage caused by stripped electrons indicate that at least a fraction of the electrons are reflected back into the vacuum chamber. This paper presents the results of a 3D computational model employed to simulate the dynamics of foil-stripped electrons, including a surface interaction model for the catcher. Simulation results are compared with experimental observations, and the implications for high power beam operations are discussed.

Stripper foil failure modes and cures at the Oak Ridge Spallation Neutron Source

The Oak Ridge Spallation Neutron Source comprises a 1 GeV, 1.5 MW linear accelerator followed by an accumulator ring and a liquid mercury target. To manage the beam loss caused by the ${H}^{0}$ excited states created during the ${H}^{\ensuremath{-}}$ charge-exchange injection into the accumulator ring, the stripper foil is located inside one of the chicane dipoles. This has some interesting consequences that were not fully appreciated until the beam power reached about 840 kW. One consequence was sudden failure of the stripper foil system due to convoy electrons stripped from the incoming ${H}^{\ensuremath{-}}$ beam, which circled around to strike the foil bracket and cause bracket failure. Another consequence is that convoy electrons can reflect back up from the electron catcher and strike the foil and bracket. An additional contributor to foil system failure is vacuum breakdown due to the charge developed on the foil by secondary electron emission. In this paper we detail these and other interesting failure mechanisms and describe the improvements we have made to mitigate them.

Measurement of longitudinal acceptance and emittance of the Oak Ridge Spallation Neutron Source Superconducting Linac

The longitudinal acceptance of the Spallation Neutron Source superconducting linac is computed with a longitudinal model. A beam current monitor and beam loss monitors are utilized in a new beam acceptance measurement technique, and the measured results show close agreement with the model. Based on the simulations and on the measurements of the superconducting linac acceptance, we developed a novel method to measure beam bunch shape, beam energy profile, and the longitudinal emittance at the entrance of the linac. The experimental measurements reveal that a large longitudinal beam halo exists in the injected beam to the superconducting linac, and the longitudinal rms emittance is approximately twice that of the nominal design. The simple measurement method is applicable to other superconducting linacs.

Electron cloud instabilities in the Proton Storage Ring and Spallation Neutron Source

Electron cloud instabilities in the Los Alamos ProtonStorage Ring (PSR) and those foreseen forthe Oak Ridge SpallationNeutron Source (SNS) are examined theoretically, numerically, andexperimentally.

Comparison of white neutron sources for nuclear astrophysics experiments using very small samples

The purpose of this paper is to compare the relative merits of four white neutron source facilities for nuclear astrophysics experiments. The four facilities are the Oak Ridge Electron Linear Accelerator, the Manuel Lujan Jr. Neutron Scattering Center, the CERN-TOF facility, and the Oak Ridge Spallation Neutron Source. The criteria for this comparison are based on experiments designed to determine astrophysical rates for neutron-induced reactions using radioactive as well as very small stable samples – areas largely unexplored at present.