Sample alignment in neutron scattering experiments is critical to ensuring high quality data for the users. This process typically involves a skilled operator or beamline scientist. Machine learning has been demonstrated as an effective tool for a wide range of automation tasks. RadiaSoft in particular has been developing ML tools for a range of accelerator applications including beamline automation. In this poster we will present recent developments for selecting and aligning multiple samples at the HB-2A powder diffractometer at HFIR.

The installation and alignment of new beamlines and beamline components is necessary at any accelerator facility. The equipment and methods used to perform these precision driven tasks must be accurate, reliable and above all, easily repeatable. Using coordinate measuring machines (CMM), laser trackers, combined with Spatial Analyzer, Autodesk Inventor and other custom tools, it is possible to rapidly and accurately take an idea from model to reality, as shown through the construction of the ATLAS Multi-User beamline.

At a heavy ion linac facility, such as ATLAS at Argonne National Laboratory, a new ion beam is tuned once or twice a week. The use of machine learning can be leveraged to streamline the tuning process, reducing the time needed to tune a given beam and allowing more beam time for the experimental program. After establishing automatic data collection and two-way communication with the control system, we have developed and deployed machine learning models to tune and control the machine. We have successfully trained online different Bayesian Optimization (BO)-based models for different sections of the linac, including the commissioning of a new beamline. We have demonstrated transfer learning from one ion beam to another allowing fast switching between different ion beams. We have also demonstrated transfer learning from a simulation-based model to an online machine model and used Neural Networks as prior-mean for BO optimization. More recently, we have succeeded in training a Reinforcement Learning (RL) model online for one beam and deployed it for the tuning of another beam. These models are being generalized to other sections of the ATLAS linac and can, in principle, be adapted to control other ion linacs and accelerators with modern control systems.

From being the first computer-controlled accelerator, through its 52-year long operational history, today the LANSCE Instrumentation and Control System (LICS) shows little resemblance of its early days. Over the past 5 decades, generations of control system engineers were faced with the challenge of maintaining the LICS. However, its maintainability depends on the ability that a failed component or system can be restored or repaired. Complicating this task is the undeniable fact that technology has significantly evolved over the last decades and that older component and systems, while still performing their function, have become obsolete and unmaintainable. When a technology migration path isn't viable to ensure LICS maintainability, the only alternative and opportunity is to upgrade to a new technology platform. Consideration needs to be given that the new technology platform needs to seamlessly integrate with the existing LICS infrastructure while allowing for technological progress. Given LICS’s technology complexity multiple dependencies make the migration and upgrade paths a challenging one. In this paper, we discuss technology choices and compromises made, technology migration and upgrade challenges still faced, and LICS vision for the future. All this under the budgetary and schedule constraints of an operating accelerator facility with an enduring mission.

Sample alignment in neutron scattering experiments is critical to ensuring high quality data for the users. This process typically involves a skilled operator or beamline scientist. Machine learning has been demonstrated as an effective tool for a wide range of automation tasks. RadiaSoft in particular has been developing ML tools for a range of accelerator applications including beamline automation. In this poster we will present recent developments for selecting and aligning multiple samples at the HB-2A powder diffractometer at HFIR.

Achieving high-quality proton beams for accelerators hinges on effective beam tuning. However, the conventional "Monkey Jump" method, widely used for tuning, proves labor-intensive and inefficient. Through harnessing Reinforcement Learning (RL), a novel beam tuning strategy can swiftly emerge, making informed decisions based on the prevailing system status and control demands, offering a promising alternative for accelerator systems. We explore novel techniques RL-based beam tuning and applying it to the beam tuning process of the CiADS Front End accelerator currently, with the aim of significantly enhancing the efficiency of the tuning process. To achieve this, we will first establish an RL-compatible environment based on dynamic simulation software. Subsequently, the policy is trained under different initial conditions. Finally, the strategy successfully trained in the simulation environment will be tested on real accelerator to verify its effectiveness.

RadiaBeam is fabricating a novel RF vacuum window for use with the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (ORNL). The window features a coaxial ceramic window between two waveguides, brazed as a single assembly. Unlike traditional pillbox window designs, this approach allows the outer diameter of the ceramic to decrease and the added benefit of water cooling the inner diameter of the ceramic. This paper covers the engineering design including details of key features, the impact of the unique RF design on manufacturability, and mechanical simulations. A status update on the fabrication is also provided with emphasis on the ceramic TiN coating and brazing process.

The International Liner Collider requires a crabbing system to increase the luminosity of the colliding electron bunches. The ILC has a large crossing angle that requires compensation in order to meet the luminosity requirements. There are several frequency options proposed for the crabbing cavity design. Two crab cavity designs were selected to be prototyped in the pre-lab phase, following the Down Selection Review on Crab Cavity Design held in April 2023. The two rf designs selected are the 1.3 GHz rf-dipole cavity and the 2.6 GHz QMiR cavity. We will be presenting the electromagnetic and mechanical design details of the two compact crabbing cavity designs.

Maintaining beam transport efficiency in the APS linac requires several feedback mechanisms to control orbit, phase, and other parameters. Presently, we apply pre-computed matrices to sets of deviations from fixed setpoints, corresponding to proportional linear feedback. This approach works most of the time but is slow and can become unstable at low charge levels. We explore two alternative machine learning (ML) methods - adaptive Bayesian optimization (ABO, developed previously) and reinforcement learning (RL). To pre-train ML methods we use a differentiable linac simulation to generate a custom kernel and policy, respectively. All 3 methods are experimentally tested using a set of simulated disturbances, and performance in terms of charge stability and recovery speed analyzed. We find that both ABO and RL techniques are more flexible than standard feedback but behave quite differently if beam degradation is large. Overall, RL appears to be the more robust long-term method for rough correction, while ABO is best for fine tuning on recent history. Based on the above results we implemented a novel hybrid scheme that dynamically combines algorithm outputs using historical and expected performance. It also restricts parameter space to the most relevant region. Preliminary results show this to be both more stable and more accurate than the standard approach. We are now exploring strategies for dynamic retraining and other advanced capabilities.

In this work, we integrate and extend an HKL computation package* into EPICS** through a PyDevice*** IOC. This provides EPICS users a generalized approach to mapping real motor rotation space to HKL reflections for a wide range of diffractometers (4-circle, 6-circle, kappa geometries). Utilizing PyDevice for EPICS IOC development allows us to bind core calculations written in C to Python, simultaneously taking advantage of the efficiency of C and readability of Python. The EPICS IOC provides an interface between beamline hardware and users through an intuitive Phoebus CSS GUI, Extensions are being developed to the original HKL package to handle inelastic scattering in addition to the original elastic scattering case for neutron and X-ray diffraction.

The utilization of large language models (LLMs) such as ChatGPT has seen a remarkable increase in various fields over the past few years. These models have demonstrated their versatility and capability in understanding and generating human-like text, making them invaluable tools in numerous applications. In this project, we explore the integration of a LLM into the Experimental Physics and Industrial Control System (EPICS). The primary focus of this integration is to employ the LLM for advanced image processing and spatial analysis on images obtained from the beamlines. By leveraging the capabilities of the LLM, we aim to enhance the accuracy and efficiency of image interpretation, enabling more precise data analysis and decision-making within the EPICS framework. This integration not only showcases the potential of LLMs in scientific and industrial applications but also sets the stage for future advancements in automated control systems.

The 65k pixel TimePix3 chip with ToA of 1.5625 [ns] nominal time resolution, allows timing and imaging studies using X-ray, neutron, and electron spectroscopies. The EPICS* ADTimePix3** areaDetector*** driver enables control and integration into the beamline acquisition system. This presentation will discuss the recent development of the beamline integration of the detector into neutron beamlines and selected results****.

The RIKEN Linear Accelerator (RILAC), one of the injectors at RIBF was upgraded by installing a superconducting RILAC (SRILAC) to search for superheavy elements with element number 119 and above. Before the SRILAC upgrade, the machine protection system in the RILAC was constructed using simple relay circuits. On the other hand, most of the accelerators at RIBF other than RILAC have been equipped with machine protection systems using Mitsubishi MELSEC-Q Programmable Logic Controllers (PLCs) since 2006. They have a mechanism that triggers an anomaly signal to drive the beam chopper to stop the beam and are called beam interlock systems (BIS). Machine protection was needed in the SRILAC project to prevent vacuum deterioration of the superconducting cavity due to changes in the beam orbit. We have developed an FA-M3 PLC-based system to realize a BIS with high response performance at a lower cost than conventional systems. This system is characterized by implementing relatively slow response and I/O requiring high response performance. For example, in the case triggered by an anomaly signal of the electromagnet power supply, simulation of the beam orbit shows that the response performance is relatively slow, a few milliseconds being sufficient. In this conference, the performance results of the constructed BIS will be reported based on the types of anomaly signals in actual SRILAC operation.

The timing system is a critical element in scientific facilities such as particle accelerator or laser ignition installations. The different subsystems that integrate these scientific facilities need to have a common notion of time. This common time reference is provided by the timing system. Thank to that, it is possible to operate the machine in a time coherent manner and to properly track the different events that occur during the operation of the machine. The timing system also provides the discrete triggering events and periodic signals requested for the different subsystems. Furthermore, it can be used also for radiofrequency distribution across the facility. In this work it is presented the timing system architecture, based on the White Rabbit technology and currently under development by Safran Electronic & Defense Spain SLU, for the distribution of synchronized triggers. The hardware, based on FPGA, will be detailed. The timing system allows total triggering configuration in terms of direction, number of pulses, pulse rate, pulse period and delay offering a resolution in the order of 5ps. The White Rabbit technology provide sub-nanosecond accuracy and picosecond precision in addition to important characteristics as the automatic link calibration. The performance achieved will be shown in this work.

A common challenge in online accelerator operations is aligning beams through a series of quadrupole magnets, especially when in situ beam position monitors are not present. Accelerator operators generally use a trial-and-error approach to solve this problem by sequentially measuring the centroid deflection of the beam as a function of quadrupole strengths. This is a challenging process that necessitates dedicated effort by operational experts, requiring significant beam time and personnel resources to configure basic accelerator operations. In this work, we use Bayesian Algorithm Execution (BAX) with virtual objectives to autonomously control steering magnets at the Argonne Wakefield Accelerator to center the beam through a quadrupole triplet. This technique uses virtual objectives to reduce the number of measurements needed to converge to an optimal solution, resulting in a turn-key algorithm for finding the optimal steering configuration for a set of accelerator magnets from scratch.

The recent development of advanced black box optimization algorithms has promised order of magnitude improvements in optimization speed when solving accelerator physics problems. These algorithms have been implemented in the python package Xopt, which has been used to solve online and offline accelerator optimization problems at a wide number of facilities, including at SLAC, Argonne, BNL, DESY, ESRF, and others. In this work, we describe updates to the Xopt framework that expand its capabilities and improves optimization performance in solving online optimization problems. We also discuss how Xopt has been incorporated into the Badger graphical user interface that allows easy access to these advanced control algorithms in the accelerator control room.

A compact, high-voltage (HV) pulser in the nanosecond regime for transverse electromagnetic (TEM) kickers is presented. TEM kickers are electromagnetic deflectors used in particle accelerators to redirect bunches of particles out of their original trajectory into a new path, such as alternate beam paths, detectors, or other instrumentation devices. The circuit proposed in this design consists of two main portions: a gate driver and a HV switch. The gate driver consists of an isolated and high-speed gate driver, powered by an isolated DC/DC converter with dual output voltages. The HV switch portion was simulated in Ansys HFSS and is composed of a SiC MOSFET, LC resonance components, and specialized diodes. When switched, the MOSFET is used to pump a high voltage into the LC circuit and diode stack, and the ultrafast diode turnoff delivers the final HV pulse to the resistor load. Careful layout techniques were implemented for the MOSFET driver to reduce pulse to pulse instability. A 1 MHz repetition rate was the target of our design.