Accelerator System Research Articles

DLS: A Fast and Flexible Neural Network Training System With Fine-grained Heterogeneous Device Orchestration

Neural network accelerators (e.g., TPUs) have become mainstream devices in computing systems. Unfortunately, the existing accelerator-based systems for neural networks fail to fully leverage the acceleration opportunities due to the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">limited flexibility</i> . Specifically, the majority of the accelerators focus on only the compute-intensive operations of neural networks (e.g., convolution and fully-connected layers). However, we identify that sub-optimal handling of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">auxiliary operations</i> such as embedding and compression can incur non-trivial loss in terms of accuracy, training speed, and adaptability to new domains. The problem persists considering that recent advancements in neural networks often come from auxiliary operations. To effectively handle rapidly evolving auxiliary operations and maximize acceleration opportunities, we propose DLS, a holistic neural network acceleration system using heterogeneous computing devices. The key idea is to distribute compute-intensive operations on highly specialized ASICs for maximum performance, and auxiliary operations on flexible devices (e.g., FPGA, GPU) for better adaptability. We emphasize that a naïve integration of different devices fails to deliver high performance due to high communication overheads. To address this communication inefficiency, we propose an efficient FPGA-based device orchestration utilizing direct device-to-device communication and fine-grained operation scheduling. In this way, our system alleviates the communication overhead between heterogeneous devices by removing expensive kernel stack traversal and leveraging computation units and communication links in parallel. The evaluation using popular neural networks with emerging auxiliary operations shows that our system achieves both flexibility and high performance for various cases from single-accelerator training to distributed training (2.6–8.9× speedup).

Material Analysis and Structural Strength of High Energy Electron Accelerator Cavity Using Finite Element Analysis (FEA)

In 2021, Indonesia planned to build high-energy accelerator electrons for the sterilization of medical instruments. With the transfer of technology from Russia to Indonesia like civil engineering, architecture, electrical, mechanical, control system, and instrumentation. But some several core technologies from accelerator electron like reviews for cavity material, strength of material cavity, vacuum system, electron accelerator system, and radiofrequency technology are not provided. For this reason, it is important to review one of the core technologies from the accelerator. This paper will review the material and strength of the material cavity of the accelerator with the software Finite Element Analysis (FEA) regarding the ILU-14 technology from Russia. Material cavity made from Oxygen free cooper with density 8,89 – 8,94 [g/cm3]. ILU-14 on the power of 100 KW can produce energy 7.5 MeV – 10 MeV. The accelerator consists of a 176 MHz acceleration structure based on a self-excited generator with a total of five GI-50A tubes. The diameter of the resonator is about 800mm with a length of 7875 – 9000 mm with a vacuum pressure of not more than 10−5 lTorr/s (0.133 x 10−2 lPa/s). The results based on the stress of the material will compare with the stress of the oxygen-free copper material.

The influence of hydrogen peroxide concentration on the chemical kinetics of photo-accelerated tooth whitening

BackgroundThe bleaching procedure consists of chemical principles of free radical release that react with chromophores, which results in an amount of energy released in this process. However, the evaluation of the electrical potential generated in these protocols has not yet been thoroughly investigated in the literature. Thus, this study aimed to examine variations in pH, mV, and temperature of different concentrations of hydrogen peroxide in the presence or absence of an intermittent LED/LASER photo acceleration system. MethodsThe study was divided into six groups (n = 9) according to the concentration of hydrogen peroxide (6%, 15%, and 35%), associated or not with the photo acceleration system LED/LASER. We followed the variation of pH, mV, and temperature at 1, 5, 10, 15, and 30 min after gel manipulation. Data were evaluated by two-way ANOVA of repeated measures (α =0.05). ResultspH, mV, and temperature of the groups showed statistical differences both in the light and bleach and in the interaction between the two factors (p < 0.0001), where pH and mV were more influenced by the bleach and light factor, while the temperature was influenced by the bleach factor associated with light. HP15 presented the most significant change in pH, mV, and temperature. ConclusionThe use of LED/laser increased the temperature of the gels and altered the pH and mV kinetics of HP6 and HP15, which did not occur in HP35, possibly due to the high ionic potential linked to the concentration.

Mathematical Modeling and Optimization of Low Energy Beam Line in 2MV Pelletron Accelerator

The 2 MV 6SDH-2 tandem pelletron accelerator installed at CASP, GC University Lahore has evolved over the past years from acceptance tests to an extremely reliable and convenient research tool in the field of particle accelerator physics, hav-ing capability of providing variety of ion beams from few keV to several MeV in energy. Many electrostatic and magnetic devices are employed to steer an ion beam in an accelerator system all the way from ion source to experimental setup. Mathematical modeling and optimization of low energy beamline has been done for this accelerator. Several ion optics and steering devices are provided in the pre-acceleration beamline for ion beam manipulation and in order to optimize the overall beam transmission. As the quality of the beam from the accelerator depends upon the design of these electrostatic and magnetic devices, optimum settings are derived by manually adjusting voltages and currents provided to low energy beam-line components. From these calibrations and optimization measurements, calculations has been done for mass analysis magnets, steering and focusing components and the results has been represented. From the results, it has been found that there exists a complex relationship between the effects of different voltages and currents in order to tune the desired output current and beam profile at the end-stations. Details of experiments and data analysis are presented.

Carbon Footprint of Clinical Photon Therapy: Initial Estimates

<h3>Purpose/Objective(s)</h3> There is an increasing concern about rising carbon dioxide (CO2) levels and its hazardous impact on human health and climate change. Radiation oncology uses high energy machines; however, their carbon footprint is not well understood. This study estimates the energy utilization of photon therapy, and estimates corresponding carbon footprint. Additionally, the study evaluates possible ways to offset CO2 with planting new trees. <h3>Materials/Methods</h3> Patients treated between 07/2020 and 06/2021 using a technology company's medical linear accelerator system at one of our satellite locations were evaluated. A technology company data sheet was used to estimate power draw of the system. BeamOn mode was specified as 39 kVA, Ready mode as 15 kVA, On mode as 11 kVA, and Standby mode as 7 kVA; these were converted to kW. This is likely the worst-case power use scenario, with power factor typically 0.75 – 1.0. Patients treated were reviewed for dose, number of fractions, number of fields, and duration of beam during each fraction. Field delivery durations were obtained from the EMR, and kWh power consumption was estimated per fraction and per treatment course. EPA web calculator was used to convert power consumption to tons of CO2, and the number of new trees required to potentially offset it. <h3>Results</h3> There were 176 patients treated for 191 treatment courses. Total of 4,517 fractions were delivered (average 23.6 fractions per treatment course). Total BeamOn time was 710,606 seconds (197.4 hours), average 157 seconds per fraction. Total BeamOn power consumption would be 7,698 kWh, or ∼1.7 kWh per fraction and ∼40 kWh per treatment course. Annual power consumption including standby mode would be ∼81,250 kWh; BeamOn consumption would account for 9.5% of total machine power usage. The corresponding carbon footprint for BeamOn time would be 5.46 metric tonnes of CO2, or 29 kg of CO2 per patient course. Carbon footprint for the annual machine power consumption would be 57.6 tonnes of CO2. Attributed footprint would thus be 57.6 / 191 = 301 kg CO2 per patient treatment course. The carbon offset required for patient BeamOn time would be 90 new trees planted (average 0.5 trees per patient course). The corresponding offset required to operate the medical linear accelerator system for a year would be 952 new trees planted. The attributed carbon offset per treatment course would thus be 5 new trees planted. <h3>Conclusion</h3> Carbon footprint of the technology company's medical linear accelerator system is likely to be ∼30 kg of CO2 equivalent for direct BeamOn treatment course energy used, ∼300 kg of CO2 equivalent per overall patient energy expenditure, and ∼5.5 metric tonnes of CO2 in total to operate per year. The corresponding number of new trees to be planted to offset the power use would be 0.5 new trees, 5 new trees, and 952 new trees respectively. We are working on measuring the actual power draw during various treatment modes, which could be 5-15% lower than the kVA specification, and calculating the actual carbon footprint of clinical radiation therapy for our next study.

New insights into the effect of calcium formate on hydration process of C3A-AF-Gp System: Formation C3A∙Ca(HCOO)2·11H2O phase

• CF and C 3 A form C 3 A∙Ca(HCOO) 2 ·11H 2 O, and the formation process is not affected by Gp and AF. • C 3 A∙Ca(HCOO) 2 ·11H 2 O phase is the main reason for the reduction of 1 d compressive strength of mortar mixed with AF. • Theoretical model of C 3 A∙Ca(HCOO) 2 ·11H 2 O phase formation process was proposed. The effect of calcium formate (CF) on the hydration products of the C 3 A-Gp(gypsum)- alkali-free liquid accelerator (AF) system was researched by XRD, Raman, and SEM. The results showed that CF and C 3 A could form a new C 3 A∙Ca(HCOO) 2 ·11H 2 O phase, and the formation of that the phase was not affected by AF and Gp. Combined with the analysis and test results, a theoretical model of C 3 A∙Ca(HCOO) 2 ·11H 2 O phase formation of the process was proposed. In addition, it was found that CF could promote the dissolution of C 3 S. However, it did not promote the 1 d compressive strength of mortar mixed with the alkali-free liquid accelerator. Analyzed and inferred that the formation of new phase C 3 A∙Ca(HCOO) 2 ·11H 2 O was the main reason for reducing the early strength of mortar. Therefore, it was not recommended to use calcium formate to promote the early strength of mortar mixed with alkali free liquid accelerator.

Dynamic Rate Neural Acceleration Using Multiprocessing Mode Support

Multiobject detection has become an integral component in various neural applications, such as autonomous driving and augmented reality. The system should be able to recognize and process multiple objects simultaneously. Moreover, the performance requirements for this system can be dynamically changed depending on the number of regions of interest (ROIs) in each frame. Consequently, the processing unit (PU) of the neural acceleration system should provide various inference rates. Therefore, we present a field-programmable gate array (FPGA)-based dynamic rate neural acceleration system called MultiLockOn to dynamically change the inference performance according to the number of ROIs per frame. It supports multiprocessing modes with different speeds through the introduction of novel multi-mode processing engines (PEs) comprising minimum reconfigurable interconnections across inference modes to minimize hardware overhead. The MultiLockOn system can provide an improvement of up to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$4\times $ </tex-math></inline-formula> in the inference performance compared to that of DNNWeaver and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$5.7\times $ </tex-math></inline-formula> compared to that of the ARM Cortex-A53 with minimum accuracy loss by supporting the multiprocessing modes.

Analysis of energy stability in 6/4 MeV dual energy RF electron linac

RF linear accelerators are used to produce high energy electron beams up to several MeV for various industrial, medical and research applications. Accelerators with switchable energy are increasingly finding application in non-destructive testing systems with material discrimination capability. These systems are based on pulsed accelerators with duration 3 to 10 micro second with repetition rate up to 400 Hz. Magnetrons are used as microwave source in relatively low power accelerator systems. This paper examines various factors affecting the pulse to pulse energy variation in such radiation sources. Contributions from individual subsystems are accounted and the cumulative performance is evaluated. Dose stability improvement with remedial measures has been discussed.

IsoDAR@Yemilab: A report on the technology, capabilities, and deployment

IsoDAR@Yemilab is a novel isotope-decay-at-rest experiment that has preliminary approval to run at the Yemi underground laboratory (Yemilab) in Jeongseon-gun, South Korea. Here, we describe in detail the considerations for installing this compact particle accelerator and neutrino target system at the Yemilab underground facility. Specifically, we describe the caverns being prepared for IsoDAR, and address installation, shielding, and utilities requirements. To give context and for completeness, we also briefly describe the physics opportunities of the IsoDAR neutrino source when paired with the Liquid Scintillator Counter (LSC) at Yemilab, and review the technical design of the neutrino source.

Design of a Curved Canted-Cosine-Theta Superconducting Magnet for a Laser Proton Radiotherapy System

Proton therapy (PT) is a precise and efficient radiotherapy method in modern medical treatment, which can be focused on the lesion location to kill the tumor cells with little affection on the normal tissues and thus largely reduce the potential side effects. However, the proton therapy instrument is usually very huge with a large occupation in the space. In order to reduce the instrument scale, a petawatt level laser proton accelerator project is being conducted in China. This paper will report a curved Canted-Cosine-Theta (CCT) magnet which is a critical component in the proposed laser proton accelerator system. The CCT superconducting magnet was designed with a curved dipole magnet that had an equivalent arc π/4 with radius 1 m, and two short quadrupole magnets were nested symmetrically at the dipole magnet ends. The warm bore of the magnet was 72 mm, where a homogeneous transverse magnetic field that amounted to 2.5 T was required along the magnet bore and homogeneous magnetic field gradients 25 T/m were required at the magnet inlet and outlet. NbTi superconducting wire will be used on the coil winding and immersed in liquid helium environment. The structural design of the magnet will be emphatically introduced in this paper.

Designing a Magnetic Measurement Data Acquisition and Control System With Reuse in Mind: A Rotating Coil System Example

Accelerator magnet test facilities frequently need to measure different magnets on differently equipped test stands and with different instrumentation. Designing a modular and highly reusable system that combines flexibility built-in at the architectural level as well as on the component level addresses this need. Specification of the backbone of the system, with the interfaces and dataflow for software components and core hardware modules, serves as a basis for building such a system. The design process and implementation of an extensible magnetic measurement data acquisition and control system are described, including techniques for maximizing the reuse of software. The discussion is supported by showing the application of this methodology to constructing two dissimilar systems for rotating coil measurements, both based on the same architecture and sharing core hardware modules and many software components. The first system is for production testing 10 m long cryo-assemblies containing two MQXFA quadrupole magnets for the high-luminosity upgrade of the Large Hadron Collider and the second for testing IQC conventional quadrupole magnets in support of the accelerator system at Fermilab.

Numerical Optimization of Electromagnet Current Distribution in Superconducting Linear Acceleration System

An enhancement of the acceleration performance of a superconducting linear accelerator when injecting pellets into a helical fusion reactor has been investigated numerically. For this purpose, a numerical code using the finite element method was developed to analyze both the time evolution of a shielding current density in a high-temperature superconductor film and the dynamic motion of the film. In the present study, to optimize the shape of double magnets, a multi-objective optimization problem is solved using the non-dominated sorting genetic Algorithm-II. The results of the computations show that the optimization of the inner electromagnet is more effective in improving the pellet velocity than that of the outer electromagnet. In addition, we found that the pellet speed and energy efficiency have a trade-off relation. Therefore, it is possible to determine the preferred solution for the design of an acceleration magnet, whether focusing on the pellet speed or the energy efficiency. In other words, this relationship can be used to choose whether to emphasize velocity or efficiency in the design of SLA systems.

Online beam orbit correction of MEBT in CiADS based on multi-agent reinforcement learning algorithm

The proton linac accelerator has been the mostly appropriate equipment applied to radioactive waste disposal and cancer treatment. However, beam offset is a particularly significant problem during proton acceleration, which not only produces a halo and causes orbital distortion of the beam to make the beam injection inefficient. In exceptionally complex accelerator systems, accurate correction of beam offset is very difficult. The traditional method is based on manual experience and manual adjustment to achieve beam trajectory correction, the automation level is low and the adjustment speed is slow. Therefore, this paper proposes a new method of beam orbit correction based on Multi-Agent Reinforcement Learning. In order to verify the feasibility of the algorithm in beam orbit correction, this method takes the MEBT section of the virtual accelerator as the control object, and uses three agents to control the beam orbit of Medium Energy Beam Transport (MEBT) in sections. Each agent is designed based on the Reinforcement Learning of the Deep Deterministic Policy Gradient (DDPG) algorithm, the deterministic strategy is used to explore the large state and action space to realize discretization, and the training model convergence is successfully realized. The experimental results show that the final calibrated average beam offset of this method is about 0.39 mm, which achieves the expected experimental goal. It can control the beam orbit of MEBT, and has application value and competitiveness in the beam orbit correction of linac.

Real-Time Regulation of Beam-Based Feedback: Implementing an FPGA Solution for a Continuous Wave Linear Accelerator.

Control applications targeting fast industrial processes rely on real-time feasible implementations. One of such applications is the stabilization of an electron bunch arrival time in the context of a linear accelerator. In the past, only the electric field accelerating the electron bunches was actively controlled in order to implicitly stabilize the accelerated electron beam. Nowadays, beam properties are specifically measured at a target position and then stabilized by a dedicated feedback loop acting on the accelerating structures. This dedicated loop is usually referred to as a beam-based feedback (BBF). Following this, the control system of the electron linear accelerator for beams with high brilliance and low emittance (ELBE) is planned to be upgraded by the BBF, and the problem of implementing a designed control algorithm becomes highly relevant. In this work, we propose a real-time feasible implementation of a high-order regulator based on a field-programmable gate array (FPGA). By presenting simulation and synthesis results made in hardware description language (HDL) VHDL, we show that the proposed digital solution is fast enough to cover the bunch repetition rates frequently used at ELBE, such as 100 . Finally, we verify the implementation by using a dedicated FPGA testbench.

Demonstration of an intense lithium beam for forward-directed pulsed neutron generation

As an alternative to research nuclear reactors, a compact accelerator-driven neutron generator that uses a lithium beam driver could be a promising candidate since it produces almost no undesired radiation. However, providing an intense lithium-ion beam has been difficult, and it has been thought that the practical application of such a device would be impossible. The most critical problem of insufficient ion fluxes has been solved by applying a direct plasma injection scheme. In this scheme, a pulsed high-density plasma from a metallic lithium foil generated by laser ablation is efficiently injected and accelerated by a radio-frequency quadrupole linear accelerator (RFQ linac). We have obtained a peak beam current of 35 mA accelerated to 1.43 MeV, which is two orders of magnitude higher than a conventional injector and accelerator system can deliver.

Characterization of the isobar separator for anions integrated into the A. E. Lalonde laboratory’s 3 MV AMS system

Accelerator Mass Spectrometry (AMS) of isotopes with abundant negative ion-forming isobars often requires the use of large accelerators to achieve high sensitivity measurements. The Isobar Separator for Anions (ISA) is a radio frequency quadrupole (RFQ) reaction cell system that can provide selective isobar suppression in the low energy system, prior to injection into an accelerator. A commercial version from Isobarex Corp. has been installed in a second injection line of the 3 MV tandem accelerator system at the A. E. Lalonde AMS Laboratory, University of Ottawa. Here, we present the characterization of the ISA for its optimization of S– suppression and Cl– transmission. He gas was selected as a cooling buffer gas, as it provided the best Cl– transmission of ∼50 % through the ISA column. These tests use NO2 as a reaction gas due to its well-known exothermic reaction with S– but endothermic reaction with Cl–. More than six orders of magnitude reduction of S to Cl has been observed.

Refrigerant flow distribution research for battery cooling coupled with cabin comfort based on dual-evaporator heat pump system for electric vehicle acceleration

Abstract The heat pump system employed with a dual evaporator for battery cooling coupled with cabin comfort is an innovative thermal management method. It can be inferred that the refrigerant thermal load distribution can trigger temperature fluctuations for thermal performance of both battery and cabin. To trade-off between the thermal management demands of battery and cabin, this study proposed a strategy to promote the decreasing of battery temperature and to ensure battery thermal uniformity with a higher priority. Hence a transient refrigerant flow rate distribution scheme with a minimum flow rate to satisfy battery thermal demands was designed. The results showed a significant reduction in the temperature fluctuations for cabin thermal comfort and BTM thermal controlling. It offers a satisfactory reference for refrigerant thermal load distribution strategy applied with the heat pump system connected to the battery and cabin by the dual evaporator.

CONDITION AND DEVELOPMENT OF UNPLANNED TOURISM BUSINESS IN TASIKMALAYA DISTRICT

Tourism activities in Tasikmalaya Regency were recorded to experience fluctuations in the number of tourist visits from year to year. There are two classifications of tourist objects in Tasikmayala Regency, namely legal and registered tourist objects and tourist objects that are not legal and not yet registered in the Disparpora Tasikmalaya Regency and the Disparbud of West Java Province. Tourism objects that are not yet legal, have not been recorded, grow from the initiative of tourists or the community, and are not yet well established in their management are referred to as unplanned tourism businesses. The study was conducted to describe the conditions and developments in the unplanned tourism business carried out in three locations, namely Pasir Kirisik, Bukit Panyangrayan, and Bukit Puji Ningrum by conducting direct observations, interviews, and questionnaires. All data collected were then analyzed descriptively covering the general condition of the research location, characteristics of tourists and local communities, perceptions, motivations, preferences, participation, and conditions of business development factors in each research location. The condition of the unplanned tourism business shows that it is running well even though it has not run optimally and is accepted by tourists and the local community. So it is important the role of stakeholders according to their capacity and capability as part of the support system for procedural acceleration of an unplanned tourism business into an established ecotourism business. &#x0D; &#x0D; Key words: acceleration, ecotourism, established, intuitive, tourism business

Synergistic combination of 2-mercaptobenzothiazole (MBT) and nitrosoamine-safe thiuram disulfide as advanced rubber vulcanizing accelerators

In this study, we developed a combination accelerator system to synergistically improve the vulcanizing activity of 2-marcapto benzothiazole (MBT) with different nitrosamine-safe thiuram disulfides (TDs), namely, bis-(N-benzyl piperazino) thiuram disulfide (BPTD), bis-(N-phenyl piperazino) thiuram disulfide (PPTD), and bis-(N-ethyl piperazino) thiuram disulfide (EPTD), which can be used as nitrosamine-safe cross-linking accelerators for sulfur based rubber vulcanization. The results were compared with that of a conventionally unsafe TD, tetramethyl thiuram disulfide (TMTD) combined with MBT, to evaluate the efficiencies of these nitrosamine-safe TD accelerators for industrial applications. The curing and mechanical properties of rubber vulcanization were investigated to understand the synergism between MBT and nitrosoamine-safe TDs. The results indicate that novel TDs combined with MBT significantly improve the curing characteristics and mechanical properties following to the TDs with smaller molecular size and higher basicity. The MBT/BPTD system had higher modulus values because BPTD has a higher molar mass, which facilitates a better distribution of accelerators in the rubber matrix. Overall, the MBT/EPTD accelerator systems with equal molar ratios can compete with the curing rates, tensile strengths, and moduli of unsafe TMTD accelerator systems in the vulcanization of rubber.

Highly Parallel Multi-FPGA System Compilation from Sequential C/C++ Code in the AWS Cloud

We present a High Level Synthesis compiler that automatically obtains a multi-chip accelerator system from a single-threaded sequential C/C++ application. Invoking the multi-chip accelerator is functionally identical to invoking the single-threaded sequential code the multi-chip accelerator is compiled from. Therefore, software development for using the multi-chip accelerator hardware is simplified, but the multi-chip accelerator can exhibit extremely high parallelism. We have implemented, tested, and verified our push-button system design model on multiple field-programmable gate arrays (FPGAs) of the Amazon Web Services EC2 F1 instances platform, using, as an example, a sequential-natured DES key search application that does not have any DOALL loops and that tries each candidate key in order and stops as soon as a correct key is found. An 8- FPGA accelerator produced by our compiler achieves 44,600 times better performance than an x86 Xeon CPU executing the sequential single-threaded C program the accelerator was compiled from. New features of our compiler system include: an ability to parallelize outer loops with loop-carried control dependences, an ability to pipeline an outer loop without fully unrolling its inner loops, and fully automated deployment, execution and termination of multi-FPGA application-specific accelerators in the AWS cloud, without requiring any manual steps.