Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Experiment Optimization”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 631 records · Page 35
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Remote Detection of Uranium Using Self-Focusing Intense Femtosecond Laser Pulses

Optical measurement techniques can address certain important challenges associated with nuclear safety and security. Detection of uranium over long distances presents one such challenge that is difficult to realize with traditional ionizing radiation detection, but may benefit from the use of techniques based on intense femtosecond laser pulses. When a high-power laser pulse propagates in air, it experiences collapse and confinement into filaments over an extended distance even without external focusing. In our experiments, we varied the initial pulse chirp to optimize the emission signal from the laser-produced uranium plasma at an extended distance. While the ablation efficiency of filaments formed by self-focusing is known to be significantly lower when compared to filaments produced by external focusing, we show that filaments formed by self-focusing can still generate luminous spectroscopic signatures of uranium detectable within seconds over a 10-m range. The intensity of uranium emission varies periodically with laser chirp, which is attributed to the interplay among self-focusing, defocusing, and multi-filament fragmentation along the beam propagation axis. Grouping of multi-filaments incident on target is found to be correlated with the uranium emission intensity. The results show promise towards long-range detection, advancing the diagnostics and analytical capabilities in ultrafast laser-based spectroscopy of high-Z elements.

42 ENGINEERING↗
DOE OSTI2023

Ducted Fuel Injection Provides Consistently Lower Soot Emissions in Sweep to Full-Load Conditions

Earlier studies have proven how ducted fuel injection (DFI) substantially reduces soot for low- and mid-load conditions in heavy-duty engines, without significant adverse effects on other emissions. Nevertheless, no comprehensive DFI study exists showing soot reductions at high- and full-load conditions. This study investigated DFI in a single-cylinder, 1.7-L, optical engine from low- to full-load conditions with a low-net-carbon fuel consisting of 80% renewable diesel and 20% biodiesel. Over the tested load range, DFI reduced engine-out soot by 38.1–63.1% compared to conventional diesel combustion (CDC). This soot reduction occurred without significant detrimental effects on other emission types. Thus, DFI reduced the severity of the soot–NO x tradeoff at all tested conditions. While DFI delivered considerable soot reductions in thepresent study, previous DFI studies at low- and mid-load conditions delivered larger soot reductions (>90%) compared to CDC operation at the same conditions. Therefore, the DFI configuration used here has been deemed nonoptimal (in terms of parameters such as the injector-spray and piston geometries), and several improvements are recommended for future studies with high-load DFI. Further, these improvements include employing better spray-duct alignment, a deeper piston bowl with a smaller injector umbrella angle, and a fuel injector that opens and closes faster. The study also suggests future research to make DFI ready for commercialization, such as metal-engine tests to ensure desirable DFI performance over an engine’s complete speed/load map. Overall, this study supports the continued development and commercialization of DFI to meet upcoming emissions regulations for heavy-duty vehicles. Specifically, multicylinder engine experiments and CFD simulations should be utilized to optimize the performance and clarify the full potential of DFI.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2020

The Cloud-resolving model Radar SIMulator (CR-SIM) Version 3.3: description and applications of a virtual observatory

Ground-based observatories use multi-sensor observations to characterize cloud and precipitation properties. One of the challenges is how to design strategies to best use these observations to understand these properties and evaluate weather and climate models. This paper introduces the Cloud resolving model Radar SIMulator (CR-SIM), which uses output from high-resolution cloud resolving models (CRMs) to emulate multi-wavelength, zenith-pointing, and scanning radar observables and multi-sensor (radar and lidar) products. CR-SIM allows direct comparison between an atmospheric model simulation and remote-sensing products using a forward-modeling framework consistent with the microphysical assumptions used in the atmospheric model. CR-SIM has the flexibility to easily incorporate additional microphysical modules, such as microphysical schemes and scattering calculations, and expand the applications to simulate multi-sensor retrieval products. In this paper, we present several applications of CR-SIM for evaluating the representativeness of cloud microphysics and dynamics in a CRM, quantifying uncertainties in radar-lidar integrated cloud products and multi-Doppler wind retrievals, and optimizing radar sampling strategy using observing system simulation experiments. These applications demonstrate the application of CR-SIM as a virtual observatory operator on high-resolution model output for a consistent comparison between model results and observations to aid interpretation of the differences and improve understanding of the representativeness errors due to the sampling limitations of the ground-based measurements. CR-SIM is licensed under the GNU GPL package and both the software and the user guide are publicly available to the scientific community.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

νBDX: a Coherent Elastic Neutrino Nucleus Scattering (CEνNS) experiment at Jefferson Lab

Particle physics is an ever-changing field seeking to explain phenomena that lie at the limits of our current knowledge. Among discovered and hypothesized interactions, Coherent Elastic Neutrino-Nucleus Scattering (CE?NS) stands out as a fascinating and elusive phenomenon, offering a unique window into the nature of neutrinos and their interactions with matter. CE?NS, a process predicted by the Standard Model, represents one of the subtle ways in which neutrinos can interact with atomic nuclei. Unlike other neutrinonucleus interactions, CE?NS occurs coherently, meaning that the entire nucleus collectively responds to the neutrino-mediated weak force, resulting in small energy transfers to the nucleus as a whole. The extremely rare nature of CE?NS events presents a formidable challenge to experimenters seeking to detect and study these interactions. Recent advances in detector technologies and the implementation of large-scale experiments, such as coherent neutrino scattering experiments at research facilities around the world, have opened up new avenues for exploring this elusive phenomenon. These experiments aim not only to observe CE?NS directly but also to extract valuable information on neutrino properties, nuclear structure, and potential deviations from the Standard Model. This thesis aims to study the feasibility of an experiment to study CE?NS at The Thomas Jefferson National Accelerator Facility, Newport News, VA, USA. This would be an experiment that exploits the production of neutrinos by an electron beam dumped on a thick target (the beam dump). The large quantity of neutrinos produced and their energy profile are well suited to the purpose of studying the CE?NS. In my thesis, I will start from previous studies on CE?NS and from theoretical considerations regarding the process, then quantify the expected events for a hypothetical experiment and possible backgrounds through simulations aimed at optimizing the geometry of the future detector. I will also show some tests on selected detector components that will be part of a first prototype.

Grazzi, Stefano↗
DOE OSTI2021

Realization of a high luminosity muon beam at the Jefferson Lab

The proton is the major component of the visible matter of our universe, so being able to determine its fundamental properties, such as its charge radius, is of vital importance. In recent years, several experiments have been performed for the accurate measurement of the proton radius with different experimental techniques. The main and most common are based on two different techniques: ep scattering and Lamb shift spectroscopy of regular hydrogen atoms. The latter can also be performed on muonic hydrogen atoms with a significant improvement in accuracy. Recent measurements that use this method have highlighted a discrepancy between the results obtained with the diffusion of electrons and those obtained with the spectroscopic technique leading to the so-called ?proton radius puzzle?. The explanation of why this difference between the measured values exists is still a matter of discussion. The last attempt to address the puzzle was the PRad Experiment. It was performed with a unique experimental setup that gave the scientists an excellent control on the systematic errors that are usually present in electron-proton scattering experiments. This has significantly increased the precision of the measurement, bringing it in the same area as the spectroscopic ones done on muonic hydrogen atoms, which are considered the most precise. However, to solve the puzzle once and for all, even more precise electron scattering measurements are needed. Alternatively, new methods can be used, such as eg. muon-proton scattering at low transferred momenta in order to lower the systematic errors that comes with the measurement. This thesis tries to address this problem by studying the set up of an intense muon beam at the Jefferson Laboratory (JLab) that can be used to carry out a high-precision m ? p scattering experiment. Using Montecarlo simulations I studied the muon production resulting from the interaction of the primary electron beam with the beam-dump at JLab Hall-A. The simulations were performed using the FLUKA package via the FLAIR graphical interface. In this study, I modified an existing geometry (built for the ?Beam Dump eXperiment?) and performed several studies in order to optimize the muon beam parameters. In order to estimate the beam intensity, I studied the muon generation and transportation through the concrete bunker surrounding the Hall-A beam-dump. Detailed studies on muon attenuation and dispersion at different depth allowed me to track the beam profile in space and momentum as a function of the distance from the beam-dump. Last, but not least, I studied a possible tagging system for the muon beam. The tagging system will allow us to know with good precision (O(1=1000)) the energy of each muon of the beam. Based on the results obtained, I created an event generator that incorporates the detailed description of the muon beam (energy, position, emittance) useful to define and optimize the experimental setup for a future measurement of the proton radius at the JLab.

Fulci, Antonino↗
DOE OSTI2024

Distributed Multi-GPU Community Detection on Exascale Computing Platforms

Community detection is a fundamental operation in graph mining, and by uncovering hidden structures and patterns within complex systems it helps solve fundamental problems pertaining to social networks, such as information diffusion, epidemics, and recommender systems. Scaling graph algorithms for massive networks becomes challenging on modern distributed-memory multi-GPU (Graphics Processing Unit) systems due to limitations such as irregular memory access patterns, load imbalances, higher communication-computation ratios, and cross-platform support. We present a novel algorithm HiPDPL-GPU (Distributed Parallel Louvain) to address these challenges. We conduct experiments involving different partitioning techniques to achieve an optimized performance of HiPDPL-GPU on the two largest supercomputers: Frontier and Summit. Remarkably, HiPDPL-GPU processes a graph with 4.2 billion edges in less than 3 minutes using 1024 GPUs. Qualitatively, the performance of HiPDPL-GPU is similar or better compared to other state-of-the-art CPU- and GPU-based implementations. While prior GPU implementations have predominantly employed CUDA, our first-of-its-kind implementation for community detection is cross-platform, accommodating both AMD and NVIDIA GPUs.

Sattar, Naw Safrin↗
DOE OSTI2025

Enhancing Post-Irradiation Examination (PIE) Efficiency with High-Throughput Experiments

Nuclear materials testing is essential for understanding changes in reactor environments, but traditional methods are costly and time-consuming. High-Throughput Experimentation (HTE) enables rapid screening and optimization of material properties by significantly increasing the number of specimens in the experiment. In this work, HTE methods were employed to irradiate novel materials at the Advanced Test Reactor (ATR) at Idaho National Laboratory. A total of 367 sub-sized tensile specimens were irradiated using the standard capsule design. To facilitate HTE, automated systems such as the Automated Mechanical Testing System were implemented for mechanical testing within hot cell environments, thereby reducing the need for manual interaction with hot cell manipulators and significantly lowering both the cost and duration of experiments. Digital image correlation (DIC) was used to measure displacement during mechanical testing, providing precise, non-contact measurement capabilities. The methodologies developed in this research will support the advancement of HTE techniques for future nuclear materials testing experiments at ATR. This approach allows for more efficient and cost-effective evaluation of new materials, ultimately leading to improved reactor performance, enhanced safety, and extended longevity within the nuclear industry.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2024

Aging heat treatment design for Haynes 282 made by wire-feed additive manufacturing using high-throughput experiments and interpretable machine learning

Wire-feed additive manufacturing (WFAM) produces superalloys with complex thermal cycles and unique microstructures, often requiring optimized heat treatments. To address this challenge, we present a hybrid approach that combines high-throughput experiments, precipitation simulation, and machine learning to design effective aging conditions for the WFAM Haynes 282 superalloy. Our results demonstrate that the γ’ radius is the critical microstructural feature for strengthening Haynes 282 during post-heat treatment compared with the matrix composition and γ’ volume fraction. New aging conditions at 770°C for 50 hours and 730°C for 200 hours were discovered based on the machine learning model and were applied to enhance yield strength, bringing it on par with the wrought counterpart. This approach has significant implications for future AM alloy production, enabling more efficient and effective heat treatment design to achieve desired properties.

CALPHAD↗
DOE OSTI2025

Optimizing the Smoothness and Thickness Uniformity of Thin-Film Parylene-N Vapor-Deposited Coatings for Inertial Confinement Fusion Experiments

Polymer coatings with submicrometer smoothness and constant thickness are a required component in a variety of inertial confinement fusion experiments. Smoothness is important for minimizing Rayleigh-Taylor-driven hydrodynamic instabilities, and uniform thickness is important for uniform shock propagation and shell convergence, both of which are critical phenomena that affect the experiment. The preferred polymer coating method is to vapor deposit the parylene-N polymer because it provides nominally smooth conformal coatings. As the coating thickness exceeds 5 µm, however, dome-shaped nodular growth defects develop and the thickness will vary by up to 17% over a distance of 3 cm. This study presents a deterministic method for achieving uniform film thicknesses with ±2% variability over 3 cm and a predictive method to control the thickness to within 5% of the desired value. A coating smoothness of ∼50 nm rms, measured over 40 000 µm 2 , was achieved by adding additional surfaces near the substrates. This additional area improved the thickness uniformity, an effect that is attributed to the low sticking coefficient of the parylene monomer.

chemical vapor deposition (CVD)↗
DOE OSTI2022

Verification of Upcoming MCNP Features For Estimating Nuclear Data Sensitivities in Fixed Source Simulations [Abstract]

Predictive simulation codes, like the Monte Carlo N-Particle (MCNP) transport code, are used throughout the nuclear community. These simulations are based on nuclear data. Maximizing the accuracy and precision of nuclear data maximizes the accuracy and precision of the overall simulation. This is imperative to applications that rely on simulations. For example, improving nuclear data for special nuclear material improves simulation accuracy in stockpile stewardship applications, which results in larger safety margins and decreased operational costs. The improvement and validation of nuclear data is completed through integral benchmark experiments. Past benchmarks have primarily been limited to focus on the effective multiplication factor ($\kappa$ eff ); broadening the purview of benchmarks beyond $\kappa$ eff -dependent nuclear data addresses nuclear data deficiencies. Different response types depend on different areas of nuclear data. This dependence is quantified as nuclear data sensitivity: the change in response due to perturbation of a contributing parameter. The larger the nuclear data sensitivity of a response, the more the experiment is influenced by the uncertainties of the nuclear data. The optimization of nuclear data sensitivities in future benchmarks would result in more detailed validation of lesser studied areas of nuclear data. Currently, direct sensitivity capabilities are not easily found for all experiment types and parameters. An MCNP tool to directly estimate the cross section sensitivities of tallied values is under development. Additionally, updates have been made to the perturbation feature of MCNP, which can be used in a less direct approach to estimating sensitivities. This work verifies these features to estimate nuclear data sensitivities in fixed source simulations of a 4.5-kg sphere of alpha- phase weapons-grade plutonium surrounded by differing amounts of copper and polyethylene. Integrated estimates made using MCNP’s tools were found to statistically agree with integrated estimates made from manual perturbation of nuclear data proving the validity of the MCNP tools.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2022

Optimization of the first CUPID detector module

CUPID will be a next generation experiment searching for the neutrinoless double β decay, whose discovery would establish the Majorana nature of the neutrino. Based on the experience achieved with the CUORE experiment, presently taking data at LNGS, CUPID aims to reach a background free environment by means of scintillating Li 2 100 MoO 4 crystals coupled to light detectors. Indeed, the simultaneous heat and light detection allows us to reject the dominant background of α particles, as proven by the CUPID-0 and CUPID-Mo demonstrators. In this work we present the results of the first test of the CUPID baseline module. In particular, we propose a new optimized detector structure and light sensors design to enhance the engineering and the light collection, respectively. We characterized the heat detectors, achieving an energy resolution of (5.9 ± 0.2) keV FWHM at the Q-value of 100 Mo (about 3034 keV). We studied the light collection of the baseline CUPID design with respect to an alternative configuration which features gravity-assisted light detectors’ mounting. In both cases we obtained an improvement in the light collection with respect to past measures and we validated the particle identification capability of the detector, which ensures an α particle rejection higher than 99.9%, fully satisfying the requirements for CUPID.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2024

Hydrothermal solubility of Dy hydroxide as a function of pH and stability of Dy hydroxyl aqueous complexes from 25 to 250 °C

The rare earth elements (REE) have important applications in green energy technologies. The formation of mineral deposits in geologic systems commonly involves hydrothermal fluids which can mobilize the REE. However, the REE speciation is not well known as a function of pH. The thermodynamic properties of REE hydroxyl complexes used in geochemical models are based on the Helgeson-Kirkham-Flowers (HKF) equation of state parameters which were derived by extrapolation of low temperature experimental and estimated data. In this study, Dy hydroxide solubility experiments are combined with available literature data to improve these models from 25 to 250 °C and optimize the thermodynamic properties of Dy 3+ and Dy hydroxyl complexes using GEMSFITS. Batch-type solubility experiments were conducted from 150 to 250 °C and at saturated water vapor pressure in perchloric acid solutions with initial pH values of 2 to 5 in 0.5 pH unit increments. The measured solubility of Dy hydroxide is retrograde with temperature and decreases with pH. The logarithm of total dissolved Dy molality ranges from –2.3 to –5.3 at 150 °C (pH 4.7–5.5), from –2.4 to –5.6 at 200 °C (pH 3.9–5.1), and from –3.7 to –6.9 at 250 °C (pH of 3.4 and 5.0). The optimized standard partial molal Gibbs energies of formation (Δ f G° T ) derived for Dy 3+ and DyOH 2+ display a close to linear relationship with temperature, fitting with previous optimizations based on DyPO 4 solubility data in the literature. A comparison of the optimized ΔfG°T values for aqueous Dy species with predictions from available HKF parameters indicates significant differences ranging from +11 to –26 kJ/mol between 25 and 250 °C. The experimental fits are used to derive the Dy hydroxide solubility products (K s0 ) and formation constants for the hydrolysis of Dy (β n with n = 1 to 3; Dy 3+ + nOH – = DyOH n 3-n ) as a function of temperature. The optimization method presented yields accurate thermodynamic properties for the Dy 3+ aqua ions and the DyOH 2+ species at the acidic to mildly acidic pH studied whereas more experimental work is needed at near-neutral and alkaline conditions to better constrain the other hydroxyl complexes. Furthermore, the optimized thermodynamic data have a significant impact on geochemical modeling of the mobility and solubility of REE minerals in acidic hydrothermal fluids.

58 GEOSCIENCES↗
DOE OSTI2022

Methods and Results for Quantum Optimal Pulse Control on Superconducting Qubit Systems

The effective use of current Noisy Intermediate-Scale Quantum (NISQ) devices is often limited by the noise which is caused by interaction with the environment and affects the fidelity of quantum gates. In transmon qubit systems, the quantum gate fidelity can be improved by applying control pulses that can minimize the effects of the environmental noise. In this work, we employ physics-guided quantum optimal control strategies to design optimal pulses driving quantum gates on superconducting qubit systems. We test our results by conducting experiments on the IBM quantum hardware using their OpenPulse API. We compare the performance of our pulse-optimized quantum gates against the default quantum gates and show that the optimized pulses improve the fidelity of the quantum gates, in particular the single-qubit gates. Here we discuss the challenges we encountered in our work and point to possible future improvements.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

ARENA: Adversary-Resistant Evolving Neural Architectures

Neural networks are becoming the cornerstone for national security prediction tasks. However, designing them requires significant research and trial/error, as they have many hyperparameters, including their computation graph (“architecture”). Neural architecture search (NAS) employs secondary optimizers to search for architectures maximizing objectives like accuracy. Evolutionary algorithms (EAs) are the most used class of optimizer for NAS. However, existing Python libraries for writing EAs limit the complexity of experiments a user can design. In this project, we built ARENA, a Python framework that encodes complex, hyper-realistic EAs. ARENA collects detailed information as it runs and is flexible enough to encode non-EA search algorithms. We tested ARENA on 4 toy optimization problems by encoding 3 search algorithms for each—random search, an EA, and simulated annealing. We also designed an EA that performs NAS on the MNIST dataset. Our experiments suggest the potential for immediate mission impact through solving lab-wide optimization problems.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Comparison of multiobjective optimization methods for the $\mathrm{LCLS-II}$ photoinjector

Particle accelerators are among some of the largest science experiments in the world and can consist of thousands of components with a wide variety of input ranges. These systems can easily become unwieldy optimization problems during design and operations studies. Starting in the early 2000s, searching for better beam dynamics configurations became synonymous with heuristic optimization methods in the accelerator physics community. Genetic algorithms and particle swarm optimization are currently the most widely used. These algorithms can take thousands of simulation evaluations to find optimal solutions for one machine prototype. For large facilities such as the Linac Coherent Light Source (LCLS) and others, this equates to a limited exploration of many possible design configurations. In this paper, the LCLS-II photoinjector is optimized with three optimization algorithms. All optimizations were started from both a uniform random and Latin hypercube sample. In all cases, the optimizations started from Latin hypercube samples outperformed optimizations started from uniform samples. All three algorithms were able to optimize the photoinjector, with the model-based methods approximating the Pareto front in fewer simulation evaluations. This work, in combination with previous optimization observations, indicates objective penalties have a strong impact on the efficiency of such methods. In general, we recommend heuristic methods for initial optimizations and model-based methods when information about the objective space is available.

43 PARTICLE ACCELERATORS↗
DOE OSTI2022

Evaluating the effect of ionomer chemical composition in silver-ionomer catalyst inks toward the oxygen evolution reaction by half-cell measurements and water electrolysis

The effect of anion exchange ionomer (AEI) chemistry on the kinetics of the oxygen evolution reaction (OER) was systematically studied in both half-cell and single-cell electrolysis experiments. OER was studied in 1 M K 2 CO 3 at 50°C using an array of ionomer-silver catalyst inks deposited on Ni foam. Different ionomer modifications were investigated to optimize the OER performance. The AEI used in this work features a block copolymer backbone of polychloromethylstyrene-b-polycyclooctene-b-polychloromethylstyrene which was functionalized with either benzyltrimethylammonium (TMA) or benzylmethylpiperidinium (MPRD) quaternary ammonium cations. Here, using an MPRD quaternary ammonium cation shifted the rate determining step at high overpotentials, ultimately providing enhanced performance. Increasing the catalyst ink dilution was also found to substantially improve mass activity and catalyst utilization (increase from 35 to 88 A g –1 Ag), likely by reducing the ionomer thickness and decreasing transport resistances. In the final ionomer modification, the Ni foam substrate was leveraged to partially hydrogenate the polycyclooctene midblock to be polyethylene-like at the Ni interface. A significant increase in the electrochemical surface area (Cdl increased from 8.8 to 20.8 mF cm –2 ) and performance (by 47 mA cm –2 or 29 A g –1 Ag) was observed with the incorporation of polyethylene. Kinetic results from half-cell experiments were validated via single-cell anion exchange membrane electrolysis experiments, where the optimized electrode displayed enhanced performance in both 1 M K 2 CO 3 and DI H 2 O.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

Combining the synergistic control capabilities of modeling and experiments: Illustration of finding a minimum-time quantum objective

A common way to manipulate a quantum system, for example spins or artificial atoms, is to use properly tailored control pulses. In order to accomplish quantum information tasks before coherence is lost, it is crucial to implement the control in the shortest possible time. Here we report the near time-optimal preparation of a Bell state with fidelity higher than 99% in an NMR experiment, which is feasible by combining the synergistic capabilities of modeling and experiments operating in tandem. In this work, the pulses preparing the Bell state are found by experiments that are recursively assisted with a gradient-based optimization algorithm working with a model. Thus, we exploit the interplay between model-based numerical optimal design and experimental-based learning control. Utilizing the balanced synergism between the dual approaches, as dictated by the case specific capabilities of each approach, should have broad applications for accelerating the search for optimal quantum controls.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2025

A Portfolio Approach to Massively Parallel Bayesian Optimization

One way to reduce the time of conducting optimization studies is to evaluate designs in parallel rather than just one-at-a-time. For expensive-to-evaluate black-boxes, batch versions of Bayesian optimization have been proposed. They work by building a surrogate model of the black-box to simultaneously select multiple designs via an infill criterion. Still, despite the increased availability of computing resources that enable large-scale parallelism, the strategies that work for selecting a few tens of parallel designs for evaluations become limiting due to the complexity of selecting more designs. It is even more crucial when the black-box is noisy, necessitating more evaluations as well as repeating experiments. Here we propose a scalable strategy that can keep up with massive batching natively, focused on the exploration/exploitation trade-off and a portfolio allocation. We compare the approach with related methods on noisy functions, for mono and multi-objective optimization tasks. These experiments show orders of magnitude speed improvements over existing methods with similar or better performance.

97 MATHEMATICS AND COMPUTING↗