Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “volume distribution”

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 793 records · Page 44
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Rotational and reflectional equivariant convolutional neural network for data-limited applications: Multiphase flow demonstration

This article deals with approximating steady-state particle-resolved fluid flow around a fixed particle of interest under the influence of randomly distributed stationary particles in a dispersed multiphase setup using convolutional neural network (CNN). The considered problem involves rotational symmetry about the mean velocity (streamwise) direction. Thus, this work enforces this symmetry using SE(3)-equivariant, special Euclidean group of dimension 3, CNN architecture, which is translation and three-dimensional rotation equivariant. This study mainly explores the generalization capabilities and benefits of a SE(3)-equivariant network. Accurate synthetic flow fields for Reynolds number and particle volume fraction combinations spanning over a range of [86.22, 172.96] and [0.11, 0.45], respectively, are produced with careful application of symmetry-aware data-driven approach.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2024

Scaling Ensembles of Data-Intensive Quantum Chemical Calculations for Millions of Molecules

Deep learning models are efficient computational tools that can accelerate the inverse design of molecules with desired functional properties by generating predictions at a fraction of the time required by traditional quantum chemical approaches. To ensure that a model maintains accuracy and transferability across broad regions of the chemical space explored during the inverse design, it must be trained on massively large volumes of simulation data. This requires running large-scale ensemble quantum chemical calculations on high-performance computing (HPC) systems for data collection. However, the efficient execution of such large ensemble calculations and the management of large volumes of output data require tools that can judiciously utilize computational resources and manage metadata overhead on the file system. Therefore, we present a high-performance, scalable, ensemble management framework for performing data-intensive quantum chemical electronic structure calculations for organic molecules. This framework provides abstractions to plug different ab initio, first principles, and first principles-based semi-empirical methods and executes them efficiently at large scale on HPC systems. It dynamically distributes tasks to resources and uses tiered storage for managing large collections of files. We employed this framework to process over ten million organic molecules and generate open-source datasets that provide UV-vis absorption spectra by running time-dependent density-functional tight-binding calculations. It is the largest database containing molecular optical spectra that were simulated with quantum chemical methods in a consistent manner.

Mehta, Kshitij↗
DOE OSTI2020

Developing of nano sized fibrous eutectic silicon in hypereutectic Al–Si alloy by laser remelting

Laser surface melting followed by rapid solidification is an effective means to produce very fine microstructures with desirable surface properties because of the high rates of cooling associated with it. In the present study, the effect of rapid cooling on the silicon particle size, distribution, and morphology of hypereutectic Al–17wt.%Si and Al–20wt.%Si alloys have been investigated. A continuous-wave CO2 laser of wavelength 10.6 μm and a Trumpf Yb-YAG disk laser of wavelength 1.030 μm were used with a beam diameter of 1 mm and scanning speeds ranging from 5 to 100 mm/s. Rapid solidification increased the solubility of silicon in aluminum to approximately 5wt% and induced non-equilibrium hypoeutectic microstructures comprising large volume fractions of primary α-Al dendrites and ultrafine Al–Si eutectic of lamellar morphology. Both α-Al dendrites and the silicon particle sizes were significantly reduced from micron to nanoscale level. The morphology of silicon particles is modified from massive polygonal and plate-like to a mixture of fine fakes with round corners, feathery and fibrous, or a coral-like and thread-like structure. The eutectic silicon size and the interlamellar spacing were reduced to 30 and 10 nm, respectively. Furthermore, most of the silicon crystals in the eutectic region and the aluminum dendrites contained a significant number of twins which were considered as an essential contributor to the mechanism of growth and branching. Microhardness values increased two to threefold due to the refinement of the microstructural constituent.

42 ENGINEERING↗
DOE OSTI2021

Distribution of U, Pu and Np in a Mo-99 Recovery Process using Sachtopore NP (titania based) Absorbent Material

A laboratory scale mockup of a prototypical Mo-99 recovery process has been operated in a radioactive materials hood at Savannah River National Laboratory. This equipment has been operated with realistic concentrations of uranium (U-238), plutonium (94% Pu-239), neptunium (Np-237), and non-radioisotopes of common fission products (including Mo). The goal of this work was to determine the amounts of U, Pu, and Np in the liquid waste streams created by the wash streams for the production process. The key question was how the initial acid wash volume will be handled, as this solution contains residual raffinate which amounted to ~ 3% ofthe mass ofthe feed solution in this experiment. If the initial 1.9 bed volumes (BV) of acid wash were recycled as in this experiment then the combined wash solutions can be expected to contain 0.4 g U/L, 2.7 pg Pu/L and < 14 pg Np/L. If, on the other extreme, the entire acid wash solution was discarded as waste, then the combined wash solutions can be expected to contain 11 g U/L, 3.8 pg Pu/L and ~ 80 pg Np/L. The Sachtopore NP absorbent (for Mo-99 recovery) retains a minuscule fraction ofthe U (2.3 mg U per g titania or 0.03% ofthe U in the feed). The uncertainties are too large to determine how close the absorbent is to saturation with U. Less mass but a higher fraction ofthe total Np was retained by the absorbent (0.7 pg Np per g titania or ~2% of the Np in the feed). As has been reported elsewhere, a large fraction ofthe Pu fed to the titania bed was retained by the absorbent (92% of Pu in the feed or 20 pg per g titania).

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2025

Comparing CRT and Pandora Tagged Tracks in ICARUS

The ICARUS Liquid Argon Time Projection Chamber (LArTPC) is used to investigate short-baseline neutrino oscillations and the possible existence of sterile neutrinos. To aid in identifying and rejecting cosmic ray backgrounds, the system is outfitted with external Cosmic Ray Taggers (CRTs). The CRT is able to measure more tracks and at a wider range of angles, useful in further detailing future calibrations. This project compared the most probable values (MPVs) of charge deposition between CRT tagged and TPC tagged tracks. The Y-Z maps of charge scale for both datasets show a strong correlation, not just in their overall structure, but also in how they reflect known detector effects. Both reconstruction paths showed similar structures in their charge distribution, such as less effective wire section or mechanical support, confirming that CRT-tagged tracks, despite being external, respond to the same calibration landscape as Pandora tracks, validating previous Pandora reconstructions and values. Overall, the analysis supports the idea that the CRT tags are a useful addition to ICARUS calibration efforts. Their broad angular range and independence from Pandora reconstruction make them a helpful secondary tool and a potential asset in extending calibration across the detector volume.

Thayer, Abigail [Colorado State U.] (ORCID:0009000↗
DOE OSTI2025

Comparing CRT and Pandora Tagged Tracks in ICARUS

The ICARUS Liquid Argon Time Projection Chamber (LArTPC) is used to investigate short-baseline neutrino oscillations and the possible existence of sterile neutrinos. To aid in identifying and rejecting cosmic ray backgrounds, the system is outfitted with external Cosmic Ray Taggers (CRTs). The CRT is able to measure more tracks and at a wider range of angles, useful in further detailing future calibrations. This project compared the most probable values (MPVs) of charge deposition between CRT tagged and TPC tagged tracks. The Y-Z maps of charge scale for both datasets show a strong correlation, not just in their overall structure, but also in how they reflect known detector effects. Both reconstruction paths showed similar structures in their charge distribution, such as less effective wire section or mechanical support, confirming that CRT-tagged tracks, despite being external, respond to the same calibration landscape as Pandora tracks, validating previous Pandora reconstructions and values. Overall, the analysis supports the idea that the CRT tags are a useful addition to ICARUS calibration efforts. Their broad angular range and independence from Pandora reconstruction make them a helpful secondary tool and a potential asset in extending calibration across the detector volume.

Thayer, Abbie [Fermilab; Colorado State U.]↗
DOE OSTI2021

Insights into storm direction effect on flood response

In this study, we investigate the directional influence of storm movement on catchment flood peak response using the synthetic circular basin. Due to the complexity in defining storm movements that require meteorological modeling, we adopt a novel approach of combining the basin rotation method (BRM) with a circular basin construct. A systematic basin rotation approach provides a proxy for the storm direction relative to the river network topology to study its hydrologic response. Using Stage-IV rainfall with 4-km by 1-hour resolution for 16 years period from 2003 to 2018 in Iowa, U.S., we systematically analyze consequent flood peak response due to storm directions (basin rotations) using a distributed hydrologic model called the Hillslope-Link Model (HLM). HLM has demonstrated the ability to reproduce observed streamflow for real river basins studied in Iowa. The hydrologic simulation results show that the BRM can quantify the relationship between the rainstorm direction and catchment hydrologic response. Also, the maximum flood peak response is strongly dependent on the interaction of the peak of the rainstorm with the peak of the width function. The results indicate significant differences in runoff volume and runoff peak and their interannual variability due to rainstorm direction. The study has important implications for flood frequency predictions and developing flood resilience and watershed management strategies, particularly due to changing storm tracks under a warming climate.

42 ENGINEERING↗
DOE OSTI2021

Overview of the Second Workshop on Scholarly Document Processing

With the ever-increasing pace of research and high volume of scholarly communication, scholars face a daunting task. Not only must they keep up with the growing literature in their own and related fields, scholars increasingly also need to rebut pseudo-science and disinformation. These needs have motivated an increasing focus on computational methods for enhancing search, summarization, and analysis of scholarly documents. However, the various strands of research on scholarly document processing remain fragmented. To reach out to the broader NLP and AI/ML community, pool distributed efforts in this area, and enable shared access to published research, we held the 2nd Workshop on Scholarly Document Processing (SDP) at NAACL 2021 as a virtual event (https://sdproc.org/2021/). The SDP workshop consisted of a research track, three invited talks, and three Shared Tasks (LongSumm 2021, SCIVER, and 3C). The program was geared towards the application of NLP, information retrieval, and data mining for scholarly documents, with an emphasis on identifying and providing solutions to open challenges.

Beltagy, Iz↗
DOE OSTI2024

Efficient Anomaly Detection Driven By Different Machine Learning Architectures And Models

The rapid growth and ubiquitous adoption of the internet and cyber-physical systems (CPS) have fundamentally transformed modern communication, work, and human-system interactions. While networks now form the backbone of critical digital ecosystems, enabling seamless data transmission across diverse, interconnected systems, this increased connectivity also expands the attack surface, making real-time detection of network intrusions and anomalies a pressing challenge. Detecting unusual activities within network infrastructure requires advanced data traffic analysis to differentiate between legitimate and malicious interactions. Traditional approaches to network anomaly detectionâ??such as rule-based and signature-based systemsâ??often depend on predefined patterns to identify known anomalies, limiting their effectiveness against emerging, stealthy, or previously unseen threats. These conventional methods suffer from high false alarm rates and fail to adapt to the ever-evolving nature of network traffic, particularly in large-scale, decentralized environments where data volume, velocity, and variety are constantly increasing. This dissertation presents artificial intelligence (AI)-driven approaches to anomaly detection that leverage graphics processing unit (GPU)-enabled high-performance computing (HPC) platforms for processing massive network traffic data and monitoring the components of cyber-physical systems (CPS) for potentially hazardous conditions. The research advances several key contributions: (1) Designing efficient machine learning techniques for CPS condition monitoring and anomaly detection; (2) enabling federated learning (FL) frameworks that enable distributed detection while preserving data privacy and system resilience; (3) exploring graph-based methodologies combining graph neural networks (GNN) and graph machine learning (ML) approaches for the Internet of Things (IoT) and automotive network security, and (4) performing distributed edge computing optimizations that integrate FL with scalable technologies for reduced communication overhead. Through extensive experiments, these methodologies demonstrate that complex anomaly detection and condition monitoring tasks can be achieved while balancing computational efficiency and detection accuracy through fine-grained network information processing. The frameworks developed in this research establish a robust foundation for network anomaly detection, providing scalable, adaptive, and privacy-preserving solutions for safeguarding CPS and IoT networks in an increasingly interconnected digital landscape. The practical implications of these research findings are significant, as they can inform the development of next-generation network security systems and contribute to the protection of critical infrastructure against sophisticated cyber attacks.

Marfo, William↗
DOE OSTI2021

Statistical Multiobjective Optimization of Thiospinel CoNi 2 S 4 Nanocrystal Synthesis via Design of Experiments

Thiospinels, such as CoNi 2 S 4 , are showing promise for numerous applications, including as catalysts for the hydrogen evolution reaction, hydrodesulfurization, and oxygen evolution and reduction reactions; however, CoNi 2 S 4 has not been synthesized as small, colloidal nanocrystals with high surface-area-to-volume ratios. Traditional optimization methods to control nanocrystal attributes such as size typically rely upon one variable at a time (OVAT) methods that are not only time and labor intensive but also lack the ability to identify higher-order interactions between experimental variables that affect target outcomes. Herein, we demonstrate that a statistical design of experiments (DoE) approach can optimize the synthesis of CoNi 2 S 4 nanocrystals, allowing for control over the responses of nanocrystal size, size distribution, and isolated yield. After implementing a 2 5–2 fractional factorial design, the statistical screening of five different experimental variables identified temperature, Co:Ni precursor ratio, Co:thiol ratio, and their higher-order interactions as the most critical factors in influencing the aforementioned responses. Second-order design with a Doehlert matrix yielded polynomial functions used to predict the reaction parameters needed to individually optimize all three responses. A multiobjective optimization, allowing for the simultaneous optimization of size, size distribution, and isolated yield, predicted the synthetic conditions needed to achieve a minimum nanocrystal size of 6.1 nm, a minimum polydispersity (σ/$\bar{d}$) of 10%, and a maximum isolated yield of 99%, with a desirability of 96%. The resulting model was experimentally verified by performing reactions under the specified conditions. Furthermore, our work illustrates the advantage of multivariate experimental design as a powerful tool for accelerating control and optimization in nanocrystal syntheses.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Scaling functions for eigenstate entanglement crossovers in harmonic lattices

For quantum matter, eigenstate entanglement entropies obey an area law or log-area law at low energies and small subsystem sizes and cross over to volume laws for high energies and large subsystems. This transition is captured by crossover functions, which assume a universal scaling form in quantum critical regimes. We demonstrate this for the harmonic lattice model, which describes quantized lattice vibrations and is a regularization for free scalar field theories, modeling, e.g., spin-0 bosonic particles. In one dimension, the ground-state entanglement obeys a log-area law. For dimensions d ≥ 2, it displays area laws, even at criticality. The distribution of excited-state entanglement entropies is found to be sharply peaked around subsystem entropies of corresponding thermodynamic ensembles in accordance with the eigenstate thermalization hypothesis. Numerically, we determine crossover scaling functions for the quantum critical regime of the model and do a large-deviation analysis. We show how infrared singularities of the system can be handled and how to access the thermodynamic limit using a perturbative trick for the covariance matrix. Eigenstates for quasi-free bosonic systems are not Gaussian. Here, we resolve this problem by considering appropriate squeezed states instead. For these, entanglement entropies can be evaluated efficiently.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2026

Characterization of ceramic fuel powder packing fractions to support INFLUX

Triply periodic minimal surface (TPMS)-based structures show marked potential in novel nuclear reactor fuel designs, as their high surface area-to-volume ratio increases the efficiency of heat transfer out of the fuel, enabling safer, more innovative reactor designs. This milestone report addresses the role of dUO 2 powder processing route on the fill behavior of TPMS-based cladding shells to understand and advance the feasibility of manufacturing TPMS-based nuclear fuel forms. dUO 2 powder was processed through either a dry granulation route, varying consolidation pressure, or through milling, varying milling time, milling method and milled size distribution. The lowest tapped bulk densities (TBD), but best powder flowabilities, were obtained when testing unprocessed dUO 2 powder which was prone to self-agglomeration and formed low-density spheroids. The highest TBD and lowest flowabilities were obtained when using powder produced by hammer-milling dUO 2 powder to pass through a 200-mesh sieve, which led to particles with angular morphologies. Powder produced by dry granulation exhibited TBD that varied according to the consolidation pressure used to form the initial pellets and exhibited improved flowabilities when compared to hammer-milled material. Because of the large span of granule sizes formed as well as the irregular shape associated with the granules, a packing fraction of 0.69 was achieved, exceeding the analytical solution for random close packing of mono-sized spheres. TPMS polymer shells were loaded with unprocessed, granulated, and hammer-milled dUO 2 powders, and their qualitative packing behaviors were analyzed using x-ray computed tomography (xCT). TBDs calculated after loading TPMS polymer shells were 10-20% lower when compared to tapped bulk density measurements taken in a glass graduated cylinder, indicating a non-trivial impact on the tapped bulk density of either the TPMS channel size, TPMS channel surface material, powder cohesiveness, or a combination of the two parameters. A metallic zircaloy-4 TPMS shell will be loaded with hammer-milled dUO 2 powder upon receipt of the shell from Oak Ridge National Laboratory (ORNL) and shipped to Idaho National Labs (INL) for subsequent hot isostatic pressing (HIP) densification experiments.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2023

Conquering Data Chaos: Research Data Management with Kubernetes

Managing massive volumes of data and effectively making it accessible to researchers poses significant challenges and is a barrier to scientific discovery. In many cases, critical data is locked up in unwieldy file formats or one-off databases and is too large to effectively process on a single machine. This talk explores the role of Kubernetes, an open-source container orchestration platform, in addressing research data management challenges. I will discuss how we are using a set of publicly available open-source and home-grown tools in the National Renewable Energy Lab (NREL) Data, Analysis, and Visualization (DAV) group to help researchers overcome data-related bottlenecks. The talk will begin by providing an overview of the data challenges faced in research data management, including data storage, processing, and analysis. I will highlight Kubernetes' ability to handle large-scale data by leveraging containerization and distributed computing, including distributed storage. Kubernetes allows researchers to encapsulate data processing infrastructure and workflows into portable containers, enabling reproducibility and ease of deployment. Kubernetes can then schedule and manage the resource allocation of these containers to enable efficient utilization of limited computing resources, leading to more efficient data processing and analysis. I will discuss some limitations of traditional, siloed approaches to dealing with data and emphasize the need for solutions which foster collaboration. I will highlight how we are using Kubernetes at NREL to facilitate data sharing and cooperation among research teams. Kubernetes' flexible architecture enables the deployment of shared computing environments, such as Apache Superset, where researchers can seamlessly access and analyze shared datasets. Providing the ability to have one research team easily consume data generated by another, utilizing Kubernetes' as a central data platform, is one of the major wins we've encountered by adopting the platform. Finally, I will showcase real-world use cases from NREL where we have used Kubernetes to solve some persistent data challenges involving large volumes of sensor and monitoring data. I will discuss the challenges we encountered when creating our cluster and making it available as a production-ready resource. I will also discuss the specific suite of tools, including Postgres and Apache Druid for columnar and timeseries data, and Redpanda Kafka for streaming data we have deployed in our infrastructure, and the process that went into the selection of these tools.

collaborative environment↗
DOE OSTI2026

W-Band ARM Scanning Cloud Radar (WSACR) 2nd Generation CF-Radial Spectral Data, Vertically-Pointing Scan, Co-Polarization Mode (a1)

ARM's scanning cloud radars are fully coherent dual-frequency, dual-polarization Doppler radars mounted on a common scanning pedestal. Each pedestal includes a Ka-band radar (2kW peak power) and the deployment location determines whether the second radar is a W-band (WSACR; 1.7 kW peak power) or X-band (XSACR; 20 kW peak power). Beamwidths at Ka and W bands are roughly matched at 0.3 degrees. Due to the narrow antenna beamwidth, ARM’s scanning cloud radars use scanning strategies that are unlike typical weather radars. Rather than focusing on plan position indicator, or PPI, scans, the Ka-SACR uses range height indicator, or RHI, scans at numerous azimuths to obtain cloud volume data. Measurements collected with the W-SACR are copolar and cross-polar radar reflectivity, Doppler velocity, spectra width and spectra when not scanning, and linear depolarization ration.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

W-Band ARM Scanning Cloud Radar (WSACR) 2nd Generation CF-Radial Spectral Data, Vertically-Pointing Scan, Cross Mode (a1)

ARM's scanning cloud radars are fully coherent dual-frequency, dual-polarization Doppler radars mounted on a common scanning pedestal. Each pedestal includes a Ka-band radar (2kW peak power) and the deployment location determines whether the second radar is a W-band (WSACR; 1.7 kW peak power) or X-band (XSACR; 20 kW peak power). Beamwidths at Ka and W bands are roughly matched at 0.3 degrees. Due to the narrow antenna beamwidth, ARM’s scanning cloud radars use scanning strategies that are unlike typical weather radars. Rather than focusing on plan position indicator, or PPI, scans, the Ka-SACR uses range height indicator, or RHI, scans at numerous azimuths to obtain cloud volume data. Measurements collected with the W-SACR are copolar and cross-polar radar reflectivity, Doppler velocity, spectra width and spectra when not scanning, and linear depolarization ration.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

W-Band ARM Scanning Cloud Radar (WSACR) 2nd Generation CF-Radial Spectral Data, Vertically-Pointing Scan, Cross-Polarization Model (a1)

ARM's scanning cloud radars are fully coherent dual-frequency, dual-polarization Doppler radars mounted on a common scanning pedestal. Each pedestal includes a Ka-band radar (2kW peak power) and the deployment location determines whether the second radar is a W-band (WSACR; 1.7 kW peak power) or X-band (XSACR; 20 kW peak power). Beamwidths at Ka and W bands are roughly matched at 0.3 degrees. Due to the narrow antenna beamwidth, ARM’s scanning cloud radars use scanning strategies that are unlike typical weather radars. Rather than focusing on plan position indicator, or PPI, scans, the Ka-SACR uses range height indicator, or RHI, scans at numerous azimuths to obtain cloud volume data. Measurements collected with the W-SACR are copolar and cross-polar radar reflectivity, Doppler velocity, spectra width and spectra when not scanning, and linear depolarization ration.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Photometric redshifts probability density estimation from recurrent neural networks in the DECam local volume exploration survey data release 2

Photometric wide-field surveys are imaging the sky in unprecedented detail. These surveys face a significant challenge in efficiently estimating galactic photometric redshifts while accurately quantifying associated uncertainties. In this work, we address this challenge by exploring the estimation of Probability Density Functions (PDFs) for the photometric redshifts of galaxies across a vast area of 17,000 square degrees, encompassing objects with a median 5 σ point-source depth of g = 24.3, r = 23 . 9 , i = 23.5, and z = 22.8 mag. Our approach uses deep learning, specifically integrating a Recurrent Neural Network architecture with a Mixture Density Network, to leverage magnitudes and colors as input features for constructing photometric redshift PDFs across the whole DECam Local Volume Exploration (DELVE) survey sky footprint. Subsequently, we rigorously evaluate the reliability and robustness of our estimation methodology, gauging its performance against other well-established machine learning methods to ensure the quality of our redshift estimations. Our best results constrain photometric redshifts with the bias of − 0 . 0013 , a scatter of 0.0293, and an outlier fraction of 5.1%. These point estimates are accompanied by well-calibrated PDFs evaluated using diagnostic tools such as Probability Integral Transform and Odds distribution. We also address the problem of the accessibility of PDFs in terms of disk space storage and the time demand required to generate their corresponding parameters.We present a novel Autoencoder model that reduces the size of PDF parameter arrays to one-sixth of their original length, significantly decreasing the time required for PDF generation to one-eighth of the time needed when generating PDFs directly from the magnitudes.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

Bayesian learning of orthogonal embeddings for multi-fidelity Gaussian Processes

Uncertainty propagation in complex engineering systems often poses significant computational challenges related to modeling and quantifying probability distributions of model outputs, as those emerge as the result of various sources of uncertainty that are inherent in the system under investigation. Gaussian Processes regression (GPs) is a robust meta-modeling technique that allows for fast model prediction and exploration of response surfaces. Multi-fidelity variations of GPs further leverage information from cheap and low fidelity model simulations in order to improve their predictive performance on the high fidelity model. In order to cope with the high volume of data required to train GPs in high dimensional design spaces, a common practice is to introduce latent design variables that are typically projections of the original input space to a lower dimensional subspace, and therefore substitute the problem of learning the initial high dimensional mapping, with that of training a GP on a low dimensional space. Here in this paper, we present a Bayesian approach to identify optimal transformations that map the input points to low dimensional latent variables. The \projection" mapping consists of an orthonormal matrix that is considered a priori unknown and needs to be inferred jointly with the GP parameters, conditioned on the available training data. The proposed Bayesian inference scheme relies on a two-step iterative algorithm that samples from the marginal posteriors of the GP parameters and the projection matrix respectively, both using Markov Chain Monte Carlo (MCMC) sampling. In order to take into account the orthogonality constraints imposed on the orthonormal projection matrix, a Geodesic Monte Carlo sampling algorithm is employed, that is suitable for exploiting probability measures on manifolds. We extend the proposed framework to multi-fidelity models using GPs including the scenarios of training multiple outputs together. We validate our framework on three synthetic problems with a known lower-dimensional subspace. The benefits of our proposed framework, are illustrated on the computationally challenging aerodynamic optimization of a last-stage blade for an industrial gas turbine, where we study the effect of an 85-dimensional shape parameterization of a three-dimensional airfoil on two output quantities of interest, specifically on the aerodynamic efficiency and the degree of reaction

42 ENGINEERING↗