Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “complex fluids”

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 775 records · Page 43
All sourcesNASA NTRSDOE OSTI
NASA NTRS2018

Validation of Heat Transfer Correlations in Line Chill-Down Tests of Cryogenic Fluid in SINDA/FLUINT

Line chill-down is an important process in cryogenic tank propellant management, storage, and usage Complex flow dynamics during these processes: boiling heat transfer (film, transition, and nucleate) Understanding boiling phenomena can lead to efficient line chill-down systems that use less propellant, propellant stored, reducing cost for space missions Line Chill-down heat transfer was modelled using SINDA/FLUINT version 5.8 (SF) Multiple chill-down tests were modelled using: heat transfer correlations readily available in SF using HTN/HTC TIES heat transfer empiricisms developed by the University of Florida (UF) based on a series of liquid nitrogen chill-down tests using SF HTU TIES Chill-down tests modelled: liquid nitrogen tests conducted by the University of Florida horizontal flow, upward flow, and downward flow (Reynolds Numbers ranging 850-231,000)liquid hydrogen tests conducted by NASA Glenn Research Center vertical upward flow (Reynolds Number range of 18,400 - 433,000)The flow rate was measured far downstream of the test section, near the system exit. Where to set the flow rate? SF was highly sensitive, and sometime unstable, setting the test flow rate downstream (the outlet) of the test section model and setting the test pressure upstream (the inlet) of the test section model higher flow rate oscillations at the entrance of the model's test section SF was more stable setting the test flow rate upstream (than the downstream flow rate set case)test pressure was used as an inlet (SF plenum) to set the thermodynamic state (temperature and quality) coming into the system setting the appropriate downstream pressure was the unknown. The pressure drops predicted by SF for the downstream set flow rate boundary condition were much smaller than test section measured pressure drops. The multiphase pressure drop correlations used internally in SF may need to be adjusted. Models with an upstream flow rate set assumed a pressure drop that was small

cryogenic modeling↗
NASA NTRS2005

CFD Multiphysics Tool

The recent bold initiatives to expand the human presence in space require innovative approaches to the design of propulsion systems whose underlying technology is not yet mature. The space propulsion community has identified a number of candidate concepts. A short list includes solar sails, high-energy-density chemical propellants, electric and electromagnetic accelerators, solar-thermal and nuclear-thermal expanders. For each of these, the underlying physics are relatively well understood. One could easily cite authoritative texts, addressing both the governing equations, and practical solution methods for, e.g. electromagnetic fields, heat transfer, radiation, thermophysics, structural dynamics, particulate kinematics, nuclear energy, power conversion, and fluid dynamics. One could also easily cite scholarly works in which complete equation sets for any one of these physical processes have been accurately solved relative to complex engineered systems. The Advanced Concepts and Analysis Office (ACAO), Space Transportation Directorate, NASA Marshall Space Flight Center, has recently released the first alpha version of a set of computer utilities for performing the applicable physical analyses relative to candidate deep-space propulsion systems such as those listed above. PARSEC, Preliminary Analysis of Revolutionary in-Space Engineering Concepts, enables rapid iterative calculations using several physics tools developed in-house. A complete cycle of the entire tool set takes about twenty minutes. PARSEC is a level-zero/level-one design tool. For PARSEC s proof-of-concept, and preliminary design decision-making, assumptions that significantly simplify the governing equation sets are necessary. To proceed to level-two, one wishes to retain modeling of the underlying physics as close as practical to known applicable first principles. This report describes results of collaboration between ACAO, and Embry-Riddle Aeronautical University (ERAU), to begin building a set of level-two design tools for PARSEC. The "CFD Multiphysics Tool" will be the propulsive element of the tool set. The name acknowledges that space propulsion performance assessment is primarily a fluid mechanics problem. At the core of the CFD Multiphysics Tool is an open-source CFD code, HYP, under development at ERAU. ERAU is renowned for its undergraduate degree program in Aerospace Engineering the largest in the nation. The strength of the program is its applications-oriented curriculum, which culminates in one of three two-course Engineering Design sequences: Aerospace Propulsion, Spacecraft, or Aircraft. This same philosophy applies to the HYP Project, albeit with fluid physics modeling commensurate with graduate research. HYP s purpose, like the Multiphysics Tool s, is to enable calculations of real (three-dimensional; geometrically complex; intended for hardware development) applications of high speed and propulsive fluid flows.

Perrell, Eric R.↗
NASA NTRS1995

Advanced Methodology for Simulation of Complex Flows Using Structured Grid Systems

Detailed simulations of viscous flows in complicated geometries pose a significant challenge to current capabilities of Computational Fluid Dynamics (CFD). To enable routine application of CFD to this class of problems, advanced methodologies are required that employ (a) automated grid generation, (b) adaptivity, (c) accurate discretizations and efficient solvers, and (d) advanced software techniques. Each of these ingredients contributes to increased accuracy, efficiency (in terms of human effort and computer time), and/or reliability of CFD software. In the long run, methodologies employing structured grid systems will remain a viable choice for routine simulation of flows in complex geometries only if genuinely automatic grid generation techniques for structured grids can be developed and if adaptivity is employed more routinely. More research in both these areas is urgently needed.

Steinthorsson, Erlendur↗
DOE OSTI2022

Validation and parameterization of a novel physics-constrained neural dynamics model applied to turbulent fluid flow

We report, in fluid physics, data-driven models to enhance or accelerate time to solution are becoming increasingly popular for many application domains, such as alternatives to turbulence closures, system surrogates, or for new physics discovery. In the context of reduced order models of high-dimensional time-dependent fluid systems, machine learning methods grant the benefit of automated learning from data, but the burden of a model lies on its reduced-order representation of both the fluid state and physical dynamics. In this work, we build a physics-constrained, data-driven reduced order model for Navier–Stokes equations to approximate spatiotemporal fluid dynamics in the canonical case of isotropic turbulence in a triply periodic box. The model design choices mimic numerical and physical constraints by, for example, implicitly enforcing the incompressibility constraint and utilizing continuous neural ordinary differential equations for tracking the evolution of the governing differential equation. We demonstrate this technique on a three-dimensional, moderate Reynolds number turbulent fluid flow. In assessing the statistical quality and characteristics of the machine-learned model through rigorous diagnostic tests, we find that our model is capable of reconstructing the dynamics of the flow over large integral timescales, favoring accuracy at the larger length scales. More significantly, comprehensive diagnostics suggest that physically interpretable model parameters, corresponding to the representations of the fluid state and dynamics, have attributable and quantifiable impact on the quality of the model predictions and computational complexity.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
NASA NTRS1982

Analysis of volatile metabolites in biological fluids as indicators of prodromal disease condition

The volatile profile cannot be defined as a single class of substances, rather it is a broad spectrum of materials of different polarities characterized by having a boiling-point in the low to medium range (up to approximately 300 C) and the fact that the compounds are suitable for gas chromatography without derivatization. The organic volatile profiles are very complex mixtures of metabolic byproducts, intermediates, and terminal products of enzymatic degradations composed mainly of alcohols, ketones, aldehydes, pyrazines, sulfides, isothiocyanates, pyrroles, and furans. The concentration of organic volatiles in biological fluids covers a wide range with many important components present at trace levels. The complexity of the organic volatile fraction requires the use of capillary columns for their separation.

Zlatkis, A.↗
DOE OSTI2020

Magneto-structural and induction heating properties of MFe2O4 (M?=?Co, Mn, Zn) MNPs for magnetic particle hyperthermia application

The thermal decomposition approach was used for the synthesis of MFe2O4 magnetic nanoparticles (MNPs) by substituting M as Co, Mn, and Zn. The obtained MNPs were characterized for magneto-structural properties using X-Ray diffraction patterns, FTIR, Raman and Mossbauer spectroscopy techniques which confirm the formation of phase pure cubic spinel ferrite with space group Fd3m and five Raman active modes. The size, morphology, and compositional analysis was performed using HRTEM and EDX where the size of MNPs was found to be less than 10 nm that attains superparamagnetism. The magnetic hyperthermia performance of obtained MNPs was evaluated by induction heating experiments at magnetic field range 13.3 to 26.7 kAm-1. The specific absorption rate (SAR) and intrinsic loss power (ILP) values were determined at different magnetic fields within the human tolerable range and correlated with magneto-structural properties to evaluate its potential for possible magnetic particle hyperthermia therapy.

Magnetic nanoparticles, complex-decomposition, mag↗
DOE OSTI2023

Fabric controls on fracture surface roughness of an architected rock material

Fluid flow through fractures is intimately linked to the fracture surfaces that define the void geometry through which fluids flow. Thus, an understanding of what controls fracture surface roughness is essential to the development of models for predicting fluid transport through fractured rock. The difficulty in predicting surface roughness arises from the complexity of rock which is inherently heterogeneous and nonuniform in composition, fabric, and structural components, even when samples are acquired from the same rock mass. Here, a benchmarked-simulation approach motivated from geo-architected 3D printed synthetic gypsum rocks is used to provide insight into the competing contributions from fabric and layering on fracture roughness formation. Simulation results from a discrete element model (Particle Flow Code, Itasca Consulting Group, Inc.) clearly indicate that the relative orientation between mineral layers and in-layer mineral fabric, and the variability in mineral bonding strengths determine whether anisotropic corrugated surfaces or isotropic surfaces are formed. Weak mineral layers oriented perpendicular to the applied load resulted in strong roughness anisotropy. Peak failure loads were found to vary up to 30% depending on the strength of the mineral fabric at the location of fracture initiation, which provides insight into the observed high variability in strength values of natural rock. The uniqueness of induced fracture roughness and peak failure load is intimately linked to layering, mineral fabric, and their distribution in the rock. These findings have important implications for any architected material fabricated through serial printing of layers with local compositional heterogeneity.

3D printed rock↗
DOE OSTI2026

Mesh-based multiphysics coupling acceleration for fusion neutronics through clustering for fusion blanket applications

Accurate modeling of particle transport within fusion blankets is essential for predicting performance metrics such as heat deposition and the tritium breeding ratio (TBR). However, high-fidelity coupling of thermal fluids from computational fluid dynamics (CFD) to neutronics simulations often incurs significant computational costs due to the complexity of surface intersection calculations in Monte Carlo codes. This paper presents an accelerated multiphysics coupling method for neutronics that utilizes hierarchical agglomerative clustering to map complex material property distributions to a neutronics model. Implemented within the fusion reactor design and assessment (FREDA) framework, the method leverages existing Python packages to automate the creation of clustered geometries for OpenMC. The approach is demonstrated on a sector model of an ARC-class tokamak with an immersion molten salt blanket, and an simple geometry with varying isotopic concentrations. Results show that the clustering method significantly reduces computational burden without compromising fidelity, providing a foundation for agile iteration of neutronics simulations involving multiple coupled material properties.

Bae, Jin Whan [ORNL] (ORCID:0000000326548907)↗
DOE OSTI2024

Application of Modified Meshgraphnets for Subsurface Prediction during CO2 Sequestration

In the face of the increasingly dire consequences of anthropogenic climate change, capturing and storing carbon dioxide is paramount. However, several impediments exist to the safe and effective subsurface storage of CO2, such as cost of transport, identification of suitable sites for subsurface storage, and assessment of long-term risk from storage in subsurface aquifers. Accurate subsurface modeling is necessary to ensure that CO2 storage is both safe and effective. Still, such modeling has traditionally required either substantial time and computational power (numerical simulation) or a substantial amount of pre-existing data for training (machine learning models). Additionally, these models lack flexibility in dealing with both changes in discretization of the input data and generalizability beyond the data on which they are trained. In order to address these issues, this research applies graph neural networks (GNNs) to predict subsurface saturation and pressure during CO₂ injection in a model of the Illinois Basin-Decatur Project (IBDP). GNNs provide a flexible, intuitive method for representing and manipulating complex unstructured data, which is often found in many practical domain problems such as fluid flow and subsurface characterization. These unstructured grids are easily represented in GNNs by representing spatially-localized features such as permeability, porosity, saturation, and pressure as nodes in a graph and relationships between these properties as edges connecting these nodes. This research applies a specific GNN model called MeshGraphNets (MGN) to model the change in CO2 saturation and pressure over a 50-month time period (36 months of injection, 14 months post-injection). The MGN model leverages a message passing process that allows the network to learn both the spatial and temporal dynamics of this system simultaneously. Additionally, training on a limited dataset (64 realizations, 20 time points each) resulted in a high degree of accuracy in saturation prediction both within the same timeframe as the training (20 months, 0.039 average RMSE) and when projecting out to the end of injection (36 months, 0.053 average RMSE). Temporal predictions such as those generated by MGNs and other similar models are prone to accumulated error over time; in order to address this, a multi-step rollout (MSR) training process was applied to calculate training loss. This method mimics the forward prediction during inference by “rolling out” multiple time points in a single training step using the previous prediction as input to the MGN model. By calculating the loss several time steps forward from the current prediction, the model is forced to find a more stable state over time. Application of MSR to the MGN model resulted in an average 15% reduction in inference error over time during forward prediction. This study showcases the immense potential of GNNs as a game-changing methodology for predicting pressure and saturation evolution in CCS projects, ultimately paving the way for more sustainable and effective carbon storage solutions. Presentation prepared for the 2024 AiChE Annual Meeting, October 27 to November 1 2024, San Diego, CA.

Holcomb, Paul↗
NASA NTRS1978

A three-dimensional model of corotating streams in the solar wind. I - Theoretical foundations

The paper is concerned with the development of the theoretical and mathematical background pertinent to the study of steady, corotating solar wind structure in all three spatial dimensions. The dynamical evolution of the plasma in interplanetary space (defined as the region beyond roughly 35 Rs where the flow is supersonic) is approximately described by the nonlinear, single-fluid, polytropic magnetohydrodynamic or hydrodynamic equations. Efficient numerical techniques are outlined for solving this complex system of coupled, hyperbolic partial differential equations. The present formulation is inviscid and nonmagnetic, but the methods used allow for the potential inclusion of both features with only modest modifications. A simple, highly idealized hydrodynamic model stream is examined to illustrate the fundamental processes involved in the three-dimensional dynamics of stream evolution. It is found that spatial variations in the rotational stream interaction mechanism produce small nonradial flows on a global scale that lead to the transport of mass, energy, and momentum away from regions of relative compression and into regions of relative rarefaction. Comparison with simpler models demonstrates the essential nonlinear, multidimensional nature of the interplanetary dynamics.

Pizzo, V.↗
NASA NTRS1993

Algorithm and code development for unsteady three-dimensional Navier-Stokes equations

In the last two decades, there have been extensive developments in computational aerodynamics, which constitutes a major part of the general area of computational fluid dynamics. Such developments are essential to advance the understanding of the physics of complex flows, to complement expensive wind-tunnel tests, and to reduce the overall design cost of an aircraft, particularly in the area of aeroelasticity. Aeroelasticity plays an important role in the design and development of aircraft, particularly modern aircraft, which tend to be more flexible. Several phenomena that can be dangerous and limit the performance of an aircraft occur because of the interaction of the flow with flexible components. For example, an aircraft with highly swept wings may experience vortex-induced aeroelastic oscillations. Also, undesirable aeroelastic phenomena due to the presence and movement of shock waves occur in the transonic range. Aeroelastically critical phenomena, such as a low transonic flutter speed, have been known to occur through limited wind-tunnel tests and flight tests. Aeroelastic tests require extensive cost and risk. An aeroelastic wind-tunnel experiment is an order of magnitude more expensive than a parallel experiment involving only aerodynamics. By complementing the wind-tunnel experiments with numerical simulations the overall cost of the development of aircraft can be considerably reduced. In order to accurately compute aeroelastic phenomenon it is necessary to solve the unsteady Euler/Navier-Stokes equations simultaneously with the structural equations of motion. These equations accurately describe the flow phenomena for aeroelastic applications. At Ames a code, ENSAERO, is being developed for computing the unsteady aerodynamics and aeroelasticity of aircraft and it solves the Euler/Navier-Stokes equations. The purpose of this contract is to continue the algorithm enhancements of ENSAERO and to apply the code to complicated geometries. During the last year, the geometric capability of the code was extended to simulate transonic flows, a wing with oscillating control surface. Single-grid and zonal approaches were tested. For the zonal approach, a new interpolation technique was introduced. The key development of the algorithm was an interface treatment between moving zones for a control surface using the virtual-zone concept. The work performed during the period, 1 Apr. 1992 through 31 Mar. 1993 is summarized. Additional details on the various aspects of the study are given in the Appendices.

Obayashi, Shigeru↗
NASA NTRS1995

Numerical simulation of the flow about the F-18 HARV at high angle of attack

This research has been aimed at validating numerical methods for computing the flow about the complete F-18 HARV at alpha = 30 deg and alpha = 45 deg. At 30 deg angle of attack, the flow about the F-18 is dominated by the formation, and subsequent breakdown, of strong vortices over the wing leading-edge extensions (LEX). As the angle of attack is increased to alpha = 45 deg, the fuselage forebody of the F-18 contains significant laminar and transitional regions which are not present at alpha = 30 deg. Further, the flow over the LEX at alpha = 45 deg is dominated by an unsteady shedding in time, rather than strong coherent vortices. This complex physics, combined with the complex geometry of a full-aircraft configuration, provides a challenge for current computational fluid dynamics (CFD) techniques. The following sections present the numerical method and grid generation scheme that was used, a review of prior research done to numerically model the F-18 HARV, and a discussion of the current research. The current research is broken into three main topics; the effect of engine-inlet mass-flow rate on the F-18 vortex breakdown position, the results using a refined F-18 computational model to compute the flow at alpha = 30 deg and alpha = 45 deg, and research done using the simplified geometry of an ogive-cylinder configuration to investigate the physics of unsteady shear-layer shedding. The last section briefly summarizes the discussion.

Murman, Scott M.↗
NASA NTRS1995

The Physics of Turbulence in the Boundary Layer

The geometry of the velocity field in a numerically simulated incompressible turbulent boundary layer over a flat plate at Re theta=670 has been studied using the invariants of the velocity gradient tensor. These invariants are computed at every grid point in the flow and used to form the discriminant. Of primary interest are those regions in the flow where the discriminant is positive; regions where, according to the characteristic equation, the eigenvalues of the velocity gradient tensor are complex. An observer moving with a frame of reference which is attached to a fluid particle lying within such a region would see a local flow pattern of the type stable-focus-stretching or unstable-focus-compressing. When the flow is visualized this way, continuous, connected, large-scale structures are revealed that extend from the point just below the buffer layer out to the beginning of the wake region. These structures are aligned with the mean shear close to the wall and arch in the cross-stream direction away from the wall. In some cases the structures observed are very similar to to the hairpin eddy vision of boundary layer structure proposed by Theodorsen. That the structure of the flow is revealed more effectively by the discriminant rather than by the vorticity is important and adds support to recent observations of the discriminant in a channel flow simulation. Of particular importance is the fact that the procedure does not require the use of an arbitrary threshold in the discriminant. Further analysis using computer flow visualization shows a high degree of spatial correlation between regions of positive discriminant, extreme negative pressure fluctuations and large instantaneous values of Reynolds shear stress.

Kline, Stephen↗
NASA NTRS2003

Direct Manipulation in Virtual Reality

Virtual Reality interfaces offer several advantages for scientific visualization such as the ability to perceive three-dimensional data structures in a natural way. The focus of this chapter is direct manipulation, the ability for a user in virtual reality to control objects in the virtual environment in a direct and natural way, much as objects are manipulated in the real world. Direct manipulation provides many advantages for the exploration of complex, multi-dimensional data sets, by allowing the investigator the ability to intuitively explore the data environment. Because direct manipulation is essentially a control interface, it is better suited for the exploration and analysis of a data set than for the publishing or communication of features found in that data set. Thus direct manipulation is most relevant to the analysis of complex data that fills a volume of three-dimensional space, such as a fluid flow data set. Direct manipulation allows the intuitive exploration of that data, which facilitates the discovery of data features that would be difficult to find using more conventional visualization methods. Using a direct manipulation interface in virtual reality, an investigator can, for example, move a data probe about in space, watching the results and getting a sense of how the data varies within its spatial volume.

Bryson, Steve↗
NASA NTRS2004

Combustion Module-2 Achieved Scientific Success on Shuttle Mission STS-107

The familiar teardrop shape of a candle is caused by hot, spent air rising and cool fresh air flowing behind it. This type of airflow obscures many of the fundamental processes of combustion and is an impediment to our understanding and modeling of key combustion controls used for manufacturing, transportation, fire safety, and pollution. Conducting experiments in the microgravity environment onboard the space shuttles eliminates these impediments. NASA Glenn Research Center's Combustion Module-2 (CM-2) and its three experiments successfully flew on STS-107/Columbia in the SPACEHAB module and provided the answers for many research questions. However, this research also opened up new questions. The CM-2 facility was the largest and most complex pressurized system ever flown by NASA and was a precursor to the Glenn Fluids and Combustion Facility planned to fly on the International Space Station. CM-2 operated three combustion experiments: Laminar Soot Processes (LSP), Structure of Flame Balls at Low Lewis-Number (SOFBALL), and Water Mist Fire Suppression Experiment (Mist). Although Columbia's mission ended in tragedy with the loss of her crew and much data, most of the CM-2 results were sent to the ground team during the mission.

Over, Ann P.↗
NASA NTRS2008

Fuel Cell Thermal Management Through Conductive Cooling Plates

An analysis was performed to evaluate the concept of utilizing conductive cooling plates to remove heat from a fuel cell stack, as opposed to a conventional internal cooling loop. The potential advantages of this type of cooling system are reduced stack complexity and weight and increased reliability through the reduction of the number of internal fluid seals. The conductive cooling plates would extract heat from the stack transferring it to an external coolant loop. The analysis was performed to determine the required thickness of these plates. The analysis was based on an energy balance between the thermal energy produced within the stack and the heat removal from the cooling plates. To accomplish the energy balance, the heat flow into and along the plates to the cooling fluid was modeled. Results were generated for various numbers of cells being cooled by a single cooling plate. The results provided cooling plate thickness, mass, and operating temperature of the plates. It was determined that utilizing high-conductivity pyrolitic graphite cooling plates can provide a specific cooling capacity (W/kg) equivalent to or potentially greater than a conventional internal cooling loop system.

Colozza, Anthony J.↗
NASA NTRS2008

NASA Tech Briefs, April 2008

Topics covered include: Gas Sensors Based on Coated and Doped Carbon Nanotubes; Tactile Robotic Topographical Mapping Without Force or Contact Sensors; Thin-Film Magnetic-Field-Response Fluid-Level Sensor for Non-Viscous Fluids; Progress in Development of Improved Ion-Channel Biosensors; Simulating Operation of a Complex Sensor Network; Using Transponders on the Moon to Increase Accuracy of GPS; Controller for Driving a Piezoelectric Actuator at Resonance; Coaxial Electric Heaters; Dual-Input AND Gate From Single-Channel Thin-Film FET; High-Density, High-Bandwidth, Multilevel Holographic Memory; Fabrication of Gate-Electrode Integrated Carbon-Nanotube Bundle Field Emitters; Hydroxide-Assisted Bonding of Ultra-Low-Expansion Glass; Photochemically Synthesized Polyimides; Optimized Carbonate and Ester-Based Li-Ion Electrolytes; Compact 6-DOF Stage for Optical Adjustments; Ultrasonic/Sonic Impacting Penetrators; Miniature, Lightweight, One-Time-Opening Valve; Supplier Management System; Improved CLARAty Functional-Layer/Decision-Layer Interface; JAVA Stereo Display Toolkit; Remote-Sensing Time Series Analysis, a Vegetation Monitoring Tool; PyPele Rewritten To Use MPI; Data Assimilation Cycling for Weather Analysis; Hydrocyclone/Filter for Concentrating Biomarkers from Soil; Activating STAT3 Alpha for Promoting Healing of Neurons; and Probing a Spray Using Frequency-Analyzed Light Scattering.

Source record↗
NASA NTRS2007

Automated Knowledge Discovery From Simulators

A computational method, SimLearn, has been devised to facilitate efficient knowledge discovery from simulators. Simulators are complex computer programs used in science and engineering to model diverse phenomena such as fluid flow, gravitational interactions, coupled mechanical systems, and nuclear, chemical, and biological processes. SimLearn uses active-learning techniques to efficiently address the "landscape characterization problem." In particular, SimLearn tries to determine which regions in "input space" lead to a given output from the simulator, where "input space" refers to an abstraction of all the variables going into the simulator, e.g., initial conditions, parameters, and interaction equations. Landscape characterization can be viewed as an attempt to invert the forward mapping of the simulator and recover the inputs that produce a particular output. Given that a single simulation run can take days or weeks to complete even on a large computing cluster, SimLearn attempts to reduce costs by reducing the number of simulations needed to effect discoveries. Unlike conventional data-mining methods that are applied to static predefined datasets, SimLearn involves an iterative process in which a most informative dataset is constructed dynamically by using the simulator as an oracle. On each iteration, the algorithm models the knowledge it has gained through previous simulation trials and then chooses which simulation trials to run next. Running these trials through the simulator produces new data in the form of input-output pairs. The overall process is embodied in an algorithm that combines support vector machines (SVMs) with active learning. SVMs use learning from examples (the examples are the input-output pairs generated by running the simulator) and a principle called maximum margin to derive predictors that generalize well to new inputs. In SimLearn, the SVM plays the role of modeling the knowledge that has been gained through previous simulation trials. Active learning is used to determine which new input points would be most informative if their output were known. The selected input points are run through the simulator to generate new information that can be used to refine the SVM. The process is then repeated. SimLearn carefully balances exploration (semi-randomly searching around the input space) versus exploitation (using the current state of knowledge to conduct a tightly focused search). During each iteration, SimLearn uses not one, but an ensemble of SVMs. Each SVM in the ensemble is characterized by different hyper-parameters that control various aspects of the learned predictor - for example, whether the predictor is constrained to be very smooth (nearby points in input space lead to similar output predictions) or whether the predictor is allowed to be "bumpy." The various SVMs will have different preferences about which input points they would like to run through the simulator next. SimLearn includes a formal mechanism for balancing the ensemble SVM preferences so that a single choice can be made for the next set of trials.

Burl, Michael↗