Search NASA⌕ Search

SEARCH · Search NASA

Results for “fitting”

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 1,729 records · Page 96

Network Sensitivity Solutions for Regional Moment Tensor Inversions

Well-resolved moment tensor solutions reveal information about the sources of seismic waves. Here we introduce a new way of assessing confidence in the regional full moment tensor inversion via the introduction of the network sensitivity solution (NSS). The NSS takes into account the unique station distribution, frequency band, and signal-to-noise ratio of a given event scenario. The NSS compares both a hypothetical pure source (for example an explosion or an earthquake) and the actual data with several thousand sets of synthetic data from a uniform distribution of all possible sources. The comparison with a hypothetical pure source provides the theoretically best-constrained source-type region for a given set of stations, and with it one can determine whether further analysis with the data is warranted. The NSS that employs the actual data gives a direct comparison of all other source-types with the best-fit source. In this way, one can choose a threshold level of fit where the solution is comfortably constrained. The method is tested for the well-recorded nuclear test, JUNCTION, at the Nevada Test Site. Sources that fit comparably well to a hypothetical pure explosion recorded with no noise at the JUNCTION data stations have a large volumetric component and are not described well by a double-couple (DC) source. The NSS using the real data from JUNCTION is even more tightly constrained to an explosion since the data contains some energy that precludes fitting with any type of deviatoric source. We also calculate the NSS for the October 2006 North Korea test and a nearby earthquake, where the station coverage is poor and the event magnitude is small. The earthquake solution is very well fit by a DC source, and the best-fit solution to the nuclear test (M{sub W}4.1) is dominantly explosion.

58 GEOSCIENCES↗

Feed-in Tariff Policy: Design, Implementation, and RPS Policy Interactions

Feed-in tariff (FIT) policies are implemented in more than 40 countries around the world and are cited as the primary reason for the success of the German and Spanish renewable energy markets. As a result of that success, FIT policy proposals are starting to gain traction in several U.S. states and municipalities. Experience from Europe is also beginning to demonstrate that properly designed FITs may be more cost-effective than renewable portfolio standards (RPS), which make use of competitive solicitations. This article explores the design and operation of feed-in tariff policies, including a FIT policy definition, payment-structure options, and payment differentiation. The article also touches on the potential interactions between FIT policies and RPS policies at the state level.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Impact of Schedule Duration on Head and Neck Radiotherapy: Accelerated Tumor Repopulation Versus Compensatory Mucosal Proliferation

Purpose: To determine how modelled maximum tumor control rates, achievable without exceeding mucositis tolerance (tcp{sub max-early}) vary with schedule duration for head and neck squamous cell carcinoma (HNSCC). Methods and materials: Using maximum-likelihood techniques, we have fitted a range of tcp models to two HNSCC datasets (Withers' and British Institute of Radiology [BIR]), characterizing the dependence of tcp on duration and equivalent dose in 2 Gy fractions (EQD{sub 2}). Models likely to best describe future data have been selected using the Akaike information criterion (AIC) and its quasi-AIC extension to overdispersed data. Setting EQD{sub 2}s in the selected tcp models to levels just tolerable for mucositis, we have plotted tcp{sub max-early} against schedule duration. Results: While BIR dataset tcp fits describe dose levels isoeffective for tumor control as rising significantly with schedule protraction, indicative of accelerated tumor repopulation, repopulation terms in fits to Withers' dataset do not reach significance after accounting for overdispersion of the data. The tcp{sub max-early} curves calculated from tcp fits to the overall Withers' and BIR datasets rise by 8% and 0-4%, respectively, between 20 and 50 days duration; likewise, tcp{sub max-early} curves calculated for stage-specific cohorts also generally rise slowly with increasing duration. However none of the increases in tcp{sub max-early} calculated from the overall or stage-specific fits reach significance. Conclusions: Local control rates modeled for treatments which lie just within mucosal tolerance rise slowly but insignificantly with increasing schedule length. This finding suggests that whereas useful gains may be made by accelerating unnecessarily slow schedules until they approach early reaction tolerance, little is achieved by shortening schedules further while reducing doses to remain within mucosal tolerance, an approach that may slightly worsen outcomes.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Assessment of Lung Tumour Motion and Volume Size Dependencies Using Various Evaluation Measures

The purpose of the study was to assess internal target volume changes through the breathing cycle and associated tumour motion for lung patients and to establish possible correlations between different parameters. Respiration-induced volume changes with breathing cycle and the associated tumour motion were analyzed for 11 patients. Selected phases were the maximum and average intensity projections and the 10 phases of equal duration and separation obtained through the respiratory cycle. Tumour centre of mass (COM) motion planes were generated using least square fitting, and their angles and orientations were then compared between the cases studied. Trajectories that are composed by the points of COM location in different phases were identified, and their interrelation was assessed through different similarity measures. The results were used to determine if there is any correlation between parameters chosen and if the margins conventionally used for the planning target volume creation successfully encompassed lung tumour motion and volume change. The results show that the extent of tumour motion was related to its volume and location. The tumour displacement was predominantly left and inferior. Planar fitting to COM motion data through respiratory phases demonstrated some correlation in best fit motion plane positions between different data sets. In the plane fit comparison, for each patient, the lower root mean square error values showed that a good planar fit can be achieved for the COM motion path. The evaluation of the inhale and exhale trajectories may allow, for certain tumour locations and size, contouring on only inhalation or exhalation phases, knowing that tumour motion will be adequately covered on the other phases. Taking all the data into account and knowing the tumour size and location, a good estimate can be made of the motion plane position in the three-dimensional space and the required dosimetric margins.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Five ab initio potential energy and dipole moment surfaces for hydrated NaCl and NaF. I. Two-body interactions

We report full-dimensional, ab initio-based potentials and dipole moment surfaces for NaCl, NaF, Na{sup +}H{sub 2}O, F{sup −}H{sub 2}O, and Cl{sup −}H{sub 2}O. The NaCl and NaF potentials are diabatic ones that dissociate to ions. These are obtained using spline fits to CCSD(T)/aug-cc-pV5Z energies. In addition, non-linear least square fits using the Born-Mayer-Huggins potential are presented, providing accurate parameters based strictly on the current ab initio energies. The long-range behavior of the NaCl and NaF potentials is shown to go, as expected, accurately to the point-charge Coulomb interaction. The three ion-H{sub 2}O potentials are permutationally invariant fits to roughly 20 000 coupled cluster CCSD(T) energies (awCVTZ basis for Na{sup +} and aVTZ basis for Cl{sup −} and F{sup −}), over a large range of distances and H{sub 2}O intramolecular configurations. These potentials are switched accurately in the long range to the analytical ion-dipole interactions, to improve computational efficiency. Dipole moment surfaces are fits to MP2 data; for the ion-ion cases, these are well described in the intermediate- and long-range by the simple point-charge expression. The performance of these new fits is examined by direct comparison to additional ab initio energies and dipole moments along various cuts. Equilibrium structures, harmonic frequencies, and electronic dissociation energies are also reported and compared to direct ab initio results. These indicate the high fidelity of the new PESs.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

De-biased populations of Kuiper belt objects from the deep ecliptic survey

The Deep Ecliptic Survey (DES) was a survey project that discovered hundreds of Kuiper Belt objects from 1998 to 2005. Extensive follow-up observations of these bodies has yielded 304 objects with well-determined orbits and dynamical classifications into one of several categories: Classical, Scattered, Centaur, or 16 mean-motion resonances with Neptune. The DES search fields are well documented, enabling us to calculate the probability on each frame of detecting an object with its particular orbital parameters and absolute magnitude at a randomized point in its orbit. The detection probabilities range from a maximum of 0.32 for the 3:2 resonant object 2002 GF {sub 32} to a minimum of 1.5 × 10{sup –7} for the faint Scattered object 2001 FU {sub 185}. By grouping individual objects together by dynamical classes, we can estimate the distributions of four parameters that define each class: semimajor axis, eccentricity, inclination, and object size. The orbital element distributions (a, e, and i) were fit to the largest three classes (Classical, 3:2, and Scattered) using a maximum likelihood fit. Using the absolute magnitude (H magnitude) as a proxy for the object size, we fit a power law to the number of objects versus H magnitude for eight classes with at least five detected members (246 objects). The Classical objects are best fit with a power-law slope of α = 1.02 ± 0.01 (observed from 5 ≤ H ≤ 7.2). Six other dynamical classes (Scattered plus five resonances) have consistent magnitude distribution slopes with the Classicals, provided that the absolute number of objects is scaled. Scattered objects are somewhat more numerous than Classical objects, while there are only a quarter as many 3:2 objects as Classicals. The exception to the power law relation is the Centaurs, which are non-resonant objects with perihelia closer than Neptune and therefore brighter and detectable at smaller sizes. Centaurs were observed from 7.5 < H < 11, and that population is best fit by a power law with α = 0.42 ± 0.02. This is consistent with a knee in the H-distribution around H = 7.2 as reported elsewhere. Based on the Classical-derived magnitude distribution, the total number of objects (H ≤ 7) in each class is: Classical (2100 ± 300 objects), Scattered (2800 ± 400), 3:2 (570 ± 80), 2:1 (400 ± 50), 5:2 (270 ± 40), 7:4 (69 ± 9), 5:3 (60 ± 8). The independent estimate for the number of Centaurs in the same H range is 13 ± 5. If instead all objects are divided by inclination into 'Hot' and 'Cold' populations, following Fraser et al., we find that α{sub Hot} = 0.90 ± 0.02, while α{sub Cold} = 1.32 ± 0.02, in good agreement with that work.

79 ASTRONOMY AND ASTROPHYSICS↗

Nonproportionality of Scintillator Detectors: Theory and Experiment

On the basis of nonproportionality data obtained for several scintillators, we have developed a model to describe the carrier dynamics to fit the light yield versus electron energy. The theory of Onsager was adapted to explain how the carriers form excitons or sequentially arrive at the activators to promote the ion to an excited state, and the theory of Birks was employed to allow for exciton-exciton annihilation. We then developed a second model to deduce the degradation in resolution that results from nonproportionality by evoking Landau fluctuations, which are essentially variations in the deposited energy density that occur as the high energy electron travels along its trajectory. Generally there is agreement with the data, in terms of fitting the nonproportionality curves and reproducing the literature values of nonproportionality's contribution to the scintillator resolution.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

TU-CD-207-11: Patient-Driven Automatic Exposure Control for Dedicated Breast CT

Purpose: To implement automatic exposure control (AEC) in dedicated breast CT (bCT) on a patient-specific basis using only the pre-scan scout views. Methods: Using a large cohort (N=153) of bCT data sets, the breast effective diameter (D) and width in orthogonal planes (Wa,Wb) were calculated from the reconstructed bCT image and pre-scan scout views, respectively. D, Wa, and Wb were measured at the breast center-of-mass (COM), making use of the known geometry of our bCT system. These data were then fit to a second-order polynomial “D=F(Wa,Wb)” in a least squares sense in order to provide a functional form for determining the breast diameter. The coefficient of determination (R{sup 2}) and mean percent error between the measured breast diameter and fit breast diameter were used to evaluate the overall robustness of the polynomial fit. Lastly, previously-reported bCT technique factors derived from Monte Carlo simulations were used to determine the tube current required for each breast diameter in order to match two-view mammographic dose levels. Results: F(Wa,Wb) provided fitted breast diameters in agreement with the measured breast diameters resulting in R{sup 2} values ranging from 0.908 to 0.929 and mean percent errors ranging from 3.2% to 3.7%. For all 153 bCT data sets used in this study, the fitted breast diameters ranged from 7.9 cm to 15.7 cm corresponding to tube current values ranging from 0.6 mA to 4.9 mA in order to deliver the same dose as two-view mammography in a 50% glandular breast with a 80 kV x-ray beam and 16.6 second scan time. Conclusion: The present work provides a robust framework for AEC in dedicated bCT using only the width measurements derived from the two orthogonal pre-scan scout views. Future work will investigate how these automatically chosen exposure levels affect the quality of the reconstructed image.

60 APPLIED LIFE SCIENCES↗

SU-G-TeP2-14: Universal Function Form for Photon Open Field In-Water Output Ratio

Purpose: In-Water-Output-Ratio (IWOR) plays a significant role in linac-based radiotherapy treatment planning, linking MUs to delivered radiation dose. For an open rectangular field, IWOR depends on both its width and length, and changes rapidly when one of them becomes small. In this study, a universal functional form is proposed to fit the open field IWOR tables in Varian TrueBeam representative datasets for all photon energies. Methods: A novel Generalized Mean formula is first used to estimate the Equivalent Square (ES) for a rectangular field. The formula’s weighting factor and power index are determined by collapsing all data points as much as possible onto a single curve in IWOR vs. ES plot. The result is then fitted with a novel universal function IWOR=1+b*Log(ES/10cm)/(ES/10cm)^c via a least-square procedure to determine the optimal values for parameters b and c. The maximum relative residual error in IWOR over the entire two-dimensional measurement table with field sizes between 3cm and 40cm is used to evaluate the quality of fit for the function. Results: The two-step fitting strategy works very well in determining the optimal parameter values for open field IWOR of each photon energies in the Varian data-set. Relative residual error ≤0.71% is achieved for all photon energies (including Flattening-Filter-Free modes) with field sizes between 3cm and 40cm. The optimal parameter values change smoothly with regular photon beam quality. Conclusion: The universal functional form fits the Varian TrueBeam open field IWOR measurement tables accurately with small relative residual errors for all photon energies. Therefore, it can be an excellent choice to represent IWOR in absolute dose and MU calculations. The functional form can also be used as a QA/commissioning tool to verify the measured data quality and consistency by checking the IWOR data behavior against the function for new photon energies with arbitrary beam quality.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

SU-G-IeP3-11: On the Utility of Pixel Variance to Characterize Noise for Image Receptors of Digital Radiography Systems

Purpose: To characterize noise for image receptors of digital radiography systems based on pixel variance. Methods: Nine calibrated digital image receptors associated with nine new portable digital radiography systems (Carestream Health, Inc., Rochester, NY) were used in this study. For each image receptor, thirteen images were acquired with RQA5 beam conditions for input detector air kerma ranging from 0 to 110 µGy, and linearized ‘For Processing’ images were extracted. Mean pixel value (MPV), standard deviation (SD) and relative noise (SD/MPV) were obtained from each image using ROI sizes varying from 2.5×2.5 to 20×20 mm{sup 2}. Variance (SD{sup 2}) was plotted as a function of input detector air kerma and the coefficients of the quadratic fit were used to derive structured, quantum and electronic noise coefficients. Relative noise was also fitted as a function of input detector air kerma to identify noise sources. The fitting functions used least-squares approach. Results: The coefficient of variation values obtained using different ROI sizes was less than 1% for all the images. The structured, quantum and electronic coefficients obtained from the quadratic fit of variance (r>0.97) were 0.43±0.10, 3.95±0.27 and 2.89±0.74 (mean ± standard deviation), respectively, indicating that overall the quantum noise was the dominant noise source. However, for one system electronic noise coefficient (3.91) was greater than quantum noise coefficient (3.56) indicating electronic noise to be dominant. Using relative noise values, the power parameter of the fitting equation (|r|>0.93) showed a mean and standard deviation of 0.46±0.02. A 0.50 value for this power parameter indicates quantum noise to be the dominant noise source whereas values around 0.50 indicate presence of other noise sources. Conclusion: Characterizing noise from pixel variance assists in identifying contributions from various noise sources that, eventually, may affect image quality. This approach may be integrated during periodic quality assessments of digital image receptors.

60 APPLIED LIFE SCIENCES↗

COVID-19 global pandemic planning: Performance and electret charge of N95 respirators after recommended decontamination methods

Shortages of N95 respirators for use by medical personnel have driven consideration of novel conservation strategies, including decontamination for reuse and extended use. Decontamination methods listed as promising by the Centers for Disease Control and Prevention (CDC) (vaporous hydrogen peroxide (VHP), wet heat, ultraviolet irradiation (UVI)) and several methods considered for low resource environments (bleach, isopropyl alcohol and detergent/soap) were studied for two commonly used surgical N95 respirators (3M™ 1860 and 1870+ Aura™). Although N95 filtration performance depends on the electrostatically charged electret filtration layer, the impact of decontamination on this layer is largely unexplored. As such, respirator performance following decontamination was assessed based on the fit, filtration efficiency, and pressure drop, along with the relationship between (1) surface charge of the electret layer, and (2) elastic properties of the straps. Decontamination with VHP, wet heat, UVI, and bleach did not degrade fit and filtration performance or electret charge. Isopropyl alcohol and soap significantly degraded fit, filtration performance, and electret charge. Pressure drop across the respirators was unchanged. Modest degradation of N95 strap elasticity was observed in mechanical fatigue testing, a model for repeated donnings and doffings. CDC recommended decontamination methods including VHP, wet heat, and UV light did not degrade N95 respirator fit or filtration performance in these tests. Extended use of N95 respirators may degrade strap elasticity, but a loss of face seal integrity should be apparent during user seal checks. NIOSH recommends performing user seal checks after every donning to detect loss of appropriate fit. Decontamination methods which degrade electret charge such as alcohols or detergents should not be used on N95 respirators. The loss of N95 performance due to electret degradation would not be apparent to a respirator user or evident during a negative pressure user seal check.

60 APPLIED LIFE SCIENCES↗

Simulations of Turbulent Flows with Strong Shocks and Density Variations

In this report, we present the research efforts made by our group at UCLA in the SciDAC project Simulations of turbulent flows with strong shocks and density variations. We use shock-fitting methodologies as an alternative to shock-capturing schemes for the problems where a well defined shock is present. In past five years, we have focused on development of high-order shock-fitting Navier-Stokes solvers for perfect gas flow and thermochemical non-equilibrium flow and simulation of shock-turbulence interaction physics for very strong shocks. Such simulation has not been possible before because the limitation of conventional shock capturing methods. The limitation of shock Mach number is removed by using our high-order shock-fitting scheme. With the help of DOE and TeraGrid/XSEDE super computing resources, we have obtained new results which show new trends of turbulence statistics behind the shock which were not known before. Moreover, we are also developing tools to consider multi-species non-equilibrium flows. The main results are in three areas: (1) development of high-order shock-fitting scheme for perfect gas flow, (2) Direct Numerical Simulation (DNS) of interaction of realistic turbulence with moderate to very strong shocks using super computing resources, and (3) development and implementation of models for computation of mutli-species non-quilibrium flows with shock-fitting codes.

97 MATHEMATICS AND COMPUTING↗

Angular analysis of the B o → K *o μ + μ – decay using 3 fb –1 of integrated luminosity

An angular analysis of the B o → K *o (→ K + π – )μ + μ – decay is presented. The dataset corresponds to an integrated luminosity of 3.0 fb –1 of pp collision data collected at the LHCb experiment. The complete angular information from the decay is used to determine CP-averaged observables and CP asymmetries, taking account of possible contamination from decays with the K + π – system in an S-wave configuration. The angular observables and their correlations are reported in bins of q 2 , the invariant mass squared of the dimuon system. The observables are determined both from an unbinned maximum likelihood fit and by using the principal moments of the angular distribution. In addition, by fitting for q 2 -dependent decay amplitudes in the region 1.1 < q 2 < 6.0 GeV 2 /c 4 , the zero-crossing points of several angular observables are computed. A global fit is performed to the complete set of CP-averaged observables obtained from the maximum likelihood fit. This fit indicates differences with predictions based on the Standard Model at the level of 3.4 standard deviations. These differences could be explained by contributions from physics beyond the Standard Model, or by an unexpectedly large hadronic effect that is not accounted for in the Standard Model predictions.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Parameter inference from event ensembles and the top-quark mass

One of the key tasks of any particle collider is measurement. In practice, this is often done by fitting data to a simulation, which depends on many parameters. Sometimes, when the effects of varying different parameters are highly correlated, a large ensemble of data may be needed to resolve parameter-space degeneracies. An important example is measuring the top-quark mass, where other physical and unphysical parameters in the simulation must be profiled when fitting the top-quark mass parameter. We compare four different methodologies for top-quark mass measurement: a classical histogram fit similar to one commonly used in experiment augmented by soft-drop jet grooming; a 2D profile likelihood fit with a nuisance parameter; a machine-learning method called DCTR; and a linear regression approach, either using a least-squares fit or with a dense linearly-activated neural network. Despite the fact that individual events are totally uncorrelated, we find that the linear regression methods work most effectively when we input an ensemble of events sorted by mass, rather than training them on individual events. Although all methods provide robust extraction of the top-quark mass parameter, the linear network does marginally best and is remarkably simple. For the top study, we conclude that the Monte-Carlo-based uncertainty on current extractions of the top-quark mass from LHC data can be reduced significantly (by perhaps a factor of 2) using networks trained on sorted event ensembles. More generally, machine learning from ensembles for parameter estimation has broad potential for collider physics measurements.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Interactions of Alumina-Based and Magnesia-Based Refractories with Iron Melts and Slags: A Review

A novel flash ironmaking technology (FIT) based on the direct reduction of iron ore concentrate with a reductant gas (such as hydrogen, natural gas, coal gas, or a combination thereof) in a flash furnace is being developed at the University of Utah. This technology which is undergoing large-scale laboratory testing aims at overcoming the limitations of blast furnace ironmaking by bypassing the problematic pelletization/sintering and cokemaking steps.[1–5] Refractory selection is expected to play an important step in the development of FIT and its proposed scale-up. For nominating an appropriate refractory for the FIT, understanding the interactions of candidate refractories with iron/iron oxide and slags under H{sub 2}/CO/CO{sub 2}/H{sub 2}O environments is necessary. This work is undertaken to review the existing literature on the interactions of important refractories with iron melts and relevant slags with an emphasis on two of the most commonly used refractories in ironmaking and steelmaking applications: the alumina-based refractories (used widely in blast furnace operations) and the magnesia-based refractories (used extensively in primary as well as secondary steelmaking). First, a comprehensive review on the interactions of alumina-based refractories with iron melts and slags has been done. Next the existing literature on the interactions of magnesia-based refractories with iron melts and relevant slags has been reviewed. Summaries have been included after each section and sub-section along with comments and critical insights from the authors. Finally, in the concluding remarks the differences in operating conditions between existing iron and steelmaking practices and the novel FIT have been highlighted. On the basis of these differences, it has been argued that the results and conclusions available from previous studies on refractory–metal–slag interactions are of little significance to flash ironmaking. Thus, there exists a need to carry out laboratory experiments for evaluating refractory performance in flash ironmaking under conditions relevant to the current process (FIT).

36 MATERIALS SCIENCE↗

SymbolFit: Automatic Parametric Modeling with Symbolic Regression

We introduce SymbolFit (API: https://github.com/hftsoi/symbolfit), a framework that automates parametric modeling by using symbolic regression to perform a machine-search for functions that fit the data while simultaneously providing uncertainty estimates in a single run. Traditionally, constructing a parametric model to accurately describe binned data has been a manual and iterative process, requiring an adequate functional form to be determined before the fit can be performed. The main challenge arises when the appropriate functional forms cannot be derived from first principles, especially when there is no underlying true closed-form function for the distribution. In this work, we develop a framework that automates and streamlines the process by utilizing symbolic regression, a machine learning technique that explores a vast space of candidate functions without requiring a predefined functional form because the functional form itself is treated as a trainable parameter, making the process far more efficient and effortless than traditional regression methods. We demonstrate the framework in high-energy physics experiments at the CERN Large Hadron Collider (LHC) using five real proton-proton collision datasets from new physics searches, including background modeling in resonance searches for high-mass dijet, trijet, paired-dijet, diphoton, and dimuon events. We show that our framework can flexibly and efficiently generate a wide range of candidate functions that fit a nontrivial distribution well using a simple fit configuration that varies only by random seed, and that the same fit configuration, which defines a vast function space, can also be applied to distributions of different shapes, whereas achieving a comparable result with traditional methods would have required extensive manual effort.

Tsoi, Ho Fung [Univ. of Pennsylvania, Philadelphia↗

A near-stationary subspace for ridge approximation

Response surfaces are common surrogates for expensive computer simulations in engineering analysis. However, the cost of fitting an accurate response surface increases exponentially as the number of model inputs increases, which leaves response surface construction intractable for high-dimensional, nonlinear models. Here we describe ridge approximation for fitting response surfaces in several variables. A ridge function is constant along several directions in its domain, so fitting occurs on the coordinates of a low-dimensional subspace of the input space. We review essential theory for ridge approximation – e.g., the best mean-squared approximation and an optimal low-dimensional subspace – and we prove that the gradient-based active subspace is near-stationary for the least-squares problem that defines an optimal subspace. Motivated by the theory, we propose a computational heuristic that uses an estimated active subspace as an initial guess for a ridge approximation fitting problem. We show a simple example where the heuristic fails, which reveals a type of function for which the proposed approach is inappropriate. We then propose a simple alternating heuristic for fitting a ridge function, and we demonstrate the effectiveness of the active subspace initial guess applied to an airfoil model of drag as a function of its 18 shape parameters.

42 ENGINEERING↗