Search NASA⌕ Search

SEARCH · Search NASA

Results for “data similarity”

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 19 records

Exploiting Data Similarity to Reduce Memory Footprints

Memory size has long limited large-scale applications on high-performance computing (HPC) systems. Since compute nodes frequently do not have swap space, physical memory often limits problem sizes. Increasing core counts per chip and power density constraints, which limit the number of DIMMs per node, have exacerbated this problem. Further, DRAM constitutes a significant portion of overall HPC system cost. Therefore, instead of adding more DRAM to the nodes, mechanisms to manage memory usage more efficiently - preferably transparently - could increase effective DRAM capacity and thus the benefit of multicore nodes for HPC systems. MPI application processes often exhibit significant data similarity. These data regions occupy multiple physical locations across the individual rank processes within a multicore node and thus offer a potential savings in memory capacity. These regions, primarily residing in heap, are dynamic, which makes them difficult to manage statically. Our novel memory allocation library, SBLLmalloc, automatically identifies identical memory blocks and merges them into a single copy. SBLLmalloc does not require application or OS changes since we implement it as a user-level library. Overall, we demonstrate that SBLLmalloc reduces the memory footprint of a range of MPI applications by 32.03% on average and up to 60.87%. Further, SBLLmalloc supports problem sizes for IRS over 21.36% larger than using standard memory management techniques, thus significantly increasing effective system size. Similarly, SBLLmalloc requires 43.75% fewer nodes than standard memory management techniques to solve an AMG problem.

97 MATHEMATICS AND COMPUTING↗

Adaptive Sampling for High Throughput Data Using Similarity Measures

The need for adaptive sampling arises in the context of high throughput data because the rates of data arrival are many orders of magnitude larger than the rates at which they can be analyzed. A very fast decision must therefore be made regarding the value of each incoming observation and its inclusion in the analysis. In this report we discuss one approach to adaptive sampling, based on the new data point’s similarity to the other data points being considered for inclusion. We present preliminary results for one real and one synthetic data set.

97 MATHEMATICS AND COMPUTING↗

A machine-learning-aided data recovery approach for predicting multi-material thermal behaviors in advanced test reactor capsules

Instrumented experiments conducted at test reactors are essential to the deployment of new advanced reactor systems. Designing new experiments and generating data on specific reactor conditions require significant investments in terms of both time and cost. Finite element analysis software can be used to create high-fidelity models of experiment environments in order to support the actual experiments, but computation time remains a concern in terms of applying outcomes to real-time usage of data (e.g., a digital twin [DT]). Here, the present research proposes a machine-learning (ML) aided approach to making temperature and displacement predictions based on the thickness of the outer gas gap on the experimental capsule used for in-pile demonstration of a novel new thermal conductivity probe in the Advanced Test Reactor (ATR). This capsule consisted of U10Zr fuel, a rodlet, sodium, and inner and outer capsules. Gas gaps existed between the fuel and the rodlet, and between the inner and the outer capsule. The learning data pertained to an experimental capsule's radial distributions of temperature and displacement, as obtained based on Abaqus and the physical features. For the first step of ML sequence, the temperature was predicted using three positional parameters. Next, the displacement was predicted using seven additional parameters. Each physical feature was normalized in order to be both nondimensional and standardized. The temperature and displacement predictions showed good agreement with the simulation results in all cases involving interpolation and extrapolation. Furthermore, data similarity enhancement increased the similarity between the training and the target data, thereby increasing the predictive accuracy of the ML models. In certain extrapolation cases involving limited original ML model accuracy, data similarity enhancement and data recovery was able to somewhat improve this accuracy.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Machine-Learning-aided Approach for Predicting the Thermal Expansion Behaviors in Advanced Test Reactor Capsules (NURETH-20 full paper)

Instrumented experiments at test reactors are essential to deploying new advanced reactor systems. Designing new experiments and generating data on specific conditions require both time and cost investment. A high-fidelity model of the experiment environment can be created using finite element analysis software to support the actual experiments, but computation time is still a concern in applying outcomes to real-time usage (e.g., a digital twin). This research proposes a machine-learning-aided approach to temperature and displacement predictions, based on the thickness of the outer gas gap on the experimental capsule used for the in-pile demonstration of a novel thermal conductivity probe in the Advanced Test Reactor. The capsule consisted of U10Zr fuel, a rodlet, sodium, and inner and outer capsules. There were gas gaps between the fuel and rodlet and between the inner and outer capsule. The learning data consisted of an experimental capsule’s radial distributions of temperature and displacement, as obtained from Abaqus and the physical features. For the first step, temperature was predicted using three positional parameters. Then the displacement was predicted using six different positional parameters. Each physical feature was normalized to be both nondimensional and standardized. The temperature and displacement predictions showed good agreement in all cases involving interpolation and extrapolation. Also, data similarity enhancement increased the similarity between training and target data increasing the predictive accuracy of machine-learning models. In some cases of extrapolation, the accuracy of the machine-learning model showed limited performance, but still data similarity enhancement improved the accuracy.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Complete fusion of {sup 9}Be with spherical targets

The complete fusion of {sup 9}Be with {sup 144}Sm and {sup 208}Pb targets is calculated in the coupled-channels approach. The calculations include couplings among the 3/2{sup -}, 5/2{sup -}, and 7/2{sup -} states in the K=3/2 ground-state rotational band of {sup 9}Be. It is shown that the B(E2) values for the excitation of these states are accurately described in the rotor model. The interaction of the strongly deformed {sup 9}Be nucleus with a spherical target is calculated using the double-folding technique and the effective M3Y interaction, which is supplemented with a repulsive term that is adjusted to optimize the fit to the data for the {sup 144}Sm target. The complete fusion is described by ingoing-wave boundary conditions. The decay of the unbound excited states in {sup 9}Be is considered explicitly in the calculations by using complex excitation energies. The model gives an excellent account of the complete fusion (CF) data for {sup 9}Be+{sup 144}Sm, and the cross sections for the decay of the excited states are in surprisingly good agreement with the incomplete fusion (ICF) data. Similar calculations for {sup 9}Be+{sup 208}Pb explain the total fusion data at high energies but fail to explain the CF data, which are suppressed by 20%, and the calculated cross section for the decay of excited states is a factor of 3 smaller than the ICF data at high energies. Possible reasons for these discrepancies are discussed.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Optimizing Kernel Machines Using Deep Learning

Building highly non-linear and non-parametric models is central to several state-of-the-art machine learning systems. Kernel methods form an important class of techniques that induce a reproducing kernel Hilbert space (RKHS) for inferring non-linear models through the construction of similarity functions from data. These methods are particularly preferred in cases where the training data sizes are limited and when prior knowledge of the data similarities is available. Despite their usefulness, they are limited by the computational complexity and their inability to support end-to-end learning with a task specific objective. On the other hand, deep neural networks have become the de facto solution for end-to-end inference in several learning paradigms. In this article, we explore the idea of using deep architectures to perform kernel machine optimization, for both computational efficiency and end-to-end inferencing. To this end, we develop the DKMO (Deep Kernel Machine Optimization) framework, that creates an ensemble of dense embeddings using Nystrõm kernel approximations and utilizes deep learning to generate task-specific representations through the fusion of the embeddings. Intuitively, the filters of the network are trained to fuse information from an ensemble of linear subspaces in the RKHS. Furthermore, we introduce the kernel dropout regularization to enable improved training convergence. Finally, we extend this framework to the multiple kernel case, by coupling a global fusion layer with pre-trained deep kernel machines for each of the constituent kernels. Using case studies with limited training data, and lack of explicit feature sources, we demonstrate the effectiveness of our framework over conventional model inferencing techniques.

97 MATHEMATICS AND COMPUTING↗

Comparison of pyrometry and thermography for thermal analysis of thermite reactions

This article examines the thermal behavior of a laser ignited thermite composed of aluminum and bismuth trioxide. Temperature data were collected during the reaction using a four-color pyrometer and a high-speed color camera modified for thermography. The two diagnostics were arranged to collect data simultaneously, with similar fields of view and with similar data acquisition rates, so that the two techniques could be directly compared. Results show that at initial and final stages of the reaction, a lower signal-to-noise ratio affects the accuracy of the measured temperatures. Both diagnostics captured the same trends in transient thermal behavior, but the average temperatures measured with thermography were about 750 K higher than those from the pyrometer. This difference was attributed to the lower dynamic range of the thermography camera’s image sensor, which was unable to resolve cooler temperatures in the field of view as well as the photomultiplier tube sensors in the pyrometer. Overall, while the camera could not accurately capture the average temperature of a scene, its ability to capture peak temperatures and spatial data make it the preferred method for tracking thermal behavior in thermite reactions.

36 MATERIALS SCIENCE↗

Air Corrosivity in U.S. Outdoor-Air-Cooled Data Centers is Similar to That in Conventional Data Centers

There is a concern that environmental-contamination caused corrosion may negatively affect Information Technology (IT) equipment reliability. Nineteen data centers in the United States and two in India were evaluated using Corrosion Classification Coupons (CCC) to assess environmental air quality as it may relate IT equipment reliability. The data centers were of two basic types: closed and outside-air cooled. A closed data center provides cool air to the IT equipment using air conditioning in which only a small percent age of the recirculation air is make-up air continuously supplied from outside to meet human health requirements. An outside-air cooled data center uses outside air directly as the primary source for IT equipment cooling. Corrosion measuring coupons containing copper and silver metal strips were placed in both closed and outside-air cooled data centers. The coupons were placed at each data center (closed and outside-air cooled types) with the location categorized into three groups: (1) Outside - coupons sheltered, located near or at the supply air inlet, but located before any filtering, (2) Supply - starting just after initial air filtering continuing inside the plenums and ducts feeding the data center rooms, and (3) Inside located inside the data center rooms near the IT equipment. Each coupon was exposed for thirty days and then sent to a laboratory for a corrosion rate measurement analysis. The goal of this research was to investigate whether gaseous contamination is a concern for U.S. data center operators as it relates to the reliability of IT equipment. More specifically, should there be an increased concern if outside air for IT equipment cooling is used To begin to answer this question limited exploratory measurements of corrosion rates in operating data centers in various locations were undertaken. This study sought to answer the following questions: (1) What is the precision of the measurements (2) What are the approximate statistical distributions of copper and silver corrosion rates in the sampled data centers(3) To what extent are copper and silver corrosion measurements related (4) What is the relationship of corrosion rate measurements between outside-air cooled data centers compared to closed data centers (5) How do corrosivity measurements relate to IT equipment failure rates The data from our limited sample size suggests that most United States data center operators should not be concerned with environmental gaseous contamination causing high IT equipment failure rates even when using outside-air cooling. The research team recommends additional basic research on how environmental conditions, specifically gaseous contamination, affect electronic equipment reliability.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

CMS Storage Performance with RNTuple

CMS is transitioning to use ROOT’s new RNTuple data storage format for the files CMS will write in the HL-LHC era. Based on initial tests, CMS expects faster I/O and smaller files compared to the present TTree storage format. This contribution will show a comprehensive performance comparison between RNTuple and TTree I/O using CMS AOD and MiniAOD data formats as test cases for both simulation and collision data corresponding to similar data taking conditions of LHC Run 3. Quantities such as the resulting file size, the memory usage of the I/O components, and the rate of events being read from a file or written to a file will be measured. CMS’ data processing relies heavily on reading files over the local or wide area networks. The file read patterns are important because the latencies have been seen to influence the total production job times. Therefore a study on the file read patterns will be conducted by recording traces of the offset, size, and timestamp of each read request for both RNTuple and TTree. The behavior of network reads will be mimicked by reading local files where artificial latency will be added to the read requests. The effect of different latency values on the job times will be studied.

Jones, Christopher D. [Fermilab]↗

Prediction of histone post-translational modifications using deep learning

Abstract Motivation Histone post-translational modifications (PTMs) are involved in a variety of essential regulatory processes in the cell, including transcription control. Recent studies have shown that histone PTMs can be accurately predicted from the knowledge of transcription factor binding or DNase hypersensitivity data. Similarly, it has been shown that one can predict PTMs from the underlying DNA primary sequence. Results In this study, we introduce a deep learning architecture called DeepPTM for predicting histone PTMs from transcription factor binding data and the primary DNA sequence. Extensive experimental results show that our deep learning model outperforms the prediction accuracy of the model proposed in Benveniste et al. (PNAS 2014) and DeepHistone (BMC Genomics 2019). The competitive advantage of our framework lies in the synergistic use of deep learning combined with an effective pre-processing step. Our classification framework has also enabled the discovery that the knowledge of a small subset of transcription factors (which are histone-PTM and cell-type-specific) can provide almost the same prediction accuracy that can be obtained using all the transcription factors data. Availabilityand implementation https://github.com/dDipankar/DeepPTM. Supplementary information Supplementary data are available at Bioinformatics online.

Baisya, Dipankar Ranjan (ORCID:0000000267847359)↗

First Look at Smoke Emissions from Prescribed Burns in Long-unburned Longleaf Pine Forests

While fire has long played a role in the longleaf pine ecosystem, there are still some stands in the southeastern United States where fire has not been reintroduced and fuels have accumulated for 50 years or more. As part of a larger study examining fuel loading and smoke emissions on Department of Defense installations in the southeastern U.S., fuels and trace emissions were measured during three prescribed burns at Ft. Jackson Army Base near Columbia, South Carolina in November 2011. These pine-forest understory fires provided valuable emissions data for fires that burned in stands that had little or no exposure to fire for decades. Smoke emissions were measured on the ground and from an aircraft by scientists from a large team of atmospheric researchers. (Akagi et al., 2013) To characterize initial emissions in the lofted plume and in point sources of residual smoldering combustion, trace-gas species were measured using an airborne FTIR and a ground-based FTIR, respectively. Whole-air sampling canisters were also collected from both ground- and airborne-based platforms. A total of 97 trace gases were quantified in this work, largely via infrared spectroscopy. Selected emissions data were compared with similar data collected from prescribed burns sampled in coastal North Carolina in 2010 in younger fuels beds of loblolly/longleaf stands near Camp Lejeune (Burling et al., 2011). The emission factors measured in this work differ by ~13-195% to EF measured from the managed stands at Camp Lejeune for organic and N-containing species, suggesting that fire emissions in similar ecosystems can exhibit large variability. Part of the differences, however, may be ascribed to burn conditions as well since the NC burns were during the wet season whereas the SC stands were burned after an extended drought. We also report the first detailed FTIR emissions data for a suite of monoterpenes. Figure 1 displays the emission factors (g/kg fuel) for several monoterpenes and isoprene as measured by the ground-based FTIR system. Due to their unsaturated structure, terpenes are highly reactive compounds emitted from plants thought to contribute to secondary organic aerosol formation (SOA) (Saathoff et al., 2009; Hennigan et al., 2011) and the formation of small oxygenated volatile organic compounds (OVOCs) (Jacob et al., 2002) in fire plumes. The known chemistry and measured abundance of monoterpenes suggests that these species impacted secondary plume processes including ozone, OVOC, and SOA formation in sampled plumes within the first few hours after emission.

biomass burning↗

The one-loop matter bispectrum in the Effective Field Theory of Large Scale Structures

With this study, given the importance of future large scale structure surveys for delivering new cosmological information, it is crucial to reliably predict their observables. The Effective Field Theory of Large Scale Structures (EFTofLSS) provides a manifestly convergent perturbative scheme to compute the clustering of dark matter in the weakly nonlinear regime in an expansion in k/k NL , where k is the wavenumber of interest and k NL is the wavenumber associated to the nonlinear scale. It has been recently shown that the EFTofLSS matches to 1% level the dark matter power spectrum at redshift zero up to k ≃ 0.3 h Mpc –1 and k ≃ 0.6 h Mpc –1 at one and two loops respectively, using only one counterterm that is fit to data. Similar results have been obtained for the momentum power spectrum at one loop. This is a remarkable improvement with respect to former analytical techniques. Here we study the prediction for the equal-time dark matter bispectrum at one loop. We find that at this order it is sufficient to consider the same counterterm that was measured in the power spectrum. Without any remaining free parameter, and in a cosmology for which kNL is smaller than in the previously considered cases (σ 8 =0.9), we find that the prediction from the EFTofLSS agrees very well with N-body simulations up to k ≃ 0.25 h Mpc –1 , given the accuracy of the measurements, which is of order a few percent at the highest k's of interest. While the fit is very good on average up to k ≃ 0.25 h Mpc –1 , the fit performs slightly worse on equilateral configurations, in agreement with expectations that for a given maximum k, equilateral triangles are the most nonlinear.

79 ASTRONOMY AND ASTROPHYSICS↗

Development of a Kalman filter estimator for simulation and control of particulate matter distribution of a diesel catalyzed particulate filter

The knowledge of the temperature and particulate matter mass distribution is essential for monitoring the performance and durability of a catalyzed particulate filter. A catalyzed particulate filter model was developed, and it showed capability to accurately predict temperature and particulate matter mass distribution and pressure drop across the catalyzed particulate filter. However, the high-fidelity model is computationally demanding. Therefore, a reduced order multi-zone particulate filter model was developed to reduce computational complexity with an acceptable level of accuracy. In order to develop a reduced order model, a parametric study was carried out to determine the number of zones necessary for aftertreatment control applications. The catalyzed particulate filter model was further reduced by carrying out a sensitivity study of the selected model assumptions. The reduced order multi-zone particulate filter model with 5 × 5 zones was selected to develop a catalyzed particulate filter state estimator considering its computational time and accuracy. Next, a Kalman filter–based catalyzed particulate filter estimator was developed to estimate unknown states of the catalyzed particulate filter such as temperature and particulate matter mass distribution and pressure drop (Δ P) using the sensor inputs to the engine electronic control unit and the reduced order multi-zone particulate filter model. A diesel oxidation catalyst estimator was also integrated with the catalyzed particulate filter estimator in order to provide estimates of diesel oxidation catalyst outlet concentrations of NO 2 and hydrocarbons and inlet temperature for the catalyzed particulate filter estimator. The combined diesel oxidation catalyst–catalyzed particulate filter estimator was validated for an active regeneration experiment. The validation results for catalyzed particulate filter temperature distribution showed that the root mean square temperature error by using the diesel oxidation catalyst–catalyzed particulate filter estimator is within 3.2 °C compared to the experimental data. Similarly, the Δ P estimator closely simulated the measured total Δ P and the estimated cake pressure drop error is within 0.2 kPa compared to the high-fidelity catalyzed particulate filter model.

Engineering↗

IMPROVED LINE DATA FOR THE SWAN SYSTEM {sup 12}C{sup 13}C ISOTOPOLOGUE

We present new, accurate predictions for rotational line positions, excitation energies, and transition probabilities of the {sup 12}C{sup 13}C isotopologue Swan d{sup 3}Π-a{sup 3}Π system 0-0, 0–1, 0–2, 1–0, 1–1, 1–2, 2–0, 2–1, and 2–2 vibrational bands. The line positions and energy levels were predicted through new analyses of published laboratory data for the {sup 12}C{sup 13}C lines. Transition probabilities were derived from recent computations of transition dipole moments and related quantities. The {sup 12}C{sup 13}C line data were combined with similar data for {sup 12}C{sub 2,} reported in a companion paper, and applied to produce synthetic spectra of carbon-rich metal-poor stars that have strong C{sub 2} Swan bands. The matches between synthesized and observed spectra were used to estimate band head positions for a few of the {sup 12}C{sup 13}C vibrational bands and to verify that the new computed line data match observed spectra. The much weaker C{sub 2} lines of the bright red giant Arcturus were also synthesized in the band head regions.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Distinguishing isotropic and anisotropic signals for X-ray total scattering using machine learning

Understanding structure–property relationships is essential for advancing technologies based on thin films. X-ray pair distribution function (PDF) analysis can access relevant atomic structure details spanning local-, mid- and long-range structure. While X-ray PDF has been adapted for thin films on amorphous substrates, measurements on single-crystal substrates are necessary to accurately determine structure origins for some thin film materials, especially those for which the substrate changes the accessible structure and properties. However, when measuring films on single-crystal substrates, high-intensity anisotropic Bragg spots saturate 2D detector images, overshadowing the thin films' isotropic scattering signal. This renders previous data processing methods for films on amorphous substrates unsuitable for films on single-crystal substrates. To address this measurement need, we developed IsoDAT2D, an innovative data processing approach using unsupervised machine learning algorithms. The program combines dimensionality reduction and clustering algorithms to separate thin film and single-crystal substrate X-ray scattering signals. We use SimDAT2D , a program we developed to generate simulated thin film data, to validate IsoDAT2D . Here we also use IsoDAT2D to isolate X-ray total scattering signal from a thin film on a single-crystal substrate. The resulting PDF data are compared with similar data processed using previous methods, especially substrate subtraction for single-crystal and amorphous substrates. PDF data from IsoDAT2D -identified X-ray total scattering data are significantly better than from single-crystal substrate subtraction, but not as reliable as PDF data from amorphous substrate subtraction. With IsoDAT2D , there are new opportunities to expand PDF to a wider variety of thin films, including those on single-crystal substrates, with which new structure–property relationships can be elucidated to enable fundamental understanding and technological advances.

36 MATERIALS SCIENCE↗

A Deterministic and Probabilistic Framework for Forecasting of Time-Series Damage States and Associated End-of-lIfe of a Pressurizer Water Reactor Surge Line under Design-Basis and Grid-Load-Following Loading Conditions

Argonne National Laboratory in support of the Department of Energy’s Light Water Reactor Sustainability (LWRS) program. The present semi-annual report covers the period between May 2018 and September 2018. In this report, we present the work performed to further improve the capability for structural integrity prediction. Part of this work focused on validating a computational fluid dynamics model for modeling thermal stratification. We also present finite element (FE) model based thermal-mechanical stress analysis of a pressurized water reactor (PWR) surge line under grid-load-following loading condition. In addition, we present work related to end-of-life prediction of a PWR surge line based on ASME code and NUREG-6909 deterministic approaches under design-basis and grid-load-following conditions. Our research in this report period also majorly involved with the development of a strategy for probabilistic fatigue life estimation. First, we applied a Weibull probabilistic modeling approach based on end-of-life fatigue data (traditional S~N data under in-air and PWR water conditions) to estimate the probabilistic life for a given strain amplitude. Second, we developed a data analytics based Markov-Chain-Monte-Carlo (MCMC) model for probabilistic modeling of time-series fatigue damage states and the associated probabilistic end-of-life of a component. Unlike the Weibull type approach, the MCMC type may not require a large number of fatigue tests for estimating the end-of-life for given loading and environmental conditions. We demonstrated the use of this framework through symmetric fatigue tests (stress- or strain-ontrolled tests either under in-air or PWR-water conditions) obtained from our earlier experiments. The MCMC model can use base time-series data either obtained from a model or fatigue experiment. Since experimental data are typically more accurate than model-calculated data, we used experiment-based time-series data to demonstrate the MCMC framework. However, when performing a complex experiment is not possible (due to time and costs involved), a model-based approach can be used to generate the time-series data. Similar model-based data are presented in our earlier report (e.g. for cyclic hardening damage states of a PWR surge line under constant amplitude and design-basis loading [1]). However, for the MCMC work in this report, we used experiment-based time-series damage states obtained from fatigue experiments under design-basis and grid-load-following loading conditions. The experimental loading inputs were based on thermal-mechanical FE model results of a typical PWR surge line. Then, we used the experimental time-series damage state data for forecasting the probabilistic time-series damage states and end-of-life of the PWR surge line. Based on this MCMC modeling strategy, and with the assumed loading, environment and thermal-mechanical boundary conditions, we have estimated that a 316SS-pure base metal PWR surge line would survive at least 159 years with zero failure probability.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Understanding the Evolution of Customer Motivations and Adoption Barriers in Residential Solar Markets: Survey Data

This research focused on accelerating solar photovoltaic (PV) diffusion by collecting new market data and developing predictive modeling frameworks to test and refine understandings of household level motivations for adopting solar. Three different household-level surveys were fielded: one for households who had installed PV on their current home or had signed a contract to do so (the Adopter survey), one for households that had seriously considered PV but had not installed it (the Considerer survey), and one for the general population who did not have PV on their current home (the general population survey or GPS). Survey respondents were from four U.S. states: New Jersey, New York, Arizona, and California. Details of recruiting and sampling are documented below. Research projects on residential PV adoption often collect data only from PV adopters or from the general population. One of the innovations of this project was the three-pronged household survey data collection. By collecting similar data from three fairly different "statuses" with respect to adoption, the surveys provide a basis for understanding how those who do not have rooftop PV differ from those who have, for how and why people do (or don't) transition from not having to having rooftop PV on their home, and for understanding the characteristics and viewpoints of households who have scarcely, or not at all, entered the "PV consideration" track. All three surveys covered single-family owner-occupied households in each of the four target states used in the project -- Arizona, California, New Jersey, and New York - allowing a comparative approach to understanding how the factors that affect PV adoption vary by geography and policy conditions. The General Population and Considerer surveys provide a basis for understanding opinions about and interest in solar, and how these relate to household demographics and other conditions. Paired with the Adopter survey, they also provide data for understanding how those who do not have rooftop PV differ from those who have, and for how and why people do (or don't) transition from not having to having rooftop PV on their home. The Adopter survey questions were designed to capture a broad range of information on what motivates and impedes households to install rooftop PV, as well as the details and timing of the decision and installation. Survey instrument development drew from existing PV adoption survey instruments, PV adoption literature, and research team experience, as well as from past work on household interest in energy efficiency, environmental attitudes, purchasing tendencies, and related knowledge. Early interviews and discussions with installers and others in the PV industry were also taken into consideration.

14 SOLAR ENERGY↗