Search NASA⌕ Search

Engineering topics

Kwan, Chiman

Publications and source records attributed to Kwan, Chiman.

At least 19 records

Deep Multitask Learning Models for Radiation Estimation at High Energy Accelerator Facility

Controlling the dose of radiation exposure in potential radioactive facilities is critical for ensuring the safety of staff and the public. Here, in this paper, we developed machine learning models to estimate radiation exposure efficiently at the Thomas Jefferson National Accelerator Facility (JLab), aiming to enhance safety in both accelerator facilities and public areas. Multiple sensors were deployed around the three experimental halls at JLab. Data on single-beam currents, energy levels, and radiation values at the sensor locations were collected during accelerator operation. We proposed a multi-task learning model for radiation estimation, utilizing either one-dimensional convolutional neural networks (1-D CNNs) or long short-term memory networks (LSTMs) as the backbone. The proposed model was trained to simultaneously estimate radiation levels at the sensor locations. Experimental results demonstrated that the proposed model with LSTM backbone achieved the best estimation performance, with an average R 2 score of 0.7557 for estimation within the same year and 0.7157 for estimation across different years. These results significantly surpassed those of competing models.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

An LSTM Deep Learning Network for ¿Background Radiation Prediction

Determination of appropriate background radiation is important in any measurement application. Environmental radiation monitors and monitors used to assess dose to individuals outside of controlled areas are particularly susceptible to changes in readings due to fluctuations in the environmental conditions. These fluctuations (e.g. radon progeny concentrations) lead to changes in the detector response in the actual radiation environment, and they need to be taken into account when extracting the net operational doses. Work has been ongoing to apply advances in Deep Learning and Artificial Intelligence to account for changes in detector responses based on environmental parameters; in particular, a Long-Short Term Memory (LSTM) Deep Learning architecture has been utilized to incorporate time-series data into a prediction model. In this work, we present the current status of the project to predict radiation measurements based on meteorological conditions and air packet trajectories extracted using the National Oceanic and Atmospheric Administration's (NOAA) Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT4) model.

Degtiarenko, Pavel↗

An Accurate Vegetation and Non-Vegetation Differentiation Approach Based on Land Cover Classification

Accurate vegetation detection is important for many applications, such as crop yield estimation, land cover land use monitoring, urban growth monitoring, drought monitoring, etc. Popular conventional approaches to vegetation detection incorporate the normalized difference vegetation index (NDVI), which uses the red and near infrared (NIR) bands, and enhanced vegetation index (EVI), which uses red, NIR, and the blue bands. Although NDVI and EVI are efficient, their accuracies still have room for further improvement. In this paper, we propose a new approach to vegetation detection based on land cover classification. That is, we first perform an accurate classification of 15 or more land cover types. The land covers such as grass, shrub, and trees are then grouped into vegetation and other land cover types such as roads, buildings, etc. are grouped into non-vegetation. Similar to NDVI and EVI, only RGB and NIR bands are needed in our proposed approach. If Laser imaging, Detection, and Ranging (LiDAR) data are available, our approach can also incorporate LiDAR in the detection process. Results using a well-known dataset demonstrated that the proposed approach is feasible and achieves more accurate vegetation detection than both NDVI and EVI. In particular, a Support Vector Machine (SVM) approach performed 6% better than NDVI and 50% better than EVI in terms of overall accuracy (OA).

54 ENVIRONMENTAL SCIENCES↗

Deep Learning for Land Cover Classification Using Only a Few Bands

There is an emerging interest in using hyperspectral data for land cover classification. The motivation behind using hyperspectral data is the notion that increasing the number of narrowband spectral channels would provide richer spectral information and thus help improve the land cover classification performance. Although hyperspectral data with hundreds of channels provide detailed spectral signatures, the curse of dimensionality might lead to degradation in the land cover classification performance. Moreover, in some practical applications, hyperspectral data may not be available due to cost, data storage, or bandwidth issues, and RGB and near infrared (NIR) could be the only image bands available for land cover classification. Light detection and ranging (LiDAR) data is another type of data to assist land cover classification especially if the land covers of interest have different heights. In this paper, we examined the performance of two Convolutional Neural Network (CNN)-based deep learning algorithms for land cover classification using only four bands (RGB+NIR) and five bands (RGB+NIR+LiDAR), where these limited number of image bands were augmented using Extended Multi-attribute Profiles (EMAP). The deep learning algorithms were applied to a well-known dataset used in the 2013 IEEE Geoscience and Remote Sensing Society (GRSS) Data Fusion Contest. With EMAP augmentation, the two deep learning algorithms were observed to achieve better land cover classification performance using only four bands as compared to that using all 144 hyperspectral bands.

58 GEOSCIENCES↗

Improving Land Cover Classification Using Extended Multi-Attribute Profiles (EMAP) Enhanced Color, Near Infrared, and LiDAR Data

Hyperspectral (HS) data have found a wide range of applications in recent years. Researchers observed that more spectral information helps land cover classification performance in many cases. However, in some practical applications, HS data may not be available, due to cost, data storage, or bandwidth issues. Instead, users may only have RGB and near infrared (NIR) bands available for land cover classification. Sometimes, light detection and ranging (LiDAR) data may also be available to assist land cover classification. A natural research problem is to investigate how well land cover classification can be achieved under the aforementioned data constraints. In this paper, we investigate the performance of land cover classification while only using four bands (RGB+NIR) or five bands (RGB+NIR+LiDAR). A number of algorithms have been applied to a well-known dataset (2013 IEEE Geoscience and Remote Sensing Society Data Fusion Contest). One key observation is that some algorithms can achieve better land cover classification performance by using only four bands as compared to that of using all 144 bands in the original hyperspectral data with the help of synthetic bands generated by Extended Multi-attribute Profiles (EMAP). Moreover, LiDAR data do improve the land cover classification performance even further.

58 GEOSCIENCES↗

Tree, Shrub, and Grass Classification Using Only RGB Images

In this work, a semantic segmentation-based deep learning method, DeepLabV3+, is applied to classify three vegetation land covers, which are tree, shrub, and grass using only three band color (RGB) images. DeepLabV3+’s detection performance has been studied on low and high resolution datasets that both contain tree, shrub, and grass and some other land cover types. The two datasets are heavily imbalanced where shrub pixels are much fewer than tree and grass pixels. A simple weighting strategy known as median frequency weighting was incorporated into DeepLabV3+ to mitigate the data imbalance issue, which originally used uniform weights. The tree, shrub, grass classification performances are compared when all land cover types are included in the classification and also when classification is limited to the three vegetation classes with both uniform and median frequency weights. Among the three vegetation types, shrub is found to be the most challenging one to classify correctly whereas correct classification accuracy was highest for tree. It is observed that even though the median frequency weighting did not improve the overall accuracy, it resulted in better classification accuracy for the underrepresented classes such as shrub in our case and it also significantly increased the average class accuracy. The classification performance and computation time comparison of DeepLabV3+ with two other pixel-based classification methods on sampled pixels of the three vegetation classes showed that DeepLabV3+ achieves significantly higher accuracy than these methods with a trade-off for longer model training time.

58 GEOSCIENCES↗

HMI Data Driven Magnetohydrodynamic Model Predicted Active Region Photospheric Heating Rates: Their Scale Invariant, Flare Like Power Law Distributions, and Their Possible Association With Flares

There are many flare forecasting models. For an excellent review and comparison of some of them see Barnes et al. (2016). All these models are successful to some degree, but there is a need for better models. We claim the most successful models explicitly or implicitly base their forecasts on various estimates of components of the photospheric current density J, based on observations of the photospheric magnetic field B. However, none of the models we are aware of compute the complete J. We seek to develop a better model based on computing the complete photospheric J. Initial results from this model are presented in this talk. We present a data driven, near photospheric, 3 D, non-force free magnetohydrodynamic (MHD) model that computes time series of the total J, and associated resistive heating rate in each pixel at the photosphere in the neutral line regions (NLRs) of 14 active regions (ARs). The model is driven by time series of B measured by the Helioseismic & Magnetic Imager (HMI) on the Solar Dynamics Observatory (SDO) satellite. Spurious Doppler periods due to SDO orbital motion are filtered out of the time series of B in every AR pixel. Errors in B due to these periods can be significant.

Goodman, Michael L.↗

HMI Data Driven Magnetohydrodynamic Model Predicted Active Region Photospheric Heating Rates: Their Scale Invariant, Flare Like Power Law Distributions, and Their Possible Association With Flares

A data driven, near photospheric, 3 D, non-force free magnetohydrodynamic model pre- dicts time series of the complete current density, and the resistive heating rate Q at the photosphere in neutral line regions (NLRs) of 14 active regions (ARs). The model is driven by time series of the magnetic field B observed by the Helioseismic & Magnetic Imager on the Solar Dynamics Observatory (SDO) satellite. Spurious Doppler periods due to SDO orbital motion are filtered out of the time series for B in every AR pixel. Errors in B due to these periods can be significant. The number of occurrences N(q) of values of Q > or = q for each AR time series is found to be a scale invariant power law distribution, N(Q) / Q−s, above an AR dependent threshold value of Q, where 0.3952 < or = s < or = 0.5298 with mean and standard deviation of 0.4678 and 0.0454, indicating little variation between ARs. Observations show that the number of occurrences N(E) of coronal flares with a total energy released > or = E obeys the same type of distribution, N(E) / E−S, above an AR dependent threshold value of E, with 0.38 < or approx. S < or approx. 0.60, also with little variation among ARs. Within error margins the ranges of s and S are nearly identical. This strong similarity between N(Q) and N(E) suggests a fundamental connection between the process that drives coronal flares and the process that drives photospheric NLR heating rates in ARs. In addition, results suggest it is plausible that spikes in Q, several orders of magnitude above background values, are correlated with times of the subsequent occurrence of M or X flares.

Goodman, Michael L.↗

HMI Data Driven Magnetohydrodynamic Model Predicted Active Region Photospheric Heating Rates: Their Scale Invariant, Flare Like Power Law Distributions, and Their Possible Association With Flares

A data driven, near photospheric, 3 D, non-force free magnetohydrodynamic model predicts time series of the complete current density, and the resistive heating rate Q at the photosphere in neutral line regions (NLRs) of 14 active regions (ARs). The model is driven by time series of the magnetic field B observed by the Helioseismic and Magnetic Imager on the Solar Dynamics Observatory (SDO) satellite. Spurious Doppler periods due to SDO orbital motion are filtered out of the time series for B in every AR pixel. Errors in B due to these periods can be significant. The number of occurrences N(q) of values of Q > or = q for each AR time series is found to be a scale invariant power law distribution, N(Q) / Q−s, above an AR dependent threshold value of Q, where 0.3952 < or = s < or = 0.5298 with mean and standard deviation of 0.4678 and 0.0454, indicating little variation between ARs. Observations show that the number of occurrences N(E) of coronal flares with a total energy released > or = E obeys the same type of distribution, N(E) / E−S, above an AR dependent threshold value of E, with 0.38 < or approx. S < or approx. 0.60, also with little variation among ARs. Within error margins the ranges of s and S are nearly identical. This strong similarity between N(Q) and N(E) suggests a fundamental connection between the process that drives coronal flares and the process that drives photospheric NLR heating rates in ARs. In addition, results suggest it is plausible that spikes in Q, several orders of magnitude above background values, are correlated with times of the subsequent occurrence of M or X flares.

Goodman, Michael L.↗

Photospheric Current Spikes as Possible Predictors of Flares

Flares involve generation of the largest current densities in the solar atmosphere. This suggests the hypothesis that prior to a large (M,X) flare there are related time dependent changes in the photospheric current distribution, and hence in the resistive heating rate in neutral line regions (NLRs). If this is true, these changes might be useful predictors of flares. Preliminary evidence supporting this hypothesis is presented. Results from a data driven, near photospheric, 3D magnetohydrodynamic type model suggest the model might be useful for predicting M and X flares several hours to several days in advance. The model takes as input the photospheric magnetic field observed by the Helioseismic and Magnetic Imager (HMI) on the Solar Dynamics Observatory (SDO) satellite. The model computes quantities in every active region (AR) pixel for 14 ARs, with spurious Doppler periods due to SDO orbital motion filtered out of the time series of the magnetic field for each pixel. Spikes in the NLR resistive heating rate Q, appearing as increases by orders of magnitude above background values in the time series of Q are found to occur, and appear to be correlated with the occurrence of M or X flares a few hours to a few days later. The subset of spikes analyzed at the pixel level are found to occur on HMI and granulation scales of 1 arcsec and 12 minutes. Spikes are found in NLRs with and without M or X flares, and outside as well as inside NLRs, but the largest spikes are localized in the NLRs of ARs with M or X flares, and associated with horizontal magnetic field strengths approximately several hG, and vertical magnetic field strengths several orders of magnitude smaller. The spikes may be signatures of horizontal current sheets associated with emerging magnetic flux.

Goodman, Michael L.↗

Photospheric Current Spikes and Their Possible Association with Flares - Results from an HMI Data Driven Model

A data driven, near photospheric magnetohydrodynamic model predicts spikes in the horizontal current density, and associated resistive heating rate per unit volume Q. The spikes appear as increases by orders of magnitude above background values in neutral line regions (NLRs) of active regions (ARs). The largest spikes typically occur a few hours to a few days prior to M or X flares. The spikes correspond to large vertical derivatives of the horizontal magnetic field. The model takes as input the photospheric magnetic field observed by the Helioseismic & Magnetic Imager (HMI) on the Solar Dynamics Observatory (SDO) satellite. This 2.5 D field is used to determine an analytic expression for a 3 D magnetic field, from which the current density, vector potential, and electric field are computed in every AR pixel for 14 ARs. The field is not assumed to be force-free. The spurious 6, 12, and 24 hour Doppler periods due to SDO orbital motion are filtered out of the time series of the HMI magnetic field for each pixel using a band pass filter. The subset of spikes analyzed at the pixel level are found to occur on HMI and granulation scales of 1 arcsec and 12 minutes. Spikes are found in ARs with and without M or X flares, and outside as well as inside NLRs, but the largest spikes are localized in the NLRs of ARs with M or X flares. The energy to drive the heating associated with the largest current spikes comes from bulk flow kinetic energy, not the electromagnetic field, and the current density is highly non-force free. The results suggest that, in combination with the model, HMI is revealing strong, convection driven, non-force free heating events on granulation scales, and that it is plausible these events are correlated with subsequent M or X flares. More and longer time series need to be analyzed to determine if such a correlation exists. Above an AR dependent threshold value of Q, the number of events N(Q) with heating rates greater than or equal to Q obeys a scale invariant power law distribution for each AR given by N(Q) varies Q(sup -s), where 0.40 less than or equal to S less than or equal to 0.53, with a mean and standard deviation across the 14 ARs of 0.47 and 0.045, showing there is little variation of S from one AR to another. These properties of N(Q) are in close agreement with those of the distribution N(E) for the total energy E of solar flares, determined from observations to be N(E) = constant x E(sup -alpha). From observations of nanoflares in the 0.7 to 4 MK range, and from observations of flares in hard X-rays, it is found that 0.51 less than or equal to alpha less than or equal to 0.57, and 0.4 less than or equal to alpha less than or equal to 0.6, respectively (Crosby et al. 1993, Sol. Phys., 143, 275; Aschwanden & Parnell 2002, ApJ, 572, 1048). Observations also show that, as is found here for the exponent S, there is little variation of alpha with AR (Wheatland 2000, ApJ, 532, 1209), indicating N(E) and N(Q) are largely independent of individual properties of ARs such as area, total magnetic flux, and distribution of current density (i.e. non-potentiality). Therefore the power law scaling of the photospheric heating rate Q computed here on granulation scales is essentially identical to that found for coronal observations of flare energies on scales 1-2 orders of magnitude larger. This suggests the physical mechanisms that cause Q and coronal flares are closely related. It seems likely that Q is the signature of a magnetic reconnection process in an energy range and volume orders of magnitude smaller than those of flares. In this context, at least the larger spikes in Q might be signatures of UV photospheric or lower chromospheric bombs in which plasma is heated to temperatures approximately 10(exp -5) K (Peter et al. 2014, Science 346, 1255726; Judge 2015, ApJ, 808, 116). In addition, lattice based avalanche simulations of flare energy release predict 0.4 less than or equal to alpha less than or equal to 0.5, while analytic, fractal-diffusive self-organized criticality models predict 0.4 less than or equal to alpha less than or equal to 0.67, in excellent agreement with observations, and the results presented here (Aschwanden & Parnell 2002, ApJ, 572, 1048; Aschwanden 2012, A&A, 539, A2; Aschwanden 2013, in "Self Organized Criticality Systems"; Aschwanden et al. 2016, SSR, 198, 47).

Goodman, Michael↗

Photospheric Current Spikes as Possible Predictors of Flares

Flares involve generation of the largest current densities in the solar atmosphere. This suggests the hypothesis that prior to a large (M,X) flare there are related time dependent changes in the photospheric current distribution, and hence in the resistive heating rate in neutral line regions (NLRs). If this is true, these changes might be useful predictors of flares. Evidence supporting this hypothesis is presented. Results from a data driven, near photospheric, 3D magnetohydrodynamic type model suggest the model might be useful for predicting M and X flares several hours to several days in advance. The model takes as input the photospheric magnetic field observed by the Helioseismic & Magnetic Imager (HMI) on the Solar Dynamics Observatory (SDO) satellite. The model computes quantities in every active region (AR) pixel for 14 ARs, with spurious Doppler periods due to SDO orbital motion filtered out of the time series of the magnetic field for each pixel. Spikes in the NLR resistive heating rate Q, appearing as increases by orders of magnitude above background values in the time series of Q are found to occur, and appear to be correlated with the occurrence of M or X flares a few hours to a few days later. The subset of spikes analyzed at the pixel level are found to occur on HMI and granulation scales of 1 arcsec and 12 minutes. Spikes are found in NLRs with and without M or X flares, and outside as well as inside NLRs, but the largest spikes are localized in the NLRs of ARs with M or X flares, and associated with horizontal magnetic field strengths several hG, and vertical magnetic field strengths several orders of magnitude smaller, suggesting that the spikes are associated with current sheets.

Goodman, Michael L.↗

E-Nose Vapor Identification Based on Dempster-Shafer Fusion of Multiple Classifiers

Electronic nose (e-nose) vapor identification is an efficient approach to monitor air contaminants in space stations and shuttles in order to ensure the health and safety of astronauts. Data preprocessing (measurement denoising and feature extraction) and pattern classification are important components of an e-nose system. In this paper, a wavelet-based denoising method is applied to filter the noisy sensor measurements. Transient-state features are then extracted from the denoised sensor measurements, and are used to train multiple classifiers such as multi-layer perceptions (MLP), support vector machines (SVM), k nearest neighbor (KNN), and Parzen classifier. The Dempster-Shafer (DS) technique is used at the end to fuse the results of the multiple classifiers to get the final classification. Experimental analysis based on real vapor data shows that the wavelet denoising method can remove both random noise and outliers successfully, and the classification rate can be improved by using classifier fusion.

Li, Winston↗