Search NASA⌕ Search

SEARCH · Search NASA

Results for “prediction algorithms”

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 127 records · Page 7

Progress Towards Interpretable Machine Learning-based Disruption Predictors Across Tokamaks

In this paper we lay the groundwork for a robust cross-device comparison of data-driven disruption prediction algorithms on DIII-D and JET tokamaks. In order to consistently carry on a comparative analysis, we define physics-based indicators of disruption precursors based on temperature, density, and radiation profiles that are currently missing for DIII-D data. These profile-based indicators are shown to well-describe impurity accumulation events in both DIII-D and JET discharges that eventually disrupt. Thanks to the univariate analysis on the features used in such data-driven applications on both tokamaks, we are able to statistically highlight differences in the dominant disruption precursors: JET with its ITER-like wall is more prone to impurity accumulation events, while DIII-D is more subject to edge cooling mechanisms that destabilize dangerous MHD modes. Even though the analyzed datasets are characterized by such intrinsic differences, we show how data-driven algorithms trained on one device can be used to predict and interpret disruptive scenarios on the other. As long as the destabilizing precursors are diagnosed in a device-independent way, the knowledge that data-driven algorithms learn on one device can be used to explain a disruptive behavior on another device.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Implementation of EPID transit dosimetry based on a through-air dosimetry algorithm

Purpose: A method to perform transit dosimetry with an electronic portal imaging device (EPID) by extending the commercial implementation of a published through-air portal dose image (PDI) prediction algorithm Van Esch et al.[Radiother. Oncol. 71, 223-234 (2004)] is proposed and validated. A detailed characterization of the attenuation, scattering, and EPID response behind objects in the beam path is used to convert through-air PDIs into transit PDIs. Methods: The EPID detector response beyond a range of water equivalent thicknesses (0-35 cm) and field sizes (3x3 to 22.2x29.6 cm{sup 2}) was analyzed. A constant air gap between the phantom exit surface and the EPID was utilized. A model was constructed that accounts for the beam's attenuation along the central axis, the presence of phantom scattered radiation, the detector's energy dependent response, and the difference in EPID off-axis pixel response relative to the central pixel. The efficacy of the algorithm was verified by comparing predicted and measured PDIs for IMRT fields delivered through phantoms of increasing complexity. Results: The expression that converts a through-air PDI to a transit PDI is dependent on the object's thickness, the irradiated field size, and the EPID pixel position. Monte Carlo derived narrow-beam linear attenuation coefficients are used to model the decrease in primary fluence incident upon the EPID due to the object's presence in the beam. This term is multiplied by a factor that accounts for the broad beam scatter geometry of the linac-phantom-EPID system and the detector's response to the incident beam quality. A 2D Gaussian function that models the nonuniformity of pixel response across the EPID detector plane is developed. For algorithmic verification, 49 IMRT fields were repeatedly delivered to homogeneous slab phantoms in 5 cm increments. Over the entire set of measurements, the average area passing a 3%/3mm gamma criteria slowly decreased from 98% for no material in the beam to 96.7% for 35 cm of material in the beam. The same 49 fields were delivered to a heterogeneous slab phantom and on average, 97.1% of the pixels passed the gamma criteria. Finally, a total of 33 IMRT fields were delivered to the anthropomorphic phantom and on average, 98.1% of the pixels passed. The likelihood of good matches was independent of anatomical site. Conclusions: A prediction of the transit PDI behind a phantom or patient can be created for the purposes of treatment verification via an extension of the Van Esch through-air PDI algorithm. The results of the verification measurements through phantoms indicate that further investigation through patients during their treatments is warranted.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

STPMiner: A Highperformance Spatiotemporal Pattern Mining Toolbox

The volume of spatiotemporal data being generated from scientific simulations and observations from sensors is growing at an astronomical rate. This data explosion is going to pose three challenges to the existing data mining infrastructure: algorithmic, computational, and I/O. Over the years we have implemented several spatiotemporal data mining algorithms including: outliers/anomalies, colocation patterns, change patterns, clustering, classification, and prediction algorithms. In this paper we briefly discuss the core spatiotemporal pattern mining algorithms along with some of the computational and I/O challenges associated with the big data.

99 GENERAL AND MISCELLANEOUS↗

Estimating the Pyrolysis Kinetic Parameters of Coal, Biomass, and Their Blends: A Comparative Study

In this work, the pyrolysis kinetic parameters of two coal ranks (DECS-25 lignite and DECS-38 sub-bituminous), two biomass materials (corn stover and switchgrass), and their respective blends were investigated at various heating rates ranging between 5 and 40 °C/min using thermogravimetric analysis. Complex models for devolatilization of the feedstocks were solved for obtaining and predicting the global kinetic parameters. Distributed activation energy model (method 1) and matrix inversion algorithm (method 2) were utilized and compared for this purpose. The results indicate that the matrix inversion algorithm predicts the kinetic parameters such that the weight loss characteristics can be best represented for both single fuels as well as that of blended materials. The algorithm can also be used for determining the number of reactions occurring in the devolatilization temperature interval. The number of reactions occurring during the devolatilization of blended materials falls between those that occur during the devolatilization of single fuels and the number of reactions gradually decrease with increase of biomass concentration in the blend. In addition, weight loss characteristics of fuel blends at unknown heating rates can be effectively predicted within 1% error through the use of this algorithm.

01 COAL, LIGNITE, AND PEAT↗

Hybrid deep learning architecture for general disruption prediction across tokamaks

In this paper, we present a new deep learning disruption prediction algorithm based on important findings from explorative data analysis which effectively allows knowledge transfer from existing devices to new ones, thereby predicting disruptions using very limited disruptive data from the new devices. The explorative data analysis conducted via unsupervised clustering techniques confirms that time-sequence data are much better separators of disruptive and non-disruptive behavior than the instantaneous plasma state data with further advantageous implications for a sequence-based predictor. Based on such important findings, we have designed a new algorithm for multi-machine disruption prediction that achieves high predictive accuracy on the C-Mod (AUC=0.801), DIII-D (AUC=0.947) and EAST (AUC=0.973). tokamaks with limited hyperparameter tuning. Through numerical experiments, we show that boosted accuracy (AUC=0.959) is achieved on EAST predictions by including in the training only 20 disruptive discharges, thousands of non-disruptive discharges from EAST, and combining this with more than a thousand discharges from DIII-D and C-Mod. The improvement of predictive ability obtained by combining disruptive data from other devices is found to be true for all permutations of the three devices. Furthermore, by comparing the predictive performance of each individual numerical experiment, we find that non-disruptive data are machine-specific while disruptive data from multiple devices contain device-independent knowledge that can be used to inform predictions for disruptions occurring on a new device.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Hybrid deep learning architecture for general disruption prediction across tokamaks

In this paper, we present a new deep learning disruption prediction algorithm based on important findings from explorative data analysis which effectively allows knowledge transfer from existing devices to new ones, thereby predicting disruptions using very limited disruptive data from the new devices. Here, the explorative data analysis conducted via unsupervised clustering techniques confirms that time-sequence data are much better separators of disruptive and non-disruptive behavior than the instantaneous plasma state data with further advantageous implications for a sequence-based predictor. Based on such important findings, we have designed a new algorithm for multi-machine disruption prediction that achieves high predictive accuracy on the C-Mod (AUC=0.801), DIII-D (AUC=0.947) and EAST (AUC=0.973) tokamaks with limited hyperparameter tuning. Through numerical experiments, we show that boosted accuracy (AUC=0.959) is achieved on EAST predictions by including in the training only 20 disruptive discharges, thousands of non-disruptive discharges from EAST, and combining this with more than a thousand discharges from DIII-D and C-Mod. The improvement of predictive ability obtained by combining disruptive data from other devices is found to be true for all permutations of the three devices. Furthermore, by comparing the predictive performance of each individual numerical experiment, we find that non-disruptive data are machine-specific while disruptive data from multiple devices contain device-independent knowledge that can be used to inform predictions for disruptions occurring on a new device.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Progress Toward Interpretable Machine Learning–Based Disruption Predictors Across Tokamaks

Here in this paper we lay the groundwork for a robust cross-device comparison of data-driven disruption prediction algorithms on DIII-D and JET tokamaks. In order to consistently carry on a comparative analysis, we define physics-based indicators of disruption precursors based on temperature, density, and radiation profiles that are currently not used in many other machine learning predictors for DIII-D data. These profile-based indicators are shown to well-describe impurity accumulation events in both DIII-D and JET discharges that eventually disrupt. The univariate analysis of the features used as input signals in the data-driven algorithms applied on the data of both tokamaks statistically highlights the differences in the dominant disruption precursors. JET with its ITER-like wall is more prone to impurity accumulation events, while DIII-D is more subject to edge-cooling mechanisms that destabilize dangerous magnetohydrodynamic modes. Even though the analyzed data sets are characterized by such intrinsic differences, we show through a few examples that the inclusion of physics-based disruption markers in data-driven algorithms is a promising path toward the realization of a uniform framework to predict and interpret disruptive scenarios across different tokamaks. As long as the destabilizing precursors are diagnosed in a device-independent way, the knowledge that data-driven algorithms learn on one device can be re-used to explain a disruptive behavior on another device.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Determining flow directions in river channel networks using planform morphology and topology

Abstract. The abundance of global, remotely sensed surface water observations has accelerated efforts toward characterizing and modeling how water moves across the Earth's surface through complex channel networks. In particular, deltas and braided river channel networks may contain thousands of links that route water, sediment, and nutrients across landscapes. In order to model flows through channel networks and characterize network structure, the direction of flow for each link within the network must be known. In this work, we propose a rapid, automatic, and objective method to identify flow directions for all links of a channel network using only remotely sensed imagery and knowledge of the network's inlet and outlet locations. We designed a suite of direction-predicting algorithms (DPAs), each of which exploits a particular morphologic characteristic of the channel network to provide a prediction of a link's flow direction. DPAs were chained together to create “recipes”, or algorithms that set all the flow directions of a channel network. Separate recipes were built for deltas and braided rivers and applied to seven delta and two braided river channel networks. Across all nine channel networks, the recipe-predicted flow directions agreed with expert judgement for 97 % of all tested links, and most disagreements were attributed to unusual channel network topologies that can easily be accounted for by pre-seeding critical links with known flow directions. Our results highlight the (non)universality of process–form relationships across deltas and braided rivers.

54 ENVIRONMENTAL SCIENCES↗

Development of a Real-Time Mobility Control and Visualization System with Predictive Vehicle Speed Control for Connected and Automated Vehicles (CAVs)

The Oak Ridge National Laboratory is developing a cloud-based platform for the deployment and evaluation of mobility- and efficiency-oriented connected and automated vehicles (CAVs) applications that will include fundamental data management, communications and visualization elements. This system will serve as a data portal, providing indirect communications that are relevant to efficiency applications, and can be used for both real world CAVs deployments and virtual systems that allow evaluation of new CAVs applications and algorithms. Predictive vehicle speed control algorithms have been developed and will be integrated with the real-time queryable transportation database and geospatial platform to achieve optimized energy efficiency and improved traffic flow in off-freeway driving. This paper describes the technology and summarizes evaluations being conducted for a corridor in Chattanooga, Tennessee, where the technology is expected to be initially deployed. To evaluate the fuel savings resulting from the vehicle speed control and the RTMCS data framework, vehicle powertrain simulations will be completed using results from a calibrated microscopic traffic model.

Laclair, Tim↗

TU-F-17A-07: Real-Time Personalized Margins

Purpose: To maximize normal tissue sparing for treatments requiring motion encompassing margins. Motion mitigation techniques including DMLC or couch tracking can freeze tumor motion within the treatment aperture potentially allowing for smaller treatment margins and thus better sparing of normal tissue. To enable for a safe application of this concept in the clinic we propose adapting margins dynamically in real-time during radiotherapy delivery based on personalized tumor localization confidence. To demonstrate technical feasibility we present a phantom study. Methods: We utilize a realistic anthropomorphic dynamic thorax phantom with a lung tumor model embedded close to the spine. The tumor, a 3D-printout of a patient's GTV, is moved 15mm peak-to-peak by diaphragm compression and monitored by continuous EPID imaging in real-time. Two treatment apertures are created for each beam, one representing ITV -based and the other GTV-based margin expansion. A soft tissue localization (STiL) algorithm utilizing the continuous EPID images is employed to freeze tumor motion within the treatment aperture by means of DMLC tracking. Depending on a tracking confidence measure (TCM), the treatment aperture is adjusted between the ITV and the GTV leaf. Results: We successfully demonstrate real-time personalized margin adjustment in a phantom study. We measured a system latency of about 250 ms which we compensated by utilizing a respiratory motion prediction algorithm (ridge regression). With prediction in place we observe tracking accuracies better than 1mm. For TCM=0 (as during startup) an ITV-based treatment aperture is chosen, for TCM=1 a GTV-based aperture and for 0<TCM<1 a linear interpolation between the two apertures is chosen. Conclusions: We have proposed and demonstrated a concept to safely shrink treatment margins during radiotherapy and adapt to tracking confidence in real-time. Normal tissue sparing is maximized. The worst case scenario results in delivering a plan with standard margins used in the clinic today.

60 APPLIED LIFE SCIENCES↗

SU-F-SPS-11: The Dosimetric Comparison of Truebeam 2.0 and Cyberknife M6 Treatment Plans for Brain SRS Treatment

Purpose: Brain stereotactic radiosurgery involves the use of precisely directed, single session radiation to create a desired radiobiologic response within the brain target with acceptable minimal effects on surrounding structures or tissues. In this study, the dosimetric comparison of Truebeam 2.0 and Cyberknife M6 treatment plans were made. Methods: For Truebeam 2.0 machine, treatment planning were done using 2 full arc VMAT technique with 6 FFF beam on the CT scan of Randophantom simulating the treatment of sterotactic treatments for one brain metastasis. The dose distribution were calculated using Eclipse treatment planning system with Acuros XB algorithm. The treatment planning of the same target were also done for Cyberknife M6 machine with Multiplan treatment planning system using Monte Carlo algorithm. Using the same film batch, the net OD to dose calibration curve was obtained using both machine by delivering 0- 800 cGy. Films were scanned 48 hours after irradiation using an Epson 1000XL flatbed scanner. Dose distribution were measured using EBT3 film dosimeter. The measured and calculated doses were compared. Results: The dose distribution in the target and 2 cm beyond the target edge were calculated on TPSs and measured using EBT3 film. For cyberknife plans, the gamma analysis passing rates between measured and calculated dose distributions were 99.2% and 96.7% for target and peripheral region of target respectively. For Truebeam plans, the gamma analysis passing rates were 99.1% and 95.5% for target and peripheral region of target respectively. Conclusion: Although, target dose distribution calculated accurately by Acuros XB and Monte Carlo algorithms, Monte carlo calculation algorithm predicts dose distribution around the peripheral region of target more accurately than Acuros algorithm.

60 APPLIED LIFE SCIENCES↗

A Sensitivity Study of the Aircraft Vortex Spacing System (AVOSS) Wake Predictor Algorithm to the Resolution of Input Meteorological Profiles

The AVOSS project demonstrated the feasibility of applying aircraft wake vortex sensing and prediction technologies to safe aircraft spacing for single runway arrivals. On average, AVOSS provided spacing recommendations that were less than the current FAA prescribed spacing rules, resulting in a potential airport efficiency gain. Subsequent efforts have included quantifying the operational specifications for future Wake Vortex Advisory Systems (WakeVAS). In support of these efforts, each of the candidate subsystems for a WakeVAS must be specified. The specifications represent a consensus between the high-level requirements and the capabilities of the candidate technologies. This report documents the beginnings of an effort to quantify the capabilities of the AVOSS Prediction Algorithm (APA). Specifically, the APA horizontal position and circulation strength output sensitivity to the resolution of its wind and turbulence inputs is examined. The results of this analysis have implications for the requirements of the meteorological sensing and prediction systems comprising a WakeVAS implementation.

Rutishauser, David K.↗

Design, Development, and Testing of a UAV Hardware-in-the-Loop Testbed for Aviation and Airspace Prognostics Research

The airspace is becoming more and more complicated, and will continue to do so in the future with the integration of Unmanned Aerial Vehicles (UAVs), autonomy, spacecraft, other forms of aviation technology into the airspace. The new technology and complexity increases the importance and difficulty of safety assurance. Additionally, testing new technologies on complex aviation systems & systems of systems can be very difficult, expensive, and sometimes unsafe in real life scenarios. Prognostic methodology provides an estimate of the health and risks of a component, vehicle, or airspace and knowledge of how that will change over time. That measure is especially useful in safety determination, mission planning, and maintenance scheduling. The developed testbed will be used to validate prediction algorithms for the real-time safety monitoring of the National Airspace System (NAS) and the prediction of unsafe events. The framework injects flight related anomalies related to ground systems, routing, airport congestion, etc. to test and verify algorithms for NAS safety. In our research work, we develop a live, distributed, hardware-in-the-loop testbed for aviation and airspace prognostics along with exploring further research possibilities to verify and validate future algorithms for NAS safety. The testbed integrates virtual aircraft using the X-Plane simulator and X-PlaneConnect toolbox, UAVs using onboard sensors and cellular communications, and hardware in the loop components. In addition, the testbed includes an additional research framework to support and simplify future research activities. It enables safe, accurate, and inexpensive experimentation and research into airspace and vehicle prognosis that would not have been possible otherwise. This paper describes the design, development, and testing of this system. Software reliability, safety and latency are some of the critical design considerations in development of the testbed. Integration of HITL elements in the development phases and veri cation/ validation are key elements to this report.

HITL↗

Prodigal: Prokaryotic Gene Recognition and Translation Initiation Site Identification

Background The quality of automated gene prediction in microbial organisms has improved steadily over the past decade, but there is still room for improvement. Increasing the number of correct identifications, both of genes and of the translation initiation sites for each gene, and reducing the overall number of false positives, are all desirable goals. Results With our years of experience in manually curating genomes for the Joint Genome Institute, we developed a new gene prediction algorithm called Prodigal (PROkaryotic DYnamic programming Gene-finding ALgorithm). With Prodigal, we focused specifically on the three goals of improved gene structure prediction, improved translation initiation site recognition, and reduced false positives. We compared the results of Prodigal to existing gene-finding methods to demonstrate that it met each of these objectives. Conclusion We built a fast, lightweight, open source gene prediction program called Prodigal (http://compbio.ornl.gov/prodigal/). Prodigal achieved good results compared to existing methods, and we believe it will be a valuable asset to automated microbial annotation pipelines.

59 BASIC BIOLOGICAL SCIENCES↗

Prodigal: prokaryotic gene recognition and translation initiation site identification

The quality of automated gene prediction in microbial organisms has improved steadily over the past decade, but there is still room for improvement. Increasing the number of correct identifications, both of genes and of the translation initiation sites for each gene, and reducing the overall number of false positives, are all desirable goals. With our years of experience in manually curating genomes for the Joint Genome Institute, we developed a new gene prediction algorithm called Prodigal (PROkaryotic DYnamic programming Gene-finding ALgorithm). With Prodigal, we focused specifically on the three goals of improved gene structure prediction, improved translation initiation site recognition, and reduced false positives. We compared the results of Prodigal to existing gene-finding methods to demonstrate that it met each of these objectives. We built a fast, lightweight, open source gene prediction program called Prodigal http://compbio.ornl.gov/prodigal/. Prodigal achieved good results compared to existing methods, and we believe it will be a valuable asset to automated microbial annotation pipelines. The goals of Prodigal were to attain greater sensitivity in identifying existing genes, to predict translation initiation sites more accurately, and to minimize the number of false positive predictions. The results of Prodigal were compared to existing methods for both purely experimentally verified genes as well as curated Genbank files for a number of genomes. Prodigal's performance was found to be comparable or better to existing methods in the prediction of genes while also predicting fewer overall genes. In the prediction of translation initiation sites, Prodigal performed competitively with existing methods. Prodigal is currently already in use at many institutions, and it has been used to annotate all finished microbial genomes submitted to Genbank by DOE-JGI in 2008 and onward (a substantial percentage of the overall finished microbial genomes at NCBI). It is run regularly at NCBI alongside GenemarkHMM and Glimmer, and it has also been incorporated into the Swiss Institute of Bioinformatics microbial genomics browser. In conclusion, Prodigal should prove to be a valuable resource for genome annotation of either draft or finished microbial sequence.

59 BASIC BIOLOGICAL SCIENCES↗

Predicting Spacecraft Trajectories by the WeavEncke Method

A combination of methods is proposed of predicting spacecraft trajectories that possibly include multiple maneuvers and/or perturbing accelerations, with greater speed, accuracy, and repeatability than were heretofore achievable. The combination is denoted the WeavEncke method because it is based on unpublished studies by Jonathan Weaver of the orbit-prediction formulation of the noted astronomer Johann Franz Encke. Weaver evaluated a number of alternatives that arise within that formulation, arriving at an orbit-predicting algorithm optimized for complex trajectory operations. In the WeavEncke method, Encke's method of prediction of perturbed orbits is enhanced by application of modern numerical methods. Among these methods are efficient Kepler s-equation time-of-flight solutions and self-starting numerical integration with time as the independent variable. Self-starting numerical integration satisfies the requirements for accuracy, reproducibility, and efficiency (and, hence, speed). Self-starting numerical integration also supports fully analytic regulation of integration step sizes, thereby further increasing speed while maintaining accuracy.

Weaver, Jonathan K.↗

Protein structure prediction guided by crosslinking restraints – A systematic evaluation of the impact of the crosslinking spacer length

Recent development of high-resolution mass spectrometry (MS) instruments enables chemical crosslinking (XL) to become a high-throughput method for obtaining structural information about proteins. Restraints derived from XL-MS experiments have been used successfully for structure refinement and protein–protein docking. However, one formidable question is under which circumstances XL-MS data might be sufficient to determine a protein’s tertiary structure de novo? Answering this question will not only include understanding the impact of XL-MS data on sampling and scoring within a de novo protein structure prediction algorithm, it must also determine an optimal crosslinker type and length for protein structure determination. While a longer crosslinker will yield more restraints, the value of each restraint for protein structure prediction decreases as the restraint is consistent with a larger conformational space. In this study, the number of crosslinks and their discriminative power was systematically analyzed in silico on a set of 2055 non-redundant protein folds considering Lys–Lys, Lys–Asp, Lys–Glu, Cys–Cys, and Arg–Arg reactive crosslinkers between 1 and 60 Å. Depending on the protein size a heuristic was developed that determines the optimal crosslinker length. Next, simulated restraints of variable length were used to de novo predict the tertiary structure of fifteen proteins using the BCL::Fold algorithm. The results demonstrate that a distinct crosslinker length exists for which information content for de novo protein structure prediction is maximized. The sampling accuracy improves on average by 1.0 Å and up to 2.2 Å in the most prominent example. XL-MS restraints enable consistently an improved selection of native-like models with an average enrichment of 2.1.

Biochemistry & Molecular Biology↗