Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Real Data”

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 1,693 records · Page 94

Application of the airborne ocean color imager for commercial fishing

The objective of the investigation was to develop a commercial remote sensing system for providing near-real-time data (within one day) in support of commercial fishing operations. The Airborne Ocean Color Imager (AOCI) had been built for NASA by Daedalus Enterprises, Inc., but it needed certain improvements, data processing software, and a delivery system to make it into a commercial system for fisheries. Two products were developed to support this effort: the AOCI with its associated processing system and an information service for both commercial and recreational fisheries to be created by Spectro Scan, Inc. The investigation achieved all technical objectives: improving the AOCI, creating software for atmospheric correction and bio-optical output products, georeferencing the output products, and creating a delivery system to get those products into the hands of commercial and recreational fishermen in near-real-time. The first set of business objectives involved Daedalus Enterprises and also were achieved: they have an improved AOCI and new data processing software with a set of example data products for fisheries applications to show their customers. Daedalus' marketing activities showed the need for simplification of the product for fisheries, but they successfully marketed the current version to an Italian consortium. The second set of business objectives tasked Spectro Scan to provide an information service and they could not be achieved because Spectro Scan was unable to obtain necessary venture capital to start up operations.

Wrigley, Robert C.↗

Photometry on Structured Backgrounds: Local Pixel-wise Infilling by Regression

Photometric pipelines struggle to estimate both the flux and flux uncertainty for stars in the presence of structured backgrounds such as filaments or clouds. However, it is exactly stars in these complex regions that are critical to understanding star formation and the structure of the interstellar medium. We develop a method, similar to Gaussian process regression, which we term local pixel-wise infilling (LPI). Using a local covariance estimate, we predict the background behind each star and the uncertainty of that prediction in order to improve estimates of flux and flux uncertainty. We show the validity of our model on synthetic data and real dust fields. We further demonstrate that the method is stable even in the crowded field limit. While we focus on optical-IR photometry, this method is not restricted to those wavelengths. We apply this technique to the 34 billion detections in the second data release of the Dark Energy Camera Plane Survey. In addition to removing many >3σ outliers and improving uncertainty estimates by a factor of ~2–3 on nebulous fields, we also show that our method is well behaved on uncrowded fields. The entirely post-processing nature of our implementation of LPI photometry allows it to easily improve the flux and flux uncertainty estimates of past as well as future surveys.

79 ASTRONOMY AND ASTROPHYSICS↗

NASA JSC EV2 Intern Spring 2016 - Jennie Chung

Exploration Mission 2 (EM-2) is a mission to resume the manned exploration of the Solar System. This mission is the first crewed mission of NASA’s Orion on the Space Launch System. The target for EM-2 is to perform a flyby of a captured asteroid in lunar orbit, which NASA plans to launch in 2023. As an intern working with EV-2 – Avionics Systems Division in Johnson Space Center, we are developing flight instrumentation systems for EM-2 (MISL & RFID). The Modular Integrated Stackable Layer (MISL) is a compact space-related computer system that is modular, scalable and reconfigurable. The RFID (radio frequency identification) sensors are used to take lower frequency (TC) type measurements and be able to stream data real-time to an RF (radio frequency) interrogator upon demand. Our job, in EV-2, is to certify, test, manufacture/assemble and deliver flight EM-2 DFI System (MISL & RFID). Our goal is to propose a development effort to design low-mass wire and wireless data acquisition and sensor solutions for EM-2 DFI (Development Flight Instrumentation). The team is tasked to provide the most effective use of 75 pounds to acquire DFI data and to collect sensor data for 100-200 high priority DFI channels (mass driven).

Chung, Jennie↗

The power of event-driven analytics in Large Scale Data Processing

FeedZai is a software company specialized in creating high-­‐throughput low-­‐latency data processing solutions. FeedZai develops a product called "FeedZai Pulse" for continuous event-­‐driven analytics that makes application development easier for end users. It automatically calculates key performance indicators and baselines, showing how current performance differ from previous history, creating timely business intelligence updated to the second. The tool does predictive analytics and trend analysis, displaying data on real-­‐time web-­‐based graphics. In 2010 FeedZai won the European EBN Smart Entrepreneurship Competition, in the Digital Models category, being considered one of the "top-­‐20 smart companies in Europe". The main objective of this seminar/workshop is to explore the topic for large-­‐scale data processing using Complex Event Processing and, in particular, the possible uses of Pulse in the scope of the data processing needs of CERN. Pulse is available as open-­‐source and can be licensed both for non-­‐commercial and commercial applications. FeedZai is interested in exploring possible synergies with CERN in high-­‐volume low-­‐latency data processing applications. The seminar will be structured in two sessions, the first one being aimed to expose the general scope of FeedZai's activities, and the second focused on Pulse itself: 10:00-11:00 FeedZai and Large Scale Data Processing; Introduction to FeedZai; FeedZai Pulse and Complex Event Processing; Demonstration; Use-Cases and Applications; Conclusion and Q&A. 11:00-11:15 Coffee break 11:15-12:30 FeedZai Pulse Under the Hood; A First FeedZai Pulse Application; PulseQL overview; Defining KPIs and Baselines; Conclusion and Q&A. About the speakers Nuno Sebastião is the CEO of FeedZai. Having worked for many years for the European Space Agency (ESA), he was responsible the overall design and development of Satellite Simulation Infrastructure of the agency. Having left ESA to found FeedZai, Nuno is currently responsible for the whole operations of the company. Nuno holds an M.Eng. in Informatics Engineering for the University of Coimbra, and an MBA from the London Business School. Paulo Marques is the CTO of FeedZai, being responsible for product development. Paulo is an Assistant Professor at the University of Coimbra, in the area of Distributed Data Processing, and an Adjunct Associated Professor at Carnegie Mellon, in the US. In the past Paulo lead a large number of projects for institutions like the ESA, Microsoft Research, SciSys, Siemens, among others, being now fully dedicated to FeedZai. Paulo holds a Ph.D. in Distributed Systems from the University of Coimbra.

97 MATHEMATICS AND COMPUTING↗

Managing Waste Inventory and License Limits at the Perma-Fix Northwest Facility to Meet CH2M Hill Plateau Remediation Company (CHPRC) American Recovery and Reinvestment Act (ARRA) Deliverables - 12335

CH2M Hill Plateau Remediation Company (CHRPC) is a prime contractor to the U.S. Department of Energy (DOE) focused on the largest ongoing environmental remediation project in the world at the DOE Hanford Site Central Plateau, i.e. the DOE Hanford Plateau Remediation Contract. The East Tennessee Materials and Energy Corporation (M and EC); a wholly owned subsidiary of Perma-Fix Environmental Services, Inc. (PESI), is a small business team member to CHPRC. Our scope includes project management; operation and maintenance of on-site storage, repackaging, treatment, and disposal facilities; and on-site waste management including waste receipt from generators and delivery to on-site and off-site treatment, storage, and disposal facilities. As part of this scope, M and EC staffs the centralized Waste Support Services organization responsible for all waste characterization and acceptance required to support CHPRC and waste generators across the Hanford Site. At the time of the CHPRC contract award (August 2008) slightly more than 9,000 cubic meters (m{sup 3}) of legacy waste was defined as 'no-path-forward waste'. A significant portion of this waste (7,650 m{sup 3}) comprised wastes with up to 50 grams of special nuclear materials (SNM) in oversized packages recovered during retrieval operations and large glove boxes removed from the Plutonium Finishing Plant (PFP). Through a collaborative effort between the DOE, CHPRC, and Perma-Fix Environmental Services, Inc. (PESI), pathways for these problematic wastes were developed that took advantage of commercial treatment capabilities at a nearby vendor facility, Perma-Fix Northwest (PFNW). In the spring of 2009, CHPRC initiated a pilot program under which they began shipping large package, low gram suspect TRU (<15 g SNM per container), and large package contact and remote handled MLLW to the off-site PFNW facility for treatment. PFNW is restricted by the SNM limits set for the total quantity of SNM allowed at the facility in accordance with the facility's radioactive materials license(s) (RML). While both CHPRC and PFNW maintain waste databases to track all waste movements, it became evident early in the process that a tool was needed that married the two systems to better track SNM inventories and sequence waste from the point of generation, through the PFNW facility, and back to the Hanford site for final disposition. This tool, known as the Treatment Integration and Planning Tool (TIPT), has become a robust planning tool that provides real-time data to support compliant and efficient waste generation, transportation, treatment, and disposition. TIPT is developing into the next generation tool that will change the way in which legacy wastes, retrieval wastes and decontamination and decommissioning operations are conducted on the Plateau Remediation Contract (PRC). The real value of the TIPT is its predictive capability. It allows the W and FMP to map out optimal windows for processing waste through the PFNW facility, or through any process that is in some way resource limited. It allows project managers to identify and focus on problem areas before shipments are affected. It has been modified for use in broader applications to predict turnaround times and identify windows of opportunity for processing higher gram wastes through PFNW and to allow waste generators, site-wide, to accurately predict scope, cost, and schedule for waste generation to optimize processing and eliminate storage, double handling, and related costs and unnecessary safety risks. The TIPT addresses the years old problem of how to effectively predict not only what needs to be done, but when. 'When' is the key planning parameter that has been ignored by the generator and processor for many years, but has proven to be the most important parameter for both parties. While further refinement is a natural part of any development process, the current improvements on the TIPT have shown that prediction is a powerful consideration. Even in lean times expected for the foreseeable future, the improved TIPT continues to play a central role in managing our way through those times to assure facilities remain viable and available. It is recommended that other major remediation projects and waste processing facilities incorporate a tool such as TIPT to improve customer-commercial supplier communications and better optimization of resources. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

WE-AB-207B-07: Dose Cloud: Generating “Big Data” for Radiation Therapy Treatment Plan Optimization Research

Purpose: To provide a tool to generate large sets of realistic virtual patient geometries and beamlet doses for treatment optimization research. This tool enables countless studies exploring the fundamental interplay between patient geometry, objective functions, weight selections, and achievable dose distributions for various algorithms and modalities. Methods: Generating realistic virtual patient geometries requires a small set of real patient data. We developed a normalized patient shape model (PSM) which captures organ and target contours in a correspondence-preserving manner. Using PSM-processed data, we perform principal component analysis (PCA) to extract major modes of variation from the population. These PCA modes can be shared without exposing patient information. The modes are re-combined with different weights to produce sets of realistic virtual patient contours. Because virtual patients lack imaging information, we developed a shape-based dose calculation (SBD) relying on the assumption that the region inside the body contour is water. SBD utilizes a 2D fluence-convolved scatter kernel, derived from Monte Carlo simulations, and can compute both full dose for a given set of fluence maps, or produce a dose matrix (dose per fluence pixel) for many modalities. Combining the shape model with SBD provides the data needed for treatment plan optimization research. Results: We used PSM to capture organ and target contours for 96 prostate cases, extracted the first 20 PCA modes, and generated 2048 virtual patient shapes by randomly sampling mode scores. Nearly half of the shapes were thrown out for failing anatomical checks, the remaining 1124 were used in computing dose matrices via SBD and a standard 7-beam protocol. As a proof of concept, and to generate data for later study, we performed fluence map optimization emphasizing PTV coverage. Conclusions: We successfully developed and tested a tool for creating customizable sets of virtual patients suitable for large-scale radiation therapy optimization research.

60 APPLIED LIFE SCIENCES↗

Progress and plan of KSTAR plasma control system upgrade

The plasma control system (PCS) has been one of essential systems in annual KSTAR plasma campaigns: starting from a single-process version in 2008, extensive upgrades are done through the previous 7 years in order to achieve major goals of KSTAR performance enhancement. Here, major implementations are explained in this paper. In consequences of successive upgrades, the present KSTAR PCS is able to achieve ~48 s of 500 kA plasma pulses with full real-time shaping controls and real-time NB power controls. It has become a huge system capable of dealing with 8 separate categories of algorithms, 26 actuators directly controllable during the shot, and real-time data communication units consisting of +180 analog channels and +600 digital input/outputs through the reflective memory (RFM) network. The next upgrade of the KSTAR PCS is planned in 2015 before the campaign. An overview of the upgrade layout will be given for this paper. The real-time system box is planned to use the CERN MRG-Realtime OS, an ITER-compatible standard operating system. New hardware is developed for faster real-time streaming system for future installations of actuators/diagnostics.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Detecting optical transients using artificial neural networks and reference images from different surveys

ABSTRACT We present a technique to detect optical transients based on an artificial neural networks method. We describe the architecture of two networks capable of comparing images of the same part of the sky taken by different telescopes. One image corresponds to the epoch in which a potential transient could exist; the other is a reference image of an earlier epoch. We use data obtained by the Dr. Cristina V. Torres Memorial Astronomical Observatory and archival reference images from the Sloan Digital Sky Survey. We trained a convolutional neural network and a dense layer network on simulated source samples and then tested the trained networks on samples created from real image data. Autonomous detection methods replace the standard process of detecting transients, which is normally achieved by source extraction of a difference image followed by human inspection of the detected candidates. Replacing the human inspection component with an entirely autonomous method would allow for a rapid and automatic follow-up of interesting targets of opportunity. The toy-model pipeline that we present here is not yet able to replace human inspection, but it might provide useful hints to identify potential candidates. The method will be further expanded and tested on telescopes participating in the Transient Optical Robotic Observatory of the South Collaboration.

79 ASTRONOMY AND ASTROPHYSICS↗

CNN-Based Phase Fault Classification in Real and Simulated Power Systems Data

This study proposes a convolutional neural network (CNN)–based two-step phase fault detection and identification method to classify anomalies in the power grid signal. Specifically, the first step checks the fault’s existence and determines the need for the second step. Subsequently, in the case of anomalies in the power grid signal, the second step identifies the type of fault, including line-to-line, single-line-to-ground, double-line-to-ground, and triple-line. Accordingly, the CNN architecture is both designed for the classification layers and trained with simulated data. To provide maximum prediction accuracy with minimum processing time, this study investigates the combinations of various feature extraction (FE) techniques, such as fast Fourier transform (FFT), amplitude and phase (AP), auto-correlation function, power spectral density, and wavelet transform (WT). Consequently, simulated and real-world results demonstrate that the proposed two-step method outperforms conventional one-step techniques, with the best performance obtained by using the combination of AP-AP, AP-WT, FFT-AP, and FFT-WT–based FE methods.

Alaca, Ozgur↗

The role of the host in a cooperating mainframe and workstation environment, volumes 1 and 2

In recent years, advancements made in computer systems have prompted a move from centralized computing based on timesharing a large mainframe computer to distributed computing based on a connected set of engineering workstations. A major factor in this advancement is the increased performance and lower cost of engineering workstations. The shift to distributed computing from centralized computing has led to challenges associated with the residency of application programs within the system. In a combined system of multiple engineering workstations attached to a mainframe host, the question arises as to how does a system designer assign applications between the larger mainframe host and the smaller, yet powerful, workstation. The concepts related to real time data processing are analyzed and systems are displayed which use a host mainframe and a number of engineering workstations interconnected by a local area network. In most cases, distributed systems can be classified as having a single function or multiple functions and as executing programs in real time or nonreal time. In a system of multiple computers, the degree of autonomy of the computers is important; a system with one master control computer generally differs in reliability, performance, and complexity from a system in which all computers share the control. This research is concerned with generating general criteria principles for software residency decisions (host or workstation) for a diverse yet coupled group of users (the clustered workstations) which may need the use of a shared resource (the mainframe) to perform their functions.

Kusmanoff, Antone↗

Signal Analysis of Helicopter Blade-Vortex-Interaction Acoustic Noise Data

Blade-Vortex-Interaction (BVI) produces annoying high-intensity impulsive noise. NASA Ames collected several sets of BVI noise data during in-flight and wind tunnel tests. The goal of this work is to extract the essential features of the BVI signals from the in-flight data and examine the feasibility of extracting those features from BVI noise recorded inside a large wind tunnel. BVI noise generating mechanisms and BVI radiation patterns an are considered and a simple mathematical-physical model is presented. It allows the construction of simple synthetic BVI events that are comparable to free flight data. The boundary effects of the wind tunnel floor and ceiling are identified and more complex synthetic BVI events are constructed to account for features observed in the wind tunnel data. It is demonstrated that improved recording of BVI events can be attained by changing the geometry of the rotor hub, floor, ceiling and microphone. The Euclidean distance measure is used to align BVI events from each blade and improved BVI signals are obtained by time-domain averaging the aligned data. The differences between BVI events for individual blades are then apparent. Removal of wind tunnel background noise by optimal Wiener-filtering is shown to be effective provided representative noise-only data have been recorded. Elimination of wind tunnel reflections by cepstral and optimal filtering deconvolution is examined. It is seen that the cepstral method is not applicable but that a pragmatic optimal filtering approach gives encouraging results. Recommendations for further work include: altering measurement geometry, real-time data observation and evaluation, examining reflection signals (particularly those from the ceiling) and performing further analysis of expected BVI signals for flight conditions of interest so that microphone placement can be optimized for each condition.

Rogers, James C.↗

Hardware in-the-Loop Demonstration of Real-Time Orbit Determination in High Earth Orbits

This paper presents results from a study conducted at Goddard Space Flight Center (GSFC) to assess the real-time orbit determination accuracy of GPS-based navigation in a number of different high Earth orbital regimes. Measurements collected from a GPS receiver (connected to a GPS radio frequency (RF) signal simulator) were processed in a navigation filter in real-time, and resulting errors in the estimated states were assessed. For the most challenging orbit simulated, a 12 hour Molniya orbit with an apogee of approximately 39,000 km, mean total position and velocity errors were approximately 7 meters and 3 mm/s respectively. The study also makes direct comparisons between the results from the above hardware in-the-loop tests and results obtained by processing GPS measurements generated from software simulations. Care was taken to use the same models and assumptions in the generation of both the real-time and software simulated measurements, in order that the real-time data could be used to help validate the assumptions and models used in the software simulations. The study makes use of the unique capabilities of the Formation Flying Test Bed at GSFC, which provides a capability to interface with different GPS receivers and to produce real-time, filtered orbit solutions even when less than four satellites are visible. The result is a powerful tool for assessing onboard navigation performance in a wide range of orbital regimes, and a test-bed for developing software and procedures for use in real spacecraft applications.

Moreau, Michael↗

Earth Independent Medical Operations (EIMO) DATASCOPE Technical Interchange Meeting 21st August 2023: Background and Summary of Discussion

An aspiration for EIMO datascope is to realize artificial intelligence-enhanced solutions for analysis of crew health & performance data and to facilitate clinical decision support for autonomous medical operations. A vision proposed to the meeting participants was that of a “system of systems,” whereby EIMO will utilize AI-supported natural language processing and machine learning techniques to synthesize embedded reference databases and real-time data streams [input vectors] from multiple data sources to continuously and seamlessly assess crew health & performance. Constituent input vectors may include environmental controls, countermeasures data, behavioral data, physiologic wearables, point-of-care laboratory tests, personalized medical records, inventory trade space risk assessments, COTS medical databases, and ground support inputs. An ideal AI capability would possess trained fusion algorithms to cross reference input vectors with medical ‘knowledge’ [cultivated database] to stratify relevant data streams for predictive and actionable capabilities. In addition, EIMO will ideally have a degree of mobility, in that it can be accessed and can push/pull data within and between multiple vehicles/habitats.

Artificial Intelligence↗

Feature Acquisition with Imbalanced Training Data

This work considers cost-sensitive feature acquisition that attempts to classify a candidate datapoint from incomplete information. In this task, an agent acquires features of the datapoint using one or more costly diagnostic tests, and eventually ascribes a classification label. A cost function describes both the penalties for feature acquisition, as well as misclassification errors. A common solution is a Cost Sensitive Decision Tree (CSDT), a branching sequence of tests with features acquired at interior decision points and class assignment at the leaves. CSDT's can incorporate a wide range of diagnostic tests and can reflect arbitrary cost structures. They are particularly useful for online applications due to their low computational overhead. In this innovation, CSDT's are applied to cost-sensitive feature acquisition where the goal is to recognize very rare or unique phenomena in real time. Example applications from this domain include four areas. In stream processing, one seeks unique events in a real time data stream that is too large to store. In fault protection, a system must adapt quickly to react to anticipated errors by triggering repair activities or follow- up diagnostics. With real-time sensor networks, one seeks to classify unique, new events as they occur. With observational sciences, a new generation of instrumentation seeks unique events through online analysis of large observational datasets. This work presents a solution based on transfer learning principles that permits principled CSDT learning while exploiting any prior knowledge of the designer to correct both between-class and withinclass imbalance. Training examples are adaptively reweighted based on a decomposition of the data attributes. The result is a new, nonparametric representation that matches the anticipated attribute distribution for the target events.

Thompson, David R.↗

Fuel consumption for various driving styles in conventional and hybrid electric vehicles: Integrating driving cycle predictions with fuel consumption optimization

Here, improving fuel economy and lowering emissions are key societal goals. Standard driving cycles, pre-designed by the US Environmental Protection Agency (EPA), have long been used to estimate vehicle fuel economy in laboratory-controlled conditions. They have also been used to test and tune different energy management strategies for hybrid electric vehicles (HEVs). This paper aims to estimate fuel consumption for a conventional vehicle and a HEV using personalized driving cycles extracted from real-world data to study the effects of different driving styles and vehicle types on fuel consumption when compared to the estimates based on standard driving cycles. To do this, we extracted driving cycles for conventional vehicles and HEVs from a large-scale U.S. survey that contains real-world GPS-based driving records. Next, the driving cycles were assigned to one of three categories: volatile, normal, or calm. Then, the driving cycles were used along with a driver-vehicle simulation that captures driver decisions (vehicle speed during a trip), powertrain, and vehicle dynamics to estimate fuel consumption for conventional vehicles and HEVs with power-split powertrain. To further optimize fuel consumption for HEVs, the Equivalent Consumption Minimization Strategy (ECMS) is applied. The results show that depending on the driving style and the driving scenario, conventional vehicle fuel consumption can vary widely compared with standard EPA driving cycles. Specifically, conventional vehicle fuel consumption was 13% lower in calm urban driving, but almost 34% higher for volatile highway driving compared with standard EPA driving cycles. Interestingly, when a driving cycle is predicted based on the application of case-based reasoning and used to tune the power distribution in a hybrid electric vehicle, its fuel consumption can be reduced by up to 12% in urban driving. Implications and limitations of the findings are discussed.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

DEEP CELLULAR RECURRENT NEURAL ARCHITECTURE FOR EFFICIENT MULTIDIMENSIONAL TIME-SERIES DATA PROCESSING

Efficient processing of time series data is a fundamental yet challenging problem in pattern recognition. Though recent developments in machine learning and deep learning have enabled remarkable improvements in processing large scale datasets in many application domains, most are designed and regulated to handle inputs that are static in time. Many real-world data, such as in biomedical, surveillance and security, financial, manufacturing and engineering applications, are rarely static in time, and demand models able to recognize patterns in both space and time. Current machine learning (ML) and deep learning (DL) models adapted for time series processing tend to grow in complexity and size to accommodate the additional dimensionality of time. Specifically, the biologically inspired learning based models known as artificial neural networks that have shown extraordinary success in pattern recognition, tend to grow prohibitively large and cumbersome in the presence of large scale multi-dimensional time series biomedical data such as EEG. Consequently, this work aims to develop representative ML and DL models for robust and efficient large scale time series processing. First, we design a novel ML pipeline with efficient feature engineering to process a large scale multi-channel scalp EEG dataset for automated detection of epileptic seizures. With the use of a sophisticated yet computationally efficient time-frequency analysis technique known as harmonic wavelet packet transform and an efficient self-similarity computation based on fractal dimension, we achieve state-of-the-art performance for automated seizure detection in EEG data. Subsequently, we investigate the development of a novel efficient deep recurrent learning model for large scale time series processing. For this, we first study the functionality and training of a biologically inspired neural network architecture known as cellular simultaneous recurrent neural network (CSRN). We obtain a generalization of this network for multiple topological image processing tasks and investigate the learning efficacy of the complex cellular architecture using several state-of-the?art training methods. Finally, we develop a novel deep cellular recurrent neural network (CDRNN) architecture based on the biologically inspired distributed processing used in CSRN for processing time series data. The proposed DCRNN leverages the cellular recurrent architecture to promote extensive weight sharing and efficient, individualized, synchronous processing of multi-source time series data. Experiments on a large scale multi-channel scalp EEG, and a machine fault detection dataset show that the proposed DCRNN offers state-of-the-art recognition performance while using substantially fewer trainable recurrent units.

Vidyaratne, Lasitha S.↗

KGML-ag: a modeling framework of knowledge-guided machine learning to simulate agroecosystems: a case study of estimating N<sub>2</sub>O emission using data from mesocosm experiments

Abstract. Agricultural nitrous oxide (N2O) emission accounts for a non-trivial fraction of global greenhouse gas (GHG) budget. To date, estimating N2O fluxes from cropland remains a challenging task because the related microbial processes (e.g., nitrification and denitrification) are controlled by complex interactions among climate, soil, plant and human activities. Existing approaches such as process-based (PB) models have well-known limitations due to insufficient representations of the processes or uncertainties of model parameters, and due to leverage recent advances in machine learning (ML) a new method is needed to unlock the “black box” to overcome its limitations such as low interpretability, out-of-sample failure and massive data demand. In this study, we developed a first-of-its-kind knowledge-guided machine learning model for agroecosystems (KGML-ag) by incorporating biogeophysical and chemical domain knowledge from an advanced PB model, ecosys, and tested it by comparing simulating daily N2O fluxes with real observed data from mesocosm experiments. The gated recurrent unit (GRU) was used as the basis to build the model structure. To optimize the model performance, we have investigated a range of ideas, including (1) using initial values of intermediate variables (IMVs) instead of time series as model input to reduce data demand; (2) building hierarchical structures to explicitly estimate IMVs for further N2O prediction; (3) using multi-task learning to balance the simultaneous training on multiple variables; and (4) pre-training with millions of synthetic data generated from ecosys and fine-tuning with mesocosm observations. Six other pure ML models were developed using the same mesocosm data to serve as the benchmark for the KGML-ag model. Results show that KGML-ag did an excellent job in reproducing the mesocosm N2O fluxes (overall r2=0.81, and RMSE=3.6 mgNm-2d-1 from cross validation). Importantly, KGML-ag always outperforms the PB model and ML models in predicting N2O fluxes, especially for complex temporal dynamics and emission peaks. Besides, KGML-ag goes beyond the pure ML models by providing more interpretable predictions as well as pinpointing desired new knowledge and data to further empower the current KGML-ag. We believe the KGML-ag development in this study will stimulate a new body of research on interpretable ML for biogeochemistry and other related geoscience processes.

54 ENVIRONMENTAL SCIENCES↗

Mass spectrometric gas composition measurements associated with jet interaction tests in a high-enthalpy wind tunnel

Knowledge of test gas composition is important in wind-tunnel experiments measuring aerothermodynamic interactions. This paper describes measurements made by sampling the top of the test section during runs of the Langley 7-Inch High-Temperature Tunnel. The tests were conducted to determine the mixing of gas injected from a flat-plate model into a combustion-heated hypervelocity test stream and to monitor the CO2 produced in the combustion. The Mass Spectrometric (MS) measurements yield the mole fraction of N2 or He and CO2 reaching the sample inlets. The data obtained for several tunnel run conditions are related to the pressures measured in the tunnel test section and at the MS ionizer inlet. The apparent distributions of injected gas species and tunnel gas (CO2) are discussed relative to the sampling techniques. The measurements provided significant real-time data for the distribution of injected gases in the test section. The jet N2 diffused readily from the test stream, but the jet He was mostly entrained. The amounts of CO2 and Ar diffusing upward in the test section for several run conditions indicated the variability of the combustion-gas test-stream composition.

Lewis, B. W.↗