Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “random variables”

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 1,621 records · Page 90

[NASA/DOD Aerospace Knowledge Diffusion Research Project. Report 5:] Summary report to phase 1 respondents including frequency distributions

Phase 1 of a four part study was undertaken to investigate the use of scientific and technical information (STI) by U.S. aerospace engineers and scientists. Specific attention was paid to institutional and sociometric variables and to the step-by-step process of information gathering used by the respondents. Data were collected by means of three self-administered mail-back questionnaires. The approximately 34,000 members of the American Institute of Aeronautics and Astronautics served as the study population. More than 65 percent of the randomly selected respondants returned the questionnaires in each of the three groups. Respondants relied more heavily on informal sources of information than formal sources and turned to librarians and other technical information specialists only when they did not obtain results via informal means or their own formal searches. The report includes frequency distributions for the questions.

Pinelli, Thomas E.↗

Ultrasoft x ray bursts: A clue to the origin of gamma ray bursts?

A search was conducted for short timescale (less than or approx. 10 s) faint X ray bursts in the complete 201,000,000 photon Einstein Observatory imaging proportional counter (IPC) data base. Some 73,000,000 events were extracted which were distributed among 11,230 observing intervals with an average observing time of approx. 1400 seconds. To exclude all sources, only those spatial pixels were searched with a count rate less than 6 x 10(exp -4) counts/sec sq arcmin. The mean count rate was much less. All files with less than 400 secs of observing time were excluded. The data was then binned into overlapping cells 4'.3 by 4'.3 by 10 secs in volume, and all the space-time cells were searched to locate cells with more than 5 counts in a 10 sec interval; i.e., a minimum flux enhancement of 50 over the mean rate for that point in space. Some 52 such events were found. An annulus was inspected around the event to see if it coincided with a general rise in the counter's event rate. Ten events did coincide with counter phenomena; these were associated with sunrise, sunset, or the approach of the South Atlantic Anomaly and were excluded. The remaining 42 events were then examined in an attempt to associate them with instrumental or environmental artifacts. The event occurrence times are uniformly distributed over the course of the mission and show no correlation with local solar time; their positions are randomly distributed in geographic, Galactic, and celestial coordinates. Their summed radial distribution is consistent with the point response function of the IPC for a soft source, indicating that the events are focused by the Observatory's mirror system. The integrated spectrum is, indeed, quite soft and is very similar to that of the cataclysmic variable star U Gem in outburst. Most of the events have risetimes of less than or approx. 1 sec and decay times ranging from 1 to 20 sec. Their positions are not associated with any class of catalogued objects. Several possible origins are discussed for the events; if they are associated with gamma ray bursters, they provide interesting constraints on the space density and repetition rate of the underlying source population.

Helfand, D. J.↗

Machine learning modeling and model predictive control of a closed-circuit reverse osmosis system

Closed-circuit reverse osmosis (CCRO) offers a flexible and energy-efficient alternative to conventional reverse osmosis by operating in a semi-batch mode that recycles brine, enabling higher recovery rates and reduced specific energy consumption (SEC). However, developing accurate, system-level dynamic models for CCRO remains challenging due to its nonlinear, multi-phase operation and sensitivity to variable feed water conditions. Traditional modeling approaches, such as NARMAX (nonlinear autoregressive moving average with exogenous inputs), often struggle to generalize across varying inlet feed concentrations, necessitating frequent parameter re-estimation and limiting their utility for real-time control applications. To address these limitations, we developed a long short-term memory (LSTM) neural network model trained on an extensive experimental data set from a CCRO pilot plant. The model accepts three inputs, feed flow rate, recirculation flow rate, and initial feed conductivity, and predicts three key outputs: reject conductivity, feed pump power draw, and recirculation pump power draw. We validated the LSTM model against experimental data, demonstrating its ability to distinguish between different feed conductivities and adapt to variable flow rates. Subsequently, we incorporated the LSTM model within a nonlinear model predictive control (MPC) scheme and conducted closed-loop simulations to optimize the integrated SEC (iSEC). In conclusion, the results project up to a 6% reduction in iSEC by using MPC to optimize performance over the entire experiment duration, without requiring any random excitation for data collection or parameter re-estimation.

Desalination↗

Five Year Results of US Intergroup/RTOG 9704 With Postoperative CA 19-9 {<=}90 U/mL and Comparison to the CONKO-001 Trial

Purpose: Radiation Therapy Oncology Group (RTOG) trial 9704 was the largest randomized trial to use adjuvant chemoradiation therapy for patients with pancreatic cancer. This report analyzes 5-year survival by serum level of tumor marker CA 19-9 of {<=}90 vs >90 U/mL and compares results to the those of the CONKO-001 trial. Methods and Materials: CA 19-9 expression was analyzed as a dichotomized variable ({<=}90 vs >90 U/mL). Cox proportional hazard models were used to identify the impact of the CA 19-9 value on overall survival (OS). Actuarial estimates of OS were calculated using the Kaplan-Meier method. Results: Both univariate (hazard ratio [HR] = 3.2; 95% confidence interval [CI], 2.3-4.3, P<.0001) and multivariate (HR = 3.1; 95% CI, 2.2-4.2, P<.0001) analyses demonstrated a statistically significant decrease in OS for CA 19-9 serum level of {>=}90 U/mL. For patients in the gemcitabine (Gem) treatment arm with CA 19-9 <90 U/mL, median survival was 21 months. For patients with CA 19-9 {>=}90 U/mL, this number dropped to 10 months. In patients with pancreatic head tumors in the Gem treatment arm with RT quality assurance per protocol and CA 19-9 of <90 U/mL, median survival and 5-year rate were 24 months and 34%. In comparison, the median survival and 5-year OS rate for patients in the Gem arm of the CONKO trial were 22 months and 21%. Conclusions: This analysis demonstrates that patients with postresection CA 19-9 values {>=}90 U/mL had a significantly worse survival. Patients with pancreatic head tumors treated with Gem with CA 19-9 serum level of <90 U/mL and per protocol RT had favorable survival compared to that seen in the CONKO trial. CA 19-9 is a stratification factor for the current RTOG adjuvant pancreas trial (0848).

62 RADIOLOGY AND NUCLEAR MEDICINE↗

A Physically-Based Correlation of Irradiation-Induced Transition Temperature Shifts for RPV Steels1

This paper presents a physically-based, empirically calibrated model for estimating irradiation-induced transition temperature shifts in reactor pressure vessel steels, based on a broader database and more complete understanding of embrittlement mechanisms than was available for earlier models. Brief descriptions of the underlying radiation damage mechanisms and the database are included, but the emphasis is on the model and the quality of its fit to U.S. power reactor surveillance data. The model is compared to a random sample of surveillance data that were set aside and not used in fitting and to selected independent data from test reactor irradiations, in both cases showing good ability to predict data that were not used for calibration. The model is a good fit to the surveillance data, with no significant residual error trends for variables included in the model or additional variables that could be included.

Eason, Ernest D.↗

AN ALTERNATIVE APPROACH TO MEASURING REVERBERATION LAGS IN ACTIVE GALACTIC NUCLEI

Motivated by recent progress in the statistical modeling of quasar variability, we develop a new approach to measuring emission-line reverberation lags to estimate the size of broad-line regions (BLRs) in active galactic nuclei. Assuming that all emission-line light curves are scaled, smoothed, and displaced versions of the continuum, this alternative approach fits the light curves directly using a damped random walk model and aligns them to recover the time lag and its statistical confidence limits. We introduce the mathematical formalism of this approach and demonstrate its ability to cope with some of the problems for traditional methods, such as irregular sampling, correlated errors, and seasonal gaps. We redetermine the lags for 87 emission lines in 31 quasars and reassess the BLR size-luminosity relationship using 60 H{beta} lags. We confirm the general results from the traditional cross-correlation methods, with a few exceptions. Our method, however, also supports a broad range of extensions. In particular, it can simultaneously fit multiple lines and continuum light curves which improves the lag estimate for the lines and provides estimates of the error correlations between them. Determining these correlations is of particular importance for interpreting emission-line velocity-delay maps. We can also include parameters for luminosity-dependent lags or line responses. We use this to detect the scaling of the BLR size with continuum luminosity in NGC 5548.

79 ASTRONOMY AND ASTROPHYSICS↗

Quantum benchmarks for the storage or transmission of quantum light from minimal resources

We investigate several recently published benchmark criteria for storage or transmission of continuous-variable quantum information. A comparison reveals that criteria based on a Gaussian distribution of coherent states are most resilient to noise. We then address the issue of experimental resources and derive an equally strong benchmark, solely based on three coherent states and homodyne detection. This benchmark is further simplified in the presence of naturally occurring random phases, which remove the need for active input-state modulation.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Discrete-Direct Model Calibration and Uncertainty Propagation Method Confirmed on Multi-Parameter Plasticity Model Calibrated to Sparse Random Field Data

A discrete direct (DD) model calibration and uncertainty propagation approach is explained and demonstrated on a 4-parameter Johnson-Cook (J-C) strain-rate dependent material strength model for an aluminum alloy. The methodology's performance is characterized in many trials involving four random realizations of strain-rate dependent material-test data curves per trial, drawn from a large synthetic population. The J-C model is calibrated to particular combinations of the data curves to obtain calibration parameter sets which are then propagated to “Can Crush” structural model predictions to produce samples of predicted response variability. These are processed with appropriate sparse-sample uncertainty quantification (UQ) methods to estimate various statistics of response with an appropriate level of conservatism. This is tested on 16 output quantities (von Mises stresses and equivalent plastic strains) and it is shown that important statistics of the true variabilities of the 16 quantities are bounded with a high success rate that is reasonably predictable and controllable. The DD approach has several advantages over other calibration-UQ approaches like Bayesian inference for capturing and utilizing the information obtained from typically small numbers of replicate experiments in model calibration situations—especially when sparse replicate functional data are involved like force–displacement curves from material tests. The DD methodology is straightforward and efficient for calibration and propagation problems involving aleatory and epistemic uncertainties in calibration experiments, models, and procedures.

42 ENGINEERING↗

Both the intratumoral immune and microbial microenvironment are linked to recurrence in human colon cancer: results from a prospective, multicenter nodal ultrastaging trial

Background: Colon cancer (CC) is the third most common cancer diagnosed in the United States and the incidence has been rising among young adults. We and others have shown a relationship between the immune infiltrate and prognosis, with improved disease-free survival (DFS) associated with a higher expression of CD8+ T cells. Additionally, numerous studies indicate the gut microbiota is linked to colon cancer and clinical outcomes. Therefore, we hypothesized a microbial signature might be associated with both the intratumoral immune cells as well as DFS. Results: Ninety-one patients were randomly selected from a prospective NCI-sponsored multicenter trial evaluating ultrastaging in CC to investigate the intratumoral microbiota by 16S rRNA gene amplicon sequencing. Operational taxonomic units (OTUs) were grouped by 97% sequence similarity. A series of clinical, immunohistochemical, and microbiota-related data were first evaluated by univariable cox regression to determine candidate variables associated with DFS. DFS was influenced by three parameters: N-stage, CD8+ labeling, and one microbiota principal component by multivariate analysis (MVA). Not only were the microbiota and CD8 significant contributors to the DFS model, but they were also significantly associated with each other. Alpha diversity showed an inverse correlation to CD8+ T cells (p=0.010, R=-0.278) and beta diversity showed an association with the CD8+ T cells (u-UniFrac p=0.026, w-UniFrac p=0.034). Further analysis at the OTU level with false discovery correction revealed one OTU, OTU_104, belonging to the order Clostridiales to be associated with increased recurrence (HR 1.21, CI 1.08 to 1.36). This OTU_104 was then found to be inversely correlated to CD8+ T cells (p=0.031, R=-0.35). Conclusions: This study is the first to demonstrate an association between the intratumoral microbiota, CD8+ T cells, and recurrence in CC. An increased relative abundance of a specific OTU_104 was inversely associated with CD8+ T cells and increased CC recurrence. The link between this microbe, CD8+ T cells and DFS has not been previously shown. Further studies are warranted to examine the role of infiltrating immune cells and the microbiota on colon cancer.

colon cancer↗

Optimization and Annual Average Power Predictions of a Backward Bent Duct Buoy Oscillating Water Column Device Using the Wells Turbine.

This Technical Report presents work completed by The Applied Research Laboratory at The Pennsylvania State University, in conjunction with Sandia National Labs, on the optimization of the power conversion chain (PCC) design to maximize the Average Annual Electric Power (AAEP) output of an Oscillating Water Column (OWC) device. The design consists of two independent stages. First, the design of a floating OWC, a Backward Bent Duct Buoy (BBDB), and second the design of the PCC. The pneumatic power output of the BBDB in random waves is optimized through the use of a hydrodynamically coupled, linear, frequency-domain, performance model that links the oscillating structure to internal air-pressure fluctuations. The PCC optimization is centered on the selection and sizing of a Wells Turbine and electric power generation equipment. The optimization of the PCC involves the following variables: the type of Wells Turbine (fixed or variable pitched, with and without guide vanes), the radius of the turbine, the optimal vent pressure, the sizing of the power electronics, and number of turbines. Also included in this Technical Report are further details on how rotor thrust and torque are estimated, along with further details on the type of variable frequency drive selected.

13 HYDRO ENERGY↗

Incidence and predictors of loss to follow-up among adult patients receiving antiretroviral therapy in Central Ethiopia: a multi-center retrospective cohort study

Background: Globally, loss to follow-up (LTFU) remains a significant public health concern despite the rapid expansion of antiretroviral medication programs. It is a significant cause of treatment failure and threatens the enhancement of HIV treatment outcomes among patients on antiretroviral therapy (ART). However, there is a paucity of evidence on its incidence and predictors in Ethiopia. Thus, this study aimed to examine the incidence and predictors of LTFU among adult HIV patients receiving ART at hospitals in Central Ethiopia. Methods: A multi-centered facility-based retrospective cohort study was conducted among 432 randomly selected adult patients who received antiretroviral therapy. Data were entered into EpiData version 3.1 and exported to Stata version 14 for analysis. The Kaplan–Meier failure function was employed to determine the overall failure estimates, and the log-rank test was used to compare the probability of failure among the different categories of variables. The Cox proportional hazard model was used to identify independent predictors of LTFU. Results: Overall, 172 (39.8%) study participants were lost to follow-up over the 10-year follow-up period with an incidence rate of 8.12 (95% CI: 7.11, 9.09) per 1,000 person-months. Undisclosed HIV status (AHR: 1.96, 95% CI: 1.14, 3.36), not able to work (AHR: 1.84, 95% CI: 1.13, 2.22), opportunistic infections (AHR: 3.13, 95% CI: 2.17, 4.52), CD4 < 200 cell/mL (AHR: 1.95, 95% CI: 1.18, 3.21), not receiving isoniazid preventive therapy (IPT) (AHR: 2.57, 95% CI: 1.62, 4.06), not participating in clubs (AHR: 1.68, 95% CI: 1.10, 2.22), side effects of drugs (AHR: 1.44, 95% CI: 1.02, 2.04), and high viral load (AHR: 3.15, 95% CI: 1.81, 5.47) were identified as significant predictors of loss to follow-up. Conclusion: In this study, the incidence of LTFU was high. The focus should be on creating awareness and prevention programs that aim to reduce loss to follow-up by continuing counseling, especially on the negative effects of loss to follow-up and the benefits of ART care.

60 APPLIED LIFE SCIENCES↗

Cascade Optimization Strategy for Aircraft and Air-Breathing Propulsion System Concepts

Design optimization for subsonic and supersonic aircraft and for air-breathing propulsion engine concepts has been accomplished by soft-coupling the Flight Optimization System (FLOPS) and the NASA Engine Performance Program analyzer (NEPP), to the NASA Lewis multidisciplinary optimization tool COMETBOARDS. Aircraft and engine design problems, with their associated constraints and design variables, were cast as nonlinear optimization problems with aircraft weight and engine thrust as the respective merit functions. Because of the diversity of constraint types and the overall distortion of the design space, the most reliable single optimization algorithm available in COMETBOARDS could not produce a satisfactory feasible optimum solution. Some of COMETBOARDS' unique features, which include a cascade strategy, variable and constraint formulations, and scaling devised especially for difficult multidisciplinary applications, successfully optimized the performance of both aircraft and engines. The cascade method has two principal steps: In the first, the solution initiates from a user-specified design and optimizer, in the second, the optimum design obtained in the first step with some random perturbation is used to begin the next specified optimizer. The second step is repeated for a specified sequence of optimizers or until a successful solution of the problem is achieved. A successful solution should satisfy the specified convergence criteria and have several active constraints but no violated constraints. The cascade strategy available in the combined COMETBOARDS, FLOPS, and NEPP design tool converges to the same global optimum solution even when it starts from different design points. This reliable and robust design tool eliminates manual intervention in the design of aircraft and of air-breathing propulsion engines where it eases the cycle analysis procedures. The combined code is also much easier to use, which is an added benefit. This paper describes COMETBOARDS and its cascade strategy and illustrates the capability of the combined design tool through the optimization of a subsonic aircraft and a high-bypass-turbofan wave-rotor-topped engine.

Patnaik, Surya N.↗

Experimental Investigation of the Behavior of Sub-Grid Scale Motions in Turbulent Shear Flow

Experiments have been carried out on a vertical jet of helium issuing into a co-flow of air at a fixed exit velocity ratio of 2.0. At all the experimental conditions studied, the flow exhibits a strong self excited periodicity. The natural frequency behavior of the jet, the underlying fine-scale flow structure, and the transition to turbulence have been studied over a wide range of flow conditions. The experiments were conducted in a variable pressure facility which made it possible to vary the Reynolds number and Richardson number independently. A stroboscopic schlieren system was used for flow visualization and single-component Laser Doppler Anemometry was used to measure the axial component of velocity. The flow exhibits several interesting features. The presence of co-flow eliminates the random meandering typical of buoyant plumes in a quiescent environment and the periodicity of the helium jet under high Richardson number conditions is striking. Under these conditions transition to turbulence consists of a rapid but highly structured and repeatable breakdown and intermingling of jet and freestream fluid. At Ri = 1.6 the three-dimensional structure of the flow is seen to repeat from cycle to cycle. The point of transition moves closer to the jet exit as either the Reynolds number or the Richardson number increases. The wavelength of the longitudinal instability increases with Richardson number. At low Richardson numbers, the natural frequency scales on an inertial time scale. At high Richardson number the natural frequency scales on a buoyancy time scale. The transition from one flow regime to another occurs over a narrow range of Richardson numbers from 0.7 to 1. A buoyancy Strouhal number is used to correlate the high Richardson number frequency behavior.

Cantwell, Brian↗

Hidden Statistics Approach to Quantum Simulations

Recent advances in quantum information theory have inspired an explosion of interest in new quantum algorithms for solving hard computational (quantum and non-quantum) problems. The basic principle of quantum computation is that the quantum properties can be used to represent structure data, and that quantum mechanisms can be devised and built to perform operations with this data. Three basic non-classical properties of quantum mechanics superposition, entanglement, and direct-product decomposability were main reasons for optimism about capabilities of quantum computers that promised simultaneous processing of large massifs of highly correlated data. Unfortunately, these advantages of quantum mechanics came with a high price. One major problem is keeping the components of the computer in a coherent state, as the slightest interaction with the external world would cause the system to decohere. That is why the hardware implementation of a quantum computer is still unsolved. The basic idea of this work is to create a new kind of dynamical system that would preserve the main three properties of quantum physics superposition, entanglement, and direct-product decomposability while allowing one to measure its state variables using classical methods. In other words, such a system would reinforce the advantages and minimize limitations of both quantum and classical aspects. Based upon a concept of hidden statistics, a new kind of dynamical system for simulation of Schroedinger equation is proposed. The system represents a modified Madelung version of Schroedinger equation. It preserves superposition, entanglement, and direct-product decomposability while allowing one to measure its state variables using classical methods. Such an optimal combination of characteristics is a perfect match for simulating quantum systems. The model includes a transitional component of quantum potential (that has been overlooked in previous treatment of the Madelung equation). The role of the transitional potential is to provide a jump from a deterministic state to a random state with prescribed probability density. This jump is triggered by blowup instability due to violation of Lipschitz condition generated by the quantum potential. As a result, the dynamics attains quantum properties on a classical scale. The model can be implemented physically as an analog VLSI-based (very-large-scale integration-based) computer, or numerically on a digital computer. This work opens a way of developing fundamentally new algorithms for quantum simulations of exponentially complex problems that expand NASA capabilities in conducting space activities. It has been illustrated that the complexity of simulations of particle interaction can be reduced from an exponential one to a polynomial one.

Zak, Michail↗

Hidden Statistics of Schroedinger Equation

Work was carried out in determination of the mathematical origin of randomness in quantum mechanics and creating a hidden statistics of Schr dinger equation; i.e., to expose the transitional stochastic process as a "bridge" to the quantum world. The governing equations of hidden statistics would preserve such properties of quantum physics as superposition, entanglement, and direct-product decomposability while allowing one to measure its state variables using classical methods.

Zak, Michail↗

Monte Carlo Mean Field Treatment of Microbunching Instability in the FERMI@Elettra First Bunch Compressor

Bunch compressors, designed to increase the peak current, can lead to a microbunching instability with detrimental effects on the beam quality. This is a major concern for free electron lasers (FELs) where very bright electron beams are required, i.e. beams with low emittance and energy spread. In this paper, we apply our self-consistent, parallel solver to study the microbunching instability in the first bunch compressor system of FERMI{at}Elettra. Our basic model is a 2D Vlasov-Maxwell system. We treat the beam evolution through a bunch compressor using our Monte Carlo mean field approximation. We randomly generate N points from an initial phase space density. We then calculate the charge density using a smooth density estimation procedure, from statistics, based on Fourier series. The electric and magnetic fields are calculated from the smooth charge/current density using a novel field formula that avoids singularities by using the retarded time as a variable of integration. The points are then moved forward in small time steps using the beam frame equations of motion, with the fields frozen during a time step, and a new charge density is determined using our density estimation procedure. We try to choose N large enough so that the charge density is a good approximation to the density that would be obtained from solving the 2D Vlasov-Maxwell system exactly. We call this method the Monte Carlo Particle (MCP) method.

43 PARTICLE ACCELERATORS↗

Laboratory Demonstration of Retroactive Influence in a Digital System

Retrocausation has been postulated in physical systems and observed in animate systems. The experiments described here extend methods used in human experiments to systems that are inanimate. One random-event generator, the controller-REG, was used to shut off a second REG, the subject-REG, at a random time. The output of the subject-REG was accumulated over several runs, each consisting of hundreds of trials, to look for a change in the randomness of its output bit stream in advance of the subject-REGs shut off. For the first three runs large changes were observed during the last second before shut off, changes of approximately 1 bit in 40 that exceeded odds against chance of 1 million to 1. Variations and later an exact replication of the early results failed to show the changes observed in the first three runs. This failure to replicate is an indication that there is an additional uncontrolled variable that must be taken into account, quite possibly the intention and enthusiasm of the experimenters. That addition leads to the question as to whether the subject-REG was subject to advance influence from its impending shut off, or instead whether its output was in a superposition of different states until the operator observed the results. The observation would then have caused a collapse of the superposition into a fixed state, like the collapse of quantum mechanical wavefunction. In either case, a retroactive influence was clearly in evidence.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Machine learning based algorithms for uncertainty quantification in numerical weather prediction models

Complex numerical weather prediction models incorporate a variety of physical processes, each described by multiple alternative physical schemes with specific parameters. The selection of the physical schemes and the choice of the corresponding physical parameters during model configuration can significantly impact the accuracy of model forecasts. There is no combination of physical schemes that works best for all times, at all locations, and under all conditions. It is therefore of considerable interest to understand the interplay between the choice of physics and the accuracy of the resulting forecasts under different conditions. This paper demonstrates the use of machine learning techniques to study the uncertainty in numerical weather prediction models due to the interaction of multiple physical processes. The first problem addressed herein is the estimation of systematic model errors in output quantities of interest at future times, and the use of this information to improve the model forecasts. The second problem considered is the identification of those specific physical processes that contribute most to the forecast uncertainty in the quantity of interest under specified meteorological conditions. In order to address these questions we employ two machine learning approaches, random forests and artificial neural networks. The discrepancies between model results and observations at past times are used to learn the relationships between the choice of physical processes and the resulting forecast errors. Numerical experiments are carried out with the Weather Research and Forecasting (WRF) model. The output quantity of interest is the model precipitation, a variable that is both extremely important and very challenging to forecast. The physical processes under consideration include various micro-physics schemes, cumulus parameterizations, short wave, and long wave radiation schemes. The experiments demonstrate the strong potential of machine learning approaches to aid the study of model errors.

97 MATHEMATICS AND COMPUTING↗