Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “conditional probability”

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 1,405 records · Page 78
All sourcesNASA NTRSDOE OSTI
NASA NTRS2010

Risk Presentation Using the Three Dimensions of Likelihood, Severity, and Level of Control

Traditional hazard analysis techniques utilize a two-dimensional representation of the results determined by relative likelihood and severity of the residual risk. These matrices present a quick-look at the Likelihood (Y-axis) and Severity (X-axis) of the probable outcome of a hazardous event. A three-dimensional method, described herein, utilizes the traditional X and Y axes, while adding a new, third dimension, shown as the Z-axis, and referred to as the Level of Control. The elements of the Z-axis are modifications of the Hazard Elimination and Control steps (also known as the Hazard Reduction Precedence Sequence). These steps are: 1. Eliminate risk through design. 2. Substitute less risky materials for more hazardous materials. 3. Install safety devices. 4. Install caution and warning devices. 5. Develop administrative controls (to include special procedures and training.) 6. Provide protective clothing and equipment. When added to the two-dimensional models, the level of control adds a visual representation of the risk associated with the hazardous condition, creating a tall-pole for the leastwell-controlled failure while establishing the relative likelihood and severity of all causes and effects for an identified hazard. Computer modeling of the analytical results, using spreadsheets and three-dimensional charting gives a visual confirmation of the relationship between causes and their controls.

Watson, Clifford↗
NASA NTRS2011

Risk Assessment Using the Three Dimensions of Probability (Likelihood), Severity, and Level of Control

Traditional hazard analysis techniques utilize a two-dimensional representation of the results determined by relative likelihood and severity of the residual risk. These matrices present a quick-look at the Likelihood (Y-axis) and Severity (X-axis) of the probable outcome of a hazardous event. A three-dimensional method, described herein, utilizes the traditional X and Y axes, while adding a new, third dimension, shown as the Z-axis, and referred to as the Level of Control. The elements of the Z-axis are modifications of the Hazard Elimination and Control steps (also known as the Hazard Reduction Precedence Sequence). These steps are: 1. Eliminate risk through design. 2. Substitute less risky materials for more hazardous materials. 3. Install safety devices. 4. Install caution and warning devices. 5. Develop administrative controls (to include special procedures and training.) 6. Provide protective clothing and equipment. When added to the two-dimensional models, the level of control adds a visual representation of the risk associated with the hazardous condition, creating a tall-pole for the least-well-controlled failure while establishing the relative likelihood and severity of all causes and effects for an identified hazard. Computer modeling of the analytical results, using spreadsheets and three-dimensional charting gives a visual confirmation of the relationship between causes and their controls.

Watson, Clifford C.↗
NASA NTRS1974

Radio occultation exploration of Mars

The present work summarizes the principal results of the radio occultation experiments carried out to date for Mars. The early measurements, made at the time of the Martian dust storm of 1971, showed greatly reduced temperature gradients in the daytime troposphere, indicating the heating effect of the dust. The temperature gradients that were measured later when the atmosphere was apparently free of dust, were still much lower than expected under conditions of radiative-convective balance, indicating that dynamics may play a large part in determining the temperature structure of the Martian troposphere. The surface atmospheric pressure was observed to vary widely with topography ranging from about 1 mb at the summit of the Middle Spot volcano (Pavonis Mons) to over 10 mb in the North circumpolar region. The radius of the planet was measured with accuracies ranging from about 0.25 to about 2.1 km over latitudes ranging from 86 to -80 deg. These measurements have shown that Mars has pronounced equatorial and north-south asymmetries, which make it difficult to represent its shape by a simple triaxial figure. The daytime ionosphere measurements indicated that the main ionization peak was similar in behavior to a terrestrial F1 layer and is probably produced by photoionization of carbon dioxide by solar extreme ultraviolet.

Kliore, A. J.↗
NASA NTRS1983

Corotating pressure waves without streams in the solar wind

Voyager 1 and 2 magnetic field and plasma data are presented which demonstrate the existence of large scale, corotating, non-linear pressure waves between 2 AU and 4 AU that are not accompanied by fast streams. The pressure waves are presumed to be generated by corotating streams near the Sun. For two of the three pressure waves that are discussed, the absence of a stream is probably a real, physical effect, viz., a consequence of deceleration of the stream by the associated compression wave. For the third pressure wave, the apparent absence of a stream may be a geometrical effect; it is likely that the stream was at latitudes just above those of the spacecraft, while the associated shocks and compression wave extended over a broader range of latitudes so that they could be observed by the spacecraft. It is suggested that the development of large-scale non-linear pressure waves at the expense of the kinetic energy of streams produces a qualitative change in the solar wind in the outer heliosphere. Within a few AU the quasi-stationary solar wind structure is determined by corotating streams whose structure is determined by the boundary conditions near the Sun.

Burlaga, L. F.↗
NASA NTRS1983

Corotating pressure waves without fast streams in the solar wind

Voyager 1 and 2 magnetic field and plasma data are presented which demonstrate the existence of large scale, corotating, non-linear pressure waves between 2 AU and 4 AU that are not accompanied by fast streams. The pressure waves are presumed to be generated by corotating streams near the sun. For two of the three pressure waves that are discussed, the absence of a stream is probably a real, physical effect, viz., a consequence of deceleration of the stream by the associated compression wave. For the third pressure wave, the apparent absence of a stream may be a geometrical effect; it is likely that the stream was at latitudes just above those of the spacecraft, while the associated shocks and compression wave extended over a broader range of latitudes so that they could be observed by the spacecraft. It is suggested that the development of large-scale non-linear pressure waves at the expense of the kinetic energy of streams produces a qualitative change in the solar wind in the outer heliosphere. Within a few AU the quasi-stationary solar wind structure is determined by corotating streams whose structure is determined by the boundary conditions near the sun. Previously announced in STAR as N83-19694

Burlaga, L. F.↗
NASA NTRS2004

Spectral Index and Quasi-Periodic Oscillation Frequency Correlation in Black Hole (BH) Sources: Observational Evidence of Two Phases and Phase Transition in BHs

Recent studies have shown that strong correlations are observed between the low frequencies (1-10 Hz) of quasiperiodic oscillations (QPOs) and the spectral power law index of several Black Hole (BH) candidate sources, in low hard states, steep power-law (soft) states and in transition between these states. The observations indicate that the X-ray spectrum of such state (phases) show the presence of a power-law component and are sometimes related to simultaneous radio emission indicated the probable presence of a jet. Strong QPOs (less than 20% rms) are present in the power density spectrum in the spectral range where the power-law component is dominant ( i.e. 60-90% ). This evidence contradicts the dominant long standing interpretation of QPOs as a signature of the thermal accretion disk. We present the data from the literature and our own data to illustrate the dominance of power-law index-QPO frequency correlations. We provide a model, that identifies and explains the origin of the QPOs and how they are imprinted on the properties of power-law flux component. We argue the existence of a bounded compact coronal region which is a natural consequence of the adjustment of Keplerian disk flow to the innermost sub-Keplerian boundary conditions near the central object and that ultimately leads to the formation of a transition layer (TL) between the adjustment radius and the innermost boundary. The model predicts two phases or states dictated by the photon upscattering produced in the TL: (1) hard state, in which the TL is optically thin and very hot (kT approx. greater than 50 keV) producing photon upscattering via thermal Componization; the photon spectrum index Gamma appprox.1.5 for this state is dictated by gravitational energy release and Compton cooling in an optically thin shock near the adjustment radius; (2) a soft state which is optically thick and relatively cold (approx. less than 5 keV); the index for this state, Gamma approx. 2.8 is determined by soft-photon upscattering and photon trapping in converging flow into BH. In the TL model for corona the QPO frequency vnu(sub high) is related to the gravitational (close to Keplerian) frequency nu(sub K) at the outer (adjustment) radius and nu(sub low) is related to the TL s normal mode (magnetoacoustic) oscillation frequency nu(sub MA). The observed correlations between index and low and high QPO frequencies are readily explained in terms of this model. We also suggest a new method for evaluation of the BH mass using the index-frequency correlation.

Titarchuk, Lev↗
NASA NTRS2012

Evaluation of Precipitation Detection over Various Surfaces from Passive Microwave Imagers and Sounders

During the middle part of this decade a wide variety of passive microwave imagers and sounders will be unified in the Global Precipitation Measurement (GPM) mission to provide a common basis for frequent (3 hr), global precipitation monitoring. The ability of these sensors to detect precipitation by discerning it from non-precipitating background depends upon the channels available and characteristics of the surface and atmosphere. This study quantifies the minimum detectable precipitation rate and fraction of precipitation detected for four representative instruments (TMI, GMI, AMSU-A, and AMSU-B) that will be part of the GPM constellation. Observations for these instruments were constructed from equivalent channels on the SSMIS instrument on DMSP satellites F16 and F17 and matched to precipitation data from NOAA's National Mosaic and QPE (NMQ) during 2009 over the continuous United States. A variational optimal estimation retrieval of non-precipitation surface and atmosphere parameters was used to determine the consistency between the observed brightness temperatures and these parameters, with high cost function values shown to be related to precipitation. The minimum detectable precipitation rate, defined as the lowest rate for which probability of detection exceeds 50%, and the detected fraction of precipitation, are reported for each sensor, surface type (ocean, coast, bare land, snow cover) and precipitation type (rain, mix, snow). The best sensors over ocean and bare land were GMI (0.22 mm/hr minimum threshold and 90% of precipitation detected) and AMSU (0.26 mm/hr minimum threshold and 81% of precipitation detected), respectively. Over coasts (0.74 mm/hr threshold and 12% detected) and snow-covered surfaces (0.44 mm/hr threshold and 23% detected), AMSU again performed best but with much lower detection skill, whereas TMI had no skill over these surfaces. The sounders (particularly over water) benefited from the use of re-analysis data (vs. climatology) to set the a-priori atmospheric state and all instruments benefit from the use of a conditional snow cover emissivity database over land. It is recommended that real-time sources of these data be used in the operational GPM precipitation algorithms.

GPM↗
NASA NTRS1988

Metal with anomalously low Ni and Ge concentrations in the Allan Hills A77081 winonaite

The Ge content of metal in the Allan Hills A77081 winonaite was determined by high-sensitivity electron microprobe analysis. By optimizing analytical conditions for Ge determination, a detection limit of about 75 ppm could be achieved. In A77081 some small kamacite grains contain less Ni and Ge and more Co than coarse-grained metal. These small grains are always associated with sulfide, raising the possibility that anomalous metal is related to eutectic melting. However, when published partition coefficients for Ni and Ge in the Fe-Ni-S system are used to model fractionation of these elements during eutectic melting, one finds that secondary metal should be enriched in Ni and depleted in Ge. Thus, the positive Ni-Ge correlation found in this study is the opposite of the expected trend. No explanation for this discrepancy has yet been found. Nonetheless, the existence of anomalous metal is an indication that A77081, and probably other winonaites as well, have undergone some fractionation. This supports the notion that the high-temperature history of winonaites is related to the formation of IAB iron meteorites, whose silicate inclusions are very similar to winonaites.

Kracher, Alfred↗
NASA NTRS2004

Product Distribution Theory for Control of Multi-Agent Systems

Product Distribution (PD) theory is a new framework for controlling Multi-Agent Systems (MAS's). First we review one motivation of PD theory, as the information-theoretic extension of conventional full-rationality game theory to the case of bounded rational agents. In this extension the equilibrium of the game is the optimizer of a Lagrangian of the (probability distribution of) the joint stare of the agents. Accordingly we can consider a team game in which the shared utility is a performance measure of the behavior of the MAS. For such a scenario the game is at equilibrium - the Lagrangian is optimized - when the joint distribution of the agents optimizes the system's expected performance. One common way to find that equilibrium is to have each agent run a reinforcement learning algorithm. Here we investigate the alternative of exploiting PD theory to run gradient descent on the Lagrangian. We present computer experiments validating some of the predictions of PD theory for how best to do that gradient descent. We also demonstrate how PD theory can improve performance even when we are not allowed to rerun the MAS from different initial conditions, a requirement implicit in some previous work.

Lee, Chia Fan↗
NASA NTRS2013

Multifunctional Low-Pressure Turbine for Core Noise Reduction, Improved Efficiency, and Nitrogen Oxide (NOx) Reduction

This work studied the feasibility of using Helmholtz resonator cavities embedded in low-pressure-turbine (LPT) airfoils to (1) reduce core noise by damping acoustic modes; (2) use the synthetic jets produced by the liner hole acoustic oscillations to improve engine efficiency by maintaining turbulent attached flow in the LPT at low-Reynolds-number cruise conditions; and (3) reduce engine nitrogen oxide emissions by lining the internal cavities with materials capable of catalytic conversion. Flat plates with embedded Helmholtz resonators, designed to resonate at either 3000 or at 400 Hz, were simulated using computational fluid dynamics. The simulations were conducted for two inlet Mach numbers, 0.25 and 0.5, corresponding to Reynolds numbers of 90 000 and 164 000 based on the effective chordwise distance to the resonator orifice. The results of this study are (1) the region of acoustic treatment may be large enough to have a benefit; (2) the jets may not possess sufficient strength to reduce flow separation (based on prior work by researchers in the flow control area); and (3) the additional catalytic surface area is not exposed to a high velocity, so it probably does not have any benefit.

Miller, Christopher J.↗
NASA NTRS1998

Volcanism, Cold Temperature, and Paucity of Sunspot Observing Days (1818-1858): A Connection?

During the interval of 1818-1858, several curious decreases in the number of sunspot observing days per year are noted in the observing record of Samuel Heinrich Schwabe, the discoverer of the sunspot cycle, and in the reconstructed record of Rudolf Wolf, the founder of the now familiar relative sunspot number. These decreases appear to be nonrandom in nature and often extended for 13 yr (or more). Comparison of these decreases with equivalent annual mean temperature (both annual means and 4-yr moving averages). as recorded at Armagh Observatory (Northern Ireland), indicates that the temperature during the years of decreased number of observing days trended downward near the start of' each decrease and upward (suggesting some sort of recovery) just before the end of each decrease. The drop in equivalent annual mean temperature associated with each decrease, as determined from the moving averages, measured about 0.1-0.7 C. The decreases in number of observing days are found to be closely related to the occurrences of large, cataclysmic volcanic eruptions in the tropics or northern hemisphere. In particular, the interval of increasing number of observing days at the beginning of the record (i.e., 1818-1819) may be related to the improving atmospheric conditions in Europe following the 1815 eruption of Tambora (Indonesia; 8 deg. S), which previously, has been linked to "the year without a summer" (in 1816) and which is the strongest eruption in recent history, while the decreases associated with the years of 1824, 1837, and 1847 may, be linked, respectively, to the large, catacivsmic volcanic eruptions of Galunggung (Indonesia; 7 deg. S) in 1822, Cosiguina (Nicaragua) in 1835, and, perhaps, Hekla (Iceland; 64 deg. N) in 1845. Surprisingly, the number of observing days per year, as recorded specifically b), SchAabe (from Dessau, Germany), is found to be linearly correlated against the yearly mean temperature at Armagh Observatory (r = 0.5 at the 2 percent level of significance); thus. years of fewer sunspot observing days in the historical record seem to indicate years of probable cooler clime, while years (if many sunspot observing days seem to indicate years of probable warmer clime (and Vice versa). Presuming this relationship to be real, one infers that the observed decrease in the number of observing days near 1830 (i.e., during "the lost record years" of 1825 to 1833) provides a strong indication that temperatures at Armagh (and, perhaps, most of Europe, as well) were correspondingly cooler. If true, then, the inferred cooling may have resulted from the eruption of Kliuchevsoi(Russia; 56 deg. N) in 1829.

Wilson, Robert M.↗
NASA NTRS1999

Lipid synthesis under hydrothermal conditions by Fischer-Tropsch-type reactions

Ever since their discovery in the late 1970's, mid-ocean-ridge hydrothermal systems have received a great deal of attention as a possible site for the origin of life on Earth (and environments analogous to mid-ocean-ridge hydrothermal systems are postulated to have been sites where life could have originated or Mars and elsewhere as well). Because no modern-day terrestrial hydrothermal systems are free from the influence of organic compounds derived from biologic processes, laboratory experiments provide the best opportunity for confirmation of the potential for organic synthesis in hydrothermal systems. Here we report on the formation of lipid compounds during Fischer-Tropsch-type synthesis from aqueous solutions of formic acid or oxalic acid. Optimum synthesis occurs in stainless steel vessels by heating at 175 degrees C for 2-3 days and produces lipid compounds ranging from C2 to > C35 which consist of n-alkanols, n-alkanoic acids, n-alkenes, n-alkanes and alkanones. The precursor carbon sources used are either formic acid or oxalic acid, which disproportionate to H2, CO2 and probably CO. Both carbon sources yield the same lipid classes with essentially the same ranges of compounds. The synthesis reactions were confirmed by using 13C labeled precursor acids.

Non-NASA Center↗
NASA NTRS2004

Spectral Index and Quasi-Periodic Oscillation Frequency Correlation in Black Hole Sources: Observational Evidence of Two Phases and Phase Transition in Black Holes

Recent studies have shown that strong correlations are observed between the low frequencies (1-10 Hz) of quasi-periodic oscillations (QPOs) and the spectral power law index of several black hole (BH) candidate sources, in low (hard) states, steep power law (soft) states, and transitions between these states. The observations indicate that the X-ray spectra of such state (phases) show the presence of a power-law component and are sometimes related to simultaneous radio emission, indicating the probable presence of a jet. Strong QPOs (>20% rms) are present in the power density spectrum in the spectral range where the power-law component is dominant (i.e., 60%90%). This evidence contradicts the dominant, long-standing interpretation of QPOs as a signature of the thermal accretion disk. We present the data from the literature and our own data to illustrate the dominance of power-law index-QPO frequency correlations. We provide a model that identifies and explains the origin of the QPOs and how they are imprinted on the properties of the power-law flux component. We argue for the existence of a bounded compact coronal region that is a natural consequence of the adjustment of the Keplerian disk flow to the innermost sub-Keplerian boundary conditions near the central object and that ultimately leads to the formation of a transition layer (TL) between the adjustment radius and the innermost boundary. The model predicts two phases or states dictated by the photon upscattering produced in the TL: (1) a hard state, in which the TL is optically thin and very hot (kT approximately greater than 50 keV), producing photon upscattering via thermal Comptonization (the photon spectrum index Gamma approximates 1.7 for this state is dictated by gravitational energy release and Compton cooling in an optically thin shock near the adjustment radius), and (2) a soft state that is optically thick and relatively cold (kT approximately less than 5 keV the index for this state, Gamma approximates 2.8, is determined by soft-photon upscattering and photon trapping in a converging flow into the BH). In the TL model for the corona, the QPO frequency V(sub high) is related to the gravitational (close to Keplerian) frequency V(sub K) at the outer (adjustment) radius and v(sub low) is related to the TL's normal mode (magnetoacoustic) oscillation frequency v(sub MA) . The observed correlations between index and low and high QPO frequencies are readily explained in terms of this model. We also suggest a new method for evaluation of the BH mass using the index-frequency correlation.

Titarchuk, Lev↗
NASA NTRS1987

Stochastic estimation of conditional eddies in turbulent channel flow

Several long-standing issues regarding linear estimation and coherent structures are addressed, and some questions that were addressed partially, but never with the benefit of full, three-dimensional information are answered. The objectives were to: determine how well linear estimates approximate the field obtained by true conditional averaging, using events such as those in quadrant analysis; determine the extent to which the three-dimensional linearly estimated fields correspond to coherent structures, and the degree and manner in which they differ; evaluate the type and nature of the structural information gained by employing several different types of events; and learn more about the 3-D structure of important coherent motions that occur in wall turbulence. The results indicate that linear stochastic estimation can be used effectively in the study of numerical data bases consisting of three-dimensional vector fields, both velocity and vorticity. Two-point stochastic estimation yields more structural information and more detail than single-point estimation. The structures observed occurred repeatedly within the flow, but much can not be said about their dominance or the probability of their occurrence without further systematic studies of their frequency.

Adrian, R. J.↗
NASA NTRS1994

Tidal disruption and the appearance of periodic comet Shoemaker-Levy 9

A unified model is formulated to interpret quantitatively the observed characteristics of the nuclear train, the two dust trails, and the tail region of P/Shoemaker-Levy 9 in terms of a collisionally modified rotation velocity distribution of the comet's debris. Since there is no evidence for the comet's out-gassing, the model does not assume activity. The discruption of the parent comet was due primarily to tidal stesses during its extremely close approach to Juptier in July 1992. The original nucleus is found to have been most probably approximately 5 km in radius or, equivalently, approximately 10(exp 17) g in mass. The dynamical separation of the debris occurred apparently approximately 2.2 hr after the perijove passage, even though the actual fragmentation of the original mass is likely to have begun before closest approach. Physical breakp was accompanied by ubiquitous low-velocity collisions among the particulates, resulting in a rearrangement of the initial rotaional velocities into a rapidly 'thermalized' distribution, characterized by a long tail of relatively high velocities for the debris that populates the dust trails far from the nuclear train. Compelling evidence is presented for secondary fragmentation events, indicative of the comet's continuing disintegration. Secondary fragmentation may be caused by rotational bursting of the massive fragments that had been cracked but unbroken during the tidal disruption. Impact conditions and possible interactions of the comet's debris with the jovian system during the encounter in July 1994 are described.

Sekanina, Z.↗
NASA NTRS1974

Vibrationally excited nitrogen in stable auroral red arcs and its effect on ionospheric recombination

The time-dependent continuity equations, including diffusion, were solved for the first six energy levels of molecular nitrogen for conditions in the thermosphere corresponding to stable auroral red (SAR) arcs. The results show that molecular nitrogen is excited vibrationally to the degree that the rate constant for the ionospheric loss process, O(+) + N2 yields NO(+) + N, is increased by as much as a factor of 7.6 at F2 region altitudes. It was found that deviations from the energetically equivalent Boltzmann distribution were large, causing the rate constant to be as much as 1.6 times the rate constant calculated for the Boltzmann distribution. These results indicate that SAR arc intensities as small as 58 R can produce noticeable increases in the ionosphere ion-atom interchange reaction rate and hence in the rate of loss of ionospheric electrons. It is suggested that the observed decrease of electron density in the F2 region in SAR arcs can probably be explained by enhanced reaction rates for ion-atom interchange between O(+) and N2 caused by vibrational excitation of molecular nitrogen by electron impact.

Newton, G. P.↗
NASA NTRSDate not supplied

Flight Performance and Stability of Space Launch System Core Stage Thrust Vector Control

The Space Launch System (SLS) Core Stage (CS) Thrust Vector Control (TVC) system is comprised of eight mechanical feedback Shuttle heritage Type III TVC actuators and four RS-25 engines, each attached to a Shuttle heritage gimbal block/bearing. The Core Stage TVC shares vehicle control authority with the SLS 5-segment Solid Rocket Boosters (SRBs) during boost phase flight, and is the sole means of vehicle flight control during in exoatmospheric flight following SRB separation. TVC responses during Green Run Hot Fire (GRHF) testing revealed that the TVC did not meet its performance specifications. Step and frequency responses exhibited unexpected departures from prior laboratory data and modeled behavior. Post-test analysis determined that the characteristics of the structure and gimbal friction are significantly influenced by the thrust-loaded conditions, and the command avionics exhibited a small but important gain nonlinearity. Using the available test data, the design team augmented the flight control TVC models to bound the observed results and include the additional fidelity needed for vehicle flight control analysis so as to build sufficient rationale for flight certification. Prior to the Green Run tests, “simplex” linear models typically used for flight control analysis did not include gimbal friction and other nonlinearities owing to long-standing assumptions that these effects were negligible in the Shuttle Orbiter TVC system. Following the Green Run findings, simulation analysis of the flight dynamics in the time and frequency domain revealed the propensity for a flight control limit cycle oscillation (LCO) if friction and structural compliances fell near the edges of test-predicted bounds. While the “most probable” models did not predict an in-flight LCO, the SLS Program conservatively proceeded with a system-wide evaluation and ultimate acceptance of the possibility for a small amplitude, low-frequency TVC LCO in flight. A final validation of the extensive test and modeling effort occurred when the first flight of SLS successfully demonstrated the fully integrated performance of the vehicle’s TVC system This paper is the final installment in a seven-paper series surveying the design, engineering, test validation, and flight performance of the Core Stage Thrust Vector Control system. In this paper, the development of flight rationale in light of the TVC responses observed in Green Run is discussed, along with a review of the flight telemetry illustrating the correlation of the preflight predictions with the observed performance.

John H. Wall↗
NASA NTRS2020

Application of Sparse Identification of Nonlinear Dynamics for Physics-Informed Learning

Advances in machine learning and deep neural networks has enabled complex engineering tasks like image recognition, anomaly detection, regression, and multi-objective optimization, to name but a few. The complexity of the algorithm architecture, e.g., the number of hidden layers in a deep neural network, typically grows with the complexity of the problems they are required to solve, leaving little room for interpreting (or explaining) the path that results in a specific solution. This drawback is particularly relevant for autonomous aerospace and aviation systems, where certifications require a complete understanding of the algorithm behavior in all possible scenarios. Including physics knowledge in such data-driven tools may improve the interpretability of the algorithms, thus enhancing model validation against events with low probability but relevant for system certification. Such events include, for example, spacecraft or aircraft sub-system failures, for which data may not be available in the training phase. This paper investigates a recent physics-informed learning algorithm for identification of system dynamics, and shows how the governing equations of a system can be extracted from data using sparse regression. The learned relationships can be utilized as a surrogate model which, unlike typical data-driven surrogate models, relies on the learned underlying dynamics of the system rather than large number of fitting parameters. The work shows that the algorithm can reconstruct the differential equations underlying the observed dynamics using a single trajectory when no uncertainty is involved. However, the training set size must increase when dealing with stochastic systems, e.g., nonlinear dynamics with random initial conditions.

Corbetta, Matteo↗