Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Data Compression”

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 757 records · Page 42

X-ray Diffraction of Cubic Gd2)3/Er under High Pressure

In this paper, we report the in situ high pressure X-ray diffraction studies on Er{sup 3+} doped Gd{sub 2}O{sub 3} in a diamond anvil cell up to 39.8 GPa at room temperature. Several phase transitions have been identified in our studies. The structural transformation from a starting cubic phase to a hexagonal phase occurred during the sample compression process, at 8.57 GPa. And the hexagonal phase was stable from 12.5 GPa up to the highest pressure in this study but was not quenchable and transformed to a monoclinic phase after pressure release. An anomalous high pressure behavior in the hexagonal type Gd{sub 2}O{sub 3} phase was observed, which might be caused by an electron transition influenced by Er{sup 3+} ions doping. By fitting the compression data to the Birch-Murnaghan equation of state, the bulk moduli of the cubic and two hexagonal (at p < 19.9 GPa and p > 27.0 GPa) Gd{sub 2}O{sub 3} phases were determined to be 164 {+-} 3, 185 {+-} 7, and 150 {+-} 10 GPa with B'{sub 0} = 4, respectively.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Ramp compression of tantalum to 330 GPa

We report on the stress–density and rate-dependent response for Ta, ramp compressed to 330 Pa with strain rates up to 5 x 108 s -1 . We employ temporally shaped laser drives to compress Ta stepped foils over several to tens of nanoseconds. Lagrangian wave-profile analysis reveals a stress–density relationship which falls below the Hugoniot, above the hydrostat, and is consistent with ramp-compression experiments at lower strain rates. We also report on the peak elastic stress prior to plastic deformation as a function of strain rate for laser-driven ramp and shock-compression data spanning the 1–50 × 107 s -1 strain-rate range. When combined with previously published lower strain data (101–107 s -1 ), we observe a change in rate dependence, suggesting a transition from thermally activated to defect-limited (phonon drag) dislocation motion occurring at a strain rate of about 105 s -1 .

36 MATERIALS SCIENCE↗

High pressure study of a highly energetic nitrogen-rich carbon nitride, cyanuric triazide

Cyanuric triazide (CTA), a nitrogen-rich energetic material, was compressed in a diamond anvil cell up to 63.2 GPa. Samples were characterized by x-ray diffraction, Raman, and infrared spectroscopy. A phase transition occurring between 29.8 and 30.7 GPa was found by all three techniques. The bulk modulus and its pressure derivative of the low pressure phase were determined by fitting the 300 K isothermal compression data to the Birch-Murnaghan equation of state. Due to the strong photosensitivity of CTA, synchrotron generated x-rays and visible laser radiation both lead to the progressive conversion of CTA into a two dimensional amorphous C=N network, starting from 9.2 GPa. As a result of the conversion, increasingly weak and broad x-ray diffraction lines were recorded from crystalline CTA as a function of pressure. Hence, a definite structure could not be obtained for the high pressure phase of CTA. Results from infrared spectroscopy carried out to 40.5 GPa suggest the high pressure formation of a lattice built of tri-tetrazole molecular units. The decompression study showed stability of the high pressure phase down to 13.9 GPa. Finally, two CTA samples, one loaded with neon and the other with nitrogen, used as pressure transmitting media, were laser-heated to approximately 1100 K and 1500 K while compressed at 37.7 GPa and 42.0 GPa, respectively. In both cases CTA decomposed resulting in amorphous compounds, as recovered at ambient conditions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Krypton storage capacity of the Earth's lower mantle

Noble gases are important geochemical tracers allowing reconstructing global volatile cycles in Earth's reservoirs. To constrain these fundamental processes, precise data on their partitioning behavior at deep Earth conditions are needed. Such data are only available at moderate pressures up to 25 GPa due to experimental challenges. We have investigated the possibility of noble gas storage in the Earth's lower mantle up to 115 GPa. We studied the incorporation of krypton in the second most abundant lower mantle mineral (Mg 1-x ,Fe x )O (ferropericlase) as well as in liquid metal-alloys by performing experiments up to 115 GPa and 3700 K using the laser-heated diamond anvil cell coupled to post-mortem EMPA analysis and X-ray absorption spectroscopy. The results reveal that, at these extreme conditions, up to 3 wt.% of krypton can be stored in (Mg 1-x ,Fe x )O and 3000 ppm in the Fe-rich liquid metal. For both phases the storage capacities increase with pressure (between 40 GPa and 60 GPa) at a constant high temperature of 2300 K. Fpc has never been considered as a NG host, despite being the second most abundant mineral in the Earth's LM. Using recent accurate compressibility data, we demonstrate that a substitution of krypton into the anion site of (Mg 1-x ,Fe x )O in form of neutral oxygen Schottky defects at diluted lower mantle conditions is possible. This noble gas incorporation mechanism is in agreement with a previous study on bridgmanite. We show that (Mg 1-x ,Fe x )O exhibits higher noble gas storage capacities than bridgmanite through the lower mantle using lattice strain modeling and including experimental solubility and thermoelastic data for neon, argon, krypton and xenon. We also demonstrate that both phases exhibit the highest solubilities for argon and krypton. We used the solubility data from lattice strain modeling to predict noble gas abundances stored in the solid lower mantle after magma ocean crystallization. The modeled abundances show apparent similarities with estimates for the deep noble gas reservoir that are based on either 3He abundances in ocean island basalts or radiogenic 40Ar abundances in the bulk Earth. This strongly indicates that the crystalline lower mantle may play an important role as deep noble gas storage reservoir. We propose, based on considerations on noble gas replenishment from the lower mantle to the atmosphere, that the lower mantle can only contribute to a small fraction of the present-day atmospheric noble gases. This suggests that the lower mantle is an un-degassed reservoir.

58 GEOSCIENCES↗

Real-Time Detection of Dust Devils from Pressure Readings

A method for real-time detection of dust devils at a given location is based on identifying the abrupt, temporary decreases in atmospheric pressure that are characteristic of dust devils as they travel through that location. The method was conceived for use in a study of dust devils on the Martian surface, where bandwidth limitations encourage the transmission of only those blocks of data that are most likely to contain information about features of interest, such as dust devils. The method, which is a form of intelligent data compression, could readily be adapted to use for the same purpose in scientific investigation of dust devils on Earth. In this method, the readings of an atmospheric- pressure sensor are repeatedly digitized, recorded, and processed by an algorithm that looks for extreme deviations from a continually updated model of the current pressure environment. The question in formulating the algorithm is how to model current normal observations and what minimum magnitude deviation can be considered sufficiently anomalous as to indicate the presence of a dust devil. There is no single, simple answer to this question: any answer necessarily entails a compromise between false detections and misses. For the original Mars application, the answer was sought through analysis of sliding time windows of digitized pressure readings. Windows of 5-, 10-, and 15-minute durations were considered. The windows were advanced in increments of 30 seconds. Increments of other sizes can also be used, but computational cost increases as the increment decreases and analysis is performed more frequently. Pressure models were defined using a polynomial fit to the data within the windows. For example, the figure depicts pressure readings from a 10-minute window wherein the model was defined by a third-degree polynomial fit to the readings and dust devils were identified as negative deviations larger than both 3 standard deviations (from the mean) and 0.05 mbar in magnitude. An algorithm embodying the detection scheme of this example was found to yield a miss rate of just 8 percent and a false-detection rate of 57 percent when evaluated on historical pressure-sensor data collected by the Mars Pathfinder lander. Since dust devils occur infrequently over the course of a mission, prioritizing observations that contain successful detections could greatly conserve bandwidth allocated to a given mission. This technique can be used on future Mars landers and rovers, such as Mars Phoenix and the Mars Science Laboratory.

Wagstaff, Kiri↗

The CCSDS Next Generation Space Data Link Protocol (NGSLP)

The CCSDS space link protocols i.e., Telemetry (TM), Telecommand (TC), Advanced Orbiting Systems (AOS) were developed in the early growth period of the space program. They were designed to meet the needs of the early missions, be compatible with the available technology and focused on the specific link environments. Digital technology was in its infancy and spacecraft power and mass issues enforced severe constraints on flight implementations. Therefore the Telecommand protocol was designed around a simple Bose, Hocquenghem, Chaudhuri (BCH) code that provided little coding gain and limited error detection but was relatively simple to decode on board. The infusion of the concatenated Convolutional and Reed-Solomon codes5 for telemetry was a major milestone and transformed telemetry applications by providing them the ability to more efficiently utilize the telemetry link and its ability to deliver user data. The ability to significantly lower the error rates on the telemetry links enabled the use of packet telemetry and data compression. The infusion of the high performance codes for telemetry was enabled by the advent of digital processing, but it was limited to earth based systems supporting telemetry. The latest CCSDS space link protocol, Proximity-1 was developed in early 2000 to meet the needs of short-range, bi-directional, fixed or mobile radio links characterized by short time delays, moderate but not weak signals, and short independent sessions. Proximity-1 has been successfully deployed on both NASA and ESA missions at Mars and is planned to be utilized by all Mars missions in development. A new age has arisen, one that now provides the means to perform advanced digital processing in spacecraft systems enabling the use of improved transponders, digital correlators, and high performance forward error correcting codes for all communications links. Flight transponders utilizing digital technology have emerged and can efficiently provide the means to make the next leap in performance for space link communications. Field Programmable Gate Arrays (FPGAs) provide the capability to incorporate high performance forward error correcting codes implemented within software transponders providing improved performance in data transfer, ranging, link security, and time correlation. Given these synergistic technological breakthroughs, the time has come to take advantage of them in applying them to both on going (e.g., command, telemetry) and emerging (e.g., space link security, optical communication) space link applications. However one of the constraining factors within the Data Link Layer in realizing these performance gains is the lack of a generic transfer frame format and common supporting services amongst the existing CCSDS link layer protocols. Currently each of the four CCSDS link layer protocols (TM, TC, AOS, and Proximity-1) have unique formats and services which prohibits their reuse across the totality of all space link applications of CCSDS member space agencies. For example, Mars missions. These missions implement their proximity data link layer using the Proximity-1 frame format and the services it supports but is still required to support the direct from Earth (TC) protocols and the Direct To Earth (AOS/TM) protocols. The prime purpose of this paper, is to describe a new general purpose CCSDS Data Link layer protocol, the NGSLP that will provide the required services along with a common transfer frame format for all the CCSDS space links (ground to/from space and space to space links) targeted for emerging missions after a CCSDS agency-wide coordinated date. This paper will also describe related options that can be included for the Coding and Synchronization sub-layer of the Data Link layer to extend the capacities of the link and additionally provide an independence of the transfer frame sub-layer from the coding sublayer. This feature will provide missions the option of running either the currently performed synchronous coding and transfer frame data link or an asynchronous coding/frame data link, in which the transfer frame length is independent of the block size of the code. The benefits from the elimination of this constraint (frame synchronized to the code block) will simplify the interface between the transponder and the data handling equipment and reduce implementation costs and complexities. The benefits include: inclusion of encoders/decoders into transmitters and receivers without regard to data link protocols, providing the ability to insert latency sensitive messages into the link to support launch, landing/docking, telerobotics. and Variable Coded Modulation (VCM). In addition the ability to transfer different sized frames can provide a backup for delivering stored anomaly engineering data simultaneously with real time data, or relaying of frames from various sources onto a trunk line for delivery to Earth.

Kazz, Greg J.↗

Quantifying uncertainty in analysis of shockless dynamic compression experiments on platinum. I. Inverse Lagrangian analysis

Absolute measurements of solid-material compressibility by magnetically driven shockless dynamic compression experiments to multi-megabar pressures have the potential to greatly improve the accuracy and precision of pressure calibration standards for use in diamond anvil cell experiments. Here, to this end, we apply characteristics-based inverse Lagrangian analysis (ILA) to 11 sets of ramp-compression data on pure platinum (Pt) metal and then reduce the resulting weighted-mean stress–strain curve to the principal isentrope and room-temperature isotherm using simple models for yield stress and Grüneisen parameter. We introduce several improvements to methods for ILA and quasi-isentrope reduction, the latter including calculation of corrections in wave speed instead of stress and pressure to render results largely independent of initial yield stress while enforcing thermodynamic consistency near zero pressure. More importantly, we quantify in detail the propagation of experimental uncertainty through ILA and model uncertainty through quasi-isentrope reduction, considering all potential sources of error except the electrode and window material models used in ILA. Compared to previous approaches, we find larger uncertainty in longitudinal stress. Monte Carlo analysis demonstrates that uncertainty in the yield-stress model constitutes by far the largest contribution to uncertainty in quasi-isentrope reduction corrections. We present a new room-temperature isotherm for Pt up to 444 GPa, with 1-sigma uncertainty at that pressure of just under ±1.2%; the latter is about a factor of three smaller than uncertainty previously reported for multi-megabar ramp-compression experiments on Pt. The result is well represented by a Vinet-form compression curve with (isothermal) bulk modulus K 0 = 270.3 ± 3.8 GPa, pressure derivative K$^{'}_{0}$= 5.66 ± 0.10, and correlation coefficient $R_{K_{0},K^{'}_{0}}$= –0.843.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

2007 Research and Engineering Annual Report

Selected research and technology activities at NASA Dryden Flight Research Center are summarized. These following activities exemplify the Center's varied and productive research efforts: Developing a Requirements Development Guide for an Automatic Ground Collision Avoidance System; Digital Terrain Data Compression and Rendering for Automatic Ground Collision Avoidance Systems; Nonlinear Flutter/Limit Cycle Oscillations Prediction Tool; Nonlinear System Identification Using Orthonormal Bases: Application to Aeroelastic/Aeroservoelastic Systems; Critical Aerodynamic Flow Feature Indicators: Towards Application with the Aerostructures Test Wing; Multidisciplinary Design, Analysis, and Optimization Tool Development Using a Genetic Algorithm; Structural Model Tuning Capability in an Object-Oriented Multidisciplinary Design, Analysis, and Optimization Tool; Extension of Ko Straight-Beam Displacement Theory to the Deformed Shape Predictions of Curved Structures; F-15B with Phoenix Missile and Pylon Assembly--Drag Force Estimation; Mass Property Testing of Phoenix Missile Hypersonic Testbed Hardware; ARMD Hypersonics Project Materials and Structures: Testing of Scramjet Thermal Protection System Concepts; High-Temperature Modal Survey of the Ruddervator Subcomponent Test Article; ARMD Hypersonics Project Materials and Structures: C/SiC Ruddervator Subcomponent Test and Analysis Task; Ground Vibration Testing and Model Correlation of the Phoenix Missile Hypersonic Testbed; Phoenix Missile Hypersonic Testbed: Performance Design and Analysis; Crew Exploration Vehicle Launch Abort System-Pad Abort-1 (PA-1) Flight Test; Testing the Orion (Crew Exploration Vehicle) Launch Abort System-Ascent Abort-1 (AA-1) Flight Test; SOFIA Flight-Test Flutter Prediction Methodology; SOFIA Closed-Door Aerodynamic Analyses; SOFIA Handling Qualities Evaluation for Closed-Door Operations; C-17 Support of IRAC Engine Model Development; Current Capabilities and Future Upgrade Plans of the C-17 Data Rack; Intelligent Data Mining Capabilities as Applied to Integrated Vehicle Health Management; STARS Flight Demonstration No. 2 IP Data Formatter; Space-Based Telemetry and Range Safety (STARS) Flight Demonstration No. 2 Range User Flight Test Results; Aerodynamic Effects of the Quiet Spike(tm) on an F-15B Aircraft; F-15 Intelligent Flight Controls-Increased Destabilization Failure; F-15 Integrated Resilient Aircraft Control (IRAC) Improved Adaptive Controller; Aeroelastic Analysis of the Ikhana/Fire Pod System; Ikhana: Western States Fire Missions Utilizing the Ames Research Center Fire Sensor; Ikhana: Fiber-Optic Wing Shape Sensors; Ikhana: ARTS III; SOFIA Closed-Door Flutter Envelope Flight Testing; F-15B Quiet Spike(TM) Aeroservoelastic Flight Test Data Analysis; and UAVSAR Platform Precision Autopilot Flight Results.

Stoliker, Patrick↗

New techniques for reversible compression of multispectral data

While spatial correlations are adequately exploited by standard lossless image compression techniques, little success has been attained in exploiting spectral correlations when dealing with multispectral image data. In this paper, we present some new lossless image compression techniques that capture spectral correlations as well as spatial correlation in a simple and elegant manner. The schemes are based on the notion of a prediction tree, which defines a non-causal prediction model for an image. We present a backward adaptive technique and a forward adaptive technique. We then give a computationally efficient way of approximating the backward adaptive technique. The approximation gives good results and is extremely easy to compute. Simulation results show that for high spectral resolution images, significant savings can be made by using spectral correlations in addition to spatial correlations. Furthermore, the increase in complexity incurred in order to make these gains is minimal.

Memon, N. D.↗

Image Provenance Analysis

The literature of multimedia forensics is mainly dedicated to the analysis of single assets (such as sole image or video files), aiming at individually assessing their authenticity. Different from this, image provenance analysis is devoted to the joint examination of multiple assets, intending to ascertain their history of edits, by evaluating pairwise relationships. Each relationship, thus, expresses the probability of one asset giving rise to the other, through either global or local operations, such as data compression, resizing, color-space modifications, content blurring, and content splicing. The principled combination of these relationships unveils the provenance of the assets, also constituting an important forensic tool for authenticity verification. This chapter introduces the problem of provenance analysis, discussing its importance and delving into the state-of-the-art techniques to solve it.

Moreira, Daniel↗

Uncovering acoustic signatures of pore formation in laser powder bed fusion

Abstract We present a machine learning workflow to discover signatures in acoustic measurements that can be utilized to create a low-dimensional model to accurately predict the location of keyhole pores formed during additive manufacturing processes. Acoustic measurements were sampled at 100 kHz during single-layer laser powder bed fusion (LPBF) experiments, and spatio-temporal registration of pore locations was obtained from post-build radiography. Power spectral density (PSD) estimates of the acoustic data were then decomposed using non-negative matrix factorization with custom $$\varvec{k}$$ k -means clustering (NMF $$\varvec{k}$$ k ) to learn the underlying spectral patterns associated with pore formation. NMF $$\varvec{k}$$ k returned a library of basis signals and matching coefficients to blindly construct a feature space based on the PSD estimates in an optimized fashion. Moreover, the NMF $$\varvec{k}$$ k decomposition led to the development of computationally inexpensive machine learning models which are capable of quickly and accurately identifying pore formation with classification accuracy of supervised and unsupervised label learning greater than 95% and 90%, respectively. The intrinsic data compression of NMF k , the relatively light computational cost of the machine learning workflow, and the high classification accuracy makes the proposed workflow an attractive candidate for edge computing toward in-situ keyhole pore prediction in LPBF.

36 MATERIALS SCIENCE↗

Crystal structure, thermal expansivity, and elasticity of OH-chondrodite: Trends among dense hydrous magnesium silicates

Here, we report the structure and thermoelastic properties of OH-chondrodite. The sample was synthesized at 12 GPa and 1523 K, coexisting with hydroxyl-clinohumite and hydrous olivine. The Fe content Fe/(Fe+Mg) is 1.1 mol%, and the monoclinic unit-cell parameters are: a = 4.7459(2) Å, b = 10.3480(7) Å, c = 7.9002(6) Å, α = 108.702(7)°, and V = 367.50(4) Å3. At ambient conditions the crystal structure was refined in space group P 2 1 /b from 1915 unique reflection intensities measured by single-crystal x-ray diffraction. The volume thermal expansion coefficient was measured between 150 and 800 K, resulting in α V = 2.8(5)×10 -9 (K -2 ) × T + 40.9(7) × 10 -6 (K -1 ) – 0.81(3)(K)/T 2 , with an average value of 38.0(9)×10 -6 K -1 . Brillouin spectroscopy was used to measure a set of acoustic velocities from which all thirteen components (C ij ) of the elastic tensor were determined. The Voigt-Reuss-Hill average of the moduli yield for the adiabatic bulk modulus, K S 0 = 117.9(12) GPa, and for shear modulus, G 0 = 70.1(5) GPa. The Reuss bound on the isothermal bulk modulus (K T0 ) is 114.2(14) GPa. From the measured thermodynamic properties, the Grüneisen parameter (γ) is calculated to be 1.66(4). Fitting previous static compression data using our independently measured bulk modulus (isothermal Reuss bound) as a fixed parameter, we refined the first pressure derivative of the bulk modulus, K T ’ = 5.5(1). Systematic trends between H 2 O content and physical properties are evaluated among dense hydrous magnesium silicate (DHMS) phases along the forsterite-brucite join.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A semi-agnostic ansatz with variable structure for variational quantum algorithms

Quantum machine learning—and specifically Variational Quantum Algorithms (VQAs)—offers a powerful, flexible paradigm for programming near-term quantum computers, with applications in chemistry, metrology, materials science, data science, and mathematics. Here, one trains an ansatz, in the form of a parameterized quantum circuit, to accomplish a task of interest. However, challenges have recently emerged suggesting that deep ansatzes are difficult to train, due to flat training landscapes caused by randomness or by hardware noise. This motivates our work, where we present a variable structure approach to build ansatzes for VQAs. Our approach, called VAns (Variable Ansatz), applies a set of rules to both grow and (crucially) remove quantum gates in an informed manner during the optimization. Consequently, VAns is ideally suited to mitigate trainability and noise-related issues by keeping the ansatz shallow. We employ VAns in the variational quantum eigensolver for condensed matter and quantum chemistry applications, in the quantum autoencoder for data compression and in unitary compilation problems showing successful results in all cases.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Self-consistent clustering analysis: an efficient multiscale scheme for inelastic heterogeneous materials

A predictive computational theory is shown for modeling complex, hierarchical materials ranging from metal alloys to polymer nanocomposites. The theory can capture complex mechanisms such as plasticity and failure that span across multiple length scales. This general multiscale material modeling theory relies on sound principles of mathematics and mechanics, and a cutting-edge reduced order modeling method named self-consistent clustering analysis (SCA) [Zeliang Liu, M.A. Bessa, Wing Kam Liu, “Self-consistent clustering analysis: An efficient multi-scale scheme for inelastic heterogeneous materials,” Comput. Methods Appl. Mech. Engrg. 306 (2016) 319–341]. SCA reduces by several orders of magnitude the computational cost of micromechanical and concurrent multiscale simulations, while retaining the microstructure information. This remarkable increase in efficiency is achieved with a data-driven clustering method. Computationally expensive operations are performed in the so-called offline stage, where degrees of freedom (DOFs) are agglomerated into clusters. The interaction tensor of these clusters is computed. In the online or predictive stage, the Lippmann-Schwinger integral equation is solved cluster-wise using a self-consistent scheme to ensure solution accuracy and avoid path dependence. To construct a concurrent multiscale model, this scheme is applied at each material point in a macroscale structure, replacing a conventional constitutive model with the average response computed from the microscale model using just the SCA online stage. A regularized damage theory is incorporated in the microscale that avoids the mesh and RVE size dependence that commonly plagues microscale damage calculations. The SCA method is illustrated with two cases: a carbon fiber reinforced polymer (CFRP) structure with the concurrent multiscale model and an application to fatigue prediction for additively manufactured metals. For the CFRP problem, a speed up estimated to be about 43,000 is achieved by using the SCA method, as opposed to FE2, enabling the solution of an otherwise computationally intractable problem. The second example uses a crystal plasticity constitutive law and computes the fatigue potency of extrinsic microscale features such as voids. This shows that local stress and strain are capture sufficiently well by SCA. This model has been incorporated in a process-structure-properties prediction framework for process design in additive manufacturing.

Liu, Z.↗

LaSDI: Parametric Latent Space Dynamics Identification

Enabling fast and accurate physical simulations with data has become an important area of computational physics to aid in inverse problems, design-optimization, uncertainty quantification, and other various decision-making applications. This paper presents a data-driven framework for parametric latent space dynamics identification procedure that enables fast and accurate simulations. The parametric model is achieved by building a set of local latent space model and designing an interaction among them. An individual local latent space dynamics model achieves accurate solution in a trust region. By letting the set of trust region to cover the whole parameter space, our model shows an increase in accuracy with an increase in training data. Herein we introduce two different types of interaction mechanisms, i.e., point-wise and region-based approach. Both linear and nonlinear data compression techniques are used. We illustrate the framework of Latent Space Dynamics Identification (LaSDI) enable a fast and accurate solution process on various partial differential equations, i.e., Burgers’ equations, radial advection problem, and nonlinear heat conduction problem, achieving 0 (100)x speed-up and 0 (1)% relative error with respect to the corresponding full order models.

97 MATHEMATICS AND COMPUTING↗

Solving high-dimensional inverse problems using amortized likelihood-free inference with noisy and incomplete data

Here, we present a likelihood-free probabilistic inversion method based on normalizing flows for high-dimensional inverse problems. The proposed method is composed of two complementary networks: a summary network for data compression and an inference network for parameter estimation. The summary network encodes raw observations into a fixed-size vector of summary features, while the inference network generates samples of the approximate posterior distribution of the model parameters based on these summary features. The posterior samples are produced in a deep generative fashion by sampling from a latent Gaussian distribution and passing these samples through an invertible transformation. We construct this invertible transformation by sequentially alternating conditional invertible neural network and conditional neural spline flow layers. The summary and inference networks are trained simultaneously. We apply the proposed method to an inversion problem in groundwater hydrology to estimate the posterior distribution of the log-conductivity field conditioned on spatially sparse time-series observations of the system’s hydraulic head responses. The conductivity field is represented with 706 degrees of freedom in the considered problem. Comparison with the likelihood-based iterative ensemble smoother PEST-IES method demonstrates that the proposed method accurately estimates the parameter posterior distribution and the observations’ predictive posterior distribution at a fraction of the inference time of PEST-IES.

conditional invertible neural network↗