Efficient gradient computation of a misfit function for FWI using the adjoint method
|
|
- Elfreda Cain
- 6 years ago
- Views:
Transcription
1 Efficient gradient computation of a misfit function for FWI using the adjoint method B. Galuzzi, department of Earth Sciences A. Desio, University of Milan E. Stucchi, department of Earth Sciences A. Desio, University of Milan 4 Novembre 6 GNGTS 6, Lecce Sessione 3.3: Modelling, aspetti teorici e tecnologie innovative
2 Agenda. Seismic inversion using FWI. Seismic modelling 3. The importance of misfit gradient for FWI 4. Computation of misfit gradient using the adjoint method 5. An example of application
3 Full Waveform Inversion (Tarantola, 986) Estimation of a geological macro-model of subsurface from active seismic data, by means of: ) Seismic modelling algorithm Numerical solution of wave equation (predicted data) ) Misfit function Difference between predicted and observed data 3) Seismic inversion algorithm Optimization of the misfit function estimation of the predicted model Starting model Predicted model modelling algorithm Predicted data Misfit function Observed data vs Predicted data inversion algorithm Stopping criteria: maximum number of iterations, minimum threshold,...
4 The D acoustic wave equation q x, t = v x ) Δp x, t + δ x x / s t p x, t = q x, t Initial conditions: p x, = q(x, ) =, x D ) BCD/) BFD/) = q@,a + dx ) Δ B The seismic modelling and its approximation by an explicit finite difference method p BCD = BCD/) + dt: time sampling dx: space sampling ΔH : approximation of Laplacian nd order in time and 4 nd in space + dtδ x / s B t, T : recording time x D(x, z) R ) space domain x / the location of the source s t the seismic wavelet v x the acoustic wave velocity
5 . Numerical stability: dt < dx v \]^ λ, Numerical considerations. λε.5, : Courant number. Numerical dispersion dx < v \@` f,. v \@` : minimum velocity of the model \]^n. f \]^ : maximum frequency of s(t) 3. n 5: points per wavelength 3. Absorbing boundary conditions (Cerjan, 985): q = p = B q + dtv dx ) Δ B BCD/) + + dtδ x / s B The Gaussian taper factor = V.9, inside the computational domain Inside the absorbing boundary layers
6 Misfit function for seismic inversion The aim of FWI is to find an optimal Earth model m, that minimizes a misfit function: m = min g h Χ(m ) = min \ h p m p / p / : observed data p m : predicted data (using seismic modelling) M : the set of geological models A classical functional for FWI is the L distance between the observed and the synthetic seismograms: Χ m = `t `s v nr : number of receivers p p q[p m, t, x s, x t, m p / (t, x s, x t )u ) ns : number of the sources dt, T : registration time twd swd / x s : position of r th receiver x t : position of the s th source Due to space discretization, the geological models mεm are discretized with components = [v z v t ), ), ] i :,, nx j :,, nz k:,, ny ) Number of unknowns = (number of grid nodes) * (number of geological parameters) ) Χ(m) is a generally a complicated non-linear functional of m
7 Optimization algorithms for seismic inversion Local iterative minimization algorithms Starting model m / m ycd = m y + γ y h y h y is the descend direction γ y > is the step length No Computation of h y = \ Χ m y Computation of γ y Χ m ycd < Χ m y Predicted model m ycd A descend direction is given by: Χ m ycd <tol D or \ Χ m ycd <tol )? h y = \ Χ m y the gradient at the k-th iteration The Misfit gradient is useful to guide the minimization algorithm to a local minimum of Χ m How compute the gradient \ Χ m y?
8 The adjoint method Powerful tool to shorten the time required for computing the gradient (Fichner, ) Acoustic equation+l misfit p is the regular solution of the wave equation p t vž Δp t = δ x x / s t with: p = p ) =, x D. Χ( Š = m q p t) p / v p v is the solution of the adjoint equation p t vž Δp t = g t with: p T = p T) =, x D. g v t = p t p / t δ x s To compute the time integral à simultaneously p and p for each time step, in all the domain D Regular solution p )Wave equation scheme, forward in time (from to T), to compute the adjoint source (g ) to store p only inside the absorbing boundaries )Wave equation scheme, reverse in time (from T to ) to re-compute p inside the domain using the storing p inside the absorbing boundaries Adjoint solution p Wave equation scheme, backward in time (from T to ) to satisfy the final conditions (t=t) after the adjoint source computation N.B The storing of p inside the absorbing boundaries to prevent numerical instability FD > ) during backward propagation
9 Scheme for the adjoint gradient computation ) Computation of the p, forward in time (Storing p inside the absorbing layers for each time step) ) Computation of g using p and p 3) Computation of p and p, backward in time (using the stored p, inside the absorbing layers) Computation of the time integral using p and p only 3 computation of wave equation is required: forward modelling+ backward modelling
10 .Observed data Example of application depth (km) Spread layout Portion of Marmousi model 9 receivers 6 sources The time grid dt =.s T=6s The modelling grid nx = 9 nz = 48 dx = dz = 4m Local optimization method m ycd = m y + γ y h y \ Χ m y is the gradient at the k-th iteration β y = \Χ m ycd ) ) \ Χ m y ) ) γ y obtained by a line search procedure on y = Χ(m y + γ y h y ) (linear or quadratic interpolation) h ycd = \ Χ m y + β y h y T \z s *(T \z (h y )+T \z (γ y )) s *(about 4-5 modellings)
11 3.The starting model (smooth version of the true model) depth (km) Starting model The gradient emphasizes the main differences between the starting and the true model Note the high values of the gradient near the sources The gradient of the misfit function can guide the inversion procedure to the true model
12 Results depth (km) depth (km) depth (km).5 True model Starting model Final model L Misfif Misfit evolution Iter Note the misfit reduction from to about.5 The final model isclose the global minimum solution The inversion procedure has highlighted the main geological structures that were not clear in the smooth model The main differences between the final and true model are at the border where the illumination is poor
13 Conclusion. We have implemented the adjoint method to compute the gradient of a misfit function in an efficient way. The low computational time required by the adjoint method and the low storage memory employed make this implementation of particular interest especially in applications such as the Full Waveform Inversion 3. As an example of application, we carried out a local FWI on a portion of the Marmousi model. The finalmodel, after the localinversion fairly match the true model 4. As a future work we plan to apply this method for real data Thank you for your attention!
14 The finite difference approximation How compute the misfit gradient? Computation of the partial derivatives by using a finite difference approximation: Š Χ m D,D + ϵδm D,D, m D,), Χ m D,D ϵδm D,D, m D,), m D,D ϵ with a small ϵ > Problems:. The choice of ϵ is not obvious. To obtain the gradient is necessary to compute two forward modelling for each spatial derivative 3. This method is impracticable when the number of unknowns is large and the forward modelling is computationally expensive
15 .The observed data Example of seismic acquisition Spread layout 9 receivers sources depth (km) Marmousi model km/s
16 .The forward modelling nx = 384 nz = dx = dz = 4m dt =.s 4.The predicted model 3.The inversion procedure Misfit function: L difference Starting model: D True model: Marmousi Optimization algorithm : steepest descend m ycd = m y γ y \ Χ m y Marmousi model D model.5.5 depth (km).5 depth (km) Velocity (km/s) Velocity (km/s) We need to compute the gradient of the misfit function for the predicted model
17 Finite difference vs adjoint method.5 Marmousi model.5 D model The two methods produce very similar solutions. depth (km) Velocity (km/s) depth (km) Velocity (km/s) The computational time of the adjoint method is considerably reduced: Adjoint method: forward modelling + backward modelling Classical gradient method: x384x forward modelling
18 Formula for the adjointstate method Misfit function: Χ v = D ) v [p t, x, v p / (t, x )u ) š / dtdx = Χ v, with Χ v = D ) [ p t, x, v p / (t, x ) ) Wave equation: p x, t v ) Δp x, t = f(x, t) L p v, x, t, v = f(x, t) With L linear operator
19 Formula for the adjointstate method Misfit function gradient: Χ v = z Χ p(v, x, t) p(v, x, t), Wave equation gradient: ) L p v, x, t, v = f(x, t) ) L p v, x, t, v + z L p v, x, t, v p v, x, t = 3) L p v, x, t, v g + z L p v, x, t, v p v, x, t g = 4) L p v, x, t, v g + z L p v, x, t, v p v, x, t g =
20 Formula for the adjointstate method Combining the two relation: \ Χ m = z Χ p m, x, t \ p m, x, t + \ L p m, x, t, m g + z L p m, x, t, m \ p m, x, t g \ Χ m = z Χ p m, x, t + z L p m, x, t, m g m p(m, x, t) + \ L p m, x, t, m g To remove the m p(m, x, t), we use the adjoint term g, imposing: z L p m, x, t, m g = z Χ p m, x, t If L is a linear operator L g m, x, t, m = z Χ p m, x, t, that is is the adjoint equation
Registration-guided least-squares waveform inversion
Registration-guided least-squares waveform inversion Hyoungsu Baek 1, Henri Calandra, Laurent Demanet 1 1 MIT Mathematics department, TOTAL S.A. January 15 013 Abstract Full waveform inversion with frequency
More informationFull Waveform Inversion (FWI) with wave-equation migration. Gary Margrave Rob Ferguson Chad Hogan Banff, 3 Dec. 2010
Full Waveform Inversion (FWI) with wave-equation migration (WEM) and well control Gary Margrave Rob Ferguson Chad Hogan Banff, 3 Dec. 2010 Outline The FWI cycle The fundamental theorem of FWI Understanding
More informationA 27-point scheme for a 3D frequency-domain scalar wave equation based on an average-derivative method
Geophysical Prospecting 04 6 58 77 doi: 0./365-478.090 A 7-point scheme for a 3D frequency-domain scalar wave equation based on an average-derivative method Jing-Bo Chen Key Laboratory of Petroleum Resources
More informationComparison between least-squares reverse time migration and full-waveform inversion
Comparison between least-squares reverse time migration and full-waveform inversion Lei Yang, Daniel O. Trad and Wenyong Pan Summary The inverse problem in exploration geophysics usually consists of two
More informationarxiv: v1 [math.na] 1 Apr 2015
Nonlinear seismic imaging via reduced order model backprojection Alexander V. Mamonov, University of Houston; Vladimir Druskin and Mikhail Zaslavsky, Schlumberger arxiv:1504.00094v1 [math.na] 1 Apr 2015
More informationMigration with Implicit Solvers for the Time-harmonic Helmholtz
Migration with Implicit Solvers for the Time-harmonic Helmholtz Yogi A. Erlangga, Felix J. Herrmann Seismic Laboratory for Imaging and Modeling, The University of British Columbia {yerlangga,fherrmann}@eos.ubc.ca
More informationNonlinear seismic imaging via reduced order model backprojection
Nonlinear seismic imaging via reduced order model backprojection Alexander V. Mamonov, Vladimir Druskin 2 and Mikhail Zaslavsky 2 University of Houston, 2 Schlumberger-Doll Research Center Mamonov, Druskin,
More informationVollständige Inversion seismischer Wellenfelder - Erderkundung im oberflächennahen Bereich
Seminar des FA Ultraschallprüfung Vortrag 1 More info about this article: http://www.ndt.net/?id=20944 Vollständige Inversion seismischer Wellenfelder - Erderkundung im oberflächennahen Bereich Thomas
More informationTwo-Scale Wave Equation Modeling for Seismic Inversion
Two-Scale Wave Equation Modeling for Seismic Inversion Susan E. Minkoff Department of Mathematics and Statistics University of Maryland Baltimore County Baltimore, MD 21250, USA RICAM Workshop 3: Wave
More informationAttenuation compensation in least-squares reverse time migration using the visco-acoustic wave equation
Attenuation compensation in least-squares reverse time migration using the visco-acoustic wave equation Gaurav Dutta, Kai Lu, Xin Wang and Gerard T. Schuster, King Abdullah University of Science and Technology
More informationSUMMARY REVIEW OF THE FREQUENCY DOMAIN L2 FWI-HESSIAN
Efficient stochastic Hessian estimation for full waveform inversion Lucas A. Willemsen, Alison E. Malcolm and Russell J. Hewett, Massachusetts Institute of Technology SUMMARY In this abstract we present
More informationAccelerating. Extended Least Squares Migration. Jie Hou TRIP 2014 Review Meeting May 1st, 2015
Accelerating Extended Least Squares Migration Jie Hou TRIP 2014 Review Meeting May 1st, 2015 Linearized Inverse Problem Fm = d Adjoint Operator m = F T d Approximate Inverse m = F d 2 Linearized Inverse
More information1.2 Derivation. d p f = d p f(x(p)) = x fd p x (= f x x p ). (1) Second, g x x p + g p = 0. d p f = f x g 1. The expression f x gx
PDE-constrained optimization and the adjoint method Andrew M. Bradley November 16, 21 PDE-constrained optimization and the adjoint method for solving these and related problems appear in a wide range of
More informationOverview Avoiding Cycle Skipping: Model Extension MSWI: Space-time Extension Numerical Examples
Overview Avoiding Cycle Skipping: Model Extension MSWI: Space-time Extension Numerical Examples Guanghui Huang Education University of Chinese Academy of Sciences, Beijing, China Ph.D. in Computational
More informationSeismic Imaging. William W. Symes. C. I. M. E. Martina Franca September
Seismic Imaging William W. Symes C. I. M. E. Martina Franca September 2002 www.trip.caam.rice.edu 1 0 offset (km) -4-3 -2-1 1 2 time (s) 3 4 5 How do you turn lots of this... (field seismogram from the
More informationThe Math, Science and Computation of Hydraulic Fracturing
The Math, Science and Computation of Hydraulic Fracturing 03/21/2013 LLNL-PRES-XXXXXX This work was performed under the auspices of the U.S. Department of Energy by under contract DE-AC52-07NA27344. Lawrence
More informationYouzuo Lin and Lianjie Huang
PROCEEDINGS, Thirty-Ninth Workshop on Geothermal Reservoir Engineering Stanford University, Stanford, California, February 24-26, 2014 SGP-TR-202 Building Subsurface Velocity Models with Sharp Interfaces
More informationUncertainty quantification for Wavefield Reconstruction Inversion
Uncertainty quantification for Wavefield Reconstruction Inversion Zhilong Fang *, Chia Ying Lee, Curt Da Silva *, Felix J. Herrmann *, and Rachel Kuske * Seismic Laboratory for Imaging and Modeling (SLIM),
More informationAcoustic Wave Equation
Acoustic Wave Equation Sjoerd de Ridder (most of the slides) & Biondo Biondi January 16 th 2011 Table of Topics Basic Acoustic Equations Wave Equation Finite Differences Finite Difference Solution Pseudospectral
More informationA projected Hessian for full waveform inversion
CWP-679 A projected Hessian for full waveform inversion Yong Ma & Dave Hale Center for Wave Phenomena, Colorado School of Mines, Golden, CO 80401, USA (c) Figure 1. Update directions for one iteration
More informationOne-step extrapolation method for reverse time migration
GEOPHYSICS VOL. 74 NO. 4 JULY-AUGUST 2009; P. A29 A33 5 FIGS. 10.1190/1.3123476 One-step extrapolation method for reverse time migration Yu Zhang 1 and Guanquan Zhang 2 ABSTRACT We have proposed a new
More informationSeismic imaging and optimal transport
Seismic imaging and optimal transport Bjorn Engquist In collaboration with Brittany Froese, Sergey Fomel and Yunan Yang Brenier60, Calculus of Variations and Optimal Transportation, Paris, January 10-13,
More informationBeam Propagation Method Solution to the Seminar Tasks
Beam Propagation Method Solution to the Seminar Tasks Matthias Zilk The task was to implement a 1D beam propagation method (BPM) that solves the equation z v(xz) = i 2 [ 2k x 2 + (x) k 2 ik2 v(x, z) =
More informationCourse on Inverse Problems Albert Tarantola
California Institute of Technology Division of Geological and Planetary Sciences March 26 - May 25, 27 Course on Inverse Problems Albert Tarantola Institut de Physique du Globe de Paris CONCLUSION OF THE
More informationCourse on Inverse Problems
California Institute of Technology Division of Geological and Planetary Sciences March 26 - May 25, 2007 Course on Inverse Problems Albert Tarantola Institut de Physique du Globe de Paris Lesson XVI CONCLUSION
More informationFinite difference modelling of the full acoustic wave equation in Matlab
Acoustic finite difference modelling Finite difference modelling of the full acoustic wave equation in Matlab Hugh D Geiger and Pat F Daley ABSTRACT Two subroutines have been added to the Matlab AFD (acoustic
More informationWaveform inversion for attenuation estimation in anisotropic media Tong Bai & Ilya Tsvankin Center for Wave Phenomena, Colorado School of Mines
Waveform inversion for attenuation estimation in anisotropic media Tong Bai & Ilya Tsvankin Center for Wave Phenomena, Colorado School of Mines SUMMARY Robust estimation of attenuation coefficients remains
More informationFull Waveform Inversion via Matched Source Extension
Full Waveform Inversion via Matched Source Extension Guanghui Huang and William W. Symes TRIP, Department of Computational and Applied Mathematics May 1, 215, TRIP Annual Meeting G. Huang and W. W. Symes
More informationAn explicit time-domain finite-element method for room acoustics simulation
An explicit time-domain finite-element method for room acoustics simulation Takeshi OKUZONO 1 ; Toru OTSURU 2 ; Kimihiro SAKAGAMI 3 1 Kobe University, JAPAN 2 Oita University, JAPAN 3 Kobe University,
More informationON ESTIMATION OF TEMPERATURE UNCERTAINTY USING THE SECOND ORDER ADJOINT PROBLEM
ON ESTIMATION OF TEMPERATURE UNCERTAINTY USING THE SECOND ORDER ADJOINT PROBLEM Aleksey K. Alekseev a, Michael I. Navon b,* a Department of Aerodynamics and Heat Transfer, RSC, ENERGIA, Korolev (Kaliningrad),
More informationWavefield Reconstruction Inversion (WRI) a new take on wave-equation based inversion Felix J. Herrmann
Wavefield Reconstruction Inversion (WRI) a new take on wave-equation based inversion Felix J. Herrmann SLIM University of British Columbia van Leeuwen, T and Herrmann, F J (2013). Mitigating local minima
More informationFinite Difference Methods for
CE 601: Numerical Methods Lecture 33 Finite Difference Methods for PDEs Course Coordinator: Course Coordinator: Dr. Suresh A. Kartha, Associate Professor, Department of Civil Engineering, IIT Guwahati.
More information5. A step beyond linearization: velocity analysis
5. A step beyond linearization: velocity analysis 1 Partially linearized seismic inverse problem ( velocity analysis ): given observed seismic data S obs, find smooth velocity v E(X), X R 3 oscillatory
More information23855 Rock Physics Constraints on Seismic Inversion
23855 Rock Physics Constraints on Seismic Inversion M. Sams* (Ikon Science Ltd) & D. Saussus (Ikon Science) SUMMARY Seismic data are bandlimited, offset limited and noisy. Consequently interpretation of
More informationNonlinear acoustic imaging via reduced order model backprojection
Nonlinear acoustic imaging via reduced order model backprojection Alexander V. Mamonov 1, Vladimir Druskin 2, Andrew Thaler 3 and Mikhail Zaslavsky 2 1 University of Houston, 2 Schlumberger-Doll Research
More informationWe E Multiparameter Full-waveform Inversion for Acoustic VTI Medium with Surface Seismic Data
We E16 4 Multiarameter Full-waveform Inversion for Acoustic VI Medium with Surface Seismic Data X. Cheng* (Schlumberger) K. Jiao (Schlumberger) D. Sun (Schlumberger) & D. Vigh (Schlumberger) SUMMARY In
More informationApproximate- vs. full-hessian in FWI: 1D analytical and numerical experiments
Approximate- vs. full-hessian in FWI: 1D analytical and numerical experiments Raul Cova and Kris Innanen ABSTRACT Feasibility of using Full Waveform Inversion (FWI) to build velocity models has been increasing
More informationComments on the Geophysics paper Multiparameter 11norm. waveform fitting: Interpretation of Gulf of Mexico reflection
,..-., Comments on the Geophysics paper Multiparameter 11norm waveform fitting: Interpretation of Gulf of Mexico reflection seismograms by H. Djikp&s6 and A. Tarantola Susan E. Minkoff * t and William
More informationAdditive Manufacturing Module 8
Additive Manufacturing Module 8 Spring 2015 Wenchao Zhou zhouw@uark.edu (479) 575-7250 The Department of Mechanical Engineering University of Arkansas, Fayetteville 1 Evaluating design https://www.youtube.com/watch?v=p
More information5. FVM discretization and Solution Procedure
5. FVM discretization and Solution Procedure 1. The fluid domain is divided into a finite number of control volumes (cells of a computational grid). 2. Integral form of the conservation equations are discretized
More informationq is large, to ensure accuracy of the spatial discretization 1 In practice, q = 6 is a good choice.
Time-stepping beyond CFL: a locally one-dimensional scheme for acoustic wave propagation Leonardo Zepeda-Núñez 1*), Russell J. Hewett 1, Minghua Michel Rao 2, and Laurent Demanet 1 1 Dept. of Mathematics
More informationNumerical Methods of Applied Mathematics -- II Spring 2009
MIT OpenCourseWare http://ocw.mit.edu 18.336 Numerical Methods of Applied Mathematics -- II Spring 2009 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.
More informationTime Adaptivity Stable Finite Difference for Acoustic Wave Equation
162 Int'l Conf Scientific Computing CSC'15 Time Adaptivity Stable Finite Difference for Acoustic Wave Equation A J M Antunes 1, R C P Leal-Toledo 2, E M Toledo 3, O T S Filho 2, and E Marques 2 1 SEEDUC-FAETEC,
More informationFull waveform inversion with wave equation migration and well control
FWI by WEM Full waveform inversion with wave equation migration and well control Gary F. Margrave, Robert J. Ferguson and Chad M. Hogan ABSTRACT We examine the key concepts in full waveform inversion (FWI)
More informationSource function estimation in extended full waveform inversion
Source function estimation in extended full waveform inversion Lei Fu The Rice Inversion Project (TRIP) May 6, 2014 Lei Fu (TRIP) Source estimation in EFWI May 6, 2014 1 Overview Objective Recover Earth
More informationAirfoil shape optimization using adjoint method and automatic differentiation
Airfoil shape optimization using adjoint method and automatic differentiation Praveen. C praveen@math.tifrbng.res.in Tata Institute of Fundamental Research Center for Applicable Mathematics Bangalore 5665
More informationChapter 7. Seismic imaging. 7.1 Assumptions and vocabulary
Chapter 7 Seismic imaging Much of the imaging procedure was already described in the previous chapters. An image, or a gradient update, is formed from the imaging condition by means of the incident and
More informationSource estimation for frequency-domain FWI with robust penalties
Source estimation for frequency-domain FWI with robust penalties Aleksandr Y. Aravkin, Tristan van Leeuwen, Henri Calandra, and Felix J. Herrmann Dept. of Earth and Ocean sciences University of British
More informationFRÉCHET KERNELS AND THE ADJOINT METHOD
PART II FRÉCHET KERNES AND THE ADJOINT METHOD 1. Setp of the tomographic problem: Why gradients? 2. The adjoint method 3. Practical 4. Special topics (sorce imaging and time reversal) Setp of the tomographic
More informationMicroseismic Event Estimation Via Full Waveform Inversion
Microseismic Event Estimation Via Full Waveform Inversion Susan E. Minkoff 1, Jordan Kaderli 1, Matt McChesney 2, and George McMechan 2 1 Department of Mathematical Sciences, University of Texas at Dallas
More informationMEMORIAL UNIVERSITY OF NEWFOUNDLAND
MEMORIAL UNIVERSITY OF NEWFOUNDLAND DEPARTMENT OF MATHEMATICS AND STATISTICS Section 5. Math 090 Fall 009 SOLUTIONS. a) Using long division of polynomials, we have x + x x x + ) x 4 4x + x + 0x x 4 6x
More information(0, 0), (1, ), (2, ), (3, ), (4, ), (5, ), (6, ).
1 Interpolation: The method of constructing new data points within the range of a finite set of known data points That is if (x i, y i ), i = 1, N are known, with y i the dependent variable and x i [x
More informationSUMMARY ANGLE DECOMPOSITION INTRODUCTION. A conventional cross-correlation imaging condition for wave-equation migration is (Claerbout, 1985)
Comparison of angle decomposition methods for wave-equation migration Natalya Patrikeeva and Paul Sava, Center for Wave Phenomena, Colorado School of Mines SUMMARY Angle domain common image gathers offer
More informationSeismic Modeling, Migration and Velocity Inversion
Seismic Modeling, Migration and Velocity Inversion Inverse Scattering Bee Bednar Panorama Technologies, Inc. 14811 St Marys Lane, Suite 150 Houston TX 77079 May 30, 2014 Bee Bednar (Panorama Technologies)
More informationThe Finite Difference Method
The Finite Difference Method Heiner Igel Department of Earth and Environmental Sciences Ludwig-Maximilians-University Munich Heiner Igel Computational Seismology 1 / 32 Outline 1 Introduction Motivation
More informationINTERPOLATION. and y i = cos x i, i = 0, 1, 2 This gives us the three points. Now find a quadratic polynomial. p(x) = a 0 + a 1 x + a 2 x 2.
INTERPOLATION Interpolation is a process of finding a formula (often a polynomial) whose graph will pass through a given set of points (x, y). As an example, consider defining and x 0 = 0, x 1 = π/4, x
More informationQuiescent Steady State (DC) Analysis The Newton-Raphson Method
Quiescent Steady State (DC) Analysis The Newton-Raphson Method J. Roychowdhury, University of California at Berkeley Slide 1 Solving the System's DAEs DAEs: many types of solutions useful DC steady state:
More informationAM 205: lecture 14. Last time: Boundary value problems Today: Numerical solution of PDEs
AM 205: lecture 14 Last time: Boundary value problems Today: Numerical solution of PDEs ODE BVPs A more general approach is to formulate a coupled system of equations for the BVP based on a finite difference
More informationPDE Solvers for Fluid Flow
PDE Solvers for Fluid Flow issues and algorithms for the Streaming Supercomputer Eran Guendelman February 5, 2002 Topics Equations for incompressible fluid flow 3 model PDEs: Hyperbolic, Elliptic, Parabolic
More informationFull waveform inversion and the inverse Hessian
Full waveform inversion and the inverse Hessian Gary Margrave, Matt Yedlin and Kris Innanen ABSTRACT FWI and the inverse Hessian Full waveform inversion involves defining an objective function, and then
More informationA Hybrid Method for the Wave Equation. beilina
A Hybrid Method for the Wave Equation http://www.math.unibas.ch/ beilina 1 The mathematical model The model problem is the wave equation 2 u t 2 = (a 2 u) + f, x Ω R 3, t > 0, (1) u(x, 0) = 0, x Ω, (2)
More informationAPPLICATIONS OF FD APPROXIMATIONS FOR SOLVING ORDINARY DIFFERENTIAL EQUATIONS
LECTURE 10 APPLICATIONS OF FD APPROXIMATIONS FOR SOLVING ORDINARY DIFFERENTIAL EQUATIONS Ordinary Differential Equations Initial Value Problems For Initial Value problems (IVP s), conditions are specified
More informationFull-Waveform Inversion with Gauss- Newton-Krylov Method
Full-Waveform Inversion with Gauss- Newton-Krylov Method Yogi A. Erlangga and Felix J. Herrmann {yerlangga,fherrmann}@eos.ubc.ca Seismic Laboratory for Imaging and Modeling The University of British Columbia
More informationSatish Singh* (IPG Paris, France, Tim Sears (British Gas, UK), Mark Roberts (IPG Paris, Summary. Introduction P - 92
P - 92 Fine-Scale P- and S-Wave Velocities From Elastic Full Waveform Inversion of Multi-Component and Time-Lapse Data: Future of Quantitative Seismic Imaging Satish Singh* (IPG Paris, France, singh@ipgp.jussieu.fr),
More informationAchieving depth resolution with gradient array survey data through transient electromagnetic inversion
Achieving depth resolution with gradient array survey data through transient electromagnetic inversion Downloaded /1/17 to 128.189.118.. Redistribution subject to SEG license or copyright; see Terms of
More informationLab 5: 16 th April Exercises on Neural Networks
Lab 5: 16 th April 01 Exercises on Neural Networks 1. What are the values of weights w 0, w 1, and w for the perceptron whose decision surface is illustrated in the figure? Assume the surface crosses the
More information8-1: Backpropagation Prof. J.C. Kao, UCLA. Backpropagation. Chain rule for the derivatives Backpropagation graphs Examples
8-1: Backpropagation Prof. J.C. Kao, UCLA Backpropagation Chain rule for the derivatives Backpropagation graphs Examples 8-2: Backpropagation Prof. J.C. Kao, UCLA Motivation for backpropagation To do gradient
More informationSpring 2014: Computational and Variational Methods for Inverse Problems CSE 397/GEO 391/ME 397/ORI 397 Assignment 4 (due 14 April 2014)
Spring 2014: Computational and Variational Methods for Inverse Problems CSE 397/GEO 391/ME 397/ORI 397 Assignment 4 (due 14 April 2014) The first problem in this assignment is a paper-and-pencil exercise
More informationNumerical Analysis: Interpolation Part 1
Numerical Analysis: Interpolation Part 1 Computer Science, Ben-Gurion University (slides based mostly on Prof. Ben-Shahar s notes) 2018/2019, Fall Semester BGU CS Interpolation (ver. 1.00) AY 2018/2019,
More informationMethod in 1D. Introduction of the Basic Concepts and Comparison to Optimal FD Operators
The Spectral Element Method in 1D Introduction of the Basic Concepts and Comparison to Optimal FD Operators Seismology and Seismics Seminar 17.06.2003 Ludwigs-Maximilians-Universität München by Bernhard
More informationImage-space wave-equation tomography in the generalized source domain
Image-space wave-equation tomography in the generalized source domain Yaxun Tang, Claudio Guerra, and Biondo Biondi ABSTRACT We extend the theory of image-space wave-equation tomography to the generalized
More informationIdentify polynomial functions
EXAMPLE 1 Identify polynomial functions Decide whether the function is a polynomial function. If so, write it in standard form and state its degree, type, and leading coefficient. a. h (x) = x 4 1 x 2
More informationPreliminary Examination in Numerical Analysis
Department of Applied Mathematics Preliminary Examination in Numerical Analysis August 7, 06, 0 am pm. Submit solutions to four (and no more) of the following six problems. Show all your work, and justify
More informationGEOPHYSICAL INVERSE THEORY AND REGULARIZATION PROBLEMS
Methods in Geochemistry and Geophysics, 36 GEOPHYSICAL INVERSE THEORY AND REGULARIZATION PROBLEMS Michael S. ZHDANOV University of Utah Salt Lake City UTAH, U.S.A. 2OO2 ELSEVIER Amsterdam - Boston - London
More informationApplication of Wavelets to N body Particle In Cell Simulations
Application of Wavelets to N body Particle In Cell Simulations Balša Terzić (NIU) Northern Illinois University, Department of Physics April 7, 2006 Motivation Gaining insight into the dynamics of multi
More informationA Brief Introduction to Numerical Methods for Differential Equations
A Brief Introduction to Numerical Methods for Differential Equations January 10, 2011 This tutorial introduces some basic numerical computation techniques that are useful for the simulation and analysis
More informationThe Inversion Problem: solving parameters inversion and assimilation problems
The Inversion Problem: solving parameters inversion and assimilation problems UE Numerical Methods Workshop Romain Brossier romain.brossier@univ-grenoble-alpes.fr ISTerre, Univ. Grenoble Alpes Master 08/09/2016
More informationTowards Modelling Elastic and Viscoelastic Seismic Wave Propagation in Boreholes
Towards Modelling Elastic and Viscoelastic Seismic Wave Propagation in Boreholes NA WANG, DONG SHI, BERND MILKEREIT Department of Physics, University of Toronto, Toronto, Canada M5S 1A7 Summary We are
More informationAttenuation compensation in viscoacoustic reserve-time migration Jianyong Bai*, Guoquan Chen, David Yingst, and Jacques Leveille, ION Geophysical
Attenuation compensation in viscoacoustic reserve-time migration Jianyong Bai*, Guoquan Chen, David Yingst, and Jacques Leveille, ION Geophysical Summary Seismic waves are attenuated during propagation.
More informationNONLINEAR DIFFUSION PDES
NONLINEAR DIFFUSION PDES Erkut Erdem Hacettepe University March 5 th, 0 CONTENTS Perona-Malik Type Nonlinear Diffusion Edge Enhancing Diffusion 5 References 7 PERONA-MALIK TYPE NONLINEAR DIFFUSION The
More informationMathematics of Seismic Imaging Part 1: Modeling and Inverse Problems
Mathematics of Seismic Imaging Part 1: Modeling and Inverse Problems William W. Symes Rice University 0. Introduction How do you turn lots of this... 0 offset (km) -4-3 -2-1 1 2 time (s) 3 4 5 (field seismogram
More informationProgress and Prospects for Wave Equation MVA
Progress and Prospects for Wave Equation MVA William W. Symes TRIP Annual Review, January 007 Agenda MVA, Nonlinear Inversion, and Extended Modeling MVA with two different extensions: the coherent noise
More informationMultiple Scenario Inversion of Reflection Seismic Prestack Data
Downloaded from orbit.dtu.dk on: Nov 28, 2018 Multiple Scenario Inversion of Reflection Seismic Prestack Data Hansen, Thomas Mejer; Cordua, Knud Skou; Mosegaard, Klaus Publication date: 2013 Document Version
More information4. Multilayer Perceptrons
4. Multilayer Perceptrons This is a supervised error-correction learning algorithm. 1 4.1 Introduction A multilayer feedforward network consists of an input layer, one or more hidden layers, and an output
More informationWenyong Pan and Lianjie Huang. Los Alamos National Laboratory, Geophysics Group, MS D452, Los Alamos, NM 87545, USA
PROCEEDINGS, 44th Workshop on Geothermal Reservoir Engineering Stanford University, Stanford, California, February 11-13, 019 SGP-TR-14 Adaptive Viscoelastic-Waveform Inversion Using the Local Wavelet
More informationWave-equation tomography for anisotropic parameters
Wave-equation tomography for anisotropic parameters Yunyue (Elita) Li and Biondo Biondi ABSTRACT Anisotropic models are recognized as more realistic representations of the subsurface where complex geological
More informationDOING PHYSICS WITH MATLAB WAVE MOTION
DOING PHYSICS WITH MATLAB WAVE MOTION THE [1D] SCALAR WAVE EQUATION THE FINITE DIFFERENCE TIME DOMAIN METHOD Ian Cooper School of Physics, University of Sydney ian.cooper@sydney.edu.au DOWNLOAD DIRECTORY
More informationOptimization schemes for Full Waveform Inversion: the preconditioned truncated Newton method
Optimization schemes for Full Waveform Inversion: the preconditioned truncated Newton method Ludovic Métivier, Romain Brossier, Jean Virieux, Stéphane Operto To cite this version: Ludovic Métivier, Romain
More informationA posteriori error estimates for the adaptivity technique for the Tikhonov functional and global convergence for a coefficient inverse problem
A posteriori error estimates for the adaptivity technique for the Tikhonov functional and global convergence for a coefficient inverse problem Larisa Beilina Michael V. Klibanov December 18, 29 Abstract
More information10.3 Steepest Descent Techniques
The advantage of the Newton and quasi-newton methods for solving systems of nonlinear equations is their speed of convergence once a sufficiently accurate approximation is known. A weakness of these methods
More informationA Nonlinear Differential Semblance Strategy for Waveform Inversion: Experiments in Layered Media Dong Sun and William W Symes, Rice University
A Nonlinear Differential Semblance Strategy for Waveform Inversion: Experiments in Layered Media Dong Sun and William W Symes, Rice University SUMMARY This paper proposes an alternative approach to the
More informationFraunhofer Institute for Computer Graphics Research Interactive Graphics Systems Group, TU Darmstadt Fraunhoferstrasse 5, Darmstadt, Germany
Scale Space and PDE methods in image analysis and processing Arjan Kuijper Fraunhofer Institute for Computer Graphics Research Interactive Graphics Systems Group, TU Darmstadt Fraunhoferstrasse 5, 64283
More informationOptimal transport for Seismic Imaging
Optimal transport for Seismic Imaging Bjorn Engquist In collaboration with Brittany Froese and Yunan Yang ICERM Workshop - Recent Advances in Seismic Modeling and Inversion: From Analysis to Applications,
More informationA novel full waveform inversion method based on a timeshift nonlinear operator
A novel full waveform inversion method based on a timeshift nonlinear operator Journal: Manuscript ID GJI-S--0.R Manuscript Type: Research Paper Date Submitted by the Author: -Dec-0 Complete List of Authors:
More informationChapter Two: Numerical Methods for Elliptic PDEs. 1 Finite Difference Methods for Elliptic PDEs
Chapter Two: Numerical Methods for Elliptic PDEs Finite Difference Methods for Elliptic PDEs.. Finite difference scheme. We consider a simple example u := subject to Dirichlet boundary conditions ( ) u
More informationMixed-grid and staggered-grid finite-difference methods for frequency-domain acoustic wave modelling
Geophys. J. Int. 2004 57, 269 296 doi: 0./j.365-246X.2004.02289.x Mixed-grid and staggered-grid finite-difference methods for frequency-domain acoustic wave modelling Bernhard Hustedt, Stéphane Operto
More informationDetailed Program of the Workshop
Detailed Program of the Workshop Inverse Problems: Theoretical and Numerical Aspects Laboratoire de Mathématiques de Reims, December 17-19 2018 December 17 14h30 Opening of the Workhop 15h00-16h00 Mourad
More informationScattering-based decomposition of sensitivity kernels for full-waveform inversion
Scattering-based decomposition of sensitivity kernels for full-waveform inversion Daniel Leal Macedo (Universidade Estadual de Campinas) and Ivan asconcelos (Schlumberger Cambridge Research) Copyright
More informationSUMMARY. (Sun and Symes, 2012; Biondi and Almomin, 2012; Almomin and Biondi, 2012).
Inversion Velocity Analysis via Differential Semblance Optimization in the Depth-oriented Extension Yujin Liu, William W. Symes, Yin Huang,and Zhenchun Li, China University of Petroleum (Huadong), Rice
More informationGeothermal Reservoir Imaging Using 2016 Walkaway VSP Data from the Raft River Geothermal Field
PROCEEDINGS, 44th Workshop on Geothermal Reservoir Engineering Stanford University, Stanford, California, February 11-13, 2019 SGP-TR-214 Geothermal Reservoir Imaging Using 2016 Walkaway VSP Data from
More information