3D frequency-domain elastic wave modeling with the spectral element method using a massively parallel direct solver

Geophysics ◽  
2020 ◽  
Vol 85 (2) ◽  
pp. T71-T88
Author(s):  
Yang Li ◽  
Romain Brossier ◽  
Ludovic Métivier

ABSTRACT Complex topography, the free-surface boundary condition, and anelastic properties of media should be accounted for in the frame of onshore geophysical prospecting imaging, such as full-waveform inversion (FWI). In this context, an accurate and efficient forward-modeling engine is mandatory. We have performed 3D frequency-domain anisotropic elastic wave modeling by using the highly accurate spectral element method and a sparse multifrontal direct solver. An efficient approach similar to computing the matrix-vector product in the time domain is used to build the matrix. We validate the numerical results by comparing with analytical solutions. A parallel direct solver, the sparse direct multifrontal massively parallel solver (MUMPS), is used to solve the linear system. We find that a hybrid implementation of message passing interface and open multiprocessing is more efficient in flops and memory cost. The influence of the deformed mesh, free-surface boundary condition, and heterogeneity of media on MUMPS performance is negligible. Complexity analysis suggests that the memory complexity of MUMPS agrees with the theoretical order [Formula: see text] (or [Formula: see text] with an efficient matrix reordering method) for an [Formula: see text] grid when nontrivial topography is considered. With the available resources, we conduct a moderate scale modeling with a subset of the SEAM Phase II Foothills model, where 60 wavelengths in the [Formula: see text]-axis are propagated. Computing one gradient of FWI based on this model using the frequency-domain modeling is shown to require similar or fewer computational resources than what would be required for a time-domain solver, depending on the number of sources, while larger memory is necessary. An estimation of the increasing trend indicates that approximately 20 Tb of memory would be required for a [Formula: see text] wavelength modeling. The limit of MUMPS scalability hinders the application to larger scale applications.

Author(s):  
M. D. Yang ◽  
B. Teng

A time-domain simulation method is developed for the coupled dynamic analysis of a spar platform with mooring lines. For the hydrodynamic loads, a time domain second order method is developed. In this approach, Taylor series expansions are applied to the body surface boundary condition and the free surface boundary condition, and Stokes perturbation procedure is then used to establish corresponding boundary value problems with time-independent boundaries. A higher order boundary element method is developed to calculate the velocity potential of the resulting flow field at each time step. The free-surface boundary condition is satisfied to the second order by 4th order Adams-Bashforth-Moultn method. An artificial damping layer is adopted on the free surface to avoid the wave reflection. For the mooring-line dynamics, a geometrically nonlinear finite element method using isoparametric cable element based on the total Lagrangian formulation is developed. In the coupled dynamic analysis, the motion equation for the hull and dynamic equations for mooring lines are solved simultaneously using Newmark method. Numerical results including motions and tensions in the mooring lines are presented.


1986 ◽  
Vol 108 (4) ◽  
pp. 645-654 ◽  
Author(s):  
R. H. Buckholz ◽  
J. F. Lin

An analysis for hydrodynamic, non-Newtonian lubrication of misaligned journal bearings is given. The hydrodynamic load-carrying capacity for partial arc journal bearings lubricated by power-law, non-Newtonian fluids is calculated for small valves of the bearing aspect ratios. These results are compared with: numerical solutions to the non-Newtonian modified Reynolds equation, with Ocvirk’s experimental results for misaligned bearings, and with other numerical simulations. The cavitation (i.e., film rupture) boundary location is calculated using the Reynolds’ free-surface, boundary condition.


1993 ◽  
Author(s):  
Bruce S. Rosen ◽  
Joseph P. Laiosa ◽  
Warren H. Davis ◽  
David Stavetski

A unique free-surface flow methodology and its application to design and analysis of IACC yachts are discussed. Numerical aspects of the inviscid panel code and details of the free-surface boundary condition are included, along with enhancements developed specifically for the '92 America's Cup defense. Extensive code validation using wind tunnel and towing tank experimental data address several areas of interest to the yacht designer. Lift and induced drag at zero Froude number are studied via a series of isolated fin/bulb/winglet appendages. An isolated surface piercing foil is used to evaluate simple lift/free­surface interactions. For complete IACC yacht models, upright wave resistance is investigated, as well as lift and induced drag at heel and yaw. The excellent correlation obtained for these cases demonstrates the value of this linear free-surface methodology for use in designing high performance sailing yachts.


2016 ◽  
Vol 811 ◽  
pp. 400-420 ◽  
Author(s):  
Yong Liang ◽  
Ahmad Zareei ◽  
Mohammad-Reza Alam

Here we show that there exist internal gravity waves that are inherently unstable, that is, they cannot exist in nature for a long time. The instability mechanism is a one-way (irreversible) harmonic-generation resonance that permanently transfers the energy of an internal wave to its higher harmonics. We show that, in fact, there are a countably infinite number of such unstable waves. For the harmonic-generation resonance to take place, the nonlinear terms in the free surface boundary condition play a pivotal role, and the instability does not occur in a linearly stratified fluid if a simplified boundary condition, such as a rigid lid or a linearized boundary condition, is employed. Harmonic-generation resonance presented here provides a mechanism for the transfer of internal wave energy to the higher-frequency part of the spectrum hence affecting, potentially significantly, the evolution of the internal waves spectrum.


2019 ◽  
Vol 145 (6) ◽  
pp. 3299-3310 ◽  
Author(s):  
Finnur Pind ◽  
Allan P. Engsig-Karup ◽  
Cheol-Ho Jeong ◽  
Jan S. Hesthaven ◽  
Mikael S. Mejling ◽  
...  

2019 ◽  
Vol 63 (4) ◽  
pp. 251-267 ◽  
Author(s):  
Zhi-Ming Yuan ◽  
Liang Li ◽  
Ronald W. Yeung

Ships often have to pass each other in proximity in harbor areas and waterways in dense shipping-traffic environment. Hydrodynamic interaction occurs when a ship is overtaking (or being overtaken) or encountering other ships. Such an interactive effect could be magnified in confined waterways, e.g., shallow and narrow rivers. Since Yeung published his initial work on ship interaction in shallow water, progress on unsteady interaction among multiple ships has been slow, though steady, over the following decades. With some exceptions, nearly all the published studies on ship-to-ship problem neglected free-surface effects, and a rigid-wall condition has often been applied on the water surface as the boundary condition. When the speed of the ships is low, this assumption is reasonably accurate as the hydrodynamic interaction is mainly induced by near-field disturbances. However, in many maneuvering operations, the encountering or overtaking speeds are actually moderately high (Froude number Fn > 0.2, where <inline-graphic xlink:href="josr10180089inf1.tif"/>, U is ship speed, g is the gravitational acceleration, and L is the ship length), especially when the lateral separation between ships is the order of ship length. Here, the far-field effects arising from ship waves can be important. The hydrodynamic interaction model must take into account the surface-wave effects. Classical potential-flow formulation is only able to deal with the boundary value problem when there is only one speed involved in the free-surface boundary condition. For multiple ships traveling with different speeds, it is not possible to express the free-surface boundary condition by a single velocity potential. Instead, a superposition method can be applied to account for the velocity field induced by each vessel with its own and unique speed. The main objective of the present article is to propose a rational superposition method to handle the unsteady free-surface boundary condition containing two or more speed terms, and validate its feasibility in predicting the hydrodynamic behavior in ship encountering. The methodology used in the present article is a three-dimensional boundary-element method based on a Rankine-type (infinite-space) source function, initially introduced by Bai and Yeung. The numerical simulations are conducted by using an in-house‐developed multibody hydrodynamic interaction program “MHydro.” Waves generated and forces (or moments) are calculated when ships are encountering or passing each other. Published model-test results are used to validate our calculations, and very good agreement has been observed. The numerical results show that free-surface effects need to be taken into account for Fn > 0.2.


Sign in / Sign up

Export Citation Format

Share Document