• Tidak ada hasil yang ditemukan

silent boundary methods· for transient wave analysis

N/A
N/A
Protected

Academic year: 2023

Membagikan "silent boundary methods· for transient wave analysis"

Copied!
213
0
0

Teks penuh

The second part of the analysis calculates Es*• the forces and moments caused by the seismic excitation. Roughness serves as the boundary force contribution to the interior finite element equations.

TABLE  OF  CONTENTS
TABLE OF CONTENTS

ORGANIZATION

In review, each of the proposed transmission boundary schemes has been shown to be effective for selected wave problems. Through the equilibrium equations it is shown that their underlying prescriptions are equivalent to those of the paraxial boundary.

CHAPTER 2

INTRODUCTION

ENERGY FORMS

Letting h be proportional to u in the boundary region to satisfy equation (4) is an obvious choice, but some alternatives exist. However, it does not indicate whether these boundary terms remove all energy impinging on the boundary.

Figure  1.  Simple  absorbing  boundary  for  one-dimensional  wave  propagation.
Figure 1. Simple absorbing boundary for one-dimensional wave propagation.

THEORY OF PARAXIAL BOUNDARIES

In this case, h becomes a convective term which sends energy in the positive x-direction and eventually out of the system. In the derivation herein, the paraxial boundary idea is best introduced using the one-dimensional wave equation.

TWO-DIMENSIONAL SCALAR WAVE EQUATION

If we substitute the positive root ckx/w into equation (13), then this equation represents waves traveling in the positive-x direction. We can determine the differential equation corresponding to the first three tens of equation (17) by inspecting equation (13) and its .

Figure  2.  Orientation  of  wave  used  in  the  scalar-wave,  and  elasticity  analysis
Figure 2. Orientation of wave used in the scalar-wave, and elasticity analysis

BOUNDARY APPROXIMATIONS FOR TWO-DIMENSIONAL, LINEAR ELASTICITY The analysis using the elasticity equations follows along lines

The third analogy to the case of scalar waves is that now the paraxial differential equations can be derived. In fact, the term negative stiffness raises the question of whether the paraaxial equations are stable.

Figure  3.  Displacements  and  c~ordinate  axes  utilized  in  the  plane-strain,  elasti:  problems
Figure 3. Displacements and c~ordinate axes utilized in the plane-strain, elasti: problems

STABILITY ANALYSIS

For cs< cd/2 (ie, Poisson's ratio greater than 1/3), instabilities may lead to the use of the paraxial equations. The rationale for this simple solution is that the stiffness terms are the least important element of the paraxial approximation, as found in the derivation of the paraxial equations.

Figure  4.  Schematic  representation  of  a  paraxial  boundary  for  plane  strain.
Figure 4. Schematic representation of a paraxial boundary for plane strain.

THEORETICAL COMPARISONS OF PARAXIAL TO VISCOUS BOUNDARIES A. The Viscous Boundary

The viscous boundary applies the ut and wt values ​​directly to the left side of the element. Substituting the potentials in equations (48) into equations (49) yields the elastic displacements propagating in the elastic region.

Figure  8.  Schematic  illustration  of  dilatational  and  shear  waves  reflecting  from  a  boundary
Figure 8. Schematic illustration of dilatational and shear waves reflecting from a boundary

ANGLE ~F INCIDENCE ANGLE ~F INCIDENCE

ANGLE ~F INCIDENCE

INCIDENCE

In fact, earlier authorsC47 •57l eliminated these amplitudes from the comparison when they multiplied them by the wave velocity time cos a (a= angle of incidence.) This new quantity measured the energy flux at the boundary; the energy propagating a 1 from the boundary was assumed to be confined there. Overall, considering all wave reflection curves, we conclude that each of the boundary schemes produces acceptable results.

Figure  12.  Absolute  amplitudes  of  reflected  waves  for  various  angles  of  incidence  (plane  strain)
Figure 12. Absolute amplitudes of reflected waves for various angles of incidence (plane strain)

ANGLE ~F INCIOENCE ANGLE ~F INCIDENCE

RAYLEIGH WAVES

Therefore, we want to determine the ability of the "body wave" paraxial equations to absorb Rayleigh wave motion. The values ​​of the coefficients for other Poisson ratios are in the same range as those presented in Tables 1 and 2. The effectiveness of the standard viscous boundary in the transmission of Rayleigh waves is largely untested.

In a practical sense, it is not clear that using the variable coefficients will improve the absorption of Rayleigh waves. The bound causes a significant reduction in the critical time step for the explicit part of the solution algorithm (which is described in Chapter 3).

Figure  18.  Schematic  diagram  of  a  Rayleigh  wave  approaching  a  boundary.
Figure 18. Schematic diagram of a Rayleigh wave approaching a boundary.

OTHER SILENT-BOUNDARY APPLICATIONS A. Spherically Symmetric Case

However, if the above considerations are relatively less important, then the boundary stresses (57) may be useful for Rayleigh wave applications. The derivation and implementation of the paraxial boundary, although not discussed here, follows from this analysis and the procedures described earlier in this chapter. If we now replace potentials (69) and (72) in equations (73) and (75), we get two equations for the solutions of two unknowns, AP and As.

For purposes of comparison with the plane strain case, we can determine the amplitudes of the potential $ used in the analysis in Sect. The vector potential,!· is related by the equation in which g x is the symbol for the convolved vector and z is the unit vector in the z direction.

Figure  20.  Comparison  of  the  Bessel  function,  J
Figure 20. Comparison of the Bessel function, J

ANGLE ~F

ANGLE DF INCIDENCE

ANGLE DF

RNGLE CTF INCIDENCE RNGLE CTF INCIDENCE

The analyzes of Chapter 2 suggest that the boundary schemes discussed earlier can reproduce most of the effects of an infinite domain. We discuss the problems of using a paraxial boundary and then implementing two viscous boundaries. We can evaluate the behavior of the extended-paraxial equations, for a one-dimensional mesh which is drawn in figure 1.

As ;1 moves to the left, less and less ripple is transferred to the right. This technique is one way of "weighting" the integration of finite elements in the flow direction.

Figure  1.  Schematic  diagram  of  an  elastic  bar  which  is  discretized  into  a  one-dimensional  mesh
Figure 1. Schematic diagram of an elastic bar which is discretized into a one-dimensional mesh

VALIDATION OF NUMERICAL PROCEDURES

In all these problems and in the test examples in Chapter IV, the silent boundaries were implemented as shown in Figure 9. In the case of the viscous boundary, the black strip denotes a set of applied stresses, crxx and 'xz', which are applied to the boundary. A load is applied to the interior of the rod during the first time step and then removed.

The wave pulses are denoted by circles whose diameters represent the magnitude and direction of the particle velocities. The circles that remain at the paraxial boundary after step number 9 do so because of the attachment of the boundary nodes on.

INTERIOR REGION

All the free nodes experience some displacement as the wave passes them, but the right-hand nodes remain stationary. After checking the code with several simple, smooth analytical solutions, we test its ability to model discontinuous waves generated by a delta function load in both time and space. A byproduct of this example is that the results indicate the effectiveness of the silent-limit methods.

Miklowitz(BS) presents a derivation, based on the Cagniard-DeHoop method, of Lamb's analytical solutions and plots them along the x and z axes. We compare these results with those produced by a grid which is coarse with respect to the wavefront.

Figure  10.  Top:  Mesh  used  for  a  one-dimensional  wave;
Figure 10. Top: Mesh used for a one-dimensional wave;

LAMB'S PA~BLEM

The comparisons between the analytical solutions and finite-element solutions using the implicit algorithm are presented in Figures 15 and 16. The analytical result shows a Rayleigh singularity where a negative infinite displacement changes instantaneously to a positive infinite displacement. When we compare the analytical and finite element solutions in the interior (x = 0.0, z = 2.75), we obtain better agreement between the methods.

The width of the initial wave pulse is still narrower than one element, but its size is finite, unlike the displacement along the surface. It leads to slightly improved accuracy inside, but adds spurious surface noise.

Figure  14.  Discretized  approximations  to  the  delta-function  load.
Figure 14. Discretized approximations to the delta-function load.

TIMECSECS)

TIME( SECS)

LAMB'S PACIBLEM

INTRODUCTION A. General Aim

CHAPTER 4

  • DIRECT INCIDENCE OF DILATATIONAL WAVES

All the examples in this chapter were solved using the silent boundaries described in Chapter 3. Both the elastic and boundary regions were discretized with four-node elements using bilinear, isoparametric shape functions. 1This lumping of the mass matrix was done to facilitate the calculations of energy. The symmetry of the load allows the problem to be reduced to that of a conveniently resolved quarter plane.

Silent boundaries are marked with a thick black bar at the outer radial edge of the grid. This boundary absorbs almost all of the radiated energy, except for a small reflection that occurs at the inner boundary, at.

RADIAL. DILATATI~NAL PULSE

STANDARD-VISCOUS BOUNDARY, POISSON'S RATI0=.33

PULSE LOADINGS - GENERAL DISCUSSION

Time histories of the response are recorded at three points, labeled A, B or C, in each of the meshes in Figure 6. The two excitations considered are a vertical pulse that mainly generates dilational waves and a tensile pulse that generates mainly shear waves. These loads were selected for their simplicity and their relevance to the vertical and horizontal loads that occur in soil-structure interaction problems.

These are the "symmetric" boundary conditions that allow us to analyze a half space with a quarter mesh.

EXTENDED MESH

Vertical-Pulse Horizontal- Pulse

HORIZONTAL-PULSE LOADING

With these values, the dilatational waves reach the limit of the smallest grid in 9 seconds; the shear waves reach the same point in 16 seconds. Figures 8 and 9 report those displacements recorded near the lateral boundary (point B in Figure 6); The arrival of the main pulse is visible in all figures. The horizontal displacement oscillations in Figures 10 and 11 appear in all calculations, including the elongated mesh.

Each of the _broadcast limits preserves the period of these high-frequency motions, which arise from the coarseness of the mesh and the nature of the load. In displacement calculations, as shown in Figures 9 and 11, the unified (optimized) viscous boundary almost duplicates those results produced by the standard viscous boundary.

Figure  10.  Horizontal  displacements  at  point  A as  a  function  of  time.  (Horizontal-pulse  loading,  undamoed  case.)
Figure 10. Horizontal displacements at point A as a function of time. (Horizontal-pulse loading, undamoed case.)

TIMEC SECS)

VERTICAL-PULSE LOAOING

Vertical load response calculations were performed without the use of numerical or material damping. Reflections from the free boundary are evident, while both quiet boundary schemes remove the outgoing waves. Vertical displacements represent the largest part of the response and are recorded at point B in Figure 23.

The same conclusion applies to the vertical displacements at the bottom of the grid, point C, as shown in Figure 24. In general, for the vertical undrained loading problem, absorbing boundaries remove most of the output energy.

Figure  19.  Stresses,  a21•  at  point  Casa  function  of  time.
Figure 19. Stresses, a21• at point Casa function of time.

VERTICAL HALF-SINE PULSE

The long-term shift observed in these figures is similar to that found in Lamb's analytical solution. The viscous and extended paraxial bounds work almost equally well for stresses, but the paraxial bound has better accuracy for vertical displacements.

FREE BOUNDARY

BOUNDARY

STEP 16

STEP 21

STEP 26

STEP 31

TIMECSECS) 30.0

DISCUSSION OF NONLINEAR WAVES

The interior of the mesh can be governed by nonlinear equations, but the region adjacent to the silent boundary must be linear. This assumption, that the governing equations are linear at the outer edges of the computational grid, is suitable for many different problems. Wave motion originating from the interaction zone and propagating to the outer boundaries can often be adequately represented by linear soil models.

However, our experience with the uniform (“optimised”) viscous limit does indicate this. that the viscous limits are relatively insensitive to the parameters a and b. The numerical results, using either the standard or the uniform viscous limit, are almost identical. We found that using the modified viscous limits, instead of the standard viscous limit, yields virtually no differences in the stresses and vertical displacements.

RAYLEIGH-WAVE EXAMPLE

The overall conclusion is that the linear boundaries can be useful in absorbing the slower moving, non-linear waves.

SMALL MESH WITH SILENT BOUNDARY

Second, the ability of the quiet boundaries to transmit Rayleigh waves must be evaluated. The horizontal and vertical displacements along the left side of the mesh are prescribed, for the entire subsequent time, according to a well-known Rayleigh-wave solution. Figures 30 and 31 illustrate most of the worst agreement between the extended-paraxial boundary and the extended mesh.

The elongated-paraxial boundary results contain spurious noise superimposed on the overall waveform, while the viscous boundary changes the period of the motions. Rayleigh mode shapes can also be used to estimate the accuracy of transmission boundaries.

Figure  29.  Extended,  finite-element  mesh  with  free  boundary  used  for  the  Rayleigh-wave  loading
Figure 29. Extended, finite-element mesh with free boundary used for the Rayleigh-wave loading

RAYLEIGH WAVE

TI ME( SECS)

VER. OISP

CHAPTER 5

In the problems we studied, all silent bounds resulted in adequate to exceptional results. Tajimi, H., "Dynamic Analysis of a Structure Embedded in an Elastic Stratum", Proceedings of the Fourth World Conference on Earthquake Engineering, Santiago, Chile, 1964. Lysmer, J., "Bumped Mass Method for Rayleigh Waves", Seismological Society of America Bulletin, Vol.

Brooks, A., and Hughes, TJR, "Streamline-Upwind/Petrov-Galerkin Methods for Advection-dominated Streams", Proceedings of the Third. Mullen, R., and Belytschko, T., "Dispersion analysis of finite element semidiscretizations of the two-dimensional wave equation".

Gambar

Figure  5.  Stability  regions  for  the  paraxial  equations,  for  different  material  properties  and  angles  of  incidence
Figure  9.  Absolute  amplitudes  of  reflected  waves  for  various  angles  of  incidence  (plane  strain)
Figure  10.  Absolute  amplitudes  of  reflected  waves  for  various  angles  of  incidence  (plane  strain)
Figure  11.  Absolute  amplitudes  of  reflected  waves  for  various  angles  of  incidence  (plane  strain)
+7

Referensi

Dokumen terkait

New boundary value problems for the wave equation and equations of mixed type... The Fifth Simposium in applied