06.08.2013 Views

Sequential Methods for Coupled Geomechanics and Multiphase Flow

Sequential Methods for Coupled Geomechanics and Multiphase Flow

Sequential Methods for Coupled Geomechanics and Multiphase Flow

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

4.2. SPECTRAL ANALYSIS 89<br />

from the fully coupled method, <strong>and</strong> U n+1,k <strong>and</strong> P n+1,k are those from the sequential methods<br />

at the k th iteration step. Let e n,niter<br />

P<br />

<strong>and</strong> e n,niter<br />

U<br />

be the difference between the solutions from<br />

the fully coupled <strong>and</strong> sequential methods at tn <strong>for</strong> pressure <strong>and</strong> displacement, respectively.<br />

Subtracting Equations 4.19 <strong>and</strong> 4.20 from Equations 4.17 <strong>and</strong> 4.18, respectively, gives<br />

− Kdr<br />

h ek+1<br />

U j− 3 2<br />

h<br />

M<br />

We set e n,niter<br />

P<br />

− h<br />

M<br />

e k+1<br />

Pj<br />

∆t<br />

e n,niter<br />

Pj<br />

∆t<br />

+ 2 Kdr<br />

h ek+1<br />

U j− 1 2<br />

h<br />

+ b<br />

∆t (−<br />

e k+1<br />

U<br />

j− 1<br />

2<br />

<strong>and</strong> e n,niter<br />

U<br />

− b h<br />

∆t (−<br />

− Kdr<br />

h ek+1<br />

U j+ 1 2<br />

− e k+1<br />

U j+ 1 2<br />

h<br />

e n,niter<br />

U j− 1 2<br />

) − kp<br />

µ<br />

− e n,niter<br />

U j+ 1 2<br />

h<br />

− b(e k Pj−1 − ekPj ) = 0, (4.21)<br />

1<br />

h (ek+1 − 2ek+1 + ek+1<br />

Pj+1 Pj Pj−1 )<br />

) = 0. (4.22)<br />

to zero in order to investigate D in Equation 4.4 only.<br />

This implies that we drop the second term in Equation 4.4. Introducing errors of the<br />

<strong>for</strong>m ek Uj = γk eei(j)θ eU ˆ <strong>and</strong> ek Pj = γk eei(j)θ eP, ˆ where γe is the amplification factor of error,<br />

e (·) = exp(·), i = √ −1, <strong>and</strong> θ ∈ [−π, π], we obtain from Equations 4.21 <strong>and</strong> 4.22<br />

⎡<br />

Kdr<br />

h 2γe(1 − cos θ) b2i sin θ<br />

2<br />

⎤⎡<br />

⎣<br />

bγe2i sin θ<br />

2 [ 1<br />

⎦⎣<br />

<br />

kp∆t 1<br />

M h + µ h2(1 − cos θ)]γe<br />

<br />

Bdr<br />

ˆ<br />

e k U<br />

ˆ<br />

e k P<br />

⎤ ⎡<br />

⎦ = ⎣ 0<br />

⎤<br />

⎦.<br />

0<br />

(4.23)<br />

Since the matrix Bdr is required to be singular (i.e., detBdr = 0), this leads to<br />

γe = 0, −<br />

Kdr<br />

1<br />

M<br />

b 2<br />

kp∆t<br />

χ = . (4.24)<br />

+ χ2(1 − cos θ), µh2 The γe is equivalent to the eigenvalue of the error amplification matrix G defined by<br />

⎡<br />

⎣ ek+1<br />

Uj<br />

e k+1<br />

Pj<br />

⎤<br />

⎦ = G<br />

⎡<br />

⎣ ek Uj<br />

e k Pj<br />

⎤<br />

⎦. (4.25)<br />

The two γe’s in Equation 4.24 are distinct, then G can be decomposed as G = PΛP −1

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!