mirror of
https://github.com/wassname/simpeg.git
synced 2026-07-02 04:42:35 +08:00
D Fournier
266 lines
7.6 KiB
Python
266 lines
7.6 KiB
Python
#%%
|
|
from SimPEG import *
|
|
import simpegPF as PF
|
|
import pylab as plt
|
|
import os
|
|
home_dir = '.\\'
|
|
inpfile = 'PYGRAV3D_inv.inp'
|
|
dsep = '\\'
|
|
os.chdir(home_dir)
|
|
plt.close('all')
|
|
|
|
#%% User input
|
|
# Initial beta
|
|
beta_in = 1e+2
|
|
|
|
# Treshold values for compact norm
|
|
eps_p = 1e-2 # Small model values
|
|
eps_q = 1e-2 # Small model gradient
|
|
|
|
# Plotting parameter
|
|
vmin = -0.1
|
|
vmax = 0.2
|
|
#%%
|
|
# Read input file
|
|
[mshfile, obsfile, topofile, mstart, mref, wgtfile, chi, alphas, bounds, lpnorms] = PF.Gravity.read_GRAVinv_inp(home_dir + dsep + inpfile)
|
|
|
|
# Load mesh file
|
|
mesh = Mesh.TensorMesh.readUBC(mshfile)
|
|
|
|
# Load in observation file
|
|
survey = PF.Gravity.readUBCgravObs(obsfile)
|
|
|
|
# Get obs location and data
|
|
rxLoc = survey.srcField.rxList[0].locs
|
|
d = survey.dobs
|
|
wd = survey.std
|
|
|
|
ndata = survey.srcField.rxList[0].locs.shape[0]
|
|
|
|
# Load in topofile or create flat surface
|
|
if topofile == 'null':
|
|
|
|
# All active
|
|
actv = np.asarray(range(mesh.nC))
|
|
|
|
else:
|
|
|
|
topo = np.genfromtxt(topofile,skip_header=1)
|
|
# Find the active cells
|
|
actv = PF.Magnetics.getActiveTopo(mesh,topo,'N')
|
|
|
|
nC = len(actv)
|
|
|
|
# Create active map to go from reduce set to full
|
|
actvMap = Maps.InjectActiveCells(mesh, actv, -100)
|
|
|
|
# Creat reduced identity map
|
|
idenMap = Maps.IdentityMap(nP = nC)
|
|
|
|
# Load starting model file
|
|
if isinstance(mstart, float):
|
|
|
|
mstart = np.ones(nC) * mstart
|
|
else:
|
|
mstart = Mesh.TensorMesh.readModelUBC(mesh,mstart)
|
|
mstart = mstart[actv]
|
|
|
|
# Load reference file
|
|
if isinstance(mref, float):
|
|
mref = np.ones(nC) * mref
|
|
else:
|
|
mref = Mesh.TensorMesh.readModelUBC(mesh,mref)
|
|
mref = mref[actv]
|
|
|
|
# Get index of the center for plotting
|
|
midx = int(mesh.nCx/2)
|
|
midy = int(mesh.nCy/2)
|
|
|
|
|
|
#%% Plot obs data
|
|
PF.Gravity.plot_obs_2D(survey,'Observed Data')
|
|
|
|
#%% Run inversion
|
|
prob = PF.Gravity.GravityIntegral(mesh, mapping = idenMap, actInd = actv)
|
|
prob.solverOpts['accuracyTol'] = 1e-4
|
|
|
|
survey.pair(prob)
|
|
|
|
# Write out the predicted file and generate the forward operator
|
|
pred = prob.fields(mstart)
|
|
|
|
PF.Gravity.writeUBCobs(home_dir + dsep + 'Pred0.dat',survey,pred)
|
|
|
|
# Make depth weighting
|
|
wr = np.sum(prob.G**2.,axis=0)**0.5 / mesh.vol[actv]
|
|
wr = ( wr/np.max(wr) )
|
|
wr_out = actvMap * wr
|
|
|
|
#%% Plot depth weighting
|
|
#plt.figure()
|
|
#ax = plt.subplot()
|
|
#mesh.plotSlice(actvMap*mstart, ax = ax, normal = 'Y', ind=midx+1 ,clim = (-1e-1, mstart.max()))
|
|
#plt.title('Distance weighting')
|
|
#plt.xlabel('x');plt.ylabel('z')
|
|
#plt.gca().set_aspect('equal', adjustable='box')
|
|
|
|
#%% Create inversion objects
|
|
|
|
# First start with an l2 regularization
|
|
reg = Regularization.Simple(mesh, indActive = actv, mapping = idenMap)
|
|
reg.mref = mref
|
|
reg.wght = wr
|
|
|
|
|
|
# Create pre-conditioner
|
|
diagA = np.sum(prob.G**2.,axis=0) + beta_in*(reg.W.T*reg.W).diagonal()*wr
|
|
PC = Utils.sdiag(diagA**-1.)
|
|
|
|
# Data misfit function
|
|
dmis = DataMisfit.l2_DataMisfit(survey)
|
|
dmis.Wd = 1./wd
|
|
opt = Optimization.ProjectedGNCG(maxIter=20,lower=bounds[0],upper=bounds[1], maxIterCG= 20, tolCG = 1e-3)
|
|
opt.approxHinv = PC
|
|
|
|
|
|
invProb = InvProblem.BaseInvProblem(dmis, reg, opt, beta = beta_in)
|
|
beta = Directives.BetaSchedule(coolingFactor=2, coolingRate=1)
|
|
target = Directives.TargetMisfit()
|
|
|
|
inv = Inversion.BaseInversion(invProb, directiveList=[beta,target])
|
|
|
|
m0 = mstart
|
|
|
|
# Run inversion
|
|
mrec = inv.run(m0)
|
|
|
|
m_out = actvMap*mrec
|
|
|
|
# Write result
|
|
Mesh.TensorMesh.writeModelUBC(mesh,'SimPEG_inv_l2l2.sus',m_out)
|
|
#Utils.meshutils.writeUBCTensorModel(home_dir+dsep+'wr.dat',mesh,wr_out)
|
|
|
|
# Plot predicted
|
|
pred = prob.fields(mrec)
|
|
#PF.Magnetics.plot_obs_2D(rxLoc,pred,wd,'Predicted Data')
|
|
#PF.Magnetics.plot_obs_2D(rxLoc,(d-pred),wd,'Residual Data')
|
|
|
|
print "Final misfit:" + str(np.sum( ((d-pred)/wd)**2. ) )
|
|
|
|
#%% Plot out sections of the smooth model
|
|
|
|
yslice = midx+1
|
|
|
|
plt.figure()
|
|
ax = plt.subplot(221)
|
|
mesh.plotSlice(m_out, ax = ax, normal = 'Z', ind=-5, clim = (vmin,vmax))
|
|
plt.plot(np.array([mesh.vectorCCx[0],mesh.vectorCCx[-1]]), np.array([mesh.vectorCCy[yslice],mesh.vectorCCy[yslice]]),c='w',linestyle = '--')
|
|
plt.title('Z: ' + str(mesh.vectorCCz[-5]) + ' m')
|
|
plt.xlabel('x');plt.ylabel('z')
|
|
plt.gca().set_aspect('equal', adjustable='box')
|
|
|
|
ax = plt.subplot(222)
|
|
mesh.plotSlice(m_out, ax = ax, normal = 'Z', ind=-8, clim = (vmin,vmax))
|
|
plt.plot(np.array([mesh.vectorCCx[0],mesh.vectorCCx[-1]]), np.array([mesh.vectorCCy[yslice],mesh.vectorCCy[yslice]]),c='w',linestyle = '--')
|
|
plt.title('Z: ' + str(mesh.vectorCCz[-8]) + ' m')
|
|
plt.xlabel('x');plt.ylabel('z')
|
|
plt.gca().set_aspect('equal', adjustable='box')
|
|
|
|
ax = plt.subplot(212)
|
|
mesh.plotSlice(m_out, ax = ax, normal = 'Y', ind=yslice, clim = (vmin,vmax))
|
|
plt.title('Cross Section')
|
|
plt.xlabel('x');plt.ylabel('z')
|
|
plt.gca().set_aspect('equal', adjustable='box')
|
|
|
|
plt.figure()
|
|
ax = plt.subplot(121)
|
|
plt.hist(mrec,100)
|
|
plt.yscale('log', nonposy='clip')
|
|
plt.title('Histogram of model values - Smooth')
|
|
ax = plt.subplot(122)
|
|
plt.hist(reg.Wsmooth*mrec,100)
|
|
plt.yscale('log', nonposy='clip')
|
|
plt.title('Histogram of model gradient values - Smooth')
|
|
#%% Run one more round for sparsity
|
|
phim = invProb.phi_m_last
|
|
phid = invProb.phi_d
|
|
|
|
reg = Regularization.Sparse(mesh, indActive = actv, mapping = idenMap)
|
|
reg.recModel = mrec
|
|
reg.mref = mref
|
|
reg.wght = wr
|
|
reg.eps_p = eps_p
|
|
reg.eps_q = eps_q
|
|
reg.p = lpnorms[0]
|
|
reg.qx = lpnorms[1]
|
|
reg.qy = lpnorms[2]
|
|
reg.qz = lpnorms[3]
|
|
|
|
diagA = np.sum(prob.G**2.,axis=0) + beta_in*(reg.W.T*reg.W).diagonal()*(wr)
|
|
PC = Utils.sdiag(diagA**-1.)
|
|
|
|
#reg.alpha_s = 1.
|
|
|
|
dmis = DataMisfit.l2_DataMisfit(survey)
|
|
dmis.Wd = 1./wd
|
|
opt = Optimization.ProjectedGNCG(maxIter=10 ,lower=bounds[0],upper=bounds[1], maxIterCG= 25, tolCG = 1e-4)
|
|
opt.approxHinv = PC
|
|
#opt.phim_last = reg.eval(mrec)
|
|
|
|
# opt = Optimization.InexactGaussNewton(maxIter=6)
|
|
invProb = InvProblem.BaseInvProblem(dmis, reg, opt, beta = invProb.beta)
|
|
beta = Directives.BetaSchedule(coolingFactor=1, coolingRate=1)
|
|
#betaest = Directives.BetaEstimate_ByEig()
|
|
target = Directives.TargetMisfit()
|
|
IRLS =Directives.Update_IRLS( phi_m_last = phim, phi_d_last = phid )
|
|
|
|
inv = Inversion.BaseInversion(invProb, directiveList=[beta,IRLS])
|
|
|
|
m0 = mrec
|
|
|
|
# Run inversion
|
|
mrec = inv.run(m0)
|
|
|
|
m_out = actvMap*mrec
|
|
|
|
Mesh.TensorMesh.writeModelUBC(mesh,'SimPEG_inv_l0l2.sus',m_out)
|
|
|
|
pred = prob.fields(mrec)
|
|
|
|
#%% Plot obs data
|
|
PF.Magnetics.plot_obs_2D(rxLoc,pred,'Predicted Data', vmin = np.min(d), vmax = np.max(d))
|
|
PF.Magnetics.plot_obs_2D(rxLoc,d,'Observed Data')
|
|
print "Final misfit:" + str(np.sum( ((d-pred)/wd)**2. ) )
|
|
#%% Plot out a section of the model
|
|
|
|
yslice = midx
|
|
plt.figure()
|
|
ax = plt.subplot(221)
|
|
mesh.plotSlice(m_out, ax = ax, normal = 'Z', ind=-5, clim = (vmin,vmax))
|
|
plt.plot(np.array([mesh.vectorCCx[0],mesh.vectorCCx[-1]]), np.array([mesh.vectorCCy[yslice],mesh.vectorCCy[yslice]]),c='w',linestyle = '--')
|
|
plt.title('Z: ' + str(mesh.vectorCCz[-5]) + ' m')
|
|
plt.xlabel('x');plt.ylabel('z')
|
|
plt.gca().set_aspect('equal', adjustable='box')
|
|
|
|
ax = plt.subplot(222)
|
|
mesh.plotSlice(m_out, ax = ax, normal = 'Z', ind=-8, clim = (vmin,vmax))
|
|
plt.plot(np.array([mesh.vectorCCx[0],mesh.vectorCCx[-1]]), np.array([mesh.vectorCCy[yslice],mesh.vectorCCy[yslice]]),c='w',linestyle = '--')
|
|
plt.title('Z: ' + str(mesh.vectorCCz[-8]) + ' m')
|
|
plt.xlabel('x');plt.ylabel('z')
|
|
plt.gca().set_aspect('equal', adjustable='box')
|
|
|
|
ax = plt.subplot(212)
|
|
mesh.plotSlice(m_out, ax = ax, normal = 'Y', ind=yslice, clim = (vmin,vmax))
|
|
plt.title('Cross Section')
|
|
plt.xlabel('x');plt.ylabel('z')
|
|
plt.gca().set_aspect('equal', adjustable='box')
|
|
|
|
plt.figure()
|
|
ax = plt.subplot(121)
|
|
plt.hist(mrec,100)
|
|
plt.yscale('log', nonposy='clip')
|
|
plt.title('Histogram of model values - Sparse lp:'+str(lpnorms[0]))
|
|
ax = plt.subplot(122)
|
|
plt.hist(reg.Wsmooth*mrec,100)
|
|
plt.yscale('log', nonposy='clip')
|
|
plt.title('Histogram of model gradient values - Sparse lqx: ' + str(lpnorms[1]) + ' lqy:'+ str(lpnorms[2]) + ' lqz:' + str(lpnorms[3])) |