def kriging3d_block(kriging_type,grid,points,data,vm,discritization=None,mean=None,mindata=1,maxdata=1,azimuth=0.0,dip=0.0,plunge=0.0,search_range=np.inf,anisotropy=[1.0,1.0],full=False,debug=1):
if mean is None:
mean = np.mean(data)
#create kd3
kdtree = KDTree3D(points,azimuth,dip,plunge,anisotropy)
if discritization is not None:
dblock = grid.discretize(discritization)
npd = len(dblock)
max_covariance = 0
for dp in dblock:
max_covariance += np.sum(vm.covariance(dp,dblock))
max_covariance -= vm.nugget*npd
max_covariance /= npd**2
else:
dblock = None
npd = 1
max_covariance = vm.max_covariance()
print("block covariance",max_covariance)
n = len(grid)
ret = np.empty((n,2))
non_estimated = 0
if full:
ret_indices = []
for i,point in enumerate(grid):
#print i,point
d,indices = kdtree.search(point,maxdata,search_range)
if debug >=1:
print('est at:',point,'neigborhs len:',len(indices))
if full:
ret_indices += [indices]
if len(indices) < mindata:
estimation = np.nan
variance = np.nan
non_estimated +=1
else:
estimation,variance,A,b,x = kriging_system(kriging_type,vm,max_covariance,point,points[indices,:],data[indices],dblock)
if debug >=3:
print("ponits:",points[indices])
print("A=",A)
print("b=",b)
print("x=",x)
if debug >=1:
print("estimation,variance:",estimation,variance)
ret[i,0] = estimation
ret[i,1] = variance
if full:
return ret,non_estimated,ret_indices
else:
return ret,non_estimated
def kriging3d_cross_validation(kriging_type,points,data,vm,mean=None,mindata=1,maxdata=10,azimuth=0.0,dip=0.0,plunge=0.0,search_range=np.inf,anisotropy=[1.0,1.0],full=False):
if mean is None:
mean = np.mean(data)
#create kd3
kdtree = KDTree3D(points,azimuth,dip,plunge,anisotropy)
ret = kriging_cross_validation(kriging_type,points,data,vm,mean,mindata,maxdata,search_range,kdtree,full)
return ret
def kriging3d_block_iterator(kriging_type,
grid,
points,
data,
vm,
discretization=None,
mean=None,
mindata=1,
maxdata=1,
azimuth=0.0,
dip=0.0,
plunge=0.0,
search_range=np.inf,
anisotropy=[1.0, 1.0],
full=False,
debug=1):
if mean is None:
mean = np.mean(data)
#create kd3
kdtree = KDTree3D(points, azimuth, dip, plunge, anisotropy)
if discretization is not None:
dblock = grid.discretize(discretization)
npd = len(dblock)
max_covariance = 0
for dp in dblock:
max_covariance += np.sum(vm.covariance(dp, dblock))
max_covariance -= vm.nugget * npd
max_covariance /= npd**2
else:
dblock = None
npd = 1
max_covariance = vm.max_covariance()
print("block covariance", max_covariance)
for i, point in enumerate(grid):
d, indices = kdtree.search(point, maxdata, search_range)
if len(indices) < mindata:
estimation = np.nan
variance = np.nan
else:
estimation, variance, A, b, x = kriging_system(
kriging_type, vm, max_covariance, point, points[indices, :],
data[indices], dblock)
if full:
yield point, estimation, variance, indices
else:
yield point, estimation, variance
def kriging3d_puntual(kriging_type,
output_points,
points,
data,
vm,
mean=None,
mindata=1,
maxdata=1,
azimuth=0.0,
dip=0.0,
plunge=0.0,
search_range=np.inf,
anisotropy=[1.0, 1.0],
full=False):
if mean is None:
mean = np.mean(data)
#create kd3
kdtree = KDTree3D(points, azimuth, dip, plunge, anisotropy)
max_covariance = vm.max_covariance()
n = len(points)
ret = np.empty((n, 2))
non_estimated = 0
if full:
ret_indices = []
for i, point in enumerate(output_points):
d, indices = kdtree.search(point, maxdata, search_range)
if full:
ret_indices += [indices]
if len(indices) < mindata:
estimation = np.nan
variance = np.nan
non_estimated += 1
else:
estimation, variance, A, b, x = kriging_system(
kriging_type, vm, max_covariance, point, points[indices, :],
data[indices])
#print A,b,x,estimation,variance,points[indices]
ret[i, 0] = estimation
ret[i, 1] = variance
if full:
return ret, non_estimated, ret_indices
else:
return ret, non_estimated
def kriging3d_block_multipasses(kriging_type,grid,points,tag_in,data,tag_out,vm,discritization=None,mean=None,search_angles=None,search_anisotropy=None,search_passes=None,search_soft=None,full=False,debug=1):
if mean is None:
mean = np.mean(data)
if discritization is not None:
dblock = grid.discretize(discritization)
npd = len(dblock)
max_covariance = 0
for dp in dblock:
max_covariance += np.sum(vm.covariance(dp,dblock))
max_covariance -= vm.nugget*npd
max_covariance /= npd**2
print("block covariance",max_covariance)
else:
dblock = None
npd = 1
max_covariance = vm.max_covariance()
print("covariance",max_covariance)
tag_values = list(set(tag_in))
tag_values.sort()
n = len(grid)
ret = np.empty((n,3))
ret_pass = np.zeros(n,dtype=np.int8)
for tag in tag_values:
#create kd3
kdtree = []
for search_params in search_passes[tag]:
kdtree += [KDTree3D(points,search_params.azimuth,search_params.dip,search_params.plunge,search_params.anisotropy)]
kdtree_soft = KDTree3D(points,search_soft.azimuth,search_soft.dip,search_soft.plunge,search_soft.anisotropy)
non_estimated = 0
if full:
ret_indices = []
for i,point in enumerate(grid):
if tag_in[i] == tag:
'''
Search is performed by multiple passes
'''
data_found = False
for j,search_params in enumerate(search_passes):
d,indices = kdtree[j].search(point,search_params.maxdata,search_params.range)
#filter by this tag
#include any softdata
for soft_tag in search_params.softdata.keys():
d_soft,indices_soft = kdtree_soft.search(point,soft_params.maxdata_soft,soft_params.range)
if len(d) >= search_params.mindata:
data_found = True
ret_pass[i] = j+1
break
if full:
ret_indices += [indices]
n_data = len(d) + len(d_soft)
if debug >=1:
print('est at:',point,'neigborhs len:',len(indices))
print('est at:',point,'soft len:',len(indices_soft))
locations_data = np.empty((n_data,3))
locations_data[:len(d),:] = points[indices,:]
locations_data[len(d):,:] = points[indices_soft,:]
value_data = np.empty(n_data)
value_data[:len(d),:] = data[indices]
value_data[len(d):,:] = data[indices_soft]
estimation,variance,A,b,x = kriging_system(kriging_type,vm,max_covariance,point,locations_data,value_data,dblock)
else:
estimation = np.nan
variance = np.nan
non_estimated +=1
if debug >=3:
print("ponits:",points[indices])
print("A=",A)
print("b=",b)
print("x=",x)
if debug >=1:
print("estimation,variance:",estimation,variance)
ret[i,0] = estimation
ret[i,1] = variance
if full:
return ret,non_estimated,ret_indices
else:
return ret,non_estimated
