Merge branch 'development' into 19-NewStylePhenopowerlaw
This commit is contained in:
commit
8bea82c72f
|
@ -158,12 +158,12 @@ Post_AverageDown:
|
|||
- master
|
||||
- release
|
||||
|
||||
Post_General:
|
||||
stage: postprocessing
|
||||
script: PostProcessing/test.py
|
||||
except:
|
||||
- master
|
||||
- release
|
||||
#Post_General:
|
||||
# stage: postprocessing
|
||||
# script: PostProcessing/test.py
|
||||
# except:
|
||||
# - master
|
||||
# - release
|
||||
|
||||
Post_GeometryReconstruction:
|
||||
stage: postprocessing
|
||||
|
@ -364,12 +364,12 @@ Phenopowerlaw_singleSlip:
|
|||
- master
|
||||
- release
|
||||
|
||||
TextureComponents:
|
||||
stage: spectral
|
||||
script: TextureComponents/test.py
|
||||
except:
|
||||
- master
|
||||
- release
|
||||
#TextureComponents:
|
||||
# stage: spectral
|
||||
# script: TextureComponents/test.py
|
||||
# except:
|
||||
# - master
|
||||
# - release
|
||||
|
||||
|
||||
###################################################################################################
|
||||
|
|
|
@ -119,6 +119,9 @@ for executable in mpirun mpiexec; do
|
|||
getDetails $executable '--version'
|
||||
done
|
||||
|
||||
firstLevel "CMake"
|
||||
getDetails cmake --version
|
||||
|
||||
firstLevel "Abaqus"
|
||||
cd installation/mods_Abaqus # to have the right environment file
|
||||
for executable in abaqus abq2017 abq2018; do
|
||||
|
|
|
@ -44,7 +44,6 @@ if ( $?prompt ) then
|
|||
echo "DAMASK $DAMASK_ROOT"
|
||||
echo "Spectral Solver $SOLVER"
|
||||
echo "Post Processing $PROCESSING"
|
||||
echo "Multithreading DAMASK_NUM_THREADS=$DAMASK_NUM_THREADS"
|
||||
if ( $?PETSC_DIR) then
|
||||
echo "PETSc location $PETSC_DIR"
|
||||
endif
|
||||
|
@ -52,8 +51,10 @@ if ( $?prompt ) then
|
|||
echo "MSC.Marc/Mentat $MSC_ROOT"
|
||||
endif
|
||||
echo
|
||||
echo "Multithreading DAMASK_NUM_THREADS=$DAMASK_NUM_THREADS"
|
||||
echo `limit datasize`
|
||||
echo `limit stacksize`
|
||||
echo
|
||||
endif
|
||||
|
||||
setenv DAMASK_NUM_THREADS $DAMASK_NUM_THREADS
|
||||
|
|
|
@ -88,6 +88,7 @@ if [ ! -z "$PS1" ]; then
|
|||
size=$(( 1024*$(ulimit -s) )); \
|
||||
print('{:.4g} {}'.format(size / (1 << ((int(math.log(size,2) / 10) if size else 0) * 10)), \
|
||||
['bytes','KiB','MiB','GiB','TiB','EiB','ZiB'][int(math.log(size,2) / 10) if size else 0]))")
|
||||
echo
|
||||
fi
|
||||
|
||||
export DAMASK_NUM_THREADS
|
||||
|
|
|
@ -81,6 +81,7 @@ if [ ! -z "$PS1" ]; then
|
|||
size=$(( 1024*$(ulimit -s) )); \
|
||||
print('{:.4g} {}'.format(size / (1 << ((int(math.log(size,2) / 10) if size else 0) * 10)), \
|
||||
['bytes','KiB','MiB','GiB','TiB','EiB','ZiB'][int(math.log(size,2) / 10) if size else 0]))")
|
||||
echo
|
||||
fi
|
||||
|
||||
export DAMASK_NUM_THREADS
|
||||
|
|
|
@ -33,7 +33,7 @@ class Quaternion:
|
|||
All methods and naming conventions based on Rowenhorst_etal2015
|
||||
Convention 1: coordinate frames are right-handed
|
||||
Convention 2: a rotation angle ω is taken to be positive for a counterclockwise rotation
|
||||
when viewing from the end point of the rotation axis unit vector towards the origin
|
||||
when viewing from the end point of the rotation axis towards the origin
|
||||
Convention 3: rotations will be interpreted in the passive sense
|
||||
Convention 4: Euler angle triplets are implemented using the Bunge convention,
|
||||
with the angular ranges as [0, 2π],[0, π],[0, 2π]
|
||||
|
@ -48,206 +48,138 @@ class Quaternion:
|
|||
"""
|
||||
|
||||
def __init__(self,
|
||||
quatArray = [1.0,0.0,0.0,0.0]):
|
||||
"""Initializes to identity if not given"""
|
||||
self.w, \
|
||||
self.x, \
|
||||
self.y, \
|
||||
self.z = quatArray
|
||||
quat = None,
|
||||
q = 1.0,
|
||||
p = np.zeros(3,dtype=float)):
|
||||
"""Initializes to identity unless specified"""
|
||||
self.q = quat[0] if quat is not None else q
|
||||
self.p = np.array(quat[1:4]) if quat is not None else p
|
||||
self.homomorph()
|
||||
|
||||
def __iter__(self):
|
||||
"""Components"""
|
||||
return iter([self.w,self.x,self.y,self.z])
|
||||
return iter(self.asList())
|
||||
|
||||
def __copy__(self):
|
||||
"""Create copy"""
|
||||
Q = Quaternion([self.w,self.x,self.y,self.z])
|
||||
return Q
|
||||
"""Copy"""
|
||||
return self.__class__(q=self.q,p=self.p.copy())
|
||||
|
||||
copy = __copy__
|
||||
|
||||
def __repr__(self):
|
||||
"""Readbable string"""
|
||||
return 'Quaternion(real=%+.6f, imag=<%+.6f, %+.6f, %+.6f>)' % \
|
||||
(self.w, self.x, self.y, self.z)
|
||||
"""Readable string"""
|
||||
return 'Quaternion(real={q:+.6f}, imag=<{p[0]:+.6f}, {p[1]:+.6f}, {p[2]:+.6f}>)'.format(q=self.q,p=self.p)
|
||||
|
||||
def __pow__(self, exponent):
|
||||
"""Power"""
|
||||
omega = math.acos(self.w)
|
||||
vRescale = math.sin(exponent*omega)/math.sin(omega)
|
||||
Q = Quaternion()
|
||||
Q.w = math.cos(exponent*omega)
|
||||
Q.x = self.x * vRescale
|
||||
Q.y = self.y * vRescale
|
||||
Q.z = self.z * vRescale
|
||||
return Q
|
||||
omega = math.acos(self.q)
|
||||
return self.__class__(q= math.cos(exponent*omega),
|
||||
p=self.p * math.sin(exponent*omega)/math.sin(omega))
|
||||
|
||||
def __ipow__(self, exponent):
|
||||
"""In-place power"""
|
||||
omega = math.acos(self.w)
|
||||
vRescale = math.sin(exponent*omega)/math.sin(omega)
|
||||
self.w = np.cos(exponent*omega)
|
||||
self.x *= vRescale
|
||||
self.y *= vRescale
|
||||
self.z *= vRescale
|
||||
omega = math.acos(self.q)
|
||||
self.q = math.cos(exponent*omega)
|
||||
self.p *= math.sin(exponent*omega)/math.sin(omega)
|
||||
return self
|
||||
|
||||
def __mul__(self, other):
|
||||
"""Multiplication"""
|
||||
# Rowenhorst_etal2015 MSMSE: value of P is selected as -1
|
||||
P = -1.0
|
||||
try: # quaternion
|
||||
Aw = self.w
|
||||
Ax = self.x
|
||||
Ay = self.y
|
||||
Az = self.z
|
||||
Bw = other.w
|
||||
Bx = other.x
|
||||
By = other.y
|
||||
Bz = other.z
|
||||
Q = Quaternion()
|
||||
Q.w = - Ax * Bx - Ay * By - Az * Bz + Aw * Bw
|
||||
Q.x = + Ax * Bw + Ay * Bz - Az * By + Aw * Bx
|
||||
Q.y = - Ax * Bz + Ay * Bw + Az * Bx + Aw * By
|
||||
Q.z = + Ax * By - Ay * Bx + Az * Bw + Aw * Bz
|
||||
return Q
|
||||
return self.__class__(q=self.q*other.q - np.dot(self.p,other.p),
|
||||
p=self.q*other.p + other.q*self.p + P * np.cross(self.p,other.p))
|
||||
except: pass
|
||||
try: # vector (perform active rotation, i.e. q*v*q.conjugated)
|
||||
w = self.w
|
||||
x = self.x
|
||||
y = self.y
|
||||
z = self.z
|
||||
Vx = other[0]
|
||||
Vy = other[1]
|
||||
Vz = other[2]
|
||||
try: # vector (perform passive rotation)
|
||||
( x, y, z) = self.p
|
||||
(Vx,Vy,Vz) = other[0:3]
|
||||
A = self.q*self.q - np.dot(self.p,self.p)
|
||||
B = 2.0 * (x*Vx + y*Vy + z*Vz)
|
||||
C = 2.0 * P*self.q
|
||||
|
||||
return np.array([\
|
||||
w * w * Vx + 2 * y * w * Vz - 2 * z * w * Vy + \
|
||||
x * x * Vx + 2 * y * x * Vy + 2 * z * x * Vz - \
|
||||
z * z * Vx - y * y * Vx,
|
||||
2 * x * y * Vx + y * y * Vy + 2 * z * y * Vz + \
|
||||
2 * w * z * Vx - z * z * Vy + w * w * Vy - \
|
||||
2 * x * w * Vz - x * x * Vy,
|
||||
2 * x * z * Vx + 2 * y * z * Vy + \
|
||||
z * z * Vz - 2 * w * y * Vx - y * y * Vz + \
|
||||
2 * w * x * Vy - x * x * Vz + w * w * Vz ])
|
||||
return np.array([
|
||||
A*Vx + B*x + C*(y*Vz - z*Vy),
|
||||
A*Vy + B*y + C*(z*Vx - x*Vz),
|
||||
A*Vz + B*z + C*(x*Vy - y*Vx),
|
||||
])
|
||||
except: pass
|
||||
try: # scalar
|
||||
Q = self.copy()
|
||||
Q.w *= other
|
||||
Q.x *= other
|
||||
Q.y *= other
|
||||
Q.z *= other
|
||||
return Q
|
||||
return self.__class__(q=self.q*other,
|
||||
p=self.p*other)
|
||||
except:
|
||||
return self.copy()
|
||||
|
||||
def __imul__(self, other):
|
||||
"""In-place multiplication"""
|
||||
# Rowenhorst_etal2015 MSMSE: value of P is selected as -1
|
||||
P = -1.0
|
||||
try: # Quaternion
|
||||
Aw = self.w
|
||||
Ax = self.x
|
||||
Ay = self.y
|
||||
Az = self.z
|
||||
Bw = other.w
|
||||
Bx = other.x
|
||||
By = other.y
|
||||
Bz = other.z
|
||||
self.w = - Ax * Bx - Ay * By - Az * Bz + Aw * Bw
|
||||
self.x = + Ax * Bw + Ay * Bz - Az * By + Aw * Bx
|
||||
self.y = - Ax * Bz + Ay * Bw + Az * Bx + Aw * By
|
||||
self.z = + Ax * By - Ay * Bx + Az * Bw + Aw * Bz
|
||||
self.q = self.q*other.q - np.dot(self.p,other.p)
|
||||
self.p = self.q*other.p + other.q*self.p + P * np.cross(self.p,other.p)
|
||||
except: pass
|
||||
return self
|
||||
|
||||
def __div__(self, other):
|
||||
"""Division"""
|
||||
if isinstance(other, (int,float)):
|
||||
w = self.w / other
|
||||
x = self.x / other
|
||||
y = self.y / other
|
||||
z = self.z / other
|
||||
return self.__class__([w,x,y,z])
|
||||
return self.__class__(q=self.q / other,
|
||||
p=self.p / other)
|
||||
else:
|
||||
return NotImplemented
|
||||
|
||||
def __idiv__(self, other):
|
||||
"""In-place division"""
|
||||
if isinstance(other, (int,float)):
|
||||
self.w /= other
|
||||
self.x /= other
|
||||
self.y /= other
|
||||
self.z /= other
|
||||
self.q /= other
|
||||
self.p /= other
|
||||
return self
|
||||
|
||||
def __add__(self, other):
|
||||
"""Addition"""
|
||||
if isinstance(other, Quaternion):
|
||||
w = self.w + other.w
|
||||
x = self.x + other.x
|
||||
y = self.y + other.y
|
||||
z = self.z + other.z
|
||||
return self.__class__([w,x,y,z])
|
||||
return self.__class__(q=self.q + other.q,
|
||||
p=self.p + other.p)
|
||||
else:
|
||||
return NotImplemented
|
||||
|
||||
def __iadd__(self, other):
|
||||
"""In-place addition"""
|
||||
if isinstance(other, Quaternion):
|
||||
self.w += other.w
|
||||
self.x += other.x
|
||||
self.y += other.y
|
||||
self.z += other.z
|
||||
self.q += other.q
|
||||
self.p += other.p
|
||||
return self
|
||||
|
||||
def __sub__(self, other):
|
||||
"""Subtraction"""
|
||||
if isinstance(other, Quaternion):
|
||||
Q = self.copy()
|
||||
Q.w -= other.w
|
||||
Q.x -= other.x
|
||||
Q.y -= other.y
|
||||
Q.z -= other.z
|
||||
return Q
|
||||
return self.__class__(q=self.q - other.q,
|
||||
p=self.p - other.p)
|
||||
else:
|
||||
return self.copy()
|
||||
return NotImplemented
|
||||
|
||||
def __isub__(self, other):
|
||||
"""In-place subtraction"""
|
||||
if isinstance(other, Quaternion):
|
||||
self.w -= other.w
|
||||
self.x -= other.x
|
||||
self.y -= other.y
|
||||
self.z -= other.z
|
||||
self.q -= other.q
|
||||
self.p -= other.p
|
||||
return self
|
||||
|
||||
def __neg__(self):
|
||||
"""Additive inverse"""
|
||||
self.w = -self.w
|
||||
self.x = -self.x
|
||||
self.y = -self.y
|
||||
self.z = -self.z
|
||||
self.q = -self.q
|
||||
self.p = -self.p
|
||||
return self
|
||||
|
||||
def __abs__(self):
|
||||
"""Norm"""
|
||||
return math.sqrt(self.w ** 2 + \
|
||||
self.x ** 2 + \
|
||||
self.y ** 2 + \
|
||||
self.z ** 2)
|
||||
return math.sqrt(self.q ** 2 + np.dot(self.p,self.p))
|
||||
|
||||
magnitude = __abs__
|
||||
|
||||
def __eq__(self,other):
|
||||
"""Equal at e-8 precision"""
|
||||
return (abs(self.w-other.w) < 1e-8 and \
|
||||
abs(self.x-other.x) < 1e-8 and \
|
||||
abs(self.y-other.y) < 1e-8 and \
|
||||
abs(self.z-other.z) < 1e-8) \
|
||||
or \
|
||||
(abs(-self.w-other.w) < 1e-8 and \
|
||||
abs(-self.x-other.x) < 1e-8 and \
|
||||
abs(-self.y-other.y) < 1e-8 and \
|
||||
abs(-self.z-other.z) < 1e-8)
|
||||
return (self-other).magnitude() < 1e-8 or (-self-other).magnitude() < 1e-8
|
||||
|
||||
def __ne__(self,other):
|
||||
"""Not equal at e-8 precision"""
|
||||
|
@ -255,31 +187,26 @@ class Quaternion:
|
|||
|
||||
def __cmp__(self,other):
|
||||
"""Linear ordering"""
|
||||
return (self.Rodrigues()>other.Rodrigues()) - (self.Rodrigues()<other.Rodrigues())
|
||||
return (1 if np.linalg.norm(self.asRodrigues()) > np.linalg.norm(other.asRodrigues()) else 0) \
|
||||
- (1 if np.linalg.norm(self.asRodrigues()) < np.linalg.norm(other.asRodrigues()) else 0)
|
||||
|
||||
def magnitude_squared(self):
|
||||
return self.w ** 2 + \
|
||||
self.x ** 2 + \
|
||||
self.y ** 2 + \
|
||||
self.z ** 2
|
||||
return self.q ** 2 + np.dot(self.p,self.p)
|
||||
|
||||
def identity(self):
|
||||
self.w = 1.
|
||||
self.x = 0.
|
||||
self.y = 0.
|
||||
self.z = 0.
|
||||
self.q = 1.
|
||||
self.p = np.zeros(3,dtype=float)
|
||||
return self
|
||||
|
||||
def normalize(self):
|
||||
d = self.magnitude()
|
||||
if d > 0.0:
|
||||
self /= d
|
||||
self.q /= d
|
||||
self.p /= d
|
||||
return self
|
||||
|
||||
def conjugate(self):
|
||||
self.x = -self.x
|
||||
self.y = -self.y
|
||||
self.z = -self.z
|
||||
self.p = -self.p
|
||||
return self
|
||||
|
||||
def inverse(self):
|
||||
|
@ -290,11 +217,9 @@ class Quaternion:
|
|||
return self
|
||||
|
||||
def homomorph(self):
|
||||
if self.w < 0.0:
|
||||
self.w = -self.w
|
||||
self.x = -self.x
|
||||
self.y = -self.y
|
||||
self.z = -self.z
|
||||
if self.q < 0.0:
|
||||
self.q = -self.q
|
||||
self.p = -self.p
|
||||
return self
|
||||
|
||||
def normalized(self):
|
||||
|
@ -310,27 +235,35 @@ class Quaternion:
|
|||
return self.copy().homomorph()
|
||||
|
||||
def asList(self):
|
||||
return [i for i in self]
|
||||
return [self.q]+list(self.p)
|
||||
|
||||
def asM(self): # to find Averaging Quaternions (see F. Landis Markley et al.)
|
||||
return np.outer([i for i in self],[i for i in self])
|
||||
return np.outer(self.asList(),self.asList())
|
||||
|
||||
def asMatrix(self):
|
||||
qbarhalf = 0.5*(self.w**2 - self.x**2 - self.y**2 - self.z**2)
|
||||
# Rowenhorst_etal2015 MSMSE: value of P is selected as -1
|
||||
P = -1.0
|
||||
qbarhalf = 0.5*(self.q**2 - np.dot(self.p,self.p))
|
||||
return 2.0*np.array(
|
||||
[[ qbarhalf + self.x**2 , self.x*self.y - self.w*self.z, self.x*self.z + self.w*self.y],
|
||||
[ self.x*self.y + self.w*self.z, qbarhalf + self.y**2 , self.y*self.z - self.w*self.x],
|
||||
[ self.x*self.z - self.w*self.y, self.y*self.z + self.w*self.x, qbarhalf + self.z**2 ],
|
||||
[[ qbarhalf + self.p[0]**2 ,
|
||||
self.p[0]*self.p[1] -P* self.q*self.p[2],
|
||||
self.p[0]*self.p[2] +P* self.q*self.p[1] ],
|
||||
[ self.p[0]*self.p[1] +P* self.q*self.p[2],
|
||||
qbarhalf + self.p[1]**2 ,
|
||||
self.p[1]*self.p[2] -P* self.q*self.p[0] ],
|
||||
[ self.p[0]*self.p[2] -P* self.q*self.p[1],
|
||||
self.p[1]*self.p[2] +P* self.q*self.p[0],
|
||||
qbarhalf + self.p[2]**2 ],
|
||||
])
|
||||
|
||||
def asAngleAxis(self,
|
||||
degrees = False):
|
||||
if self.w > 1:
|
||||
if self.q > 1.:
|
||||
self.normalize()
|
||||
|
||||
s = math.sqrt(1. - self.w**2)
|
||||
x = 2*self.w**2 - 1.
|
||||
y = 2*self.w * s
|
||||
s = math.sqrt(1. - self.q**2)
|
||||
x = 2*self.q**2 - 1.
|
||||
y = 2*self.q * s
|
||||
|
||||
angle = math.atan2(y,x)
|
||||
if angle < 0.0:
|
||||
|
@ -338,26 +271,28 @@ class Quaternion:
|
|||
s *= -1.
|
||||
|
||||
return (np.degrees(angle) if degrees else angle,
|
||||
np.array([1.0, 0.0, 0.0] if np.abs(angle) < 1e-6 else [self.x / s, self.y / s, self.z / s]))
|
||||
np.array([1.0, 0.0, 0.0] if np.abs(angle) < 1e-6 else self.p / s))
|
||||
|
||||
def asRodrigues(self):
|
||||
return np.inf*np.ones(3) if self.w == 0.0 else np.array([self.x, self.y, self.z])/self.w
|
||||
return np.inf*np.ones(3) if self.q == 0.0 else self.p/self.q
|
||||
|
||||
def asEulers(self,
|
||||
degrees = False):
|
||||
"""Orientation as Bunge-Euler angles."""
|
||||
q03 = self.w**2+self.z**2
|
||||
q12 = self.x**2+self.y**2
|
||||
# Rowenhorst_etal2015 MSMSE: value of P is selected as -1
|
||||
P = -1.0
|
||||
q03 = self.q**2 + self.p[2]**2
|
||||
q12 = self.p[0]**2 + self.p[1]**2
|
||||
chi = np.sqrt(q03*q12)
|
||||
|
||||
if abs(chi) < 1e-10 and abs(q12) < 1e-10:
|
||||
eulers = np.array([math.atan2(-2*self.w*self.z,self.w**2-self.z**2),0,0])
|
||||
eulers = np.array([math.atan2(-2*P*self.q*self.p[2],self.q**2-self.p[2]**2),0,0])
|
||||
elif abs(chi) < 1e-10 and abs(q03) < 1e-10:
|
||||
eulers = np.array([math.atan2( 2*self.x*self.y,self.x**2-self.y**2),np.pi,0])
|
||||
eulers = np.array([math.atan2( 2 *self.p[0]*self.p[1],self.p[0]**2-self.p[1]**2),np.pi,0])
|
||||
else:
|
||||
eulers = np.array([math.atan2((self.x*self.z-self.w*self.y)/chi,(-self.w*self.x-self.y*self.z)/chi),
|
||||
eulers = np.array([math.atan2((self.p[0]*self.p[2]-P*self.q*self.p[1])/chi,(-P*self.q*self.p[0]-self.p[1]*self.p[2])/chi),
|
||||
math.atan2(2*chi,q03-q12),
|
||||
math.atan2((self.w*self.y+self.x*self.z)/chi,( self.y*self.z-self.w*self.x)/chi),
|
||||
math.atan2((P*self.q*self.p[1]+self.p[0]*self.p[2])/chi,( self.p[1]*self.p[2]-P*self.q*self.p[0])/chi),
|
||||
])
|
||||
|
||||
return np.degrees(eulers) if degrees else eulers
|
||||
|
@ -371,25 +306,26 @@ class Quaternion:
|
|||
|
||||
@classmethod
|
||||
def fromRandom(cls,randomSeed = None):
|
||||
import binascii
|
||||
if randomSeed is None:
|
||||
randomSeed = int(os.urandom(4).encode('hex'), 16)
|
||||
randomSeed = int(binascii.hexlify(os.urandom(4)),16)
|
||||
np.random.seed(randomSeed)
|
||||
r = np.random.random(3)
|
||||
w = math.cos(2.0*math.pi*r[0])*math.sqrt(r[2])
|
||||
x = math.sin(2.0*math.pi*r[1])*math.sqrt(1.0-r[2])
|
||||
y = math.cos(2.0*math.pi*r[1])*math.sqrt(1.0-r[2])
|
||||
z = math.sin(2.0*math.pi*r[0])*math.sqrt(r[2])
|
||||
return cls([w,x,y,z])
|
||||
return cls(quat=[w,x,y,z])
|
||||
|
||||
|
||||
@classmethod
|
||||
def fromRodrigues(cls, rodrigues):
|
||||
if not isinstance(rodrigues, np.ndarray): rodrigues = np.array(rodrigues)
|
||||
halfangle = math.atan(np.linalg.norm(rodrigues))
|
||||
norm = np.linalg.norm(rodrigues)
|
||||
halfangle = math.atan(norm)
|
||||
s = math.sin(halfangle)
|
||||
c = math.cos(halfangle)
|
||||
w = c
|
||||
x,y,z = rodrigues/c
|
||||
return cls([w,x,y,z])
|
||||
return cls(q=c,p=s*rodrigues/norm)
|
||||
|
||||
|
||||
@classmethod
|
||||
|
@ -397,22 +333,19 @@ class Quaternion:
|
|||
angle,
|
||||
axis,
|
||||
degrees = False):
|
||||
if not isinstance(axis, np.ndarray): axis = np.array(axis,dtype='d')
|
||||
if not isinstance(axis, np.ndarray): axis = np.array(axis,dtype=float)
|
||||
axis = axis.astype(float)/np.linalg.norm(axis)
|
||||
angle = np.radians(angle) if degrees else angle
|
||||
s = math.sin(0.5 * angle)
|
||||
w = math.cos(0.5 * angle)
|
||||
x = axis[0] * s
|
||||
y = axis[1] * s
|
||||
z = axis[2] * s
|
||||
return cls([w,x,y,z])
|
||||
c = math.cos(0.5 * angle)
|
||||
return cls(q=c,p=axis*s)
|
||||
|
||||
|
||||
@classmethod
|
||||
def fromEulers(cls,
|
||||
eulers,
|
||||
degrees = False):
|
||||
if not isinstance(eulers, np.ndarray): eulers = np.array(eulers,dtype='d')
|
||||
if not isinstance(eulers, np.ndarray): eulers = np.array(eulers,dtype=float)
|
||||
eulers = np.radians(eulers) if degrees else eulers
|
||||
|
||||
sigma = 0.5*(eulers[0]+eulers[2])
|
||||
|
@ -420,11 +353,13 @@ class Quaternion:
|
|||
c = np.cos(0.5*eulers[1])
|
||||
s = np.sin(0.5*eulers[1])
|
||||
|
||||
# Rowenhorst_etal2015 MSMSE: value of P is selected as -1
|
||||
P = -1.0
|
||||
w = c * np.cos(sigma)
|
||||
x = -s * np.cos(delta)
|
||||
y = -s * np.sin(delta)
|
||||
z = -c * np.sin(sigma)
|
||||
return cls([w,x,y,z])
|
||||
x = -P * s * np.cos(delta)
|
||||
y = -P * s * np.sin(delta)
|
||||
z = -P * c * np.sin(sigma)
|
||||
return cls(quat=[w,x,y,z])
|
||||
|
||||
|
||||
# Modified Method to calculate Quaternion from Orientation Matrix,
|
||||
|
@ -435,16 +370,18 @@ class Quaternion:
|
|||
if m.shape != (3,3) and np.prod(m.shape) == 9:
|
||||
m = m.reshape(3,3)
|
||||
|
||||
# Rowenhorst_etal2015 MSMSE: value of P is selected as -1
|
||||
P = -1.0
|
||||
w = 0.5*math.sqrt(1.+m[0,0]+m[1,1]+m[2,2])
|
||||
x = 0.5*math.sqrt(1.+m[0,0]-m[1,1]-m[2,2])
|
||||
y = 0.5*math.sqrt(1.-m[0,0]+m[1,1]-m[2,2])
|
||||
z = 0.5*math.sqrt(1.-m[0,0]-m[1,1]+m[2,2])
|
||||
x = P*0.5*math.sqrt(1.+m[0,0]-m[1,1]-m[2,2])
|
||||
y = P*0.5*math.sqrt(1.-m[0,0]+m[1,1]-m[2,2])
|
||||
z = P*0.5*math.sqrt(1.-m[0,0]-m[1,1]+m[2,2])
|
||||
|
||||
x *= -1 if m[2,1] < m[1,2] else 1
|
||||
y *= -1 if m[0,2] < m[2,0] else 1
|
||||
z *= -1 if m[1,0] < m[0,1] else 1
|
||||
|
||||
return cls( np.array([w,x,y,z])/math.sqrt(w**2 + x**2 + y**2 + z**2))
|
||||
return cls(quat=np.array([w,x,y,z])/math.sqrt(w**2 + x**2 + y**2 + z**2))
|
||||
|
||||
|
||||
@classmethod
|
||||
|
@ -458,36 +395,30 @@ class Quaternion:
|
|||
assert isinstance(q1, Quaternion) and isinstance(q2, Quaternion)
|
||||
Q = cls()
|
||||
|
||||
costheta = q1.w * q2.w + q1.x * q2.x + q1.y * q2.y + q1.z * q2.z
|
||||
costheta = q1.q*q2.q + np.dot(q1.p,q2.p)
|
||||
if costheta < 0.:
|
||||
costheta = -costheta
|
||||
q1 = q1.conjugated()
|
||||
elif costheta > 1:
|
||||
costheta = 1
|
||||
elif costheta > 1.:
|
||||
costheta = 1.
|
||||
|
||||
theta = math.acos(costheta)
|
||||
if abs(theta) < 0.01:
|
||||
Q.w = q2.w
|
||||
Q.x = q2.x
|
||||
Q.y = q2.y
|
||||
Q.z = q2.z
|
||||
Q.q = q2.q
|
||||
Q.p = q2.p
|
||||
return Q
|
||||
|
||||
sintheta = math.sqrt(1.0 - costheta * costheta)
|
||||
if abs(sintheta) < 0.01:
|
||||
Q.w = (q1.w + q2.w) * 0.5
|
||||
Q.x = (q1.x + q2.x) * 0.5
|
||||
Q.y = (q1.y + q2.y) * 0.5
|
||||
Q.z = (q1.z + q2.z) * 0.5
|
||||
Q.q = (q1.q + q2.q) * 0.5
|
||||
Q.p = (q1.p + q2.p) * 0.5
|
||||
return Q
|
||||
|
||||
ratio1 = math.sin((1 - t) * theta) / sintheta
|
||||
ratio1 = math.sin((1.0 - t) * theta) / sintheta
|
||||
ratio2 = math.sin( t * theta) / sintheta
|
||||
|
||||
Q.w = q1.w * ratio1 + q2.w * ratio2
|
||||
Q.x = q1.x * ratio1 + q2.x * ratio2
|
||||
Q.y = q1.y * ratio1 + q2.y * ratio2
|
||||
Q.z = q1.z * ratio1 + q2.z * ratio2
|
||||
Q.q = q1.q * ratio1 + q2.q * ratio2
|
||||
Q.p = q1.p * ratio1 + q2.p * ratio2
|
||||
return Q
|
||||
|
||||
|
||||
|
@ -513,7 +444,7 @@ class Symmetry:
|
|||
|
||||
def __repr__(self):
|
||||
"""Readbable string"""
|
||||
return '%s' % (self.lattice)
|
||||
return '{}'.format(self.lattice)
|
||||
|
||||
|
||||
def __eq__(self, other):
|
||||
|
@ -737,10 +668,10 @@ class Symmetry:
|
|||
if color: # have to return color array
|
||||
if inSST:
|
||||
rgb = np.power(theComponents/np.linalg.norm(theComponents),0.5) # smoothen color ramps
|
||||
rgb = np.minimum(np.ones(3,'d'),rgb) # limit to maximum intensity
|
||||
rgb = np.minimum(np.ones(3,dtype=float),rgb) # limit to maximum intensity
|
||||
rgb /= max(rgb) # normalize to (HS)V = 1
|
||||
else:
|
||||
rgb = np.zeros(3,'d')
|
||||
rgb = np.zeros(3,dtype=float)
|
||||
return (inSST,rgb)
|
||||
else:
|
||||
return inSST
|
||||
|
@ -780,8 +711,9 @@ class Orientation:
|
|||
self.quaternion = Quaternion.fromRodrigues(Rodrigues)
|
||||
elif isinstance(quaternion, Quaternion): # based on given quaternion
|
||||
self.quaternion = quaternion.homomorphed()
|
||||
elif isinstance(quaternion, np.ndarray) and quaternion.shape == (4,): # based on given quaternion-like array
|
||||
self.quaternion = Quaternion(quaternion).homomorphed()
|
||||
elif (isinstance(quaternion, np.ndarray) and quaternion.shape == (4,)) or \
|
||||
(isinstance(quaternion, list) and len(quaternion) == 4 ): # based on given quaternion-like array
|
||||
self.quaternion = Quaternion(quat=quaternion).homomorphed()
|
||||
|
||||
self.symmetry = Symmetry(symmetry)
|
||||
|
||||
|
@ -794,10 +726,12 @@ class Orientation:
|
|||
|
||||
def __repr__(self):
|
||||
"""Value as all implemented representations"""
|
||||
return 'Symmetry: %s\n' % (self.symmetry) + \
|
||||
'Quaternion: %s\n' % (self.quaternion) + \
|
||||
'Matrix:\n%s\n' % ( '\n'.join(['\t'.join(map(str,self.asMatrix()[i,:])) for i in range(3)]) ) + \
|
||||
'Bunge Eulers / deg: %s' % ('\t'.join(map(str,self.asEulers(degrees=True))) )
|
||||
return '\n'.join([
|
||||
'Symmetry: {}'.format(self.symmetry),
|
||||
'Quaternion: {}'.format(self.quaternion),
|
||||
'Matrix:\n{}'.format( '\n'.join(['\t'.join(list(map(str,self.asMatrix()[i,:]))) for i in range(3)]) ),
|
||||
'Bunge Eulers / deg: {}'.format('\t'.join(list(map(str,self.asEulers(degrees=True)))) ),
|
||||
])
|
||||
|
||||
def asQuaternion(self):
|
||||
return self.quaternion.asList()
|
||||
|
@ -927,7 +861,7 @@ class Orientation:
|
|||
M = closest.quaternion.asM() * n if i == 0 else M + closest.quaternion.asM() * n # noqa add (multiples) of this orientation to average noqa
|
||||
eig, vec = np.linalg.eig(M/N)
|
||||
|
||||
return Orientation(quaternion = Quaternion(quatArray = np.real(vec.T[eig.argmax()])),
|
||||
return Orientation(quaternion = Quaternion(quat = np.real(vec.T[eig.argmax()])),
|
||||
symmetry = reference.symmetry.lattice)
|
||||
|
||||
|
||||
|
|
|
@ -200,9 +200,9 @@ for name in filenames:
|
|||
if gID != -1 and gID not in alreadyChecked: # indexed point belonging to a grain not yet tested?
|
||||
alreadyChecked[gID] = True # remember not to check again
|
||||
disorientation = o.disorientation(orientations[gID],SST = False)[0] # compare against other orientation
|
||||
if disorientation.quaternion.w > cos_disorientation: # within threshold ...
|
||||
if disorientation.quaternion.q > cos_disorientation: # within threshold ...
|
||||
candidates.append(gID) # remember as potential candidate
|
||||
if disorientation.quaternion.w >= bestDisorientation.w: # ... and better than current best?
|
||||
if disorientation.quaternion.q >= bestDisorientation.q: # ... and better than current best?
|
||||
matched = True
|
||||
matchedID = gID # remember that grain
|
||||
bestDisorientation = disorientation.quaternion
|
||||
|
|
|
@ -64,11 +64,11 @@ if options.dimension is None:
|
|||
parser.error('no dimension specified.')
|
||||
if options.angleaxis is not None:
|
||||
options.angleaxis = list(map(float,options.angleaxis))
|
||||
rotation = damask.Quaternion().fromAngleAxis(np.radians(options.angleaxis[0]) if options.degrees else options.angleaxis[0],
|
||||
rotation = damask.Quaternion.fromAngleAxis(np.radians(options.angleaxis[0]) if options.degrees else options.angleaxis[0],
|
||||
options.angleaxis[1:4])
|
||||
elif options.quaternion is not None:
|
||||
options.quaternion = map(float,options.quaternion)
|
||||
rotation = damask.Quaternion(options.quaternion)
|
||||
options.quaternion = list(map(float,options.quaternion))
|
||||
rotation = damask.Quaternion(quat=options.quaternion)
|
||||
else:
|
||||
rotation = damask.Quaternion()
|
||||
|
||||
|
|
|
@ -43,7 +43,7 @@ parser.add_option('-e', '--eulers',
|
|||
parser.add_option('-d', '--degrees',
|
||||
dest = 'degrees',
|
||||
action = 'store_true',
|
||||
help = 'angles are given in degrees [%default]')
|
||||
help = 'all angles are in degrees')
|
||||
parser.add_option('-m', '--matrix',
|
||||
dest = 'matrix',
|
||||
type = 'string', metavar = 'string',
|
||||
|
@ -71,7 +71,7 @@ parser.add_option('--axes',
|
|||
parser.add_option('-s', '--symmetry',
|
||||
dest = 'symmetry',
|
||||
action = 'extend', metavar = '<string LIST>',
|
||||
help = 'crystal symmetry %default {{{}}} '.format(', '.join(damask.Symmetry.lattices[1:])))
|
||||
help = 'crystal symmetry of each phase %default {{{}}} '.format(', '.join(damask.Symmetry.lattices[1:])))
|
||||
parser.add_option('--homogenization',
|
||||
dest = 'homogenization',
|
||||
type = 'int', metavar = 'int',
|
||||
|
@ -234,7 +234,7 @@ for name in filenames:
|
|||
o = damask.Orientation(Eulers = myData[colOri:colOri+3]*toRadians,
|
||||
symmetry = mySym)
|
||||
elif inputtype == 'matrix':
|
||||
o = damask.Orientation(matrix = myData[colOri:colOri+9].reshape(3,3).transpose(),
|
||||
o = damask.Orientation(matrix = myData[colOri:colOri+9].reshape(3,3),
|
||||
symmetry = mySym)
|
||||
elif inputtype == 'frame':
|
||||
o = damask.Orientation(matrix = np.hstack((myData[colOri[0]:colOri[0]+3],
|
||||
|
@ -246,7 +246,7 @@ for name in filenames:
|
|||
o = damask.Orientation(quaternion = myData[colOri:colOri+4],
|
||||
symmetry = mySym)
|
||||
|
||||
cos_disorientations = -np.ones(1,dtype='f') # largest possible disorientation
|
||||
cos_disorientations = -np.ones(1,dtype=float) # largest possible disorientation
|
||||
closest_grain = -1 # invalid neighbor
|
||||
|
||||
if options.tolerance > 0.0: # only try to compress orientations if asked to
|
||||
|
@ -258,7 +258,7 @@ for name in filenames:
|
|||
|
||||
if len(grains) > 0: # check immediate neighborhood first
|
||||
cos_disorientations = np.array([o.disorientation(orientations[grainID],
|
||||
SST = False)[0].quaternion.w \
|
||||
SST = False)[0].quaternion.q \
|
||||
for grainID in grains]) # store disorientation per grainID
|
||||
closest_grain = np.argmax(cos_disorientations) # grain among grains with closest orientation to myself
|
||||
match = 'local'
|
||||
|
@ -269,7 +269,7 @@ for name in filenames:
|
|||
|
||||
if len(grains) > 0:
|
||||
cos_disorientations = np.array([o.disorientation(orientations[grainID],
|
||||
SST = False)[0].quaternion.w \
|
||||
SST = False)[0].quaternion.q \
|
||||
for grainID in grains]) # store disorientation per grainID
|
||||
closest_grain = np.argmax(cos_disorientations) # grain among grains with closest orientation to myself
|
||||
match = 'global'
|
||||
|
|
|
@ -244,7 +244,7 @@ for name in filenames:
|
|||
continue
|
||||
damask.util.report(scriptName,name)
|
||||
|
||||
randomSeed = int(os.urandom(4).encode('hex'), 16) if options.randomSeed is None else options.randomSeed # random seed per file for second phase
|
||||
randomSeed = int(os.urandom(4).hex(), 16) if options.randomSeed is None else options.randomSeed # random seed per file for second phase
|
||||
random.seed(randomSeed)
|
||||
|
||||
# ------------------------------------------ read header and data ---------------------------------
|
||||
|
|
10
src/IO.f90
10
src/IO.f90
|
@ -191,7 +191,9 @@ recursive function IO_recursiveRead(fileName,cnt) result(fileContent)
|
|||
l,i, &
|
||||
myStat
|
||||
|
||||
if (merge(cnt,0_pInt,present(cnt))>10_pInt) call IO_error(106_pInt,ext_msg=trim(fileName))
|
||||
if (present(cnt)) then
|
||||
if (cnt>10_pInt) call IO_error(106_pInt,ext_msg=trim(fileName))
|
||||
endif
|
||||
|
||||
!--------------------------------------------------------------------------------------------------
|
||||
! read data as stream
|
||||
|
@ -684,7 +686,11 @@ function IO_stringValue(string,chunkPos,myChunk,silent)
|
|||
|
||||
logical :: warn
|
||||
|
||||
warn = merge(silent,.false.,present(silent))
|
||||
if (present(silent)) then
|
||||
warn = silent
|
||||
else
|
||||
warn = .false.
|
||||
endif
|
||||
|
||||
IO_stringValue = ''
|
||||
valuePresent: if (myChunk > chunkPos(1) .or. myChunk < 1_pInt) then
|
||||
|
|
|
@ -513,8 +513,12 @@ character(len=65536) function getString(this,key,defaultVal,raw)
|
|||
type(tPartitionedStringList), pointer :: item
|
||||
logical :: found, &
|
||||
whole
|
||||
if (present(raw)) then
|
||||
whole = raw
|
||||
else
|
||||
whole = .false.
|
||||
endif
|
||||
|
||||
whole = merge(raw,.false.,present(raw)) ! whole string or white space splitting
|
||||
found = present(defaultVal)
|
||||
if (found) then
|
||||
getString = trim(defaultVal)
|
||||
|
@ -661,7 +665,11 @@ function getStrings(this,key,defaultVal,requiredShape,raw)
|
|||
cumulative
|
||||
|
||||
cumulative = (key(1:1) == '(' .and. key(len_trim(key):len_trim(key)) == ')')
|
||||
whole = merge(raw,.false.,present(raw))
|
||||
if (present(raw)) then
|
||||
whole = raw
|
||||
else
|
||||
whole = .false.
|
||||
endif
|
||||
found = .false.
|
||||
|
||||
item => this
|
||||
|
|
|
@ -525,9 +525,9 @@ subroutine constitutive_LpAndItsTangents(Lp, dLp_dS, dLp_dFi, S6, Fi, ipc, ip, e
|
|||
dLp_dMp = math_Plain99to3333(dLp_dMp99) ! ToDo: We revert here the last statement in plastic_xx_LpAndItsTanget
|
||||
|
||||
case (PLASTICITY_DISLOTWIN_ID) plasticityType
|
||||
call plastic_dislotwin_LpAndItsTangent (Lp,dLp_dMp99, math_Mandel33to6(Mp), &
|
||||
temperature(ho)%p(tme),ipc,ip,el)
|
||||
dLp_dMp = math_Plain99to3333(dLp_dMp99) ! ToDo: We revert here the last statement in plastic_xx_LpAndItsTanget
|
||||
of = phasememberAt(ipc,ip,el)
|
||||
instance = phase_plasticityInstance(material_phase(ipc,ip,el))
|
||||
call plastic_dislotwin_LpAndItsTangent (Lp,dLp_dMp,Mp,temperature(ho)%p(tme),instance,of)
|
||||
|
||||
case (PLASTICITY_DISLOUCLA_ID) plasticityType
|
||||
call plastic_disloucla_LpAndItsTangent (Lp,dLp_dMp99, math_Mandel33to6(Mp), &
|
||||
|
@ -922,8 +922,9 @@ subroutine constitutive_collectDotState(S6, FeArray, Fi, FpArray, subdt, subfrac
|
|||
call plastic_kinehardening_dotState(math_Mandel33to6(Mp),ipc,ip,el)
|
||||
|
||||
case (PLASTICITY_DISLOTWIN_ID) plasticityType
|
||||
call plastic_dislotwin_dotState (math_Mandel33to6(Mp),temperature(ho)%p(tme), &
|
||||
ipc,ip,el)
|
||||
of = phasememberAt(ipc,ip,el)
|
||||
instance = phase_plasticityInstance(material_phase(ipc,ip,el))
|
||||
call plastic_dislotwin_dotState (Mp,temperature(ho)%p(tme),instance,of)
|
||||
|
||||
case (PLASTICITY_DISLOUCLA_ID) plasticityType
|
||||
call plastic_disloucla_dotState (math_Mandel33to6(Mp),temperature(ho)%p(tme), &
|
||||
|
@ -1135,20 +1136,27 @@ function constitutive_postResults(S6, Fi, FeArray, ipc, ip, el)
|
|||
case (PLASTICITY_ISOTROPIC_ID) plasticityType
|
||||
constitutive_postResults(startPos:endPos) = &
|
||||
plastic_isotropic_postResults(S6,ipc,ip,el)
|
||||
|
||||
case (PLASTICITY_PHENOPOWERLAW_ID) plasticityType
|
||||
of = phasememberAt(ipc,ip,el)
|
||||
instance = phase_plasticityInstance(material_phase(ipc,ip,el))
|
||||
constitutive_postResults(startPos:endPos) = &
|
||||
plastic_phenopowerlaw_postResults(Mp,instance,of)
|
||||
|
||||
case (PLASTICITY_KINEHARDENING_ID) plasticityType
|
||||
constitutive_postResults(startPos:endPos) = &
|
||||
plastic_kinehardening_postResults(S6,ipc,ip,el)
|
||||
|
||||
case (PLASTICITY_DISLOTWIN_ID) plasticityType
|
||||
of = phasememberAt(ipc,ip,el)
|
||||
instance = phase_plasticityInstance(material_phase(ipc,ip,el))
|
||||
constitutive_postResults(startPos:endPos) = &
|
||||
plastic_dislotwin_postResults(S6,temperature(ho)%p(tme),ipc,ip,el)
|
||||
plastic_dislotwin_postResults(Mp,temperature(ho)%p(tme),instance,of)
|
||||
|
||||
case (PLASTICITY_DISLOUCLA_ID) plasticityType
|
||||
constitutive_postResults(startPos:endPos) = &
|
||||
plastic_disloucla_postResults(S6,temperature(ho)%p(tme),ipc,ip,el)
|
||||
|
||||
case (PLASTICITY_NONLOCAL_ID) plasticityType
|
||||
constitutive_postResults(startPos:endPos) = &
|
||||
plastic_nonlocal_postResults (S6,FeArray,ip,el)
|
||||
|
|
|
@ -225,7 +225,7 @@ subroutine kinematics_thermal_expansion_LiAndItsTangent(Li, dLi_dTstar3333, ipc,
|
|||
+ lattice_thermalExpansion33(1:3,1:3,2,phase)*(T - TRef)**1 & ! linear coefficient
|
||||
+ lattice_thermalExpansion33(1:3,1:3,3,phase)*(T - TRef)**2 & ! quadratic coefficient
|
||||
) / &
|
||||
(1.0_pReal \
|
||||
(1.0_pReal &
|
||||
+ lattice_thermalExpansion33(1:3,1:3,1,phase)*(T - TRef)**1 / 1. &
|
||||
+ lattice_thermalExpansion33(1:3,1:3,2,phase)*(T - TRef)**2 / 2. &
|
||||
+ lattice_thermalExpansion33(1:3,1:3,3,phase)*(T - TRef)**3 / 3. &
|
||||
|
|
|
@ -72,7 +72,7 @@ module lattice
|
|||
!--------------------------------------------------------------------------------------------------
|
||||
! face centered cubic
|
||||
integer(pInt), dimension(LATTICE_maxNslipFamily), parameter, public :: &
|
||||
LATTICE_fcc_NslipSystem = int([12, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],pInt) !< # of slip systems per family for fcc
|
||||
LATTICE_fcc_NslipSystem = int([12, 6, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],pInt) !< # of slip systems per family for fcc
|
||||
|
||||
integer(pInt), dimension(LATTICE_maxNtwinFamily), parameter, public :: &
|
||||
LATTICE_fcc_NtwinSystem = int([12, 0, 0, 0],pInt) !< # of twin systems per family for fcc
|
||||
|
@ -104,11 +104,19 @@ module lattice
|
|||
1, 1, 0, 1,-1,-1, & ! A6
|
||||
0, 1, 1, -1, 1,-1, & ! D1
|
||||
1, 0,-1, -1, 1,-1, & ! D4
|
||||
-1,-1, 0, -1, 1,-1 & ! D6
|
||||
-1,-1, 0, -1, 1,-1, & ! D6
|
||||
! Slip system <110>{110}
|
||||
1, 1, 0, 1,-1, 0, &
|
||||
1,-1, 0, 1, 1, 0, &
|
||||
1, 0, 1, 1, 0,-1, &
|
||||
1, 0,-1, 1, 0, 1, &
|
||||
0, 1, 1, 0, 1,-1, &
|
||||
0, 1,-1, 0, 1, 1 &
|
||||
],pReal),shape(LATTICE_FCC_SYSTEMSLIP)) !< Slip system <110>{111} directions. Sorted according to Eisenlohr & Hantcherli
|
||||
|
||||
character(len=*), dimension(1), parameter, public :: LATTICE_FCC_SLIPFAMILY_NAME = &
|
||||
['<0 1 -1>{1 1 1}']
|
||||
character(len=*), dimension(2), parameter, public :: LATTICE_FCC_SLIPFAMILY_NAME = &
|
||||
['<0 1 -1>{1 1 1}', &
|
||||
'<0 1 -1>{0 1 1}']
|
||||
|
||||
real(pReal), dimension(3+3,LATTICE_fcc_Ntwin), parameter, private :: &
|
||||
LATTICE_fcc_systemTwin = reshape(real( [&
|
||||
|
@ -166,25 +174,38 @@ module lattice
|
|||
|
||||
integer(pInt), dimension(LATTICE_fcc_Nslip,lattice_fcc_Nslip), parameter, public :: &
|
||||
LATTICE_fcc_interactionSlipSlip = reshape(int( [&
|
||||
1,2,2,4,6,5,3,5,5,4,5,6, & ! ---> slip
|
||||
2,1,2,6,4,5,5,4,6,5,3,5, & ! |
|
||||
2,2,1,5,5,3,5,6,4,6,5,4, & ! |
|
||||
4,6,5,1,2,2,4,5,6,3,5,5, & ! v slip
|
||||
6,4,5,2,1,2,5,3,5,5,4,6, &
|
||||
5,5,3,2,2,1,6,5,4,5,6,4, &
|
||||
3,5,5,4,5,6,1,2,2,4,6,5, &
|
||||
5,4,6,5,3,5,2,1,2,6,4,5, &
|
||||
5,6,4,6,5,4,2,2,1,5,5,3, &
|
||||
4,5,6,3,5,5,4,6,5,1,2,2, &
|
||||
5,3,5,5,4,6,6,4,5,2,1,2, &
|
||||
6,5,4,5,6,4,5,5,3,2,2,1 &
|
||||
],pInt),shape(LATTICE_FCC_INTERACTIONSLIPSLIP),order=[2,1]) !< Slip--slip interaction types for fcc
|
||||
1, 2, 2, 4, 6, 5, 3, 5, 5, 4, 5, 6, 9,10, 9,10,11,12, & ! ---> slip
|
||||
2, 1, 2, 6, 4, 5, 5, 4, 6, 5, 3, 5, 9,10,11,12, 9,10, & ! |
|
||||
2, 2, 1, 5, 5, 3, 5, 6, 4, 6, 5, 4, 11,12, 9,10, 9,10, & ! |
|
||||
4, 6, 5, 1, 2, 2, 4, 5, 6, 3, 5, 5, 9,10,10, 9,12,11, & ! v slip
|
||||
6, 4, 5, 2, 1, 2, 5, 3, 5, 5, 4, 6, 9,10,12,11,10, 9, &
|
||||
5, 5, 3, 2, 2, 1, 6, 5, 4, 5, 6, 4, 11,12,10, 9,10, 9, &
|
||||
3, 5, 5, 4, 5, 6, 1, 2, 2, 4, 6, 5, 10, 9,10, 9,11,12, &
|
||||
5, 4, 6, 5, 3, 5, 2, 1, 2, 6, 4, 5, 10, 9,12,11, 9,10, &
|
||||
5, 6, 4, 6, 5, 4, 2, 2, 1, 5, 5, 3, 12,11,10, 9, 9,10, &
|
||||
4, 5, 6, 3, 5, 5, 4, 6, 5, 1, 2, 2, 10, 9, 9,10,12,11, &
|
||||
5, 3, 5, 5, 4, 6, 6, 4, 5, 2, 1, 2, 10, 9,11,12,10, 9, &
|
||||
6, 5, 4, 5, 6, 4, 5, 5, 3, 2, 2, 1, 12,11, 9,10,10, 9, &
|
||||
|
||||
9, 9,11, 9, 9,11,10,10,12,10,10,12, 1, 7, 8, 8, 8, 8, &
|
||||
10,10,12,10,10,12, 9, 9,11, 9, 9,11, 7, 1, 8, 8, 8, 8, &
|
||||
9,11, 9,10,12,10,10,12,10, 9,11, 9, 8, 8, 1, 7, 8, 8, &
|
||||
10,12,10, 9,11, 9, 9,11, 9,10,12,10, 8, 8, 7, 1, 8, 8, &
|
||||
11, 9, 9,12,10,10,11, 9, 9,12,10,10, 8, 8, 8, 8, 1, 7, &
|
||||
12,10,10,11, 9, 9,12,10,10,11, 9, 9, 8, 8, 8, 8, 7, 1 &
|
||||
],pInt),[LATTICE_fcc_Nslip,LATTICE_fcc_Nslip],order=[2,1]) !< Slip--slip interaction types for fcc
|
||||
!< 1: self interaction
|
||||
!< 2: coplanar interaction
|
||||
!< 3: collinear interaction
|
||||
!< 4: Hirth locks
|
||||
!< 5: glissile junctions
|
||||
!< 6: Lomer locks
|
||||
!< 7: crossing (similar to Hirth locks in <110>{111} for two {110} planes)
|
||||
!< 8: similar to Lomer locks in <110>{111} for two {110} planes
|
||||
!< 9: similar to Lomer locks in <110>{111} btw one {110} and one {111} plane
|
||||
!<10: similar to glissile junctions in <110>{111} btw one {110} and one {111} plane
|
||||
!<11: crossing btw one {110} and one {111} plane
|
||||
!<12: collinear btw one {110} and one {111} plane
|
||||
integer(pInt), dimension(LATTICE_fcc_Nslip,LATTICE_fcc_Ntwin), parameter, public :: &
|
||||
LATTICE_fcc_interactionSlipTwin = reshape(int( [&
|
||||
1,1,1,3,3,3,2,2,2,3,3,3, & ! ---> twin
|
||||
|
@ -198,7 +219,14 @@ module lattice
|
|||
3,3,3,3,3,3,1,1,1,2,2,2, &
|
||||
3,3,3,2,2,2,3,3,3,1,1,1, &
|
||||
2,2,2,3,3,3,3,3,3,1,1,1, &
|
||||
3,3,3,3,3,3,2,2,2,1,1,1 &
|
||||
3,3,3,3,3,3,2,2,2,1,1,1, &
|
||||
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4 &
|
||||
],pInt),shape(LATTICE_FCC_INTERACTIONSLIPTWIN),order=[2,1]) !< Slip--twin interaction types for fcc
|
||||
!< 1: coplanar interaction
|
||||
!< 2: screw trace between slip system and twin habit plane (easy cross slip)
|
||||
|
@ -235,7 +263,14 @@ module lattice
|
|||
3,3,3,3,3,3,1,1,1,2,2,2, &
|
||||
3,3,3,2,2,2,3,3,3,1,1,1, &
|
||||
2,2,2,3,3,3,3,3,3,1,1,1, &
|
||||
3,3,3,3,3,3,2,2,2,1,1,1 &
|
||||
3,3,3,3,3,3,2,2,2,1,1,1, &
|
||||
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4, &
|
||||
4,4,4,4,4,4,4,4,4,4,4,4 &
|
||||
],pInt),shape(LATTICE_FCCTOHEX_INTERACTIONSLIPTRANS),order=[2,1]) !< Slip--trans interaction types for fcc
|
||||
|
||||
integer(pInt), dimension(LATTICE_fcc_Ntrans,LATTICE_fcc_Nslip), parameter, public :: &
|
||||
|
|
|
@ -2625,12 +2625,9 @@ real(pReal) pure function math_clip(a, left, right)
|
|||
real(pReal), intent(in) :: a
|
||||
real(pReal), intent(in), optional :: left, right
|
||||
|
||||
|
||||
math_clip = min ( &
|
||||
max (merge(left, -huge(a), present(left)), a), &
|
||||
merge(right, huge(a), present(right)) &
|
||||
)
|
||||
|
||||
math_clip = a
|
||||
if (present(left)) math_clip = max(left,math_clip)
|
||||
if (present(right)) math_clip = min(right,math_clip)
|
||||
if (present(left) .and. present(right)) &
|
||||
math_clip = merge (IEEE_value(1.0_pReal,IEEE_quiet_NaN),math_clip, left>right)
|
||||
|
||||
|
|
File diff suppressed because it is too large
Load Diff
|
@ -120,8 +120,8 @@ subroutine spectral_damage_init()
|
|||
trim(snes_type) == 'vinewtonssls') then
|
||||
call DMGetGlobalVector(damage_grid,lBound,ierr); CHKERRQ(ierr)
|
||||
call DMGetGlobalVector(damage_grid,uBound,ierr); CHKERRQ(ierr)
|
||||
call VecSet(lBound,0.0,ierr); CHKERRQ(ierr)
|
||||
call VecSet(uBound,1.0,ierr); CHKERRQ(ierr)
|
||||
call VecSet(lBound,0.0_pReal,ierr); CHKERRQ(ierr)
|
||||
call VecSet(uBound,1.0_pReal,ierr); CHKERRQ(ierr)
|
||||
call SNESVISetVariableBounds(damage_snes,lBound,uBound,ierr) !< variable bounds for variational inequalities like contact mechanics, damage etc.
|
||||
call DMRestoreGlobalVector(damage_grid,lBound,ierr); CHKERRQ(ierr)
|
||||
call DMRestoreGlobalVector(damage_grid,uBound,ierr); CHKERRQ(ierr)
|
||||
|
@ -134,7 +134,7 @@ subroutine spectral_damage_init()
|
|||
xend = xstart + xend - 1
|
||||
yend = ystart + yend - 1
|
||||
zend = zstart + zend - 1
|
||||
call VecSet(solution,1.0,ierr); CHKERRQ(ierr)
|
||||
call VecSet(solution,1.0_pReal,ierr); CHKERRQ(ierr)
|
||||
allocate(damage_current(grid(1),grid(2),grid3), source=1.0_pReal)
|
||||
allocate(damage_lastInc(grid(1),grid(2),grid3), source=1.0_pReal)
|
||||
allocate(damage_stagInc(grid(1),grid(2),grid3), source=1.0_pReal)
|
||||
|
|
Loading…
Reference in New Issue