def __init__(self, v, t, b, Mach):
self.mesh = Mesh(v, t, b)
self.Mach = float(Mach)
self.ode = Ode(self.ddt, self.J)
class Euler:
def __init__(self, v, t, b, Mach):
self.mesh = Mesh(v, t, b)
self.Mach = float(Mach)
self.ode = Ode(self.ddt, self.J)
def integrate(self, *args, **argv):
return self.ode.integrate(*args, **argv)
@property
def nt(self):
return self.mesh.nt
@property
def time(self):
return self.ode.t
@property
def soln(self):
return reshape(self.ode.y, [-1,4])
@property
def Wref(self):
Wr = self.freeStream(1)[0]
Wr[2] = Wr[1]
return Wr
def freeStream(self, nT=None):
if nT is None: nT = self.nt
R, gamma, rho0, T0 = 8.314 / 29E-3, 1.4, 1.225, 288.75
u0 = self.Mach * sqrt(gamma * R * T0)
E0 = T0 * R / (gamma - 1) + 0.5 * u0**2
return array([rho0, rho0 * u0, 0, rho0 * E0]) + zeros([nT, 1])
def ddt(self, W):
assert W.size == self.nt * 4
shp = W.shape
W = W.reshape([-1, 4])
m = self.mesh
gradW = m.gradTri(W)
xt = m.xt()
dxt = m.dxt
# boundary condition categories
xeBnd = m.v[m.e[m.ieBnd,:2],:].mean(1)
isWall = ~m.isFar(xeBnd)
# interior flux
#WL, WR = W[m.e[:,2],:], W[m.e[:,3],:]
WL, WR = m.leftRightTri(W)
WL[m.ieBnd[~isWall]] = self.freeStream(1)
WE = 0.5*(WL+WR)
dW = WR - WL
#dWL = (gradW[m.e[:,2],:] * dxt[:,:,newaxis]).sum(1)
#dWR = (gradW[m.e[:,3],:] * dxt[:,:,newaxis]).sum(1)
gWL, gWR = m.leftRightTri(gradW)
dWL = einsum( 'nij, ni -> nj', gWL, dxt )
dWR = einsum( 'nij, ni -> nj', gWR, dxt )
flux = fluxE(WE, m.n) + fluxD(WE, dW, dWL, dWR, m.n)
# boundary flux
WBnd, nBnd = WL[m.ieBnd,:], m.n[m.ieBnd,:]
flux[m.ieBnd[isWall]] = wallBc(WBnd[isWall,:], nBnd[isWall,:])
# accumunate flux to cells
return (m.distributeFlux(flux) / m.a[:,newaxis]).reshape(shp)
def J(self, W):
if not self.__dict__.has_key('matJacDistFlux'):
self.prepareJacMatrices()
assert W.size == self.nt * 4
shp = W.shape
W = W.reshape([-1, 4])
m = self.mesh
# prepare values at edge
gradW = m.gradTri(W)
xt = m.xt();
dxt = m.dxt
WL, WR = m.leftRightTri(W)
isFar = m.isFar(m.v[m.e[m.ieBnd,:2],:].mean(1))
WL[m.ieBnd[isFar]] = self.freeStream(1)
WE, dW = 0.5 * (WL + WR), WR - WL
JE_WE = jacE(WE, m.n)
gWL, gWR = m.leftRightTri(gradW)
dWL = einsum( 'nij, ni -> nj', gWL, dxt )
dWR = einsum( 'nij, ni -> nj', gWR, dxt )
JD_WE, JD_dW, JD_dWL, JD_dWR = jacD(WE, dW, dWL, dWR, m.n)
# modify for wall BC
J_WE = JE_WE + JD_WE
ieWall = m.ieBnd[~isFar]
J_WE[ieWall,:] = jacWall(WE[ieWall], m.n[ieWall])
J_flux = block_diags(J_WE) * self.matJacWE \
+ block_diags(JD_dW) * self.matJacdW \
+ block_diags(JD_dWL) * self.matJacdWL \
+ block_diags(JD_dWR) * self.matJacdWR
return self.matJacDistFlux * J_flux
def prepareJacMatrices(self):
m = self.mesh
matL = accumarray(m.e[:,2], m.t.shape[0]).mat.T.tolil()
matR = accumarray(m.e[:,3], m.t.shape[0]).mat.T.tolil()
# WL on far bc is freestream
isFar = m.isFar(m.v[m.e[m.ieBnd,:2],:].mean(1))
matL[m.ieBnd[isFar],:] = 0
# average and difference
self.matJacWL = spkron(matL, eye(4)).tocsr()
self.matJacWR = spkron(matR, eye(4)).tocsr()
self.matJacWE = spkron(0.5 * (matL + matR), eye(4)).tocsr()
self.matJacdW = spkron(matR - matL, eye(4)).tocsr()
# distribute flux matrix
D = block_diags(1./ m.a[:,newaxis,newaxis])
self.matJacDistFlux = spkron(D * m.matDistFlux, eye(4)).tocsr()
# Tri to edg gradient with boundary copy
matTriToBnd = accumarray(m.e[m.ieBnd,2], m.t.shape[0]).mat.T
matGradTriEdg = m.matGradTriEdg + m.bcmatGradTriEdg * matTriToBnd
# left and right gradient in cells
edgOfTri = invertMap(m.e[:,2:])[1].reshape([-1, 3])
avgEdgToTri = (accumarray(edgOfTri[:,0], m.e.shape[0]).mat.T + \
accumarray(edgOfTri[:,1], m.e.shape[0]).mat.T + \
accumarray(edgOfTri[:,2], m.e.shape[0]).mat.T) / 3.
matGradTri = spkron(avgEdgToTri, eye(2)) * matGradTriEdg
xt = m.xt(); dxt = xt[m.e[:,3],:] - xt[m.e[:,2],:]
matL = spkron(accumarray(m.e[:,2], m.t.shape[0]).mat.T, eye(2))
matR = spkron(accumarray(m.e[:,3], m.t.shape[0]).mat.T, eye(2))
matJacdWL = block_diags(dxt[:,newaxis,:]) * matL * matGradTri
matJacdWR = block_diags(dxt[:,newaxis,:]) * matR * matGradTri
self.matJacdWL = spkron(matJacdWL, eye(4)).tocsr()
self.matJacdWR = spkron(matJacdWR, eye(4)).tocsr()
def metric(self, W=None):
if W is None: W = self.soln
gradV = self.mesh.gradTriVrt(W / self.Wref)
gradV /= ((gradV**2).sum(1)[:,newaxis,:])**.25
return (gradV[:,newaxis,:,:] * gradV[:,:,newaxis,:]).sum(-1)
# return hessian.sum(-1)
def checkJacobian(self):
'''
Generate a flow with small linear variation (supress numerical diss.),
check adjoint ddt vs Euler ddt
'''
xt = self.mesh.xt()
R = 0.2 * self.mesh.diameter()
decay = exp(-(xt**2).sum(1) / R**2)[:,newaxis]
nT = self.mesh.t.shape[0]
# random flow field
variation = 0.1 * (random.rand(nT, 4) - 0.5)
W0 = self.freeStream() + self.Wref * variation
# random perturbation
EPS = 0.000001
variation = EPS * (random.rand(nT, 4) - 0.5)
W1 = W0 + self.Wref * variation
# compare ddt with Jacobian
deltaFD = self.ddt(ravel(W1)) - self.ddt(ravel(W0))
deltaJac = self.J(0.5 * (W0 + W1)) * ravel(W1 - W0)
# difference
print sqrt(((deltaFD - deltaJac)**2).sum()), \
sqrt((deltaJac**2).sum()), sqrt((deltaFD**2).sum())
