def __init__(self, T, P, mezcla):
self.T = unidades.Temperature(T)
self.P = unidades.Pressure(P, "atm")
self.mezcla = mezcla
self.componente = mezcla.componente
self.fraccion = mezcla.fraccion
self.B = self.b * self.P.atm / R_atml / self.T
self.Tita = self.tita * self.P.atm / (R_atml * self.T)**2
delta = self.delta * self.P.atm / R_atml / self.T
epsilon = self.epsilon * (self.P.atm / R_atml / self.T)**2
eta = self.eta * self.P.atm / R_atml / self.T
Z = roots([
1, delta - self.B - 1, self.Tita + epsilon - delta * (self.B + 1),
-epsilon * (self.B + 1) - self.Tita * eta
])
self.Z = r_[Z[0].real, Z[2].real]
self.V = self.Z * R_atml * self.T / self.P.atm #mol/l
self.x, self.xi, self.yi, self.Ki = self._Flash()
self.H_exc = -(self.tita + self.dTitadT) / R_atml / self.T / (
self.delta**2 - 4 * self.epsilon)**0.5 * log(
(2 * self.V + self.delta -
(self.delta**2 - 4 * self.epsilon)**0.5) /
(2 * self.V + self.delta +
(self.delta**2 - 4 * self.epsilon)**0.5)) + 1 - self.Z
def dispersion_relation_extraordinary(kx, ky, k, nO, nE, c):
"""Dispersion relation for the extraordinary wave.
NOTE
See eq. 16 in Glytsis, "Three-dimensional (vector) rigorous
coupled-wave analysis of anisotropic grating diffraction",
JOSA A, 7(8), 1990 Always give positive real or negative
imaginary.
"""
if kx.shape != ky.shape or c.size != 3:
raise ValueError('kx and ky must have the same length and c must have 3 components')
kz = S.empty_like(kx)
for ii in xrange(0, kx.size):
alpha = nE**2 - nO**2
beta = kx[ii]/k * c[0] + ky[ii]/k * c[1]
# coeffs
C = S.array([nO**2 + c[2]**2 * alpha, \
2. * c[2] * beta * alpha, \
nO**2 * (kx[ii]**2 + ky[ii]**2) / k**2 + alpha * beta**2 - nO**2 * nE**2])
# two solutions of type +x or -x, purely real or purely imag
tmp_kz = k * S.roots(C)
# get the negative imaginary part or the positive real one
if S.any(S.isreal(tmp_kz)):
kz[ii] = S.absolute(tmp_kz[0])
else:
kz[ii] = -1j * S.absolute(tmp_kz[0])
return kz
def __init__(self, T, P, mezcla):
P_atm = P/101325
self.T = unidades.Temperature(T)
self.P = unidades.Pressure(P)
self.mezcla = mezcla
self.componente = mezcla.componente
self.zi = mezcla.fraccion
self.B = self.b*P/R/T
self.Tita = self.tita*P/(R*T)**2
delta = self.delta*P/R/T
epsilon = self.epsilon*(P/R/T)**2
# δ1, δ2 calculated from polynomial factorization
self.delta1 = ((self.delta**2-4*self.epsilon)**0.5-self.delta)/2/self.b
self.delta2 = ((self.delta**2-4*self.epsilon)**0.5+self.delta)/2/self.b
# Eq 4-6.3 in [1]_
coeff = [1, delta-self.B-1, self.Tita+epsilon-delta*(self.B+1),
-epsilon*(self.B+1)-self.Tita*self.B]
Z = roots(coeff)
# print("Z", Z)
# TODO: use the anallycal solution, Span, pag 50
# Set the minimum and maximum root values as liquid and gas Z values
self.Z = r_[Z[0].real, Z[2].real]
self.Zl = min(Z).real
self.Zg = max(Z).real
self.V = self.Z*R_atml*T/P_atm # l/mol
self.rho = 1/self.V
self.Vl = unidades.MolarVolume(self.Zl*R*T/P, "m3mol") # l/mol
self.Vg = unidades.MolarVolume(self.Zg*R*T/P, "m3mol") # l/mol
def npred(cev):
rts = sp.roots([p[3] / 4., p[2] / 3., p[1] / 2., p[0] + cev, -C])
nmax = max([sp.real(i) for i in rts if sp.isreal(i)])
cbar = p[0] + p[1] * nmax / 2. + p[2] * (nmax**2) / 3. + p[3] * (
nmax**3) / 4.
return nmax, cbar
def vanderwaals(phase, P, T, Pc, Tc, MW, **kwargs):
r"""
Uses Van der Waals equation of state to calculate the density of a real gas
Parameters
----------
P, T, Pc, Tc, MW: float, array_like
P pressure of the gas in [Pa]
T temperature of the gas in [K]
Pc critical pressure of the gas in [Pa]
T critical temperature of the gas in [K]
MW molecular weight of the gas in [kg/mol]
Returns
-------
rho, the density in [mol/m3]
"""
P = phase['pore.pressure']
T = phase['pore.temperature']
Pc = phase['pore.criticalpressure']
Tc = phase['pore.criticaltemperature']
R = 8.314
a = 27 * (R**2) * (Tc**2) / (64 * Pc)
b = R * Tc / (8 * Pc)
a0 = 1
a1 = -1 / b
a2 = (R * T + b * P) / (a * b)
a3 = -P / (a * b)
density = sp.roots([a0, a1, a2, a3])
value = sp.real(density[2])
return value
def _lib(self, cmp, T):
if cmp.id in dat:
# Use the compound specific parameters values
xic, f = dat[cmp.id]
else:
# Use the generalization correlations, Eq 20-21
f = 0.452413 + 1.30982 * cmp.f_acent - 0.295937 * cmp.f_acent**2
xic = 0.329032 - 0.076799 * cmp.f_acent + 0.0211947 * cmp.f_acent**2
# Eq 8
c = (1 - 3 * xic) * R * cmp.Tc / cmp.Pc
# Eq 10
b = roots([1, 2 - 3 * xic, 3 * xic**2, -xic**3])
Bpositivos = []
for i in b:
if i > 0:
Bpositivos.append(i)
Omegab = min(Bpositivos)
b = Omegab * R * cmp.Tc / cmp.Pc
# Eq 9
Omegaa = 3 * xic**2 + 3 * (1 -
2 * xic) * Omegab + Omegab**2 + 1 - 3 * xic
if cmp.id in PT2:
# Using improved alpha correlation from [2]_
c1, c2, c3, n = PT2[cmp.id]
alfa = 1 + c1*(T/cmp.Tc-1) + c2*((T/cmp.Tc)**0.5-1) + \
c3*((T/cmp.Tc)**n-1)
else:
alfa = (1 + f * (1 - (T / cmp.Tc)**0.5))**2
a = Omegaa * alfa * R**2 * cmp.Tc**2 / cmp.Pc
return a, b, c
def quadfit(lowerLimit, upperLimit, x,y): import scipy from scipy.optimize import leastsq y = y[lowerLimit:upperLimit] x = x[lowerLimit:upperLimit] maxY = max(y) p0 = array([-276025.,1051.,1.]) plsq = leastsq(quadresiduals, p0, args=(y, x), maxfev=2000) rplsq = [] i=1 while i <= len(plsq[0]): rplsq.append(plsq[0][-i]) i = i+1 diff = scipy.polyder(rplsq,1) roots = scipy.roots(diff) peak = 0 for root in roots: if root < x[-1] and root > x[0]: peak = root fit = [] for i in x: fit.append((i,quadfunction(i,plsq[0]))) return float(real(peak))
def polyint(lowerLimit, upperLimit, x, y): import scipy from scipy.optimize import leastsq y = y[lowerLimit:upperLimit] x = x[lowerLimit:upperLimit] maxY = max(y) p0 = array([-276025,1051,-1,1,100]) plsq = leastsq(polyresiduals, p0, args=(y, x), maxfev=2000) rplsq = [] i=1 while i <= len(plsq[0]): rplsq.append(plsq[0][-i]) i = i+1 diff = scipy.polyder(rplsq,1) roots = scipy.roots(diff) peak = 0 for root in roots: if root < x[-1] and root > x[0]: peak = root peakIntensity = polyfunction(peak,plsq[0]) return float(real(peakIntensity))
def test_roots(self):
polinomio = [1.3, 4, .6, -1] # polinomio = 1.3 x^3 + 4 x^2 + 0.6 x - 1
x = sp.arange(-4, 1, .05)
y = sp.polyval(polinomio, x)
raices = sp.roots(polinomio)
self.assertEqual(raices[0], -2.816023020827658)
self.assertEqual(raices[1], -0.6691329703979497)
self.assertEqual(raices[2], 0.4082329143025312)
def getOmegas(self, typeeos, OmegaB=False):
acc = self.acc
if typeeos in ['srk']:
OmegaC = 0.0
onethird = 1.0 / 3.0
OmegaB = onethird * (2**onethird - 1)
OmegaD = OmegaB + 0.0
OmegaA = onethird**2 / (2**onethird - 1)
elif typeeos == 'pr':
OmegaB = 0.077796
OmegaA = 0.457235
OmegaC = OmegaB + 0.0
OmegaD = OmegaB + 0.0
elif typeeos == 'pt':
Zc = 0.3272 - 0.0537 * acc - 0.0147 * acc**2 #Note: NOT experimental Zc!
A = 1.0
B = 2 - 3 * Zc
C = 3 * Zc**2
D = -Zc**3
rts = scipy.roots([A, B, C, D])
OmegaB = min([
r.real for r in rts if (r.real > 0.0) and (abs(r.imag) < 1e-9)
])
OmegaD = OmegaB + 0.0
OmegaC = 1 - 3 * Zc
OmegaA = 3 * Zc**2 + OmegaB * OmegaC + OmegaC * OmegaD + OmegaB * OmegaD + OmegaC + OmegaD
elif typeeos in ['hkm2']:
Zc = 0.3175 - 0.0364 * acc - 0.0245 * acc**2 #Note: NOT experimental Zc!
A = 1.0
B = 1.25 - 3 * Zc
C = 3 * Zc**2 - 1.5 * Zc + 0.5
D = -Zc**3
rts = scipy.roots([A, B, C, D])
OmegaB = min([
r.real for r in rts if (r.real > 0.0) and (abs(r.imag) < 1e-9)
])
OmegaE = 6 * Zc - 3 * OmegaB - 2
OmegaA = 3 * Zc**2 - 0.5 * OmegaB**2 - 0.75 * OmegaB * OmegaE - 0.5 * (
OmegaB + OmegaE)
n = -0.5
m = -0.5
nbplusme = n * OmegaB + m * OmegaE
nbme = n * OmegaB * m * OmegaE
OmegaD = 0.5 * (nbplusme - scipy.sqrt(nbplusme**2 + 4 * nbme))
OmegaC = -nbme / OmegaD
return OmegaA, OmegaB, OmegaC, OmegaD
def solve_composition(a, potentials, P=1., T=500., verbose=False):
"""
Calculate the equilibrium composition and free energy of a homonuclear gas mixture
Arguments
a: iterable of number of atoms in homonuclear molecule
e.g. [2,8] if only considering S2, S8
potentials: chemical potentials at reference pressure and given T of species studied, in J mol-1
P: Pressure in units of reference pressure
e.g. P=0.001 if study pressure is 1 mbar and reference pressure for potentials is 1 bar
T: Temperature in K. This temperature must also be used when generating input potentials.
verbose: Boolean flag. Print intermediate values.
Returns
n, mu_S:
n: Amount of component in gas mixture, in units of mol molecule / mole atoms.
To obtain units of atom%, n_atom_pc = [100 * a[i] * n_i for i, n_i in enumerate(n)]
mu_S: Chemical potential of gas mixture in J mol-1
"""
def vprint(*args):
if verbose:
for arg in args:
print arg,
else:
pass
from scipy.constants import R
a = np.array(a)
vprint("mu values: ", potentials, "J mol-1\n")
vprint("RT = {0} J mol-1\n".format(R*T))
vprint("mu / RT: ", np.array(potentials)/(R*T), '\n')
# Form empty polynomial
poly = [0]*(max(a)+1)
poly[0] = -P
for i, ai in enumerate(a):
poly[ai] += np.exp((-potentials[i])/(R*T))
poly.reverse()
vprint("terms of polynomial: " + str(poly) + '\n')
roots = scipy.roots(poly)
good_root = False
for root in roots:
if root == root.real and root >= 0 :
good_root = root
# Root = exp(lambda/RT)
lam = np.log(good_root.real) * R * T
vprint("Lagrange multiplier = free energy/atom = {0}\n".format(lam))
mu_S = lam
# Now to recover n values
vprint("P=", P, "T=", T, "a=",a, '\n')
phi = np.exp((a*lam - potentials)/(R*T))
vprint(phi)
b = 1. # 1 mole of S atoms basis
n = b * phi / (sum(a*phi))
return n, mu_S
def S_eq(self,T,P):
c0,c1,c2 = self.c_factors(T)
coeff = [c2, c1, c0-np.log(P)]
S = np.real(sp.roots(coeff))
S_temp = S[S>=0.0]
S_out = S_temp[S_temp<200]
if len(S_out)==0:
S_out = 0.0
return S_out/self.S_multiply[self.S_unit]
def R_of_l(V, l):
'''
radius of a spherocylinder with segment length l, such that
total volume is V.
'''
R = scipy.roots([(4.0 / 3.0) * np.pi, np.pi * l, 0.0, -V])
R_phys = np.real(R[np.logical_and(np.isreal(R), R > 0.0)])
if len(R_phys) != 1:
raise Exception('More or less than one physical radius found')
return R_phys[0]
def calculate_time_to_collision_single(self, other_particle):
"""
Find the time until this particle will intersect with other_particle
using linear extrapolation.
This is done by analytically solving the equation
(s_1 - s_2) ** 2 = (r_1 +- r_2) ** 2, which is the point where both
circles make contact.
The - in the right hand side of the equation is there for the case where
one of the circles is currently inside the other.
Arguments:
other_particle: another Particle object.
Returns: a float representing the time until these objects
intersect, or NaN if these particles are not on a
(future) collision trajectory.
"""
# Calculate all of the coefficients.
velocity_diff = self._velocity - other_particle.get_current_velocity()
position_diff = self._position - other_particle.get_current_position()
dt_sq_coefficient = sp.dot(velocity_diff, velocity_diff)
dt_coefficient = 2 * sp.dot(position_diff, velocity_diff)
constant_coefficient_lhs = sp.dot(position_diff, position_diff)
# (R1 +- R2) ** 2
if self.intersects(other_particle):
rhs = (self._radius - other_particle._radius)**2
else:
rhs = (self._radius + other_particle._radius)**2
constant_coefficient = constant_coefficient_lhs - rhs
discriminant = dt_coefficient ** 2 \
- (4 * dt_sq_coefficient * constant_coefficient)
if discriminant < 0:
return sp.nan
#sqrt_discriminant = sp.sqrt(discriminant)
roots = sp.roots(
[dt_sq_coefficient, dt_coefficient, constant_coefficient])
if sp.any(roots > Particle.tol):
# We're only interested in the first collision, that has a dt which
# is greater than zero, so return the smallest positive root:
return min(root for root in roots if root > Particle.tol)
return sp.nan
def vanderwaals(
phase,
pore_P="pore.pressure",
pore_T="pore.temperature",
pore_Pc="pore.critical_pressure",
pore_Tc="pore.critical_temperature",
**kwargs
):
r"""
Uses Van der Waals equation of state to calculate the density of a real gas
Parameters
----------
pore_P : string
The dictionary key containing the pressure values in Pascals (Pa)
pore_T : string
The dictionary key containing the temperature values in Kelvin (K)
pore_Pc : string
The dictionary key containing the critical pressure values in Pascals
(Pa)
pore_Tc : string
The dictionary key containing the critical temperature values in Kelvin
(K)
Returns
-------
rho, the density in [mol/m3]
Notes
-----
This equation and its constant coefficients are taken [1]_ which uses the
cgs units system. All input parameters are expected in SI, then converted
in the method.
"""
P = phase[pore_P] / 100000
T = phase[pore_T]
Pc = phase[pore_Pc] / 100000 # convert to bars
Tc = phase[pore_Tc]
R = 83.1447
a = 27 * (R ** 2) * (Tc ** 2) / (64 * Pc)
b = R * Tc / (8 * Pc)
a1 = -1 / b
a2 = (R * T + b * P) / (a * b)
a3 = -P / (a * b)
a0 = sp.ones(sp.shape(a1))
coeffs = sp.vstack((a0, a1, a2, a3)).T
density = sp.array([sp.roots(C) for C in coeffs])
value = sp.real(density[:, 2]) * 1e6 # Convert it to mol/m3
return value
def Z(self,t,p): # we pass p because in this case the pressure is the independent variable hence we don't use self.p
#Z3+AZ2+BZ+C=0
eps = self.typeeos.epsilon
sig = self.typeeos.sigma
pb = (p*self.b)/(self.r*t)
ap = (p*self.a*self.alpha(t))/((self.r*t)**2)
A = (eps+sig-1)*pb -1
B= sig*eps*(pb**2)-(sig+eps)*(pb**2)-(sig+eps)*pb+ap
C= -sig*eps*pb**3-sig*eps*pb**2-ap*pb
roots = scipy.roots([1.0,A,B,C]) # gives all roots, even complex ones
roots = [x.real for x in roots if abs(x.imag) < 1e-16] #.real and .imag are keywords coded in python
return roots
def complex_basins(func, dfunc, coeffs, interval, npoints):
"""Plot func over the complex interval and the
basins of attraction with npoints resolution"""
seeds = sp.zeros((2, npoints, npoints), dtype='complex128')
seeds[0:,:,] = sp.ones((npoints, 1))*sp.linspace(interval[0], interval[1], npoints)
seeds[0:,:,] = seeds[0:,:,] + 1j*seeds[0:,:,]
colors = pyplot.cm.colors.cnames.keys()
colors.remove('black')
clen = len(colors)
for i in xrange(npoints):
for j in xrange(npoints):
seeds[1,i,j] = newton(func, dfunc, seeds[0,i,j])
#return seeds
#find the unique roots of seed values
roots, indices = sp.unique(seeds[1])
#for each root, associate a color
col = sp.zeros_like(roots.shape)
for i in range(len(roots)):
col[i] = colors.pop(sp.random.randint(clen))
#seeds[1] = newtonv(func, )
#set the color for each
#tol = 1e-7
#valcol = [[]]
#for valx in range(npoints):
#for valy in range(npoints):
#if valx and valy == 0:
#valcol.append(colors[0])
#continue
#if not(abs(seeds[1,valx,valy]-seeds[1,valx,valy-1]) <= tol):
#valcol.append(colors.pop(sp.random.randint(1,len(colors))))
#else:
#valcol.append(valcol[0][0])
#for col in range(npoints):
#for col1 in range(npoints):
#pyplot.plot(seeds[0,col,col1], seeds[1,col,col1], marker=',', color=valcol[col][col1])
tol = 1e-07
roots = sp.roots(coeffs)
pyplot.hold(True)
#for i in range(len(roots)):
#pyplot.plot(seeds[abs(seeds[1:,:,]-roots[i])<tol], color=colors.pop(sp.random.randint(len(colors))), linestyle='None', marker='.')
pyplot.pcolor(seeds[0], seeds[1])
pyplot.hold(False)
pyplot.show()
def calculo(self):
entrada = self.kwargs["entrada"]
self.rendimientoCalculado = Dimensionless(self.kwargs["rendimiento"])
if self.kwargs["Pout"]:
DeltaP = Pressure(self.kwargs["Pout"] - entrada.P)
elif self.kwargs["deltaP"]:
DeltaP = Pressure(self.kwargs["deltaP"])
elif self.kwargs["Carga"]:
DeltaP = Pressure(self.kwargs["Carga"] * entrada.Liquido.rho * g)
else:
DeltaP = Pressure(0)
if self.kwargs["usarCurva"]:
b1 = self.kwargs["diametro"] != self.kwargs["curvaCaracteristica"][
0] # noqa
b2 = self.kwargs["velocidad"] != self.kwargs[
"curvaCaracteristica"][1] # noqa
if b1 or b2:
self.curvaActual = self.calcularCurvaActual()
else:
self.curvaActual = self.kwargs["curvaCaracteristica"]
self.Ajustar_Curvas_Caracteristicas()
if not self.kwargs["usarCurva"]:
head = Length(DeltaP / g / entrada.Liquido.rho)
power = Power(head * g * entrada.Liquido.rho * entrada.Q /
self.rendimientoCalculado)
P_freno = Power(power * self.rendimientoCalculado)
elif not self.kwargs["incognita"]:
head = Length(polyval(self.CurvaHQ, entrada.Q))
DeltaP = Pressure(head * g * entrada.Liquido.rho)
power = Power(entrada.Q * DeltaP)
P_freno = Power(polyval(self.CurvaPotQ, entrada.Q))
self.rendimientoCalculado = Dimensionless(power / P_freno)
else:
head = Length(self.DeltaP / g / entrada.Liquido.rho)
poli = [self.CurvaHQ[0], self.CurvaHQ[1], self.CurvaHQ[2] - head]
Q = roots(poli)[0]
power = Power(Q * self.DeltaP)
entrada = entrada.clone(split=Q / entrada.Q)
P_freno = Power(polyval(self.CurvaPotQ, Q))
self.rendimientoCalculado = Dimensionless(power / P_freno)
self.deltaP = DeltaP
self.headCalculada = head
self.power = power
self.P_freno = P_freno
self.salida = [entrada.clone(P=entrada.P + DeltaP)]
self.Pin = entrada.P
self.PoutCalculada = self.salida[0].P
self.volflow = entrada.Q
self.cp_cv = entrada.Liquido.cp_cv
def getOmegas(self, typeeos, OmegaB = False):
acc = self.acc
if typeeos in ['srk']:
OmegaC = 0.0
onethird = 1.0/3.0
OmegaB = onethird*(2**onethird-1)
OmegaD = OmegaB + 0.0
OmegaA = onethird**2/(2**onethird-1)
elif typeeos == 'pr':
OmegaB = 0.077796
OmegaA = 0.457235
OmegaC = OmegaB + 0.0
OmegaD = OmegaB + 0.0
elif typeeos == 'pt':
Zc = 0.3272 - 0.0537*acc - 0.0147*acc**2 #Note: NOT experimental Zc!
A = 1.0
B = 2 - 3*Zc
C = 3*Zc**2
D = -Zc**3
rts = scipy.roots([A, B, C, D])
OmegaB = min([r.real for r in rts if (r.real > 0.0) and (abs(r.imag) < 1e-9)])
OmegaD = OmegaB + 0.0
OmegaC = 1 - 3*Zc
OmegaA = 3*Zc**2 + OmegaB*OmegaC + OmegaC*OmegaD + OmegaB*OmegaD + OmegaC + OmegaD
elif typeeos in ['hkm2']:
Zc = 0.3175 - 0.0364*acc -0.0245*acc**2 #Note: NOT experimental Zc!
A = 1.0
B = 1.25 - 3*Zc
C = 3*Zc**2 - 1.5*Zc + 0.5
D = -Zc**3
rts = scipy.roots([A, B, C, D])
OmegaB = min([r.real for r in rts if (r.real > 0.0) and (abs(r.imag) < 1e-9)])
OmegaE = 6*Zc - 3*OmegaB - 2
OmegaA = 3*Zc**2 - 0.5*OmegaB**2 - 0.75*OmegaB*OmegaE - 0.5*(OmegaB+OmegaE)
n = -0.5; m = -0.5
nbplusme = n*OmegaB + m*OmegaE
nbme = n*OmegaB*m*OmegaE
OmegaD = 0.5*(nbplusme - scipy.sqrt(nbplusme**2+4*nbme))
OmegaC = -nbme/OmegaD
return OmegaA, OmegaB, OmegaC, OmegaD
def vanderwaals(phase,
pore_P='pore.pressure',
pore_T='pore.temperature',
pore_Pc='pore.critical_pressure',
pore_Tc='pore.critical_temperature',
**kwargs):
r"""
Uses Van der Waals equation of state to calculate the density of a real gas
Parameters
----------
pore_P : string
The dictionary key containing the pressure values in Pascals (Pa)
pore_T : string
The dictionary key containing the temperature values in Kelvin (K)
pore_Pc : string
The dictionary key containing the critical pressure values in Pascals
(Pa)
pore_Tc : string
The dictionary key containing the critical temperature values in Kelvin
(K)
Returns
-------
rho, the density in [mol/m3]
Notes
-----
This equation and its constant coefficients are taken [1]_ which uses the
cgs units system. All input parameters are expected in SI, then converted
in the method.
"""
P = phase[pore_P] / 100000
T = phase[pore_T]
Pc = phase[pore_Pc] / 100000 # convert to bars
Tc = phase[pore_Tc]
R = 83.1447
a = 27 * (R**2) * (Tc**2) / (64 * Pc)
b = R * Tc / (8 * Pc)
a1 = -1 / b
a2 = (R * T + b * P) / (a * b)
a3 = -P / (a * b)
a0 = sp.ones(sp.shape(a1))
coeffs = sp.vstack((a0, a1, a2, a3)).T
density = sp.array([sp.roots(C) for C in coeffs])
value = sp.real(density[:, 2]) * 1e6 # Convert it to mol/m3
return value
def complex_basins(func, dfunc, coeffs, interval, npoints):
"""Plot func over the complex interval and the
basins of attraction with npoints resolution"""
seeds = sp.zeros((2, npoints, npoints), dtype='complex128')
seeds[0:, :, ] = sp.ones(
(npoints, 1)) * sp.linspace(interval[0], interval[1], npoints)
seeds[0:, :, ] = seeds[0:, :, ] + 1j * seeds[0:, :, ]
colors = pyplot.cm.colors.cnames.keys()
colors.remove('black')
clen = len(colors)
for i in xrange(npoints):
for j in xrange(npoints):
seeds[1, i, j] = newton(func, dfunc, seeds[0, i, j])
#return seeds
#find the unique roots of seed values
roots, indices = sp.unique(seeds[1])
#for each root, associate a color
col = sp.zeros_like(roots.shape)
for i in range(len(roots)):
col[i] = colors.pop(sp.random.randint(clen))
#seeds[1] = newtonv(func, )
#set the color for each
#tol = 1e-7
#valcol = [[]]
#for valx in range(npoints):
#for valy in range(npoints):
#if valx and valy == 0:
#valcol.append(colors[0])
#continue
#if not(abs(seeds[1,valx,valy]-seeds[1,valx,valy-1]) <= tol):
#valcol.append(colors.pop(sp.random.randint(1,len(colors))))
#else:
#valcol.append(valcol[0][0])
#for col in range(npoints):
#for col1 in range(npoints):
#pyplot.plot(seeds[0,col,col1], seeds[1,col,col1], marker=',', color=valcol[col][col1])
tol = 1e-07
roots = sp.roots(coeffs)
pyplot.hold(True)
#for i in range(len(roots)):
#pyplot.plot(seeds[abs(seeds[1:,:,]-roots[i])<tol], color=colors.pop(sp.random.randint(len(colors))), linestyle='None', marker='.')
pyplot.pcolor(seeds[0], seeds[1])
pyplot.hold(False)
pyplot.show()
def vanderwaals(target,
pressure='pore.pressure',
temperature='pore.temperature',
critical_pressure='pore.critical_pressure',
critical_temperature='pore.critical_temperature'):
r"""
Uses Van der Waals equation of state to calculate the density of a real gas
Parameters
----------
target : OpenPNM Object
The object for which these values are being calculated. This
controls the length of the calculated array, and also provides
access to other necessary thermofluid properties.
pressure : string
The dictionary key containing the pressure values in Pascals (Pa)
temperature : string
The dictionary key containing the temperature values in Kelvin (K)
critical_pressure : string
The dictionary key containing the critical pressure values in Pascals
(Pa)
critical_temperature : string
The dictionary key containing the critical temperature values in Kelvin
(K)
Returns
-------
value : NumPy ndarray
Array containing molar density values [mol/m3]
"""
P = target[pressure] / 100000
T = target[temperature]
Pc = target[critical_pressure] / 100000 # convert to bars
Tc = target[critical_temperature]
R = 83.1447
a = 27 * (R**2) * (Tc**2) / (64 * Pc)
b = R * Tc / (8 * Pc)
a1 = -1 / b
a2 = (R * T + b * P) / (a * b)
a3 = -P / (a * b)
a0 = sp.ones(sp.shape(a1))
coeffs = sp.vstack((a0, a1, a2, a3)).T
density = sp.array([sp.roots(C) for C in coeffs])
value = sp.real(density[:, 2]) * 1e6 # Convert it to mol/m3
return value
def calculo(self):
entrada = self.kwargs["entrada"]
self.rendimientoCalculado = Dimensionless(self.kwargs["rendimiento"])
if self.kwargs["Pout"]:
DeltaP = Pressure(self.kwargs["Pout"]-entrada.P)
elif self.kwargs["deltaP"]:
DeltaP = Pressure(self.kwargs["deltaP"])
elif self.kwargs["Carga"]:
DeltaP = Pressure(self.kwargs["Carga"]*entrada.Liquido.rho*g)
else:
DeltaP = Pressure(0)
if self.kwargs["usarCurva"]:
b1 = self.kwargs["diametro"] != self.kwargs["curvaCaracteristica"][0] # noqa
b2 = self.kwargs["velocidad"] != self.kwargs["curvaCaracteristica"][1] # noqa
if b1 or b2:
self.curvaActual = self.calcularCurvaActual()
else:
self.curvaActual = self.kwargs["curvaCaracteristica"]
self.Ajustar_Curvas_Caracteristicas()
if not self.kwargs["usarCurva"]:
head = Length(DeltaP/g/entrada.Liquido.rho)
power = Power(head*g*entrada.Liquido.rho*entrada.Q /
self.rendimientoCalculado)
P_freno = Power(power*self.rendimientoCalculado)
elif not self.kwargs["incognita"]:
head = Length(polyval(self.CurvaHQ, entrada.Q))
DeltaP = Pressure(head*g*entrada.Liquido.rho)
power = Power(entrada.Q*DeltaP)
P_freno = Power(polyval(self.CurvaPotQ, entrada.Q))
self.rendimientoCalculado = Dimensionless(power/P_freno)
else:
head = Length(self.DeltaP/g/entrada.Liquido.rho)
poli = [self.CurvaHQ[0], self.CurvaHQ[1], self.CurvaHQ[2]-head]
Q = roots(poli)[0]
power = Power(Q*self.DeltaP)
entrada = entrada.clone(split=Q/entrada.Q)
P_freno = Power(polyval(self.CurvaPotQ, Q))
self.rendimientoCalculado = Dimensionless(power/P_freno)
self.deltaP = DeltaP
self.headCalculada = head
self.power = power
self.P_freno = P_freno
self.salida = [entrada.clone(P=entrada.P+DeltaP)]
self.Pin = entrada.P
self.PoutCalculada = self.salida[0].P
self.volflow = entrada.Q
self.cp_cv = entrada.Liquido.cp_cv
def getPborder(self,t): # this is used to get V where the dP/dv = 0. the input equation is the differentiated equation
e = self.typeeos.epsilon
s= self.typeeos.sigma
A= -self.r*t*(self.b*self.typeeos.epsilon-self.typeeos.sigma)
B = self.a*self.alpha(t)
b = self.b
C = -1
a5 =A*b**4*e**2*s**2 + B*b**4*e**2 + C*b**4*s**2
a1 = (A + B + C)
a2 = 2*A*b*e + 2*A*b*s + 2*B*b*e - 2*B*b + 2*C*b*s - 2*C*b
a3 = A*b**2*e**2 + 4*A*b**2*e*s + A*b**2*s**2 + B*b**2*e**2 - 4*B*b**2*e + B*b**2 + C*b**2*s**2 - 4*C*b**2*s + C*b**2
a4 = (2*A*b**3*e**2*s + 2*A*b**3*e*s**2 - 2*B*b**3*e**2 + 2*B*b**3*e - 2*C*b**3*s**2 + 2*C*b**3*s)
vsol = scipy.roots([a1,a2,a3,a4,a5])
return vsol
def LGR(self,figura):
"""
Função para traçado do Lugar Geométrico das Raízes
kvect: Vetor dos ganhos;
figura: referência a uma figura do Matplotlib.
O LGR é traçado sempre com ganho K = 1.
"""
num = self.polyDnum
den = self.polyDden
# Criando vetor de ganhos (sem os pontos críticos).
# Kmin, Kmax e numero de pontos são atributos desta classe.
delta_k = (self.Kmax-self.Kmin) / self.Kpontos
kvect = numpy.arange(self.Kmin,self.Kmax,delta_k)
# Geracao dos pontos de separacao
# Fazendo d(-1/G(s))/ds = 0
deriv = scipy.polyder(den)*num - scipy.polyder(num)*den
cpss = scipy.roots(deriv) # candidatos a ponto de separacao
# Verificacao de quais os candidatos pertinentes
for raiz in cpss:
aux = num(raiz)
if aux != 0:
Kc = -den(raiz) / num(raiz)
if (numpy.isreal(Kc)) and (Kc <= self.Kmax) and (Kc >= self.Kmin):
kvect = numpy.append(kvect,Kc)
# Reordena o kvect:
kvect = numpy.sort(kvect);
# Calculate the roots:
root_vector = MyRootLocus(num,den,kvect)
# Ploting:
figura.clf()
ax = figura.add_subplot(111)
# Open loop poles:
poles = numpy.array(den.r)
ax.plot(numpy.real(poles), numpy.imag(poles), 'x')
# Open loop zeros:
zeros = numpy.array(num.r)
if zeros.any():
ax.plot(numpy.real(zeros), numpy.imag(zeros), 'o')
for col in root_vector.T:
# Ploting the root locus.
ax.plot(numpy.real(col), numpy.imag(col), '-')
return root_vector
def LGR(self,figura):
"""
Função para traçado do Lugar Geométrico das Raízes
kvect: Vetor dos ganhos;
figura: referência a uma figura do Matplotlib.
O LGR é traçado sempre com ganho K = 1.
"""
num = self.polyDnum
den = self.polyDden
# Criando vetor de ganhos (sem os pontos críticos).
# Kmin, Kmax e numero de pontos são atributos desta classe.
delta_k = (self.Kmax-self.Kmin) / self.Kpontos
kvect = numpy.arange(self.Kmin,self.Kmax,delta_k)
# Geracao dos pontos de separacao
# Fazendo d(-1/G(s))/ds = 0
deriv = scipy.polyder(den)*num - scipy.polyder(num)*den
cpss = scipy.roots(deriv) # candidatos a ponto de separacao
# Verificacao de quais os candidatos pertinentes
for raiz in cpss:
aux = num(raiz)
if aux != 0:
Kc = -den(raiz) / num(raiz)
if (numpy.isreal(Kc)) and (Kc <= self.Kmax) and (Kc >= self.Kmin):
kvect = numpy.append(kvect,Kc)
# Reordena o kvect:
kvect = numpy.sort(kvect);
# Calculate the roots:
root_vector = MyRootLocus(num,den,kvect)
# Ploting:
figura.clf()
ax = figura.add_subplot(111)
# Open loop poles:
poles = numpy.array(den.r)
ax.plot(numpy.real(poles), numpy.imag(poles), 'x')
# Open loop zeros:
zeros = numpy.array(num.r)
if zeros.any():
ax.plot(numpy.real(zeros), numpy.imag(zeros), 'o')
for col in root_vector.T:
# Ploting the root locus.
ax.plot(numpy.real(col), numpy.imag(col), '-')
return root_vector
def vanderwaals(target, pressure='pore.pressure',
temperature='pore.temperature',
critical_pressure='pore.critical_pressure',
critical_temperature='pore.critical_temperature'):
r"""
Uses Van der Waals equation of state to calculate the density of a real gas
Parameters
----------
target : OpenPNM Object
The object for which these values are being calculated. This
controls the length of the calculated array, and also provides
access to other necessary thermofluid properties.
pressure : string
The dictionary key containing the pressure values in Pascals (Pa)
temperature : string
The dictionary key containing the temperature values in Kelvin (K)
critical_pressure : string
The dictionary key containing the critical pressure values in Pascals
(Pa)
critical_temperature : string
The dictionary key containing the critical temperature values in Kelvin
(K)
Returns
-------
value : NumPy ndarray
Array containing molar density values [mol/m3]
"""
P = target[pressure]/100000
T = target[temperature]
Pc = target[critical_pressure]/100000 # convert to bars
Tc = target[critical_temperature]
R = 83.1447
a = 27*(R**2)*(Tc**2)/(64*Pc)
b = R*Tc/(8*Pc)
a1 = -1/b
a2 = (R*T+b*P)/(a*b)
a3 = -P/(a*b)
a0 = sp.ones(sp.shape(a1))
coeffs = sp.vstack((a0, a1, a2, a3)).T
density = sp.array([sp.roots(C) for C in coeffs])
value = sp.real(density[:, 2])*1e6 # Convert it to mol/m3
return value
def low_accuracy_recompute_alpha_varying_fields(
sph_start, sph_end, t_start, t_end, mag_params):
"""
Compute effective damping from change in magnetisation and change in
applied field.
From Nonlinear magnetization dynamics in nanosystems eqn (2.15).
See notes 30/7/13.
Derivatives are estimated using BDF1 finite differences.
"""
# Only for normalised problems!
assert(mag_params.Ms == 1)
# Get some values
dt = t_end - t_start
m_cart_end = utils.sph2cart(sph_end)
h_eff_end = heff(mag_params, t_end, m_cart_end)
mxh = sp.cross(m_cart_end, h_eff_end)
# Finite difference derivatives
dhadt = (mag_params.Hvec(t_start) - mag_params.Hvec(t_end))/dt
assert(all(dhadt == 0)) # no field for now
dedt = (llg_state_energy(sph_end, mag_params, t_end)
- llg_state_energy(sph_start, mag_params, t_start)
)/dt
sigma = sp.dot(mxh, mxh) / (dedt + sp.dot(m_cart_end, dhadt))
possible_alphas = sp.roots([1, sigma, 1])
a = (-sigma + sqrt(sigma**2 - 4))/2
b = (-sigma - sqrt(sigma**2 - 4))/2
possible_alphas2 = [a, b]
utils.assert_list_almost_equal(possible_alphas, possible_alphas2)
print(sigma, possible_alphas)
def real_and_positive(x):
return sp.isreal(x) and x > 0
alphas = filter(real_and_positive, possible_alphas)
assert(len(alphas) == 1)
return sp.real(alphas[0])
def getvborder(self, T, alpha = False):
if not alpha:
alpha = self.alpha(T)
RT = self.R*T
a, b, c, d = self.a*alpha, self.b, self.c, self.d
a4 = -1.0
a3 = -2*(c+d) + 2*a/RT
a2 = -c**2 - d**2 - 4*a*b/RT + a*c/RT + a*d/RT
a1 = 2*c**2*d + 2*c*d**2 + 2*a*b**2/RT - 2*a*b*c/RT - 2*a*b*d/RT
a0 = -c**2*d**2 + a*b**2*c/RT + a*b**2*d/RT
roots = scipy.roots([a4.real, a3.real, a2.real, a1.real, a0.real])
roots = [x.real for x in roots if abs(x.imag) < 1e-16 and x.real > b]
if len(roots) == 2:
vborder = [roots[0], roots[1]]
return True, min(vborder), max(vborder)
else:
return False, None, None
def Z(self, T, P, alpha = False):
P = 1e-20 if P == 0.0 else P
if not alpha:
alpha = self.alpha(T)
PRT = P/(self.R*T)
a, b, c, d = self.a*alpha, self.b, self.c, self.d
A = PRT*(c + d - b) - 1
B = PRT**2*(a/P - b*c - c*d - d*b) - PRT*(c+d)
C = PRT**3*(b*c*d+c*d/PRT-a*b/P)
roots = scipy.roots([1, A.real, B.real, C.real])
roots = [x.real for x in roots if abs(x.imag) < 1e-16]
if len(roots) == 3:
return [min(roots), max(roots)]
elif len(roots) == 1:
return [roots[0]]
else:
return []
def Z(self, T, P, alpha=False):
P = 1e-20 if P == 0.0 else P
if not alpha:
alpha = self.alpha(T)
PRT = P / (self.R * T)
a, b, c, d = self.a * alpha, self.b, self.c, self.d
A = PRT * (c + d - b) - 1
B = PRT**2 * (a / P - b * c - c * d - d * b) - PRT * (c + d)
C = PRT**3 * (b * c * d + c * d / PRT - a * b / P)
roots = scipy.roots([1, A.real, B.real, C.real])
roots = [x.real for x in roots if abs(x.imag) < 1e-16]
if len(roots) == 3:
return [min(roots), max(roots)]
elif len(roots) == 1:
return [roots[0]]
else:
return []
def getvborder(self, T, alpha=False):
if not alpha:
alpha = self.alpha(T)
RT = self.R * T
a, b, c, d = self.a * alpha, self.b, self.c, self.d
a4 = -1.0
a3 = -2 * (c + d) + 2 * a / RT
a2 = -c**2 - d**2 - 4 * a * b / RT + a * c / RT + a * d / RT
a1 = 2 * c**2 * d + 2 * c * d**2 + 2 * a * b**2 / RT - 2 * a * b * c / RT - 2 * a * b * d / RT
a0 = -c**2 * d**2 + a * b**2 * c / RT + a * b**2 * d / RT
roots = scipy.roots([a4.real, a3.real, a2.real, a1.real, a0.real])
roots = [x.real for x in roots if abs(x.imag) < 1e-16 and x.real > b]
if len(roots) == 2:
vborder = [roots[0], roots[1]]
return True, min(vborder), max(vborder)
else:
return False, None, None
def __lib(self, cmp, T):
Tr = T / cmp.Tc
if cmp.id in dat:
Xc, m, p, d = dat[cmp.id]
else:
Zc = cmp.Pc.kPa * cmp.Vc * cmp.M / R / cmp.Tc
d = cmp.Vc.ccg * cmp.M / 3 # Eq 11
Xc = 1.063 * Zc # Eq 12
# Eq 13
m = 0.662 + 3.12 * cmp.f_acent - 0.854 * cmp.f_acent**2 + 9.3 * (
Zc - 0.3)
if cmp.M <= 128:
p = 0.475 + 2 * cmp.f_acent # Eq 14a
else:
p = 0.613 + 0.62 * cmp.f_acent + 4.06 * cmp.f_acent**2 # Eq 14b
alfa = 1 + m * (1 - Tr**0.5) + p * (0.7**0.5 - Tr**0.5) * (1 - Tr**0.5)
# From here common functionality of original TB EoS
# Convert d from cc/mol to m3/mol
d /= 1e6
Cc = 1 - 3 * Xc # Eq 6
Dc = d * cmp.Pc / R / cmp.Tc # Eq 12
# Bc calculated ad the smallest positive root of Eq 8
B = roots([1, 2 - 3 * Xc, 3 * Xc**2, -Dc**2 - Xc**3])
Bpositivos = []
for Bi in B:
if Bi > 0:
Bpositivos.append(Bi)
Bc = min(Bpositivos)
Ac = 3 * Xc**2 + 2 * Bc * Cc + Bc + Cc + Bc**2 + Dc**2 # Eq 7
ac = Ac * R**2 * cmp.Tc**2 / cmp.Pc
a = ac * alfa # Eq 23
b = Bc * R * cmp.Tc / cmp.Pc
c = R * cmp.Tc / cmp.Pc * (1 - 3 * Xc) # Eq 10
return a, b, c, d
def tmin(self, r, z, trace=False):
"""Calculates and returns the s value along the norm vector
which minimizes the distance from the ray to a circle defined by
input (r,z).
Args:
r: value, iterable or scipy.array, radius in meters
z: value, iterable or scipy.array, z value in meters
Kwargs:
trace: bool if set true, the ray is assumed to be traced within a
tokamak. A further evaluation reduces the value to one within
the bounds of the vacuum vessel/limiter.
Returns:
numpy array of s values in meters
"""
r = scipy.atleast_1d(r)
z = scipy.atleast_1d(z)
params = _beam.lineCirc(self(0).x(),
self.norm.unit,
r,
z)
sout = scipy.zeros(r.shape)
for i in xrange(len(params)):
temp = scipy.roots(params[i])
# only positive real solutions are taken
temp = temp[scipy.imag(temp) == 0]
temp = scipy.real(temp[temp > 0])
test = self(temp).r()
# must decide between local and global minima
sout[i] = temp[((test[0]-r[i])**2
+ (test[2] - z[i])**2).argmin()]
if trace and ((sout[i] > self.norm.s[-1]) or (sout[i] < self.norm.s[-2])):
#need to implement this such that it searches only in area of interest
sout[i] = None
return sout
def tmin(self, r, z, trace=False):
"""Calculates and returns the s value along the norm vector
which minimizes the distance from the ray to a circle defined by
input (r,z).
Args:
r: value, iterable or scipy.array, radius in meters
z: value, iterable or scipy.array, z value in meters
Kwargs:
trace: bool if set true, the ray is assumed to be traced within a
tokamak. A further evaluation reduces the value to one within
the bounds of the vacuum vessel/limiter.
Returns:
numpy array of s values in meters
"""
r = scipy.atleast_1d(r)
z = scipy.atleast_1d(z)
params = _beam.lineCirc(self(0).x(), self.norm.unit, r, z)
sout = scipy.zeros(r.shape)
for i in range(len(params)):
temp = scipy.roots(params[i])
# only positive real solutions are taken
temp = temp[scipy.imag(temp) == 0]
temp = scipy.real(temp[temp > 0])
test = self(temp).r()
# must decide between local and global minima
sout[i] = temp[((test[0] - r[i])**2 +
(test[2] - z[i])**2).argmin()]
if trace and ((sout[i] > self.norm.s[-1]) or
(sout[i] < self.norm.s[-2])):
#need to implement this such that it searches only in area of interest
sout[i] = None
return sout
def Z(self, T, P):
R = self.R; a = self.a; b = self.b
sigma = self.typeEOS.sigma; epsilon = self.typeEOS.epsilon
alpha = self.alpha(T)
p_p = P*b/R/T; a_p = P*a*alpha/R**2/T**2
a3 = 1.0
a2 = epsilon*p_p + p_p*sigma - p_p - 1.0
a1 = a_p + epsilon*p_p**2*sigma - epsilon*p_p**2 - epsilon*p_p - p_p**2*sigma - p_p*sigma
a0 = -a_p*p_p - epsilon*p_p**3*sigma - epsilon*p_p**2*sigma
roots = scipy.roots([a3, a2, a1, a0])
roots = [rt.real for rt in roots if abs(rt.imag) < 1e-16]
if len(roots) == 3:
return scipy.array([min(roots), max(roots)])
elif len(roots) == 1:
return scipy.array(roots)
else:
return []
def getPborder(self, T):
a = self.a; b = self.b
alpha = self.alpha(T)
R = self.R; s = self.typeEOS.sigma; e = self.typeEOS.epsilon
A = a*alpha/(s-e)/b/R/T
a4 = 1.0
a3 = 2*A*b*e - 2*A*b*s + 2*b*e + 2*b*s
a2 = A*b**2*e**2 - 4*A*b**2*e - A*b**2*s**2 + 4*A*b**2*s + b**2*e**2 + 4*b**2*e*s + b**2*s**2
a1 = -2*A*b**3*e**2 + 2*A*b**3*e + 2*A*b**3*s**2 - 2*A*b**3*s + 2*b**3*e**2*s + 2*b**3*e*s**2
a0 = A*b**4*e**2 - A*b**4*s**2 + b**4*e**2*s**2
roots = scipy.roots([a4, a3, a2, a1, a0])
roots = [x.real for x in roots if abs(x.imag) < 1e-16 and x.real > b]
if len(roots) == 2:
PP = [self.P(T, roots[0]), self.P(T, roots[1])]
return True, min(PP), max(PP)
else:
return False, None, None
def dispersion_relation_extraordinary(kx, ky, k, nO, nE, c):
"""Dispersion relation for the extraordinary wave.
NOTE
See eq. 16 in Glytsis, "Three-dimensional (vector) rigorous
coupled-wave analysis of anisotropic grating diffraction",
JOSA A, 7(8), 1990 Always give positive real or negative
imaginary.
"""
if kx.shape != ky.shape or c.size != 3:
raise ValueError(
"kx and ky must have the same length and c must have 3 components"
)
kz = S.empty_like(kx)
for ii in range(0, kx.size):
alpha = nE ** 2 - nO ** 2
beta = kx[ii] / k * c[0] + ky[ii] / k * c[1]
# coeffs
C = S.array(
[
nO ** 2 + c[2] ** 2 * alpha,
2.0 * c[2] * beta * alpha,
nO ** 2 * (kx[ii] ** 2 + ky[ii] ** 2) / k ** 2
+ alpha * beta ** 2
- nO ** 2 * nE ** 2,
]
)
# two solutions of type +x or -x, purely real or purely imag
tmp_kz = k * S.roots(C)
# get the negative imaginary part or the positive real one
if S.any(S.isreal(tmp_kz)):
kz[ii] = S.absolute(tmp_kz[0])
else:
kz[ii] = -1j * S.absolute(tmp_kz[0])
return kz
def __init__(self, T, P, mezcla):
self.T=unidades.Temperature(T)
self.P=unidades.Pressure(P, "atm")
self.mezcla=mezcla
self.componente=mezcla.componente
self.fraccion=mezcla.fraccion
self.B=self.b*self.P.atm/R_atml/self.T
self.Tita=self.tita*self.P.atm/(R_atml*self.T)**2
delta=self.delta*self.P.atm/R_atml/self.T
epsilon=self.epsilon*(self.P.atm/R_atml/self.T)**2
eta=self.eta*self.P.atm/R_atml/self.T
Z=roots([1, delta-self.B-1, self.Tita+epsilon-delta*(self.B+1), -epsilon*(self.B+1)-self.Tita*eta])
self.Z=r_[Z[0].real, Z[2].real]
self.V=self.Z*R_atml*self.T/self.P.atm #mol/l
self.x, self.xi, self.yi, self.Ki=self._Flash()
self.H_exc=-(self.tita+self.dTitadT)/R_atml/self.T/(self.delta**2-4*self.epsilon)**0.5*log((2*self.V+self.delta-(self.delta**2-4*self.epsilon)**0.5)/(2*self.V+self.delta+(self.delta**2-4*self.epsilon)**0.5))+1-self.Z
def PTC_lib(compuesto, T):
"""Librería de cálculo de la ecuación de estado de Patel-Teja-Crause"""
if compuesto.Tc!=0 and compuesto.Pc!=0 and compuesto.vc!=0:
Zc=compuesto.Pc.atm*compuesto.vc*compuesto.peso_molecular/R_atml/compuesto.Tc
else:
Zc=0.253168556+0.09253329*exp(-0.0048018*compuesto.peso_molecular)
c=(1-3*Zc)*R_atml*compuesto.Tc/compuesto.Pc.atm
omega=roots([1, 2-3*Zc, 3*Zc**2, -Zc**3])
Bpositivos=[]
for i in omega:
if i>0:
Bpositivos.append(i)
omegab=min(Bpositivos)
b=omegab*R_atml*compuesto.Tc/compuesto.Pc.atm
f=0.002519*compuesto.peso_molecular+0.70647
alfa=(1+f*(1-sqrt(T/compuesto.Tc)))**2
omegaa=3*Zc**2+3*(1-2*Zc)*omegab+omegab**2+1-3*Zc
a=omegaa*R_atml**2*compuesto.Tc**2/compuesto.Pc.atm**2*alfa
return a, b, c
def PTC_lib(compuesto, T):
"""Librería de cálculo de la ecuación de estado de Patel-Teja-Crause"""
if compuesto.Tc!=0 and compuesto.Pc!=0 and compuesto.vc!=0:
Zc=compuesto.Pc.atm*compuesto.vc*compuesto.peso_molecular/R_atml/compuesto.Tc
else:
Zc=0.253168556+0.09253329*exp(-0.0048018*compuesto.peso_molecular)
c=(1-3*Zc)*R_atml*compuesto.Tc/compuesto.Pc.atm
omega=roots([1, 2-3*Zc, 3*Zc**2, -Zc**3])
Bpositivos=[]
for i in omega:
if i>0:
Bpositivos.append(i)
omegab=min(Bpositivos)
b=omegab*R_atml*compuesto.Tc/compuesto.Pc.atm
f=0.002519*compuesto.peso_molecular+0.70647
alfa=(1+f*(1-sqrt(T/compuesto.Tc)))**2
omegaa=3*Zc**2+3*(1-2*Zc)*omegab+omegab**2+1-3*Zc
a=omegaa*R_atml**2*compuesto.Tc**2/compuesto.Pc.atm**2*alfa
return a, b, c
def PT_lib(compuesto, T):
"""Librería de cálculo de la ecuación de estado de Patel-Teja"""
if compuesto.Tc != 0 and compuesto.Pc != 0 and compuesto.vc != 0:
Zc = compuesto.Pc.atm * compuesto.vc * compuesto.peso_molecular / R_atml / compuesto.Tc
else:
Zc = 0.329032 + 0.076799 * compuesto.f_acent - 0.0211947 * compuesto.f_acent**2
c = (1 - 3 * Zc) * R_atml * compuesto.Tc / compuesto.Pc.atm
omega = roots([1, 2 - 3 * Zc, 3 * Zc**2, -Zc**3])
Bpositivos = []
for i in omega:
if i > 0:
Bpositivos.append(i)
omegab = min(Bpositivos)
b = omegab * R_atml * compuesto.Tc / compuesto.Pc.atm
f = 0.452413 + 1.30982 * compuesto.f_acent - 0.295937 * compuesto.f_acent**2
alfa = (1 + f * (1 - sqrt(T / compuesto.Tc)))**2
omegaa = 3 * Zc**2 + 3 * (1 - 2 * Zc) * omegab + omegab**2 + 1 - 3 * Zc
a = omegaa * R_atml**2 * compuesto.Tc**2 / compuesto.Pc.atm**2 * alfa
return a, b, c
def PT_lib(compuesto, T):
"""Librería de cálculo de la ecuación de estado de Patel-Teja"""
if compuesto.Tc!=0 and compuesto.Pc!=0 and compuesto.vc!=0:
Zc=compuesto.Pc.atm*compuesto.vc*compuesto.peso_molecular/R_atml/compuesto.Tc
else:
Zc=0.329032+0.076799*compuesto.f_acent-0.0211947*compuesto.f_acent**2
c=(1-3*Zc)*R_atml*compuesto.Tc/compuesto.Pc.atm
omega=roots([1, 2-3*Zc, 3*Zc**2, -Zc**3])
Bpositivos=[]
for i in omega:
if i>0:
Bpositivos.append(i)
omegab=min(Bpositivos)
b=omegab*R_atml*compuesto.Tc/compuesto.Pc.atm
f=0.452413+1.30982*compuesto.f_acent-0.295937*compuesto.f_acent**2
alfa=(1+f*(1-sqrt(T/compuesto.Tc)))**2
omegaa=3*Zc**2+3*(1-2*Zc)*omegab+omegab**2+1-3*Zc
a=omegaa*R_atml**2*compuesto.Tc**2/compuesto.Pc.atm**2*alfa
return a, b, c
def AproxBessel(Retardo, Wrg, Tol, N=0, Nmin=0, Nmax=15):
s = sympy.Symbol('s')
if N == 0:
#Trato de obtener la funcion de bessel que cumpla con lo pedido
N = 1
#Si tengo un valor de Nmin voy a empezar desde ahi
if Nmin != 0:
N = Nmin
while (N < Nmax) and (Nmin <= N):
Polinomio = PoliBessel(N)
Tranferencia = Polinomio.subs(s, 0) / Polinomio.subs(s, 1j * Wrg)
Temp = ((Wrg**(2 * N)) *
(abs(Tranferencia)**2)) / (Polinomio.subs(s, 0)**2)
if (Temp < Tol):
break
else:
N += 1
#Chequeo para ver si estoy en el Nmax
if N == Nmax:
Polinomio = PoliBessel(N)
print("Se alcanzo el limite")
elif N == Nmin:
print("Se alconzo el minimo")
else:
Polinomio = PoliBessel(N)
#obtengo la funcion tranferencia
Tranferencia = Polinomio.subs(s, 0) / Polinomio
#Busco los polos
Num, Den = sympy.fraction(Tranferencia)
Den_Coef = sympy.Poly(Den)
Den_Coef = Den_Coef.all_coeffs()
Polos = scipy.roots(Den_Coef)
for i in range(len(Polos)):
Polos[i] = Polos[i]
#Devuelvo el orden, los polos y la constante
Num = 1
return N, Polos, Num
def PontosSeparacao(self):
"""
Calcula os pontos de separação e retorna só os pertinentes.
"""
num = self.polyDnum
den = self.polyDden
pontos = []
ganhos = []
# Geracao dos pontos de separacao
# Fazendo d(-1/G(s))/ds = 0
deriv = scipy.polyder(den)*num - scipy.polyder(num)*den
cpss = scipy.roots(deriv) # candidatos a ponto de separacao
# Verificacao de quais os candidatos pertinentes
for raiz in cpss:
aux = num(raiz)
if aux != 0:
Kc = -den(raiz) / num(raiz)
if (numpy.isreal(Kc)):# and (Kc <= self.Kmax) and (Kc >= self.Kmin):
pontos.append(raiz)
ganhos.append(Kc)
return pontos, ganhos
def _BOpntsLoc(sys):
"""Breakout point locater
This returns the location of breakout points on the root locus and
the associated gains at those locations.
Parameters
----------
sys: StateSpace or TransferFunction
Linear system
Returns
-------
KatBO:
list of gains at which a breakout point occurs
BOpnts:
list of breakout points
"""
(nump, denp) = _systopoly1d(sys)
# Breakout points potentially occur where N*D' - N'*D = 0
numpD = polyder(nump)
denpD = polyder(denp)
BOpnts = roots((nump*denpD) - (numpD*denp))
# Selects only the Breakout points on the real line
BOpnts = BOpnts[imag(BOpnts) == 0]
# Finds gains from breakout points
fgain = vectorize(lambda s: polyval(-denp,s)/polyval(nump,s))
KatBO = fgain(BOpnts)
# Selects only the positive gains at the Breakout points
KatBO = KatBO[KatBO >= 0]
return KatBO, BOpnts
def PontosSeparacao(self):
"""
Calcula os pontos de separação e retorna só os pertinentes.
"""
num = self.polyDnum
den = self.polyDden
pontos = []
ganhos = []
# Geracao dos pontos de separacao
# Fazendo d(-1/G(s))/ds = 0
deriv = scipy.polyder(den)*num - scipy.polyder(num)*den
cpss = scipy.roots(deriv) # candidatos a ponto de separacao
# Verificacao de quais os candidatos pertinentes
for raiz in cpss:
aux = num(raiz)
if aux != 0:
Kc = -den(raiz) / num(raiz)
if (numpy.isreal(Kc)):# and (Kc <= self.Kmax) and (Kc >= self.Kmin):
pontos.append(raiz)
ganhos.append(Kc)
return pontos, ganhos
gss1 = ss(A,B,C,D) gss = ss(A,B,C2,D2) gz = c2d(gss,ts,'zoh') # Control design wn = 10 xi1 = np.sqrt(2)/2 xi2 = np.sqrt(3)/2 xi2 = 0.85 cl_p1 = [1,2*xi1*wn,wn**2] cl_p2 = [1,2*xi2*wn,wn**2] cl_p3 = [1,wn] cl_poly1 = sp.polymul(cl_p1,cl_p2) cl_poly = sp.polymul(cl_poly1,cl_p3) cl_poles = sp.roots(cl_poly) # Desired continous poles cl_polesd = sp.exp(cl_poles*ts) # Desired discrete poles # Add discrete integrator for steady state zero error Phi_f = np.vstack((gz.A,-gz.C*ts)) Phi_f = np.hstack((Phi_f,[[0],[0],[0],[0],[1]])) G_f = np.vstack((gz.B,zeros((1,1)))) # Pole placement k = place(Phi_f,G_f,cl_polesd) # Observer design - reduced order observer poli_of = 5*sp.roots(cl_poly1) # Desired continous poles poli_o = 5*cl_poles[0:2] poli_oz = sp.exp(poli_o*ts) poli_ozf = sp.exp(poli_of*ts)
def real_roots(expr):
import scipy as sc
r = sc.roots(coeffs(expr))
return np.real( r[np.imag(r)==0] )
def roots(expr):
import scipy as sc
return sc.roots(coeffs(expr))
def lars_regression_noise_old(Yp, X, positive, noise, verbose=False):
"""
Run LARS for regression problems with LASSO penalty, with optional positivity constraints
Author: Andrea Giovannucci. Adapted code from Eftychios Pnevmatikakis
Parameters:
-------
Yp: Yp[:,t] is the observed data at time t
X: the regresion problem is Yp=X*W + noise
maxcomps: maximum number of active components to allow
positive: a flag to enforce positivity
noise: the noise of the observation equation. if it is not
provided as an argument, the noise is computed from the
variance at the end point of the algorithm. The noise is
used in the computation of the Cp criterion.
Returns:
-------
Ws: weights from each iteration
lambdas: lambda_ values at each iteration
Cps: C_p estimates
last_break: last_break(m) == n means that the last break with m non-zero weights is at Ws(:,:,n)
See Also:
-------
LARS : https://en.wikipedia.org/wiki/Least-angle_regression
group Lasso :
"""
# INITAILIZATION
T = len(Yp) # of time steps
k = 1
Yp = np.squeeze(np.asarray(Yp))
# necessary for matrix multiplications
Yp = np.expand_dims(Yp, axis=1)
_, T = np.shape(Yp) # of time steps
_, N = np.shape(X) # of compartments
# 1 : Start with all Weights equal to zero.
maxcomps = N
W = np.zeros((N, k))
active_set = np.zeros((N, k))
visited_set = np.zeros((N, k))
lambdas = []
# Just preallocation. Ws may end with more or less than maxcomp columns
Ws = []
r = np.expand_dims(np.dot(X.T, Yp.flatten()), axis=1) # N-dim vector
M = np.dot(-X.T, X) # N x N matrix
# %% begin main loop
i = 0
flag = 0
while 1:
if flag == 1:
W_lam = 0
break
# % calculate new gradient component if necessary
if i > 0 and new >= 0 and visited_set[new] == 0: # AG NOT CLEAR HERE
visited_set[new] = 1 # % remember this direction was computed
# % Compute full gradient of Q
dQ = r + np.dot(M, W)
# % Compute new W
if i == 0:
if positive:
dQa = dQ
else:
dQa = np.abs(dQ)
lambda_, new = np.max(dQa), np.argmax(dQa)
if lambda_ < 0:
print('All negative directions!')
break
else: # 2 : Find the predictor x_{j} most correlated with y
# % calculate vector to travel along
avec, gamma_plus, gamma_minus = calcAvec(
new, dQ, W, lambda_, active_set, M, positive)
# % calculate time of travel and next new direction
if new == -1: # % if we just dropped a direction we don't allow it to emerge
if dropped_sign == 1: # % with the same sign
gamma_plus[dropped] = np.inf
else:
gamma_minus[dropped] = np.inf
# 3 : Increase the coefficient W in the direction of the sign of its correlation with y
# % don't consider active components
gamma_plus[active_set == 1] = np.inf
# % or components outside the range [0, lambda_]
gamma_plus[gamma_plus <= 0] = np.inf
gamma_plus[gamma_plus > lambda_] = np.inf
gp_min, gp_min_ind = np.min(gamma_plus), np.argmin(gamma_plus)
if positive:
gm_min = np.inf # % don't consider new directions that would grow negative
else:
gamma_minus[active_set == 1] = np.inf
gamma_minus[gamma_minus > lambda_] = np.inf
gamma_minus[gamma_minus <= 0] = np.inf
gm_min, gm_min_ind = np.min(
gamma_minus), np.argmin(gamma_minus)
[g_min, which] = np.min(gp_min), np.argmin(gp_min)
# % if there are no possible new components, try move to the end
if g_min == np.inf:
g_min = lambda_
# % This happens when all the components are already active or, if positive==1,
# when there are no new positive directions
# % LARS check (is g_min*avec too large?)
gamma_zero = old_div(-W[active_set == 1], np.squeeze(avec))
gamma_zero_full = np.zeros((N, k))
gamma_zero_full[active_set == 1] = gamma_zero
gamma_zero_full[gamma_zero_full <= 0] = np.inf
gz_min, gz_min_ind = np.min(
gamma_zero_full), np.argmin(gamma_zero_full)
# 4: Increase Wj,Wk in their joint least squares direction, until some other predictor x_{m}
# has as much correlation with the residual r. (see 5)
if gz_min < g_min:
if verbose:
print(('DROPPING active weight:' + str(gz_min_ind)))
active_set[gz_min_ind] = 0
dropped = gz_min_ind
dropped_sign = np.sign(W[dropped])
W[gz_min_ind] = 0
avec = avec[gamma_zero != gz_min]
g_min = gz_min
new = -1
elif g_min < lambda_:
if which == 0:
new = gp_min_ind
if verbose:
print(('new positive component:' + str(new)))
else:
new = gm_min_ind
print(('new negative component:' + str(new)))
W[active_set == 1] = W[active_set == 1] + \
np.dot(g_min, np.squeeze(avec))
if positive:
if any(W < 0):
flag = 1
lambda_ = lambda_ - g_min
# % Update weights and lambdas
lambdas.append(lambda_)
Ws.append(W.copy())
# 5 : Take residuals r=y-y_ along the way. Stop when some other predictor x_{k}
# has as much correlation with r as x_{j} has.
if len((Yp - np.dot(X, W)).shape) > 2:
res = scipy.linalg.norm(np.squeeze(Yp - np.dot(X, W)), 'fro') ** 2
else:
res = scipy.linalg.norm(Yp - np.dot(X, W), 'fro') ** 2
# % Check finishing conditions
if lambda_ == 0 or (new >= 0 and np.sum(active_set) == maxcomps) or (res < noise):
if verbose:
print('end. \n')
break
# 6: Continue until: all predictors are in the model
if new >= 0:
active_set[new] = 1
i = i + 1
Ws_old = Ws
# end main loop
#%% final calculation of mus
Ws = np.asarray(np.swapaxes(np.swapaxes(Ws_old, 0, 1), 1, 2))
if flag == 0:
if i > 0:
Ws = np.squeeze(Ws[:, :, :len(lambdas)])
w_dir = old_div(-(Ws[:, i] - Ws[:, i - 1]),
(lambdas[i] - lambdas[i - 1]))
Aw = np.dot(X, w_dir)
y_res = np.squeeze(
Yp) - np.dot(X, Ws[:, i - 1] + w_dir * lambdas[i - 1])
ld = scipy.roots([scipy.linalg.norm(Aw) ** 2, -2 * np.dot(Aw.T, y_res),
np.dot(y_res.T, y_res) - noise])
lam = ld[np.intersect1d(
np.where(ld > lambdas[i]), np.where(ld < lambdas[i - 1]))]
if len(lam) == 0 or np.any(lam) < 0 or np.any(~np.isreal(lam)):
lam = np.array([lambdas[i]])
W_lam = Ws[:, i - 1] + np.dot(w_dir, lambdas[i - 1] - lam[0])
else:
warn('LARS REGRESSION NOT SOLVABLE, USING NN LEAST SQUARE')
W_lam = scipy.optimize.nnls(X, np.ravel(Yp))[0]
lam = 10
else:
W_lam = 0
Ws = 0
lambdas = 0
lam = 0
return Ws, lambdas, W_lam, lam, flag
def TB_lib(compuesto, T):
"""Librería de cálculo de la ecuación de estado de Trebble-Bishnoi"""
if compuesto.Tc!=0.0 and compuesto.Pc.atm!=0.0 and compuesto.vc!=0.0:
Zc=compuesto.Pc.atm*compuesto.vc*compuesto.peso_molecular/R_atml/compuesto.Tc
else:
if compuesto.f_acent<0.225:
TcPc=775.9+12009*compuesto.f_acent-57335*compuesto.f_acent**2+91393*compuesto.f_acent**3
elif compuesto.f_acent<=1.:
TcPc=1876-1160*compuesto.f_acent
else:
TcPc=compuesto.Tc*compuesto.Pc.MPa
if compuesto.f_acent>=-0.14:
Zc=0.29-0.0885*compuesto.f_acent-0.0005/((compuesto.Tc*compuesto.Pc.MPa)**0.5-TcPc**0.5)
else:
Zc=0.3024
d=0.341*compuesto.vc*compuesto.peso_molecular-0.005
Xc=1.075*Zc
Dc=d*compuesto.Pc.atm/R_atml/compuesto.Tc
Cc=1-3*Xc
B=roots([1, 2-3*Xc, 3*Xc**2, -Dc**2-Xc**3])
Bpositivos=[]
for i in range(3):
if B[i]>0:
Bpositivos.append(B[i])
Bc=min(Bpositivos)
Ac=3*Xc**2+2*Bc*Cc+Bc+Cc+Bc**2+Dc**2
if compuesto.indice==212 and compuesto.tr(T)<=1:
q1=-0.31913
elif compuesto.f_acent<-0.1:
q1=0.66208+4.63961*compuesto.f_acent+7.45183*compuesto.f_acent**2
elif compuesto.f_acent<=0.4:
q1=0.35+0.7924*compuesto.f_acent+0.1875*compuesto.f_acent**2-28.93*(0.3-Zc)**2
elif compuesto.f_acent>0.4:
q1=0.32+0.9424*compuesto.f_acent-28.93*(0.3-Zc)**2
if compuesto.f_acent<-0.0423:
q2=0
elif compuesto.f_acent<=0.3:
q2=0.05246+1.15058*compuesto.f_acent-1.99348*compuesto.f_acent**2+1.5949*compuesto.f_acent**3-1.39267*compuesto.f_acent**4
else:
q2=0.17959+0.23471*compuesto.f_acent
alfa=exp(q1*(1-compuesto.tr(T)))
if T<=compuesto.Tc:
beta=1.+q2*(1-compuesto.tr(T)+log(compuesto.tr(T)))
else:
beta=1
ac=Ac*R_atml**2*compuesto.Tc**2/compuesto.Pc.atm
bc=Bc*R_atml*compuesto.Tc/compuesto.Pc.atm
c=Cc*R_atml*compuesto.Tc/compuesto.Pc.atm
return ac*alfa, bc*beta, c, d
def TB_Fugacidad(self, T, P):
"""Método de cálculo de la fugacidad mediante la ecuación de estado de Trebble-Bishnoi"""
a, b, c, d, q1, q2=self.TB_lib(T, P)
z=self.TB_Z(T, P)
A=a*P/R_atml**2/T**2
B=b*P/R_atml/T
u=1+c/b
t=1+6*c/b+c**2/b**2+4*d**2/b**2
tita=abs(t)**0.5
if t>=0:
lamda=log((2*z+B*(u-tita))/(2*z+B*(u+tita)))
else:
lamda=2*arctan((2*z+u*B)/B/tita)-pi
fi=z-1-log(z-B)+A/B/tita*lamda
return unidades.Pressure(P*exp(fi), "atm")
def TB_U_exceso(self, T, P):
"""Método de cálculo de la energía interna de exceso mediante la ecuación de estado de Trebble-Bishnoi"""
a, b, c, d, q1, q2=self.TB_lib(T, P)
v=self.TB_V(T, P)
z=P*v/R_atml/T
A=a*P/R_atml**2/T**2
B=b*P/R_atml/T
u=1+c/b
t=1+6*c/b+c**2/b**2+4*d**2/b**2
tita=abs(t)**0.5
if t>=0:
lamda=log((2*z+B*(u-tita))/(2*z+B*(u+tita)))
else:
lamda=2*arctan((2*z+u*B)/B/tita)-pi
delta=v**2+(b+c)*v-b*c-d**2
beta=1.+q2*(1-self.tr(T)+log(self.tr(T)))
da=-q1*a/self.Tc
if self.tr(T)<=1.0:
db=b/beta*(1/T-1/self.Tc)
else:
db=0
U=lamda/b/tita*(a-da*T)+db*T*(-R_atml*T/(v-b)+a/b**2/t*((v*(3*c+b)-b*c+c**2-2*d**2)/delta+(3*c+b)*lamda/b/tita)) #atm*l/mol
return unidades.Enthalpy(U*101325/1000/self.peso_molecular, "Jkg")
def TB_H_exceso(self, T, P):
"""Método de cálculo de la entalpía de exceso mediante la ecuación de estado de Trebble-Bishnoi"""
p=unidades.Pressure(P, "atm")
U=self.TB_U_exceso(T, P)
a, b, c, d, q1, q2=self.TB_lib(T, P)
t=1+6*c/b+c**2/b**2+4*d**2/b**2
if t>=0:
v=self.TB_V(T, P).m3g*self.peso_molecular
return unidades.Enthalpy(p*v-R/T/self.peso_molecular+U.Jg, "Jg")
else:
Z=self.TB_Z(T, P)
return unidades.Enthalpy(R/self.peso_molecular*T*(Z-1)+U.Jg, "Jg")
def TB_Entalpia(self, T, P):
"""Método de cálculo de la entalpía mediante la ecuación de estado de Trebble-Bishnoi"""
Ho=self._Ho(T)
Delta=self.TB_H_exceso(T, P)
return unidades.Enthalpy(Delta+Ho)
def TB_S_exceso(self, T, P):
"""Método de cálculo de la entropía de exceso mediante la ecuación de estado de Trebble-Bishnoi"""
H=self.TB_H_exceso(T, P)
f=self.TB_Fugacidad(T, P)
return unidades.SpecificHeat(H.Jg/T-R/self.peso_molecular*log(f.atm/P), "JgK")
def TB_Entropia(self, T, P):
"""Método de cálculo de la entropía mediante la ecuación de estado de Trebble-Bishnoi"""
So=self._so(T)
Delta=self.TB_S_exceso(T, P)
return unidades.SpecificHeat(Delta+So)
def TB_Cv_exceso(self, T, P):
"""Método de cálculo de la capacidad calorífica a volumen constante de exceso mediante la ecuación de estado de Trebble-Bishnoi"""
a, b, c, d, q1, q2=self.TB_lib(T, P)
v=self.TB_V(T, P)
z=P*v/R_atml/T
t=1+6*c/b+c**2/b**2+4*d**2/b**2
tita=abs(t)**0.5
A=a*P/R_atml**2/T**2
B=b*P/R_atml/T
u=1+c/b
delta=v**2+(b+c)*v-b*c-d**2
beta=1.+q2*(1-self.tr(T)+log(self.tr(T)))
da=-q1*a/self.Tc
dda=q1**2*a/self.Tc**2
if self.tr(T)<=1.0:
db=b/beta*(1/T-1/self.Tc)
ddb=-q2*b/beta/T**2
else:
db=0
ddb=0
dt=-db/b**2*(6*c+2*c**2/b+8*d**2/b)
dtita=abs(dt)/20
if t>=0:
lamda=log((2*z+B*(u-tita))/(2*z+B*(u+tita)))
dlamda=(db-db*tita-b*dtita)/(2*v+b+c-b*tita)-(db+db*tita+b*dtita)/((2*v+b+c+b*tita))
else:
lamda=2*arctan((2*z+u*B)/B/tita)-pi
dlamda=2/(1+((2*v+b+c)/b/tita)**2)*(db/b/tita-(2*v+b+c)*(db/b**2/tita+dtita/b/tita**2))
Cv=1/b/tita*(dlamda*(a-da*T)-lamda*dda*T-lamda*(a-da*T)*(db/b+dtita/tita))+(ddb*T+db)*(-R_atml*T/(v-b)+a/b**2/t*((v*(3*c+b)-b*c+c**2-2*d**2)/delta+(3*c+b)*lamda/b/tita))+db*T*(-R_atml/(v-b)-R_atml*T*db/(v-b)**2+1/b**2/t*(da-2*a*db/b-a*dt/t)*((v*(3*c+b)-b*c+c**2-2*d**2)/delta+(3*c+b)*lamda/b/tita)+a/b**2/t*(db*(v-c)*(v**2-2*c*v-c**2+d**2)/delta**2+db*lamda/b/tita+(3*c+b)/b/tita*(dlamda-lamda*(db/b+dtita/tita))))
return unidades.SpecificHeat(Cv*101325/1000/self.peso_molecular, "JkgK")
def TB_Cv(self, T, P):
"""Método de cálculo de la capacidad calorifica a volumen constante mediante la ecuación de estado de Trebble-Bishnoi"""
Cvo=self.Cv_ideal(T)
Delta=self.TB_Cv_exceso(T, P)
return unidades.SpecificHeat(Delta+Cvo)
def TB_Cp_exceso(self, T, P):
"""Método de cálculo de la capacidad calorífica a presión constante de exceso mediante la ecuación de estado de Trebble-Bishnoi"""
a, b, c, d, q1, q2=self.TB_lib(T, P)
v=self.TB_V(T, P)
Cv=self.TB_Cv_exceso(T, P)
beta=1.+q2*(1-self.tr(T)+log(self.tr(T)))
delta=v**2+(b+c)*v-b*c-d**2
da=-q1*a/self.Tc
if self.tr(T)<=1.0:
db=b/beta*(1/T-1/self.Tc)
else:
db=0
dpdt=R_atml/(v-b)+R*T*db/(v-b)**2-da/delta+a*(v-c)*db/delta**2
dpdv=-R_atml*T/(v-b)**2+a*(2*v+b+c)/delta**2
Cp=-R-T*dpdt**2/dpdv*101325/1000+Cv.JgK*self.peso_molecular
return unidades.SpecificHeat(Cp/self.peso_molecular, "JgK")
def TB_Cp(self, T, P):
"""Método de cálculo de la capacidad calorifica a presión constante mediante la ecuación de estado de Trebble-Bishnoi"""
Cpo=self.Cp_ideal(T)
Delta=self.TB_Cp_exceso(T, P)
return unidades.SpecificHeat(Delta+Cpo)
def TB_Joule_Thomson(self, T, P):
"""Método de cálculo del coeficiente de Joule-Thomson mediante la ecuación de estado de Trebble-Bishnoi"""
v=self.TB_V(T, P)
Cp=self.TB_Cp(T, P)
a, b, c, d, q1, q2=self.TB_lib(T, P)
beta=1.+q2*(1-self.tr(T)+log(self.tr(T)))
delta=v**2+(b+c)*v-b*c-d**2
da=-q1*a/self.Tc
if self.tr(T)<=1.0:
db=b/beta*(1/T-1/self.Tc)
else:
db=0
dpdt=R_atml/(v-b)+R*T*db/(v-b)**2-da/delta+a*(v-c)*db/delta**2
dpdv=-R_atml*T/(v-b)**2+a*(2*v+b+c)/delta**2
return -(T*dpdt+v*dpdv)/Cp/dpdv
def lars_regression_noise_old(Yp, X, positive, noise, verbose=False):
"""
Run LARS for regression problems with LASSO penalty, with optional positivity constraints
Author: Andrea Giovannucci. Adapted code from Eftychios Pnevmatikakis
Input Parameters:
Yp: Yp[:,t] is the observed data at time t
X: the regresion problem is Yp=X*W + noise
maxcomps: maximum number of active components to allow
positive: a flag to enforce positivity
noise: the noise of the observation equation. if it is not
provided as an argument, the noise is computed from the
variance at the end point of the algorithm. The noise is
used in the computation of the Cp criterion.
Output Parameters:
Ws: weights from each iteration
lambdas: lambda_ values at each iteration
TODO: W_lam, lam, flag
Cps: C_p estimates
last_break: last_break(m) == n means that the last break with m non-zero weights is at Ws(:,:,n)
"""
#%%
# verbose=true;
# do NNLS first
# if hard noise constraint problem infeasible we are done, otherwise good for warm-starting
# W_lam, RSS = scipy.optimize.nnls(X, np.ravel(Yp))
# RSS = RSS * RSS
# if RSS > noise: # hard noise constraint problem infeasible
# return 0, 0, W_lam, 0, 0
#
# while 1:
# eliminate = []
# for i in np.where(W_lam[:-1] > 0)[0]: # W_lam[:-1] to skip background
# mask = W_lam > 0
# mask[i] = 0
# Wtmp, tmp = scipy.optimize.nnls(X * mask, np.ravel(Yp))
# if tmp * tmp < noise:
# eliminate.append([i, tmp])
# if eliminate == []:
# return 0, 0, W_lam, 0, 0
# else:
# W_lam[eliminate[np.argmin(np.array(eliminate)[:, 1])][0]] = 0
# obsolete stuff below
# solve hard noise constraint problem
T = len(Yp) # of time steps
# A = Yp.dot(X)
# B = X.T.dot(X)
# dB = 1 / np.diag(B)
# import pdb
# pdb.set_trace()
# z = np.sqrt(1 / dB)
# lam = np.max(1 * np.sqrt(noise / T) * z)
# counter = 0
# res = X.dot(W_lam) - Yp
# tmp = X.dot(dB)
# aa = tmp.dot(tmp)
# while (RSS < noise * .9999 or RSS > noise * 1.0001) and counter < 20:
# counter += 1
# bb = tmp.dot(res)
# cc = RSS - noise
# if counter > 1:
# lam += (bb + (0 if bb * bb <= aa * cc else np.sqrt(bb * bb - aa * cc))) / aa
# for _ in range(3):
# for i in range(len(W_lam)):
# W_lam[i] = np.clip(W_lam[i] + (A[i] - W_lam.dot(B)[i] - lam) * dB[i], 0, np.inf)
# res = X.dot(W_lam) - Yp
# RSS = res.dot(res)
# return 1, 1, W_lam, 1, 1
k = 1
Yp = np.squeeze(np.asarray(Yp))
Yp = np.expand_dims(Yp, axis=1) # necessary for matrix multiplications
_, T = np.shape(Yp) # of time steps
_, N = np.shape(X) # of compartments
maxcomps = N
W = np.zeros((N, k))
active_set = np.zeros((N, k))
visited_set = np.zeros((N, k))
lambdas = []
# =np.zeros((W.shape[0],W.shape[1],maxcomps)); # Just preallocation. Ws may end with more or less than maxcomp columns
Ws = []
r = np.expand_dims(np.dot(X.T, Yp.flatten()), axis=1) # N-dim vector
M = np.dot(-X.T, X) # N x N matrix
#%% begin main loop
i = 0
flag = 0
while 1:
if flag == 1:
W_lam = 0
break
#% calculate new gradient component if necessary
if i > 0 and new >= 0 and visited_set[new] == 0: # AG NOT CLEAR HERE
visited_set[new] = 1 # % remember this direction was computed
#% Compute full gradient of Q
dQ = r + np.dot(M, W)
#% Compute new W
if i == 0:
if positive:
dQa = dQ
else:
dQa = np.abs(dQ)
lambda_, new = np.max(dQa), np.argmax(dQa)
if lambda_ < 0:
print('All negative directions!')
break
else:
#% calculate vector to travel along
avec, gamma_plus, gamma_minus = calcAvec(new, dQ, W, lambda_, active_set, M, positive)
# % calculate time of travel and next new direction
if new == -1: # % if we just dropped a direction we don't allow it to emerge
if dropped_sign == 1: # % with the same sign
gamma_plus[dropped] = np.inf
else:
gamma_minus[dropped] = np.inf
gamma_plus[active_set == 1] = np.inf # % don't consider active components
gamma_plus[gamma_plus <= 0] = np.inf # % or components outside the range [0, lambda_]
gamma_plus[gamma_plus > lambda_] = np.inf
gp_min, gp_min_ind = np.min(gamma_plus), np.argmin(gamma_plus)
if positive:
gm_min = np.inf # % don't consider new directions that would grow negative
else:
gamma_minus[active_set == 1] = np.inf
gamma_minus[gamma_minus > lambda_] = np.inf
gamma_minus[gamma_minus <= 0] = np.inf
gm_min, gm_min_ind = np.min(gamma_minus), np.argmin(gamma_minus)
[g_min, which] = np.min(gp_min), np.argmin(gp_min)
if g_min == np.inf: # % if there are no possible new components, try move to the end
g_min = lambda_ # % This happens when all the components are already active or, if positive==1, when there are no new positive directions
#% LARS check (is g_min*avec too large?)
gamma_zero = old_div(-W[active_set == 1], np.squeeze(avec))
gamma_zero_full = np.zeros((N, k))
gamma_zero_full[active_set == 1] = gamma_zero
gamma_zero_full[gamma_zero_full <= 0] = np.inf
gz_min, gz_min_ind = np.min(gamma_zero_full), np.argmin(gamma_zero_full)
if gz_min < g_min:
# print 'check_here'
if verbose:
print(('DROPPING active weight:' + str(gz_min_ind)))
active_set[gz_min_ind] = 0
dropped = gz_min_ind
dropped_sign = np.sign(W[dropped])
W[gz_min_ind] = 0
avec = avec[gamma_zero != gz_min]
g_min = gz_min
new = -1 # new = 0;
elif g_min < lambda_:
if which == 0:
new = gp_min_ind
if verbose:
print(('new positive component:' + str(new)))
else:
new = gm_min_ind
print(('new negative component:' + str(new)))
W[active_set == 1] = W[active_set == 1] + np.dot(g_min, np.squeeze(avec))
if positive:
if any(W < 0):
# min(W);
flag = 1
#%error('negative W component');
lambda_ = lambda_ - g_min
#% Update weights and lambdas
lambdas.append(lambda_)
Ws.append(W.copy())
# print Ws
if len((Yp - np.dot(X, W)).shape) > 2:
res = scipy.linalg.norm(np.squeeze(Yp - np.dot(X, W)), 'fro')**2
else:
res = scipy.linalg.norm(Yp - np.dot(X, W), 'fro')**2
#% Check finishing conditions
if lambda_ == 0 or (new >= 0 and np.sum(active_set) == maxcomps) or (res < noise):
if verbose:
print('end. \n')
break
#%
if new >= 0:
active_set[new] = 1
i = i + 1
Ws_old = Ws
# end main loop
#%% final calculation of mus
Ws = np.asarray(np.swapaxes(np.swapaxes(Ws_old, 0, 1), 1, 2))
if flag == 0:
if i > 0:
Ws = np.squeeze(Ws[:, :, :len(lambdas)])
w_dir = old_div(-(Ws[:, i] - Ws[:, i - 1]), (lambdas[i] - lambdas[i - 1]))
Aw = np.dot(X, w_dir)
y_res = np.squeeze(Yp) - np.dot(X, Ws[:, i - 1] + w_dir * lambdas[i - 1])
ld = scipy.roots([scipy.linalg.norm(Aw)**2, -2 * np.dot(Aw.T, y_res),
np.dot(y_res.T, y_res) - noise])
lam = ld[np.intersect1d(np.where(ld > lambdas[i]), np.where(ld < lambdas[i - 1]))]
if len(lam) == 0 or np.any(lam) < 0 or np.any(~np.isreal(lam)):
lam = np.array([lambdas[i]])
W_lam = Ws[:, i - 1] + np.dot(w_dir, lambdas[i - 1] - lam[0])
else:
warn('LARS REGRESSION NOT SOLVABLE, USING NN LEAST SQUARE')
W_lam = scipy.optimize.nnls(X, np.ravel(Yp))[0]
# problem = picos.Problem(X,Yp)
# W_lam = problem.add_variable('W_lam', X.shape[1])
# problem.set_objective('min', 1|W_lam)
# problem.add_constraint(W_lam >= 0)
# problem.add_constraint(picos.norm(matrix(Yp.astype(np.float))-matrix(X.astype(np.float))*W_lam,2)<=np.sqrt(noise))
# sel_solver = []
# problem.solver_selection()
# problem.solve(verbose=True)
# cvx_begin quiet
# variable W_lam(size(X,2));
# minimize(sum(W_lam));
# subject to
# W_lam >= 0;
# norm(Yp-X*W_lam)<= sqrt(noise);
# cvx_end
lam = 10
else:
W_lam = 0
Ws = 0
lambdas = 0
lam = 0
return Ws, lambdas, W_lam, lam, flag
def lars_regression_noise_old(Yp, X, positive, noise, verbose=False):
"""
Run LARS for regression problems with LASSO penalty, with optional positivity constraints
Author: Andrea Giovannucci. Adapted code from Eftychios Pnevmatikakis
Input Parameters:
Yp: Yp[:,t] is the observed data at time t
X: the regresion problem is Yp=X*W + noise
maxcomps: maximum number of active components to allow
positive: a flag to enforce positivity
noise: the noise of the observation equation. if it is not
provided as an argument, the noise is computed from the
variance at the end point of the algorithm. The noise is
used in the computation of the Cp criterion.
Output Parameters:
Ws: weights from each iteration
lambdas: lambda_ values at each iteration
TODO: W_lam, lam, flag
Cps: C_p estimates
last_break: last_break(m) == n means that the last break with m non-zero weights is at Ws(:,:,n)
"""
#%%
# verbose=true;
# do NNLS first
# if hard noise constraint problem infeasible we are done, otherwise good for warm-starting
# W_lam, RSS = scipy.optimize.nnls(X, np.ravel(Yp))
# RSS = RSS * RSS
# if RSS > noise: # hard noise constraint problem infeasible
# return 0, 0, W_lam, 0, 0
#
# while 1:
# eliminate = []
# for i in np.where(W_lam[:-1] > 0)[0]: # W_lam[:-1] to skip background
# mask = W_lam > 0
# mask[i] = 0
# Wtmp, tmp = scipy.optimize.nnls(X * mask, np.ravel(Yp))
# if tmp * tmp < noise:
# eliminate.append([i, tmp])
# if eliminate == []:
# return 0, 0, W_lam, 0, 0
# else:
# W_lam[eliminate[np.argmin(np.array(eliminate)[:, 1])][0]] = 0
# obsolete stuff below
# solve hard noise constraint problem
T = len(Yp) # of time steps
# A = Yp.dot(X)
# B = X.T.dot(X)
# dB = 1 / np.diag(B)
# import pdb
# pdb.set_trace()
# z = np.sqrt(1 / dB)
# lam = np.max(1 * np.sqrt(noise / T) * z)
# counter = 0
# res = X.dot(W_lam) - Yp
# tmp = X.dot(dB)
# aa = tmp.dot(tmp)
# while (RSS < noise * .9999 or RSS > noise * 1.0001) and counter < 20:
# counter += 1
# bb = tmp.dot(res)
# cc = RSS - noise
# if counter > 1:
# lam += (bb + (0 if bb * bb <= aa * cc else np.sqrt(bb * bb - aa * cc))) / aa
# for _ in range(3):
# for i in range(len(W_lam)):
# W_lam[i] = np.clip(W_lam[i] + (A[i] - W_lam.dot(B)[i] - lam) * dB[i], 0, np.inf)
# res = X.dot(W_lam) - Yp
# RSS = res.dot(res)
# return 1, 1, W_lam, 1, 1
k = 1
Yp = np.squeeze(np.asarray(Yp))
Yp = np.expand_dims(Yp, axis=1) # necessary for matrix multiplications
_, T = np.shape(Yp) # of time steps
_, N = np.shape(X) # of compartments
maxcomps = N
W = np.zeros((N, k))
active_set = np.zeros((N, k))
visited_set = np.zeros((N, k))
lambdas = []
# =np.zeros((W.shape[0],W.shape[1],maxcomps)); # Just preallocation. Ws may end with more or less than maxcomp columns
Ws = []
r = np.expand_dims(np.dot(X.T, Yp.flatten()), axis=1) # N-dim vector
M = np.dot(-X.T, X) # N x N matrix
#%% begin main loop
i = 0
flag = 0
while 1:
if flag == 1:
W_lam = 0
break
#% calculate new gradient component if necessary
if i > 0 and new >= 0 and visited_set[new] == 0: # AG NOT CLEAR HERE
visited_set[new] = 1 # % remember this direction was computed
#% Compute full gradient of Q
dQ = r + np.dot(M, W)
#% Compute new W
if i == 0:
if positive:
dQa = dQ
else:
dQa = np.abs(dQ)
lambda_, new = np.max(dQa), np.argmax(dQa)
if lambda_ < 0:
print('All negative directions!')
break
else:
#% calculate vector to travel along
avec, gamma_plus, gamma_minus = calcAvec(new, dQ, W, lambda_, active_set, M, positive)
# % calculate time of travel and next new direction
if new == -1: # % if we just dropped a direction we don't allow it to emerge
if dropped_sign == 1: # % with the same sign
gamma_plus[dropped] = np.inf
else:
gamma_minus[dropped] = np.inf
gamma_plus[active_set == 1] = np.inf # % don't consider active components
gamma_plus[gamma_plus <= 0] = np.inf # % or components outside the range [0, lambda_]
gamma_plus[gamma_plus > lambda_] = np.inf
gp_min, gp_min_ind = np.min(gamma_plus), np.argmin(gamma_plus)
if positive:
gm_min = np.inf # % don't consider new directions that would grow negative
else:
gamma_minus[active_set == 1] = np.inf
gamma_minus[gamma_minus > lambda_] = np.inf
gamma_minus[gamma_minus <= 0] = np.inf
gm_min, gm_min_ind = np.min(gamma_minus), np.argmin(gamma_minus)
[g_min, which] = np.min(gp_min), np.argmin(gp_min)
if g_min == np.inf: # % if there are no possible new components, try move to the end
g_min = lambda_ # % This happens when all the components are already active or, if positive==1, when there are no new positive directions
#% LARS check (is g_min*avec too large?)
gamma_zero = old_div(-W[active_set == 1], np.squeeze(avec))
gamma_zero_full = np.zeros((N, k))
gamma_zero_full[active_set == 1] = gamma_zero
gamma_zero_full[gamma_zero_full <= 0] = np.inf
gz_min, gz_min_ind = np.min(gamma_zero_full), np.argmin(gamma_zero_full)
if gz_min < g_min:
# print 'check_here'
if verbose:
print(('DROPPING active weight:' + str(gz_min_ind)))
active_set[gz_min_ind] = 0
dropped = gz_min_ind
dropped_sign = np.sign(W[dropped])
W[gz_min_ind] = 0
avec = avec[gamma_zero != gz_min]
g_min = gz_min
new = -1 # new = 0;
elif g_min < lambda_:
if which == 0:
new = gp_min_ind
if verbose:
print(('new positive component:' + str(new)))
else:
new = gm_min_ind
print(('new negative component:' + str(new)))
W[active_set == 1] = W[active_set == 1] + np.dot(g_min, np.squeeze(avec))
if positive:
if any(W < 0):
# min(W);
flag = 1
#%error('negative W component');
lambda_ = lambda_ - g_min
#% Update weights and lambdas
lambdas.append(lambda_)
Ws.append(W.copy())
# print Ws
if len((Yp - np.dot(X, W)).shape) > 2:
res = scipy.linalg.norm(np.squeeze(Yp - np.dot(X, W)), 'fro')**2
else:
res = scipy.linalg.norm(Yp - np.dot(X, W), 'fro')**2
#% Check finishing conditions
if lambda_ == 0 or (new >= 0 and np.sum(active_set) == maxcomps) or (res < noise):
if verbose:
print('end. \n')
break
#%
if new >= 0:
active_set[new] = 1
i = i + 1
Ws_old = Ws
# end main loop
#%% final calculation of mus
Ws = np.asarray(np.swapaxes(np.swapaxes(Ws_old, 0, 1), 1, 2))
if flag == 0:
if i > 0:
Ws = np.squeeze(Ws[:, :, :len(lambdas)])
w_dir = old_div(-(Ws[:, i] - Ws[:, i - 1]), (lambdas[i] - lambdas[i - 1]))
Aw = np.dot(X, w_dir)
y_res = np.squeeze(Yp) - np.dot(X, Ws[:, i - 1] + w_dir * lambdas[i - 1])
ld = scipy.roots([scipy.linalg.norm(Aw)**2, -2 * np.dot(Aw.T, y_res),
np.dot(y_res.T, y_res) - noise])
lam = ld[np.intersect1d(np.where(ld > lambdas[i]), np.where(ld < lambdas[i - 1]))]
if len(lam) == 0 or np.any(lam) < 0 or np.any(~np.isreal(lam)):
lam = np.array([lambdas[i]])
W_lam = Ws[:, i - 1] + np.dot(w_dir, lambdas[i - 1] - lam[0])
else:
warn('LARS REGRESSION NOT SOLVABLE, USING NN LEAST SQUARE')
W_lam = scipy.optimize.nnls(X, np.ravel(Yp))[0]
# problem = picos.Problem(X,Yp)
# W_lam = problem.add_variable('W_lam', X.shape[1])
# problem.set_objective('min', 1|W_lam)
# problem.add_constraint(W_lam >= 0)
# problem.add_constraint(picos.norm(matrix(Yp.astype(np.float))-matrix(X.astype(np.float))*W_lam,2)<=np.sqrt(noise))
# sel_solver = []
# problem.solver_selection()
# problem.solve(verbose=True)
# cvx_begin quiet
# variable W_lam(size(X,2));
# minimize(sum(W_lam));
# subject to
# W_lam >= 0;
# norm(Yp-X*W_lam)<= sqrt(noise);
# cvx_end
lam = 10
else:
W_lam = 0
Ws = 0
lambdas = 0
lam = 0
return Ws, lambdas, W_lam, lam, flag
def lars_regression_noise_old(Yp, X, positive, noise, verbose=False):
"""
Run LARS for regression problems with LASSO penalty, with optional positivity constraints
Author: Andrea Giovannucci. Adapted code from Eftychios Pnevmatikakis
Parameters:
-------
Yp: Yp[:,t] is the observed data at time t
X: the regresion problem is Yp=X*W + noise
maxcomps: maximum number of active components to allow
positive: a flag to enforce positivity
noise: the noise of the observation equation. if it is not
provided as an argument, the noise is computed from the
variance at the end point of the algorithm. The noise is
used in the computation of the Cp criterion.
Returns:
-------
Ws: weights from each iteration
lambdas: lambda_ values at each iteration
Cps: C_p estimates
last_break: last_break(m) == n means that the last break with m non-zero weights is at Ws(:,:,n)
See Also:
-------
LARS : https://en.wikipedia.org/wiki/Least-angle_regression
group Lasso :
"""
# INITAILIZATION
T = len(Yp) # of time steps
k = 1
Yp = np.squeeze(np.asarray(Yp))
# necessary for matrix multiplications
Yp = np.expand_dims(Yp, axis=1)
_, T = np.shape(Yp) # of time steps
_, N = np.shape(X) # of compartments
# 1 : Start with all Weights equal to zero.
maxcomps = N
W = np.zeros((N, k))
active_set = np.zeros((N, k))
visited_set = np.zeros((N, k))
lambdas = []
# Just preallocation. Ws may end with more or less than maxcomp columns
Ws = []
r = np.expand_dims(np.dot(X.T, Yp.flatten()), axis=1) # N-dim vector
M = np.dot(-X.T, X) # N x N matrix
# %% begin main loop
i = 0
flag = 0
while 1:
if flag == 1:
W_lam = 0
break
# % calculate new gradient component if necessary
if i > 0 and new >= 0 and visited_set[new] == 0: # AG NOT CLEAR HERE
visited_set[new] = 1 # % remember this direction was computed
# % Compute full gradient of Q
dQ = r + np.dot(M, W)
# % Compute new W
if i == 0:
if positive:
dQa = dQ
else:
dQa = np.abs(dQ)
lambda_, new = np.max(dQa), np.argmax(dQa)
if lambda_ < 0:
print('All negative directions!')
break
else: # 2 : Find the predictor x_{j} most correlated with y
# % calculate vector to travel along
avec, gamma_plus, gamma_minus = calcAvec(
new, dQ, W, lambda_, active_set, M, positive)
# % calculate time of travel and next new direction
if new == -1: # % if we just dropped a direction we don't allow it to emerge
if dropped_sign == 1: # % with the same sign
gamma_plus[dropped] = np.inf
else:
gamma_minus[dropped] = np.inf
# 3 : Increase the coefficient W in the direction of the sign of its correlation with y
# % don't consider active components
gamma_plus[active_set == 1] = np.inf
# % or components outside the range [0, lambda_]
gamma_plus[gamma_plus <= 0] = np.inf
gamma_plus[gamma_plus > lambda_] = np.inf
gp_min, gp_min_ind = np.min(gamma_plus), np.argmin(gamma_plus)
if positive:
gm_min = np.inf # % don't consider new directions that would grow negative
else:
gamma_minus[active_set == 1] = np.inf
gamma_minus[gamma_minus > lambda_] = np.inf
gamma_minus[gamma_minus <= 0] = np.inf
gm_min, gm_min_ind = np.min(
gamma_minus), np.argmin(gamma_minus)
[g_min, which] = np.min(gp_min), np.argmin(gp_min)
# % if there are no possible new components, try move to the end
if g_min == np.inf:
g_min = lambda_
# % This happens when all the components are already active or, if positive==1,
# when there are no new positive directions
# % LARS check (is g_min*avec too large?)
gamma_zero = old_div(-W[active_set == 1], np.squeeze(avec))
gamma_zero_full = np.zeros((N, k))
gamma_zero_full[active_set == 1] = gamma_zero
gamma_zero_full[gamma_zero_full <= 0] = np.inf
gz_min, gz_min_ind = np.min(
gamma_zero_full), np.argmin(gamma_zero_full)
# 4: Increase Wj,Wk in their joint least squares direction, until some other predictor x_{m}
# has as much correlation with the residual r. (see 5)
if gz_min < g_min:
if verbose:
print(('DROPPING active weight:' + str(gz_min_ind)))
active_set[gz_min_ind] = 0
dropped = gz_min_ind
dropped_sign = np.sign(W[dropped])
W[gz_min_ind] = 0
avec = avec[gamma_zero != gz_min]
g_min = gz_min
new = -1
elif g_min < lambda_:
if which == 0:
new = gp_min_ind
if verbose:
print(('new positive component:' + str(new)))
else:
new = gm_min_ind
print(('new negative component:' + str(new)))
W[active_set == 1] = W[active_set == 1] + \
np.dot(g_min, np.squeeze(avec))
if positive:
if any(W < 0):
flag = 1
lambda_ = lambda_ - g_min
# % Update weights and lambdas
lambdas.append(lambda_)
Ws.append(W.copy())
# 5 : Take residuals r=y-y_ along the way. Stop when some other predictor x_{k}
# has as much correlation with r as x_{j} has.
if len((Yp - np.dot(X, W)).shape) > 2:
res = scipy.linalg.norm(np.squeeze(Yp - np.dot(X, W)), 'fro') ** 2
else:
res = scipy.linalg.norm(Yp - np.dot(X, W), 'fro') ** 2
# % Check finishing conditions
if lambda_ == 0 or (new >= 0 and np.sum(active_set) == maxcomps) or (res < noise):
if verbose:
print('end. \n')
break
# 6: Continue until: all predictors are in the model
if new >= 0:
active_set[new] = 1
i = i + 1
Ws_old = Ws
# end main loop
#%% final calculation of mus
Ws = np.asarray(np.swapaxes(np.swapaxes(Ws_old, 0, 1), 1, 2))
if flag == 0:
if i > 0:
Ws = np.squeeze(Ws[:, :, :len(lambdas)])
w_dir = old_div(-(Ws[:, i] - Ws[:, i - 1]),
(lambdas[i] - lambdas[i - 1]))
Aw = np.dot(X, w_dir)
y_res = np.squeeze(
Yp) - np.dot(X, Ws[:, i - 1] + w_dir * lambdas[i - 1])
ld = scipy.roots([scipy.linalg.norm(Aw) ** 2, -2 * np.dot(Aw.T, y_res),
np.dot(y_res.T, y_res) - noise])
lam = ld[np.intersect1d(
np.where(ld > lambdas[i]), np.where(ld < lambdas[i - 1]))]
if len(lam) == 0 or np.any(lam) < 0 or np.any(~np.isreal(lam)):
lam = np.array([lambdas[i]])
W_lam = Ws[:, i - 1] + np.dot(w_dir, lambdas[i - 1] - lam[0])
else:
warn('LARS REGRESSION NOT SOLVABLE, USING NN LEAST SQUARE')
W_lam = scipy.optimize.nnls(X, np.ravel(Yp))[0]
lam = 10
else:
W_lam = 0
Ws = 0
lambdas = 0
lam = 0
return Ws, lambdas, W_lam, lam, flag
def main(argv):
##################################################################
# Parse arguments
fitopt = {0:'gaussian',1:'2nd order polynome'}
from optparse import OptionParser
parser = OptionParser()
parser.add_option( '--fittype',
help='Which type of fit will be performed. Options are %r.'%fitopt,
type='int',
default=0)
opt,args = parser.parse_args(argv)
print 'Will find minimum by %s fitting...'%(fitopt[opt.fittype])
##################################################################
# Parameters
HJD0 = 2450111.5144
Porb = 0.44267714
#
##################################################################
# Input definition
_path = os.path.expanduser('~/Documents/analise/CAL87/emap_0305keV_rateall')
#_cldat = 'cal87_0305keV_rateall.dat.122' #'cal87_0305keV_rateall.dat.122'
#_cldat = '../cal87_0153250101_b250s_051keV_master_pnm1m2.tfits' #'cal87_0305keV_rateall.dat.122'
_cldat = '../cal87_0153250101_b250s_0514keV_master_pnm1m2.tfits' #'cal87_0305keV_rateall.dat.122'
#_clmod = 'cal87_0305keV_rateall.prd.122'
#
##################################################################
# Reading data
#cldat = np.loadtxt(os.path.join(_path,_cldat),unpack=True)
#clmod = np.loadtxt(os.path.join(_path,_clmod),unpack=True)
table = pyfits.open(os.path.join(_path,_cldat))
time_hjd = table[1].data['TIME']/86400.+2450814.5000
rate_all = table[1].data['RATE_ALL']
error_all = table[1].data['ERROR_ALL']
#
##################################################################
# Fitting
if opt.fittype == 0:
phase = (time_hjd - HJD0) / Porb
phase = phase-np.floor(phase)
phase[phase > 0.5] -= 1.0
sort = phase.argsort()
cldat = np.array([phase[sort],rate_all[sort],error_all[sort]])
func = lambda p,x: p[0] + p[1]*np.exp(-((x-p[2])/p[3])**2.0)
fitfunc = lambda p,x,y: y - func(p,x)
p0 = [np.mean(cldat[1]),np.min(cldat[1])-np.mean(cldat[1]),1e-1,0.15]
sol,cov,info,mesg,ier = leastsq(fitfunc,
p0,
args=(cldat[0],cldat[1]),
full_output=True)
print sol
py.errorbar( cldat[0],
cldat[1],
cldat[2],
fmt='rs:',
capsize=0)
py.plot(cldat[0],
func(sol,cldat[0]),'k-')
if opt.fittype == 1:
# Fit parabola
phase = (time_hjd - HJD0) / Porb
ciclos = np.unique(np.floor(phase))
print '# - %i cycles, fitting %i eclipses...'%(len(ciclos),len(ciclos)-1)
for i in range(len(ciclos)-1):
print '## - Fitting from phases [%f:%f]'%(ciclos[i+1]-0.5,ciclos[i+1]+0.5)
mask = np.bitwise_and(phase >ciclos[i+1]-0.5,phase < ciclos[i+1]+0.5)
cldat = np.array([phase[mask]-ciclos[i+1],rate_all[mask],error_all[mask]])
mask = np.bitwise_and(cldat[0] > -0.1, cldat[0]<0.1)
p,residuals,rank,singular_values,rcond = polyfit(cldat[0][mask],
cldat[1][mask],
deg=2,full=True)
print '# - Bootstraping...'
ndata = MC_sim(cldat[1][mask],cldat[2][mask],100)
bootpar = np.array([p])
bootroots = np.array([roots(p).real])
for j in range(len(ndata)):
bp = polyfit( cldat[0][mask],
ndata[j],
deg=2,full=False)
bootpar = np.append(bootpar,bp)
bootroots = np.append(bootroots,np.roots(bp).real)
bootpar = bootpar.reshape(-1,3)
print bootpar.shape,bootpar[0]
print p,singular_values
print roots(p)
print bootroots.mean(),'+/-',bootroots.std()
py.subplot(2,len(ciclos)-1,i+1)
py.errorbar( cldat[0][mask],
cldat[1][mask],
cldat[2][mask],
fmt='rs:',
capsize=0)
py.plot(cldat[0][mask],
polyval(p,cldat[0][mask]),'k-')
py.subplot(2,len(ciclos)-1,len(ciclos)-1+i+1)
py.hist(bootroots,bins=40)
py.show()
return 0
mask = np.bitwise_and(cldat[0] > -0.05, cldat[0]<0.09)
p,residuals,rank,singular_values,rcond = polyfit(cldat[0][mask],
cldat[1][mask],
deg=2,full=True)
py.errorbar( cldat[0][mask],
cldat[1][mask],
cldat[2][mask],
fmt='rs:',
capsize=0)
py.plot(cldat[0][mask],
polyval(p,cldat[0][mask]),'k-')
print p,singular_values
print roots(p)
#
##################################################################
# End
py.show()
return 0
def duff_amp_solve(mu = 0.01, k = 0.013, alpha = .2, sigma = (-0.5,.5)):
sigma = sp.linspace(sigma[0],sigma[1],1000)
#print(sigma.size)
#print(sigma)
a = sp.zeros((sigma.size,3))*1j
first = 1
#print(a)
for idx, sig in enumerate(sigma):
#print(idx)
p = sp.array([alpha**2, 0, -16./3.*alpha*sig, 0, 64./9.*(mu**2 + sig**2),0,-64./9.*k**2])
soln = sp.roots(p)
#print('original soln')
#print(soln)
#print(soln)
#print(sp.sort(soln)[0:5:2])
sorted_indices = sp.argsort(sp.absolute(soln))
a[idx,:] = soln[sorted_indices][0:5:2]
if sum(sp.isreal(a[idx,:])) == 3 and first == 1:
first = 0
#if sp.absolute(a[idx,2] - a[idx,1]) < sp.absolute(a[idx,1] - a[idx,0]):
solns = sp.sort(sp.absolute(a[idx,:]))
#print(solns)
if (solns[2] - solns[1]) > (solns[1]-solns[0]):
ttl = 'Hardening spring'
softening = False
else:
ttl = 'Softening spring'
softening = True
#print(solns)
first_bif_index = idx
if first == 0 and sum(sp.isreal(a[idx,:])) == 1:
first = 2
second_bif_index = idx
if softening == True:
low_sig = sigma[0:second_bif_index]
low_amp = sp.zeros(second_bif_index)
low_amp[0:first_bif_index] = sp.absolute(sp.sum(sp.isreal(a[0:first_bif_index,:])*a[0:first_bif_index,:],axis = 1))
low_amp[first_bif_index:second_bif_index] = sp.absolute(a[first_bif_index:second_bif_index,:]).min(axis = 1)
med_sig = sigma[first_bif_index:second_bif_index]
med_amp = sp.sort(sp.absolute(a[first_bif_index:second_bif_index,:]),axis = 1)[:,1]
high_sig = sigma[first_bif_index:]
high_amp = sp.zeros(sigma.size - first_bif_index)
high_amp[0:second_bif_index - first_bif_index] = sp.absolute(a[first_bif_index:second_bif_index,:]).max(axis = 1)
high_amp[second_bif_index - first_bif_index:] = sp.absolute(sp.sum(sp.isreal(a[second_bif_index:,:])*a[second_bif_index:,:],axis = 1))
else:
high_sig = sigma[0:second_bif_index]
high_amp = sp.zeros(second_bif_index)
high_amp[0:first_bif_index] = sp.absolute(sp.sum(sp.isreal(a[0:first_bif_index,:])*a[0:first_bif_index,:],axis = 1))
high_amp[first_bif_index:second_bif_index] = sp.absolute(a[first_bif_index:second_bif_index,:]).max(axis = 1)
med_sig = sigma[first_bif_index:second_bif_index]
med_amp = sp.sort(sp.absolute(a[first_bif_index:second_bif_index,:]),axis = 1)[:,1]
low_sig = sigma[first_bif_index:]
low_amp = sp.zeros(sigma.size - first_bif_index)
low_amp[0:second_bif_index - first_bif_index] = sp.absolute(a[first_bif_index:second_bif_index,:]).min(axis = 1)
low_amp[second_bif_index - first_bif_index:] = sp.absolute(sp.sum(sp.isreal(a[second_bif_index:,:])*a[second_bif_index:,:],axis = 1))
plt.plot(low_sig,low_amp,'-b')
plt.plot(med_sig, med_amp, '--g')
plt.plot(high_sig,high_amp,'-r')
plt.title(ttl)
plt.xlabel('$\sigma$')
plt.ylabel('a')
return
def lars_regression_noise(Yp, X, positive, noise,verbose=False):
"""
Run LARS for regression problems with LASSO penalty, with optional positivity constraints
Author: Andrea Giovannucci. Adapted code from Eftychios Pnevmatikakis
Input Parameters:
Yp: Yp[:,t] is the observed data at time t
X: the regresion problem is Yp=X*W + noise
maxcomps: maximum number of active components to allow
positive: a flag to enforce positivity
noise: the noise of the observation equation. if it is not
provided as an argument, the noise is computed from the
variance at the end point of the algorithm. The noise is
used in the computation of the Cp criterion.
Output Parameters:
Ws: weights from each iteration
lambdas: lambda_ values at each iteration
TODO: W_lam, lam, flag
Cps: C_p estimates
last_break: last_break(m) == n means that the last break with m non-zero weights is at Ws(:,:,n)
"""
#%%
#verbose=true;
k=1;
Yp=np.squeeze(np.asarray(Yp))
Yp=np.expand_dims(Yp,axis=1) #necessary for matrix multiplications
_,T = np.shape(Yp); # of time steps
_,N = np.shape(X); # of compartments
maxcomps = N;
W = np.zeros((N,k));
active_set = np.zeros((N,k));
visited_set = np.zeros((N,k));
lambdas = [];
Ws=[] #=np.zeros((W.shape[0],W.shape[1],maxcomps)); # Just preallocation. Ws may end with more or less than maxcomp columns
r = np.expand_dims(np.dot(X.T,Yp.flatten()),axis=1) # N-dim vector
M = np.dot(-X.T,X); # N x N matrix
#%% begin main loop
i = 0;
flag = 0;
while 1:
if flag == 1:
W_lam = 0
break;
#% calculate new gradient component if necessary
if i>0 and new>=0 and visited_set[new] ==0: # AG NOT CLEAR HERE
visited_set[new] =1; #% remember this direction was computed
#% Compute full gradient of Q
dQ = r + np.dot(M,W);
#% Compute new W
if i == 0:
if positive:
dQa = dQ
else:
dQa = np.abs(dQ)
lambda_, new = np.max(dQa),np.argmax(dQa)
if lambda_ < 0:
print 'All negative directions!'
break
else:
#% calculate vector to travel along
avec, gamma_plus, gamma_minus = calcAvec(new, dQ, W, lambda_, active_set, M, positive)
# % calculate time of travel and next new direction
if new==-1: # % if we just dropped a direction we don't allow it to emerge
if dropped_sign == 1: # % with the same sign
gamma_plus[dropped] = np.inf;
else:
gamma_minus[dropped] = np.inf;
gamma_plus[active_set == 1] = np.inf #% don't consider active components
gamma_plus[gamma_plus <= 0] = np.inf #% or components outside the range [0, lambda_]
gamma_plus[gamma_plus> lambda_] = np.inf
gp_min, gp_min_ind = np.min(gamma_plus),np.argmin(gamma_plus)
if positive:
gm_min = np.inf; #% don't consider new directions that would grow negative
else:
gamma_minus[active_set == 1] = np.inf
gamma_minus[gamma_minus> lambda_] =np.inf
gamma_minus[gamma_minus <= 0] = np.inf
gm_min, gm_min_ind = np.min(gamma_minus),np.argmin(gamma_minus)
[g_min, which] = np.min(gp_min),np.argmin(gp_min)
if g_min == np.inf: #% if there are no possible new components, try move to the end
g_min = lambda_; #% This happens when all the components are already active or, if positive==1, when there are no new positive directions
#% LARS check (is g_min*avec too large?)
gamma_zero = -W[active_set == 1] / np.squeeze(avec);
gamma_zero_full = np.zeros((N,k));
gamma_zero_full[active_set == 1] = gamma_zero;
gamma_zero_full[gamma_zero_full <= 0] = np.inf;
gz_min, gz_min_ind = np.min(gamma_zero_full),np.argmin(gamma_zero_full)
if gz_min < g_min:
# print 'check_here'
if verbose:
print 'DROPPING active weight:' + str(gz_min_ind)
active_set[gz_min_ind] = 0;
dropped = gz_min_ind;
dropped_sign = np.sign(W[dropped]);
W[gz_min_ind] = 0;
avec = avec[gamma_zero != gz_min];
g_min = gz_min;
new=-1 # new = 0;
elif g_min < lambda_:
if which == 0:
new = gp_min_ind;
if verbose:
print 'new positive component:' + str(new)
else:
new = gm_min_ind;
print 'new negative component:' + str(new)
W[active_set == 1] = W[active_set == 1] + np.dot(g_min,np.squeeze(avec));
if positive:
if any(W<0):
#min(W);
flag = 1;
#%error('negative W component');
lambda_ = lambda_ - g_min;
#% Update weights and lambdas
lambdas.append(lambda_);
Ws.append(W.copy())
# print Ws
if len((Yp-np.dot(X,W)).shape)>2:
res = scipy.linalg.norm(np.squeeze(Yp-np.dot(X,W)),'fro')**2;
else:
res = scipy.linalg.norm(Yp-np.dot(X,W),'fro')**2;
#% Check finishing conditions
if lambda_ ==0 or (new>=0 and np.sum(active_set) == maxcomps) or (res < noise):
if verbose:
print 'end. \n'
break
#%
if new>=0:
active_set[new] = 1
i = i + 1
Ws_old=Ws
# end main loop
#%% final calculation of mus
Ws=np.asarray(np.swapaxes(np.swapaxes(Ws_old,0,1),1,2))
if flag == 0:
if i > 0:
Ws= np.squeeze(Ws[:,:,:len(lambdas)]);
w_dir = -(Ws[:,i] - Ws[:,i-1])/(lambdas[i]-lambdas[i-1]);
Aw = np.dot(X,w_dir);
y_res = np.squeeze(Yp) - np.dot(X,Ws[:,i-1] + w_dir*lambdas[i-1]);
ld = scipy.roots([scipy.linalg.norm(Aw)**2,-2*np.dot(Aw.T,y_res),np.dot(y_res.T,y_res)-noise]);
lam = ld[np.intersect1d(np.where(ld>lambdas[i]),np.where(ld<lambdas[i-1]))];
if len(lam) == 0 or np.any(lam)<0 or np.any(~np.isreal(lam)):
lam = np.array([lambdas[i]]);
W_lam = Ws[:,i-1] + np.dot(w_dir,lambdas[i-1]-lam[0]);
else:
warn('LARS REGRESSION NOT SOLVABLE, USING NN LEAST SQUARE')
W_lam=scipy.optimize.nnls(X,Yp)
# problem = picos.Problem(X,Yp)
# W_lam = problem.add_variable('W_lam', X.shape[1])
# problem.set_objective('min', 1|W_lam)
# problem.add_constraint(W_lam >= 0)
# problem.add_constraint(picos.norm(matrix(Yp.astype(np.float))-matrix(X.astype(np.float))*W_lam,2)<=np.sqrt(noise))
# sel_solver = []
# problem.solver_selection()
# problem.solve(verbose=True)
# cvx_begin quiet
# variable W_lam(size(X,2));
# minimize(sum(W_lam));
# subject to
# W_lam >= 0;
# norm(Yp-X*W_lam)<= sqrt(noise);
# cvx_end
lam = 10;
else:
W_lam = 0;
Ws = 0;
lambdas = 0;
lam = 0;
return Ws, lambdas, W_lam, lam, flag
def find_roots(wl, epsilon, alpha, beta):
"""Find roots of characteristic equation.
Given a wavelength, a 3x3 tensor epsilon and the tangential components
of the wavevector k = (alpha,beta,gamma_i), returns the 4 possible
gamma_i, i = 1,2,3,4 that satisfy the boundary conditions.
"""
omega = 2.0 * S.pi * c / wl
K = omega ** 2 * mu0 * epsilon
k0 = 2.0 * S.pi / wl
K /= k0 ** 2
alpha /= k0
beta /= k0
alpha2 = alpha ** 2
alpha3 = alpha ** 3
alpha4 = alpha ** 4
beta2 = beta ** 2
beta3 = beta ** 3
beta4 = beta ** 4
coeff = [
K[2, 2],
alpha * (K[0, 2] + K[2, 0]) + beta * (K[1, 2] + K[2, 1]),
alpha2 * (K[0, 0] + K[2, 2])
+ alpha * beta * (K[1, 0] + K[0, 1])
+ beta2 * (K[1, 1] + K[2, 2])
+ (K[0, 2] * K[2, 0] + K[1, 2] * K[2, 1] - K[0, 0] * K[2, 2] - K[1, 1] * K[2, 2]),
alpha3 * (K[0, 2] + K[2, 0])
+ beta3 * (K[1, 2] + K[2, 1])
+ alpha2 * beta * (K[1, 2] + K[2, 1])
+ alpha * beta2 * (K[0, 2] + K[2, 0])
+ alpha * (K[0, 1] * K[1, 2] + K[1, 0] * K[2, 1] - K[0, 2] * K[1, 1] - K[2, 0] * K[1, 1])
+ beta * (K[0, 1] * K[2, 0] + K[1, 0] * K[0, 2] - K[0, 0] * K[1, 2] - K[0, 0] * K[2, 1]),
alpha4 * (K[0, 0])
+ beta4 * (K[1, 1])
+ alpha3 * beta * (K[0, 1] + K[1, 0])
+ alpha * beta3 * (K[0, 1] + K[1, 0])
+ alpha2 * beta2 * (K[0, 0] + K[1, 1])
+ alpha2 * (K[0, 1] * K[1, 0] + K[0, 2] * K[2, 0] - K[0, 0] * K[2, 2] - K[0, 0] * K[1, 1])
+ beta2 * (K[0, 1] * K[1, 0] + K[1, 2] * K[2, 1] - K[0, 0] * K[1, 1] - K[1, 1] * K[2, 2])
+ alpha * beta * (K[0, 2] * K[2, 1] + K[2, 0] * K[1, 2] - K[0, 1] * K[2, 2] - K[1, 0] * K[2, 2])
+ K[0, 0] * K[1, 1] * K[2, 2]
- K[0, 0] * K[1, 2] * K[2, 1]
- K[1, 0] * K[0, 1] * K[2, 2]
+ K[1, 0] * K[0, 2] * K[2, 1]
+ K[2, 0] * K[0, 1] * K[1, 2]
- K[2, 0] * K[0, 2] * K[1, 1],
]
gamma = S.roots(coeff)
tmp = S.sort_complex(gamma)
gamma = tmp[[3, 0, 2, 1]] # convention
k = k0 * S.array([alpha * S.ones(gamma.shape), beta * S.ones(gamma.shape), gamma]).T
v = S.zeros((4, 3), dtype=complex)
for i, g in enumerate(gamma):
H = K + [
[-beta2 - g ** 2, alpha * beta, alpha * g],
[alpha * beta, -alpha2 - g ** 2, beta * g],
[alpha * g, beta * g, -alpha2 - beta2],
]
v[i, :] = [
(K[1, 1] - alpha2 - g ** 2) * (K[2, 2] - alpha2 - beta2) - (K[1, 2] + beta * g) ** 2,
(K[1, 2] + beta * g) * (K[2, 0] + alpha * g)
- (K[0, 1] + alpha * beta) * (K[2, 2] - alpha2 - beta2),
(K[0, 1] + alpha * beta) * (K[1, 2] + beta * g)
- (K[0, 2] + alpha * g) * (K[1, 1] - alpha2 - g ** 2),
]
p3 = v[0, :]
p3 /= norm(p3)
p4 = v[1, :]
p4 /= norm(p4)
p1 = S.cross(p3, k[0, :])
p1 /= norm(p1)
p2 = S.cross(p4, k[1, :])
p2 /= norm(p2)
p = S.array([p1, p2, p3, p4])
q = wl / (2.0 * S.pi * mu0 * c) * S.cross(k, p)
return k, p, q