def test_FuzzyAnalysis(self):
from scipy.optimize import rosen, rosen_der
def SysEq2(p, x):
return rosen(p)
def SensEq2(p, r, g, x):
return rosen_der(p)
pFuzz1 = pu.UncertainNumber([1, 2, 3, 4], Form='trapazoid', nalpha=3)
pFuzz2 = pu.UncertainNumber([1, 2, 3, 4], Form='trapazoid', nalpha=3)
pFuzz = [pFuzz1, pFuzz2]
Prob = pu.UncertainAnalysis(SysEq2, pUnc=pFuzz, SensEq=SensEq2)
Prob.Alg = 'NLPQLP'
Prob.nAlpha = 3
Prob.paraNorm = 0
Prob.epsStop = 1e-6
Prob.para = 1
Prob.calculate()
rUncTarget = np.array([
[1.01000000e02, 4.90400000e03],
[2.50000000e-01, 1.15625000e04],
[0, 2.25090000e04],
])
# self.assertAlmostEqual(Prob.rUnc.Value.tolist(),
# rUncTarget.tolist())
self.assertAlmostEqual(Prob.rUnc.Value.tolist()[0],
rUncTarget.tolist()[0])
self.assertAlmostEqual(Prob.rUnc.Value.tolist()[1],
rUncTarget.tolist()[1])
self.assertAlmostEqual(Prob.rUnc.Value.tolist()[2][0],
rUncTarget.tolist()[2][0])
self.assertAlmostEqual(Prob.rUnc.Value.tolist()[2][1],
rUncTarget.tolist()[2][1])
def test_IntervalAnalysis(self):
pInt = pu.UncertainNumber([1, 5])
def SysEq1(p, x):
return p - p
UncertainProblem = pu.UncertainAnalysis()
UncertainProblem.pUnc = pInt
UncertainProblem.SysEq = SysEq1
UncertainProblem.calculate()
rUncTarget = np.array([[0.0, 0.0]])
self.assertAlmostEqual(UncertainProblem.rUnc.Value.tolist(),
rUncTarget.tolist())
def FunHockettSherby(pUnc):
Prob = pu.UncertainAnalysis(HockettSherby, pUnc)
Prob.deltax = 1e-3
Prob.epsStop = 1e-3
nS = 251
epsilonMax = 0.5
epsilon = np.linspace(0, epsilonMax, nS)
rFnUnc = [[]] * nS
nEvaluation = 0
for i, val in enumerate(epsilon):
Prob.para = val
Prob.calculate()
rFnUnc[i] = Prob.rUnc
nEvaluation += Prob.nEval
Prob.rFnUnc = rFnUnc
Prob.nEval = nEvaluation
Prob.epsilon = epsilon
return Prob
ux2 = 0.001
# Function of estimated damping ratio based on two measures
def DampEst(p, x):
x1 = p[0]
x2 = p[1]
Lambda = np.log(x1 / x2)
zeta = np.sqrt(Lambda**2 / (4 * np.pi**2 + Lambda**2))
return zeta
# Interval uncertainty problem: Worst-case interval defined as +/-6sigma
x1Unc = pu.UncertainNumber([x1 - 6 * ux1, x1 + 6 * ux1])
x2Unc = pu.UncertainNumber([x2 - 6 * ux2, x2 + 6 * ux2])
Prob = pu.UncertainAnalysis(DampEst, pUnc=[x1Unc, x2Unc])
Prob.nr = 1
Prob.deltax = 1e-6
Prob.epsStop = 1e-6
Prob.Alg = 'NLPQLP'
# Calculate uncertain response
Prob.calculate()
# Calculation with standard uncertainty using first-order Taylor series
zeta = DampEst([x1, x2], [])
dzetadx1 = 1 / (2 * np.pi * x1)
dzetadx2 = 1 / (2 * np.pi * x2)
uzeta = (np.sqrt(ux1**2 * dzetadx1**2 + ux2**2 * dzetadx2**2) / 2 / np.pi)
zeta1sigma = [zeta - uzeta, zeta + uzeta]
zeta2sigma = [zeta - 2 * uzeta, zeta + 2 * uzeta]
def test_Robustness(self):
UncertainProblem = pu.UncertainAnalysis()
UncertainProblem.pUnc = pu.UncertainNumber([1, 2])
UncertainProblem.rUnc = pu.UncertainNumber([3, 6])
UncertainProblem.calcRobustness()
self.assertAlmostEqual(UncertainProblem.SystemRobustness, 1 / 3)
import numpy as np
def Eigenfrequency1DoF(p, x):
m = p[0]
k = p[1]
omega0 = np.sqrt(k / m)
f0 = omega0 / 2 / np.pi
return f0
m = pu.UncertainNumber([2.0, 2.5])
k = pu.UncertainNumber([40000, 60000])
pUnc = [m, k]
Prob = pu.UncertainAnalysis(Eigenfrequency1DoF, pUnc)
Prob.deltax = 1e-3
Prob.epsStop = 1e-3
Prob.calculate()
m.printValue()
k.printValue()
plt, _ = pu.plotIntervals([m.Value, k.Value],
labels=['mass $m$ [kg]', 'stiffness $k$ [N/mm]'])
plt.show()
f0Unc = Prob.rUnc
f0Unc.printValue()
f0Unc.plotValue(color='r', xlabel='eigenfrequency $f_0$ [Hz]')
PyUncertainties = True
except:
PyUncertainties = False
print("Package 'uncertainties' not installed!")
def SysEq(p, x):
return p - p
nAlpha = 1
pL = 1.0
pU = 5.0
pInt = pu.UncertainNumber([1, 5])
Prob = pu.UncertainAnalysis()
Prob.SysEq = SysEq
Prob.pUnc = pInt
Prob.calculate()
if IntPy:
pInt1 = intpy.IReal(pL, pU)
rInt1 = pInt1 - pInt1
if PyInterval:
pInt2 = interval.interval(pL, pU)
rInt2 = pInt2 - pInt2
if PyUncertainties:
pMean = (pU - pL) / 2.0
pDelta = pMean - pL
pUnc = uncertainties.ufloat(pMean, pDelta)
rUnc = pUnc - pUnc
import numpy as np
import pyUngewiss as pu
def HockettSherby(p, x):
sigma = p[0] + p[1] - p[1] * np.exp(-p[2] * x**p[3])
return [sigma]
sigmaY = pu.UncertainNumber([240, 260])
sigmaP = pu.UncertainNumber([40, 60])
cHS = pu.UncertainNumber([8, 12])
nHS = pu.UncertainNumber([0.7, 0.8])
pUnc = [sigmaY, sigmaP, cHS, nHS]
Prob = pu.UncertainAnalysis(HockettSherby, pUnc)
Prob.deltax = 1e-3
Prob.epsStop = 1e-3
nS = 100
epsilonMax = 0.5
epsilon = np.linspace(0, epsilonMax, nS)
rFnInt = [[]] * nS
nEvaluation = 0
for i, val in enumerate(epsilon):
Prob.para = val
Prob.calculate()
rFnInt[i] = Prob.rUnc
nEvaluation += Prob.nEval
pu.plotUncertainFn(
rFnInt,
import pyUngewiss as pu
import numpy as np
def SysEq(p, x):
return (x + p)**2
def SensEq(p, r, g, x):
return 2 * x + 2 * p
nAlpha = 11
pFuzz = pu.UncertainNumber([8, 9, 11, 12], Form='trapazoid', nalpha=nAlpha)
x = 1.0
ProbFD = pu.UncertainAnalysis(SysEq, pUnc=pFuzz)
ProbFD.Alg = 'NLPQLP'
ProbFD.nAlpha = nAlpha
ProbFD.deltax = (1e-6, )
ProbFD.paraNorm = 0
ProbFD.SBFA = False
ProbFD.Surr = (False, )
ProbFD.epsStop = 1e-6
ProbFD.para = x
ProbFD.SensCalc = 'FD'
ProbFD.calculate()
ProbAS = pu.UncertainAnalysis(SysEq, pUnc=pFuzz, SensEq=SensEq)
ProbAS.Alg = 'NLPQLP'
ProbAS.nAlpha = nAlpha
ProbAS.paraNorm = 0
ProbAS.SBFA = False
import numpy as np
import pyUngewiss as pu
def SysEq(p, x):
y = x**2 + p
return y
pInt = pu.UncertainNumber([-10, 10])
nS = 200
x = np.linspace(-10, 10, nS)
rFnInt = [[]] * nS
Prob = pu.UncertainAnalysis(SysEq, pInt)
Prob.deltax = 1e-6
Prob.epsStop = 1e-6
for ii in range(len(x)):
Prob.para = x[ii]
Prob.calculate()
rFnInt[ii] = Prob.rUnc
plt, _ = pu.plotUncertainFn(rFnInt, x)
plt.show()
