mirror of
https://github.com/JuliaFEM/JuliaFEM.jl.git
synced 2026-08-06 04:21:33 +00:00
test(verification): add symbolic Tet4 stiffness script
This commit is contained in:
Jukka Aho
@@ -0,0 +1,282 @@
|
|||||||
|
#!/usr/bin/env python3
|
||||||
|
"""
|
||||||
|
Symbolic computation of Tet4 element stiffness matrix using SymPy.
|
||||||
|
|
||||||
|
This computes the exact analytical stiffness matrix for a linear
|
||||||
|
tetrahedron element in 3D using symbolic integration.
|
||||||
|
|
||||||
|
This script validates the stiffness matrix values from the benchmark problem
|
||||||
|
in Professor Carlos A. Felippa's "Advanced Finite Element Method (AFEM)",
|
||||||
|
Chapter 15: The Linear Tetrahedron.
|
||||||
|
|
||||||
|
Reference:
|
||||||
|
- Felippa, C. A. "Advanced Finite Element Method (AFEM)", Chapter 15
|
||||||
|
University of Colorado Boulder - Center for Aerospace Structures
|
||||||
|
https://www.colorado.edu/engineering/CAS/courses.d/AFEM.d/
|
||||||
|
https://www.colorado.edu/engineering/CAS/courses.d/AFEM.d/AFEM.Ch15.pdf
|
||||||
|
|
||||||
|
The benchmark uses an arbitrary tetrahedral element (not reference element)
|
||||||
|
with nodes at:
|
||||||
|
- Node 1: (2, 3, 4)
|
||||||
|
- Node 2: (6, 3, 2)
|
||||||
|
- Node 3: (2, 5, 1)
|
||||||
|
- Node 4: (4, 3, 6)
|
||||||
|
|
||||||
|
Material: E = 96, ν = 1/3
|
||||||
|
|
||||||
|
This script:
|
||||||
|
1. Shows the reference element computation (for understanding)
|
||||||
|
2. Computes the stiffness matrix for the actual benchmark geometry
|
||||||
|
3. Validates against expected values from Felippa's chapter
|
||||||
|
|
||||||
|
The original chapter includes Mathematica verification modules for exact
|
||||||
|
computation. This Python script provides an independent verification using
|
||||||
|
SymPy symbolic integration.
|
||||||
|
"""
|
||||||
|
|
||||||
|
import sympy as sp
|
||||||
|
import numpy as np
|
||||||
|
from sympy import symbols, Matrix, simplify, integrate, lambdify
|
||||||
|
|
||||||
|
print("=" * 70)
|
||||||
|
print("Symbolic Tet4 Stiffness Matrix Computation")
|
||||||
|
print("=" * 70)
|
||||||
|
|
||||||
|
print("\n" + "=" * 70)
|
||||||
|
print("PART 1: Reference Element (for understanding)")
|
||||||
|
print("=" * 70)
|
||||||
|
|
||||||
|
# Define symbolic variables for reference coordinates
|
||||||
|
xi, eta, zeta = symbols("xi eta zeta", real=True, nonnegative=True)
|
||||||
|
|
||||||
|
# Material properties (symbolic)
|
||||||
|
E, nu = symbols("E nu", real=True, positive=True)
|
||||||
|
|
||||||
|
# Lamé parameters
|
||||||
|
lam = E * nu / ((1 + nu) * (1 - 2 * nu))
|
||||||
|
mu = E / (2 * (1 + nu))
|
||||||
|
|
||||||
|
print("\nMaterial properties:")
|
||||||
|
print(f" E = Young's modulus")
|
||||||
|
print(f" ν = Poisson's ratio")
|
||||||
|
print(f" λ = {lam}")
|
||||||
|
print(f" μ = {mu}")
|
||||||
|
|
||||||
|
# Shape functions for linear tetrahedron (reference element)
|
||||||
|
N1 = 1 - xi - eta - zeta
|
||||||
|
N2 = xi
|
||||||
|
N3 = eta
|
||||||
|
N4 = zeta
|
||||||
|
|
||||||
|
N = [N1, N2, N3, N4]
|
||||||
|
|
||||||
|
print("\nShape functions (reference element):")
|
||||||
|
for i, Ni in enumerate(N, 1):
|
||||||
|
print(f" N{i} = {Ni}")
|
||||||
|
|
||||||
|
print("\n" + "=" * 70)
|
||||||
|
print("PART 2: Actual Geometry from OLD API Test")
|
||||||
|
print("=" * 70)
|
||||||
|
|
||||||
|
# Actual node coordinates from OLD API test
|
||||||
|
X1 = sp.Matrix([2.0, 3.0, 4.0])
|
||||||
|
X2 = sp.Matrix([6.0, 3.0, 2.0])
|
||||||
|
X3 = sp.Matrix([2.0, 5.0, 1.0])
|
||||||
|
X4 = sp.Matrix([4.0, 3.0, 6.0])
|
||||||
|
|
||||||
|
print("\nNode coordinates:")
|
||||||
|
print(f" Node 1: {X1.T}")
|
||||||
|
print(f" Node 2: {X2.T}")
|
||||||
|
print(f" Node 3: {X3.T}")
|
||||||
|
print(f" Node 4: {X4.T}")
|
||||||
|
|
||||||
|
# Physical coordinates as function of reference coordinates
|
||||||
|
x = N1 * X1[0] + N2 * X2[0] + N3 * X3[0] + N4 * X4[0]
|
||||||
|
y = N1 * X1[1] + N2 * X2[1] + N3 * X3[1] + N4 * X4[1]
|
||||||
|
z = N1 * X1[2] + N2 * X2[2] + N3 * X3[2] + N4 * X4[2]
|
||||||
|
|
||||||
|
print("\nPhysical coordinates (x, y, z) as functions of (ξ, η, ζ):")
|
||||||
|
print(f" x = {x}")
|
||||||
|
print(f" y = {y}")
|
||||||
|
print(f" z = {z}")
|
||||||
|
|
||||||
|
# Jacobian matrix: J[i,j] = ∂x_i/∂ξ_j
|
||||||
|
J = sp.Matrix(
|
||||||
|
[
|
||||||
|
[sp.diff(x, xi), sp.diff(x, eta), sp.diff(x, zeta)],
|
||||||
|
[sp.diff(y, xi), sp.diff(y, eta), sp.diff(y, zeta)],
|
||||||
|
[sp.diff(z, xi), sp.diff(z, eta), sp.diff(z, zeta)],
|
||||||
|
]
|
||||||
|
)
|
||||||
|
|
||||||
|
print("\nJacobian matrix J = ∂(x,y,z)/∂(ξ,η,ζ):")
|
||||||
|
print(J)
|
||||||
|
|
||||||
|
detJ = J.det()
|
||||||
|
print(f"\ndet(J) = {detJ}")
|
||||||
|
print(f"Element volume = det(J) / 6 = {detJ / 6}")
|
||||||
|
|
||||||
|
# Inverse Jacobian
|
||||||
|
J_inv = J.inv()
|
||||||
|
print("\nInverse Jacobian J^(-1):")
|
||||||
|
print(sp.simplify(J_inv))
|
||||||
|
|
||||||
|
# Shape function derivatives in physical coordinates
|
||||||
|
# ∂N/∂x = J^(-T) · ∂N/∂ξ
|
||||||
|
dN_dxi_ref = sp.Matrix([-1, 1, 0, 0])
|
||||||
|
dN_deta_ref = sp.Matrix([-1, 0, 1, 0])
|
||||||
|
dN_dzeta_ref = sp.Matrix([-1, 0, 0, 1])
|
||||||
|
|
||||||
|
# For each node
|
||||||
|
dN_dx_list = []
|
||||||
|
dN_dy_list = []
|
||||||
|
dN_dz_list = []
|
||||||
|
|
||||||
|
for i in range(4):
|
||||||
|
dN_dref = sp.Matrix([dN_dxi_ref[i], dN_deta_ref[i], dN_dzeta_ref[i]])
|
||||||
|
dN_dphys = J_inv.T * dN_dref
|
||||||
|
dN_dx_list.append(dN_dphys[0])
|
||||||
|
dN_dy_list.append(dN_dphys[1])
|
||||||
|
dN_dz_list.append(dN_dphys[2])
|
||||||
|
|
||||||
|
print("\nShape function derivatives in physical coordinates:")
|
||||||
|
print(f" ∂N/∂x = {dN_dx_list}")
|
||||||
|
print(f" ∂N/∂y = {dN_dy_list}")
|
||||||
|
print(f" ∂N/∂z = {dN_dz_list}")
|
||||||
|
|
||||||
|
# Build B-matrix
|
||||||
|
B = sp.zeros(6, 12)
|
||||||
|
|
||||||
|
for i in range(4):
|
||||||
|
dN_dx = dN_dx_list[i]
|
||||||
|
dN_dy = dN_dy_list[i]
|
||||||
|
dN_dz = dN_dz_list[i]
|
||||||
|
|
||||||
|
# Node i, DOF u_x (column 3*i)
|
||||||
|
B[0, 3 * i] = dN_dx # ε_xx
|
||||||
|
B[3, 3 * i] = dN_dy # γ_xy
|
||||||
|
B[5, 3 * i] = dN_dz # γ_xz
|
||||||
|
|
||||||
|
# Node i, DOF u_y (column 3*i+1)
|
||||||
|
B[1, 3 * i + 1] = dN_dy # ε_yy
|
||||||
|
B[3, 3 * i + 1] = dN_dx # γ_xy
|
||||||
|
B[4, 3 * i + 1] = dN_dz # γ_yz
|
||||||
|
|
||||||
|
# Node i, DOF u_z (column 3*i+2)
|
||||||
|
B[2, 3 * i + 2] = dN_dz # ε_zz
|
||||||
|
B[4, 3 * i + 2] = dN_dy # γ_yz
|
||||||
|
B[5, 3 * i + 2] = dN_dx # γ_xz
|
||||||
|
|
||||||
|
print("\nB-matrix (6×12) constructed for actual geometry")
|
||||||
|
|
||||||
|
# Material stiffness matrix D (6×6)
|
||||||
|
D = sp.zeros(6, 6)
|
||||||
|
D[0, 0] = D[1, 1] = D[2, 2] = 2 * mu + lam
|
||||||
|
D[3, 3] = D[4, 4] = D[5, 5] = mu
|
||||||
|
D[0, 1] = D[1, 0] = D[1, 2] = D[2, 1] = D[0, 2] = D[2, 0] = lam
|
||||||
|
|
||||||
|
# Stiffness matrix: K = ∫_Ω B^T D B dV
|
||||||
|
# For the actual element, we need to account for the Jacobian
|
||||||
|
# K = ∫ B^T D B × det(J) dV_ref
|
||||||
|
# Since B and det(J) are CONSTANT for Tet4:
|
||||||
|
# K = B^T D B × det(J) × (1/6)
|
||||||
|
# where 1/6 is the volume of the reference element
|
||||||
|
|
||||||
|
print("\nComputing stiffness matrix K = ∫ B^T D B dV...")
|
||||||
|
print(" For Tet4: B and Jacobian are CONSTANT")
|
||||||
|
print(f" K = B^T D B × det(J) × (1/6)")
|
||||||
|
print(f" det(J) = {detJ}")
|
||||||
|
print(f" Element volume = det(J)/6 = {detJ/6}")
|
||||||
|
|
||||||
|
K_symbolic = B.T * D * B
|
||||||
|
K = K_symbolic * detJ * sp.Rational(1, 6)
|
||||||
|
|
||||||
|
print("\nStiffness matrix computed!")
|
||||||
|
print(f" K is {K.shape[0]}×{K.shape[1]} symbolic matrix")
|
||||||
|
|
||||||
|
# Simplify (this should be fast since entries are already simple)
|
||||||
|
K = sp.simplify(K)
|
||||||
|
|
||||||
|
print("\nSimplifying expressions...")
|
||||||
|
|
||||||
|
# Substitute numerical values for verification
|
||||||
|
# Use the same values as OLD API test: E = 96, ν = 1/3
|
||||||
|
E_val = 96.0
|
||||||
|
nu_val = 1.0 / 3.0
|
||||||
|
|
||||||
|
print("\nSubstituting numerical values:")
|
||||||
|
print(f" E = {E_val}")
|
||||||
|
print(f" ν = {nu_val}")
|
||||||
|
|
||||||
|
# Compute Lamé parameters
|
||||||
|
lam_val = E_val * nu_val / ((1 + nu_val) * (1 - 2 * nu_val))
|
||||||
|
mu_val = E_val / (2 * (1 + nu_val))
|
||||||
|
|
||||||
|
print(f" λ = {lam_val}")
|
||||||
|
print(f" μ = {mu_val}")
|
||||||
|
|
||||||
|
# Substitute into K
|
||||||
|
K_numerical = K.subs([(E, E_val), (nu, nu_val)])
|
||||||
|
|
||||||
|
# Convert to float matrix
|
||||||
|
K_float = np.array(K_numerical).astype(np.float64)
|
||||||
|
|
||||||
|
print("\n" + "=" * 70)
|
||||||
|
print("NUMERICAL STIFFNESS MATRIX (E=96, ν=1/3)")
|
||||||
|
print("=" * 70)
|
||||||
|
print("\nK =")
|
||||||
|
print(K_float)
|
||||||
|
|
||||||
|
print("\n" + "=" * 70)
|
||||||
|
print("COMPARISON WITH OLD API TEST VALUES")
|
||||||
|
print("=" * 70)
|
||||||
|
|
||||||
|
# Expected values from OLD API test
|
||||||
|
K_expected = np.array(
|
||||||
|
[
|
||||||
|
[149, 108, 24, -1, 6, 12, -54, -48, 0, -94, -66, -36],
|
||||||
|
[108, 344, 54, -24, 104, 42, -24, -216, -12, -60, -232, -84],
|
||||||
|
[24, 54, 113, 0, 30, 35, 0, -24, -54, -24, -60, -94],
|
||||||
|
[-1, -24, 0, 29, -18, -12, -18, 24, 0, -10, 18, 12],
|
||||||
|
[6, 104, 30, -18, 44, 18, 12, -72, -12, 0, -76, -36],
|
||||||
|
[12, 42, 35, -12, 18, 29, 0, -24, -18, 0, -36, -46],
|
||||||
|
[-54, -24, 0, -18, 12, 0, 36, 0, 0, 36, 12, 0],
|
||||||
|
[-48, -216, -24, 24, -72, -24, 0, 144, 0, 24, 144, 48],
|
||||||
|
[0, -12, -54, 0, -12, -18, 0, 0, 36, 0, 24, 36],
|
||||||
|
[-94, -60, -24, -10, 0, 0, 36, 24, 0, 68, 36, 24],
|
||||||
|
[-66, -232, -60, 18, -76, -36, 12, 144, 24, 36, 164, 72],
|
||||||
|
[-36, -84, -94, 12, -36, -46, 0, 48, 36, 24, 72, 104],
|
||||||
|
]
|
||||||
|
)
|
||||||
|
|
||||||
|
print("\nExpected K (from OLD API test) =")
|
||||||
|
print(K_expected)
|
||||||
|
|
||||||
|
print("\nDifference (Symbolic - Expected) =")
|
||||||
|
diff = K_float - K_expected
|
||||||
|
print(diff)
|
||||||
|
|
||||||
|
print("\nMax absolute difference:", np.max(np.abs(diff)))
|
||||||
|
print("Max relative error:", np.max(np.abs(diff / (K_expected + 1e-10))))
|
||||||
|
|
||||||
|
if np.allclose(K_float, K_expected, rtol=1e-6):
|
||||||
|
print("\n✅ MATCH! Symbolic computation agrees with OLD API test values!")
|
||||||
|
else:
|
||||||
|
print("\n❌ MISMATCH! Symbolic computation differs from OLD API test values!")
|
||||||
|
print(" This means either:")
|
||||||
|
print(" 1. The OLD API test uses different geometry")
|
||||||
|
print(" 2. The OLD API has a bug")
|
||||||
|
print(" 3. This symbolic computation has an error")
|
||||||
|
|
||||||
|
print("\n" + "=" * 70)
|
||||||
|
print("SUMMARY")
|
||||||
|
print("=" * 70)
|
||||||
|
print("\nFor reference Tet4 element with E=96, ν=1/3:")
|
||||||
|
print(f" Volume = 1/6")
|
||||||
|
print(f" λ = {lam_val}")
|
||||||
|
print(f" μ = {mu_val}")
|
||||||
|
print(f" Stiffness matrix K is 12×12 symmetric")
|
||||||
|
print(f" All entries are rational multiples of E")
|
||||||
|
print(f" K_symbolic available for arbitrary E, ν")
|
||||||
|
|
||||||
|
print("\n" + "=" * 70)
|
||||||
Reference in New Issue
Block a user