Principal stresses and stress invariants

In this article we will discuss the derivation of the principal stresses and the stress invariants from the Cauchy stress tensor. The principal stresses and the stress invariants are important parameters that are used in failure criteria, plasticity, Mohr’s circle, etc.

For every point inside a body under static equilibrium there are at least three mutually perpendicular planes, called the principal planes, where the stress vector is normal to the plane and there is no shear component (see also: Calculation of normal and shear stress on a plane). The normal stresses acting on these planes are called principal stresses. From the mathematical point of view, the derivation of the principal stresses and their directions is a problem of determining the eigenvalues and the corresponding eigenvectors of a symmetric matrix. Fig. 1 illustrates the principal stresses and their directions for a point inside the body, compared to the initial coordinate system and the stress tensor.

Stress components on an element and the principal stresses on the element rotated to the principal directions
Figure 1: Principal stresses and their direction with respect to the initial coordinate system.

On a principal plane with unit normal nin_{i}, the stress vector is parallel to the normal, i.e. Ti=σniT_{i}=\sigma n_{i}, where σ\sigma is the normal stress on the plane. Combined with Cauchy’s formula, Ti=σijnjT_{i}=\sigma_{ij}n_{j}, this leads to the homogeneous system:

(σij−σδij)nj=0(1)(\sigma_{ij}-\sigma\delta_{ij})n_{j}=0 \tag{1}

where σij\sigma_{ij} is the stress tensor, σ\sigma is the principal stress (eigenvalue), δij\delta_{ij} is the Kronecker delta (with δij=1\delta_{ij}=1 if i=ji=j and δij=0\delta_{ij}=0 if i≠ji\neq j) and njn_{j} are the components of the unknown eigenvector. To obtain a non-trivial solution for the above system (1), the determinant of the coefficients must vanish:

∣σij−σδij∣=∣σ11−σσ12σ13σ21σ22−σσ23σ31σ32σ33−σ∣=0(2)|\sigma_{ij}-\sigma\delta_{ij}|=\left|\begin{array}{ccc}\sigma_{11}-\sigma & \sigma_{12} & \sigma_{13}\\ \sigma_{21} & \sigma_{22}-\sigma & \sigma_{23}\\ \sigma_{31} & \sigma_{32} & \sigma_{33}-\sigma\\ \end{array}\right|=0 \tag{2}

The above determinant leads to the so-called characteristic equation:

σ3−I1σ2+I2σ−I3=0(3)\sigma^{3}-I_{1}\sigma^{2}+I_{2}\sigma-I_{3}=0 \tag{3}

with (the expanded forms take into account the symmetry of the stress tensor, σij=σji\sigma_{ij}=\sigma_{ji}):

I1=σ11+σ22+σ33=σkkI2=∣σ22σ23σ32σ33∣+∣σ11σ13σ31σ33∣+∣σ11σ12σ21σ22∣=σ11σ22+σ22σ33+σ11σ33−σ122−σ232−σ132=12(σiiσjj−σijσji)I3=∣σ11σ12σ13σ21σ22σ23σ31σ32σ33∣=σ11σ22σ33+2σ12σ23σ31−σ122σ33−σ232σ11−σ132σ22(4)\begin{array}{rl}I_{1}=&\sigma_{11}+\sigma_{22}+\sigma_{33}\\=&\sigma_{kk}\\I_{2}=&\left|\begin{array}{cc}\sigma_{22} & \sigma_{23}\\ \sigma_{32} & \sigma_{33}\end{array}\right|+\left|\begin{array}{cc}\sigma_{11} & \sigma_{13}\\ \sigma_{31} & \sigma_{33}\end{array}\right|+\left|\begin{array}{cc}\sigma_{11} & \sigma_{12}\\ \sigma_{21} & \sigma_{22}\end{array}\right|\\=&\sigma_{11}\sigma_{22}+\sigma_{22}\sigma_{33}+\sigma_{11}\sigma_{33}-\sigma_{12}^2-\sigma_{23}^2-\sigma_{13}^2\\=&\frac{1}{2}\left(\sigma_{ii}\sigma_{jj}-\sigma_{ij}\sigma_{ji}\right)\\I_{3}=&\left|\begin{array}{ccc}\sigma_{11} & \sigma_{12} & \sigma_{13}\\\sigma_{21} & \sigma_{22} & \sigma_{23}\\ \sigma_{31} & \sigma_{32} & \sigma_{33}\end{array}\right|\\=&\sigma_{11}\sigma_{22}\sigma_{33}+2\sigma_{12}\sigma_{23}\sigma_{31}-\sigma_{12}^{2}\sigma_{33}-\sigma_{23}^{2}\sigma_{11}-\sigma_{13}^{2}\sigma_{22}\end{array} \tag{4}
I1=σ11+σ22+σ33=σkkI2=∣σ22σ23σ32σ33∣+∣σ11σ13σ31σ33∣+∣σ11σ12σ21σ22∣=σ11σ22+σ22σ33+σ11σ33−σ122−σ232−σ132=12(σiiσjj−σijσji)I3=∣σ11σ12σ13σ21σ22σ23σ31σ32σ33∣=σ11σ22σ33+2σ12σ23σ31−σ122σ33−σ232σ11−σ132σ22(4)\begin{array}{rl}I_{1}=&\sigma_{11}+\sigma_{22}+\sigma_{33}\\=&\sigma_{kk}\\I_{2}=&\left|\begin{array}{cc}\sigma_{22} & \sigma_{23}\\ \sigma_{32} & \sigma_{33}\end{array}\right|+\left|\begin{array}{cc}\sigma_{11} & \sigma_{13}\\ \sigma_{31} & \sigma_{33}\end{array}\right|\\&+\left|\begin{array}{cc}\sigma_{11} & \sigma_{12}\\ \sigma_{21} & \sigma_{22}\end{array}\right|\\=&\sigma_{11}\sigma_{22}+\sigma_{22}\sigma_{33}+\sigma_{11}\sigma_{33}\\&-\sigma_{12}^2-\sigma_{23}^2-\sigma_{13}^2\\=&\frac{1}{2}\left(\sigma_{ii}\sigma_{jj}-\sigma_{ij}\sigma_{ji}\right)\\I_{3}=&\left|\begin{array}{ccc}\sigma_{11} & \sigma_{12} & \sigma_{13}\\\sigma_{21} & \sigma_{22} & \sigma_{23}\\ \sigma_{31} & \sigma_{32} & \sigma_{33}\end{array}\right|\\=&\sigma_{11}\sigma_{22}\sigma_{33}+2\sigma_{12}\sigma_{23}\sigma_{31}\\&-\sigma_{12}^{2}\sigma_{33}-\sigma_{23}^{2}\sigma_{11}-\sigma_{13}^{2}\sigma_{22}\end{array} \tag{4}

The coefficients I1I_{1}, I2I_{2} and I3I_{3}, called the first, second and third stress invariants, respectively, are constant and do not depend on the orientation of the coordinate system. Since the stress tensor is symmetric, Eq. (3) has three real roots, σ1\sigma_{1}, σ2\sigma_{2} and σ3\sigma_{3}, which are the principal stresses or eigenvalues; they are usually ordered so that σ1≥σ2≥σ3\sigma_{1}\ge\sigma_{2}\ge\sigma_{3}. For each principal stress there is a non-trivial solution for njn_{j} in Eq. (1), which is the eigenvector, i.e. the direction cosines of the principal plane on which the principal stress acts. Each eigenvector is defined only up to its sign (±n⃗\pm\vec{n}). If the three principal stresses are distinct, the three principal directions are mutually perpendicular; if two of them are equal, every direction in the plane normal to the third principal direction is a principal direction, so there are infinitely many principal planes.

In the coordinate system of the principal directions the stress tensor is diagonal, and the invariants take the simple form:

I1=σ1+σ2+σ3I2=σ1σ2+σ2σ3+σ3σ1I3=σ1σ2σ3(5)\begin{array}{l}I_{1}=\sigma_{1}+\sigma_{2}+\sigma_{3}\\I_{2}=\sigma_{1}\sigma_{2}+\sigma_{2}\sigma_{3}+\sigma_{3}\sigma_{1}\\I_{3}=\sigma_{1}\sigma_{2}\sigma_{3}\end{array} \tag{5}

Example

Calculate the principal stresses and their directions for the following stress tensor:

σij=[531320104](6)\sigma_{ij}=\left[\begin{array}{ccc}5 & 3 & 1 \\ 3 & 2 & 0 \\ 1 & 0 & 4\end{array}\right] \tag{6}

Solution

From Eqs. (4) we evaluate the stress invariants:

I1=11I2=28I3=2(7)\begin{array}{l}I_{1}=11\\I_{2}=28\\I_{3}=2\end{array} \tag{7}

and we obtain the characteristic equation:

σ3−11σ2+28σ−2=0(8)\sigma^{3}-11\sigma^{2}+28\sigma-2=0 \tag{8}

The principal stresses are the roots of the polynomial (approximate solution):

σi=(7.091, 3.835, 0.074)(9)\sigma_{i}=\left(7.091,\ 3.835,\ 0.074\right) \tag{9}

As a check, Eqs. (5) give σ1+σ2+σ3=11\sigma_{1}+\sigma_{2}+\sigma_{3}=11, σ1σ2+σ2σ3+σ3σ1≈28\sigma_{1}\sigma_{2}+\sigma_{2}\sigma_{3}+\sigma_{3}\sigma_{1}\approx 28 and σ1σ2σ3≈2\sigma_{1}\sigma_{2}\sigma_{3}\approx 2, in agreement with the invariants calculated above.

For each principal stress we calculate the direction cosines of its principal plane from Eq. (1). For the principal stress σ1=7.091\sigma_{1}=7.091 we have the following system of equations:

(5−7.091)n1(1)+3n2(1)+n3(1)=03n1(1)+(2−7.091)n2(1)=0n1(1)+(4−7.091)n3(1)=0(10)\begin{array}{l}(5-7.091)n_{1}^{(1)}+3n_{2}^{(1)}+n_{3}^{(1)}=0\\3n_{1}^{(1)}+(2-7.091)n_{2}^{(1)}=0\\n_{1}^{(1)}+(4-7.091)n_{3}^{(1)}=0\end{array} \tag{10}

where the superscript in ni(1)n_{i}^{(1)} indicates that these direction cosines correspond to σ1\sigma_{1} (similarly, we will have ni(2)n_{i}^{(2)} for σ2\sigma_{2} and ni(3)n_{i}^{(3)} for σ3\sigma_{3}). According to linear algebra, if the three principal stresses are distinct, only two of the three equations (10) are independent. To solve the system, we also need the condition that n⃗(1)\vec{n}^{(1)} is a unit vector:

ni(1)ni(1)=(n1(1))2+(n2(1))2+(n3(1))2=1(11)n_{i}^{(1)}n_{i}^{(1)}=\left(n_{1}^{(1)}\right)^{2}+\left(n_{2}^{(1)}\right)^{2}+\left(n_{3}^{(1)}\right)^{2}=1 \tag{11}

Then, choosing for example the first two equations of (10) together with Eq. (11), the system of equations takes the form:

(5−7.091)n1(1)+3n2(1)+n3(1)=03n1(1)+(2−7.091)n2(1)=0(n1(1))2+(n2(1))2+(n3(1))2=1(12)\begin{array}{l}(5-7.091)n_{1}^{(1)}+3n_{2}^{(1)}+n_{3}^{(1)}=0\\3n_{1}^{(1)}+(2-7.091)n_{2}^{(1)}=0\\\left(n_{1}^{(1)}\right)^{2}+\left(n_{2}^{(1)}\right)^{2}+\left(n_{3}^{(1)}\right)^{2}=1\end{array} \tag{12}

By solving the above system we get the three direction cosines of σ1\sigma_{1}:

ni(1)=(0.83, 0.49, 0.27)(13)n_{i}^{(1)}=(0.83,\ 0.49,\ 0.27) \tag{13}

Alternatively, and more easily, the eigenvector can be found without solving a nonlinear system. Since n⃗(1)\vec{n}^{(1)} is perpendicular to every row of the matrix (σij−σ1δij)(\sigma_{ij}-\sigma_{1}\delta_{ij}), it is parallel to the cross product of any two independent rows. Using the permutation symbol εijk\varepsilon_{ijk}, the cross product of the first two rows is written in index notation as:

vi=εijk(σ1j−σ1δ1j)(σ2k−σ1δ2k),ni(1)=vivkvk(14)\begin{array}{rl}&v_{i}=\varepsilon_{ijk}\left(\sigma_{1j}-\sigma_{1}\delta_{1j}\right)\left(\sigma_{2k}-\sigma_{1}\delta_{2k}\right),\qquad n_{i}^{(1)}=\dfrac{v_{i}}{\sqrt{v_{k}v_{k}}}\end{array} \tag{14}
vi=εijk(σ1j−σ1δ1j)(σ2k−σ1δ2k),ni(1)=vivkvk(14)\begin{array}{rl}&v_{i}=\varepsilon_{ijk}\left(\sigma_{1j}-\sigma_{1}\delta_{1j}\right)\left(\sigma_{2k}-\sigma_{1}\delta_{2k}\right),\qquad\\& n_{i}^{(1)}=\dfrac{v_{i}}{\sqrt{v_{k}v_{k}}}\end{array} \tag{14}

For the first two rows of Eqs. (10), this gives:

(−2.091, 3, 1)×(3, −5.091, 0)=(5.091, 3, 1.645)(15)\begin{array}{rl}&(-2.091,\ 3,\ 1)\times(3,\ -5.091,\ 0)=(5.091,\ 3,\ 1.645)\end{array} \tag{15}
(−2.091, 3, 1)×(3, −5.091, 0)=(5.091, 3, 1.645)(15)\begin{array}{rl}&(-2.091,\ 3,\ 1)\times(3,\ -5.091,\ 0)\\&=(5.091,\ 3,\ 1.645)\end{array} \tag{15}

and normalizing this vector to unit length (its magnitude is ≈6.134\approx 6.134) gives again ni(1)=(0.83, 0.49, 0.27)n_{i}^{(1)}=(0.83,\ 0.49,\ 0.27).

Similarly, we derive the direction cosines for σ2\sigma_{2} and σ3\sigma_{3}:

ni(2)=(0.16, 0.26, −0.95)ni(3)=(0.54, −0.83, −0.14)(16)\begin{array}{l}n_{i}^{(2)}=(0.16,\ 0.26,\ -0.95)\\n_{i}^{(3)}=(0.54,\ -0.83,\ -0.14)\end{array} \tag{16}

Finally, we can verify that the three principal directions are mutually perpendicular: the dot products ni(1)ni(2)n_{i}^{(1)}n_{i}^{(2)}, ni(1)ni(3)n_{i}^{(1)}n_{i}^{(3)} and ni(2)ni(3)n_{i}^{(2)}n_{i}^{(3)} are all zero (up to rounding). The three unit vectors therefore form a new coordinate system, and the matrix whose rows are ni(1)n_{i}^{(1)}, ni(2)n_{i}^{(2)} and ni(3)n_{i}^{(3)} is the rotation matrix that transforms the stress tensor into this system (see Transformation of a tensor to a new coordinate system), where it becomes diagonal with the principal stresses on its diagonal.

Tags: algebraeigenvaluesinvariantsmechanicstensors