Skip to main content

Chapter 8 · 6 hours

Vibration of Multi Degree of Freedom Systems

Practice questions

Practice questions and answers

3 exam-style questions on this chapter, written for this site from the official syllabus. We haven’t found past IOE papers for this subject yet; if you have some, share them in the community.

  • Practice · 6 marks

For a three-degree-of-freedom spring-mass system, write the equations of motion in matrix form. Define the stiffness and flexibility influence coefficients, show that [a]=[k]−1[a] = [k]^{-1}, and state and prove Maxwell's reciprocity theorem.

Answer

Equations of motion

A system of three masses m1,m2,m3m_1, m_2, m_3 in a line, connected by springs k1k_1 (wall to m1m_1), k2k_2 and k3k_3 in series, has

[m1000m2000m3]{x¨1x¨2x¨3}+[k1+k2−k20−k2k2+k3−k30−k3k3]{x1x2x3}={F1F2F3}\begin{bmatrix} m_1 & 0 & 0 \\ 0 & m_2 & 0 \\ 0 & 0 & m_3 \end{bmatrix}\begin{Bmatrix} \ddot{x}_1 \\ \ddot{x}_2 \\ \ddot{x}_3 \end{Bmatrix} + \begin{bmatrix} k_1+k_2 & -k_2 & 0 \\ -k_2 & k_2+k_3 & -k_3 \\ 0 & -k_3 & k_3 \end{bmatrix}\begin{Bmatrix} x_1 \\ x_2 \\ x_3 \end{Bmatrix} = \begin{Bmatrix} F_1 \\ F_2 \\ F_3 \end{Bmatrix}

i.e. [M]{x¨}+[K]{x}={F}[M]\{\ddot{x}\} + [K]\{x\} = \{F\}.

Influence coefficients

  • Stiffness influence coefficient kijk_{ij}: the force at coordinate ii required to produce a unit displacement at coordinate jj, with all other displacements held at zero.
  • Flexibility influence coefficient aija_{ij}: the displacement at coordinate ii due to a unit force at coordinate jj (other forces zero).

Therefore {F}=[k]{x}\{F\} = [k]\{x\} and {x}=[a]{F}\{x\} = [a]\{F\}.

Relation between them

Substituting {x}=[a]{F}\{x\} = [a]\{F\} in {F}=[k]{x}\{F\} = [k]\{x\} gives {F}=[k][a]{F}\{F\} = [k][a]\{F\} for every {F}\{F\}, so

[k][a]=[I]  ⇒  [a]=[k]−1[k][a] = [I] \;\Rightarrow\; [a] = [k]^{-1}

The flexibility matrix exists only if the system is restrained (no rigid-body motion). For an unrestrained system [k][k] is singular.

Maxwell's reciprocity theorem

Statement: The displacement at point ii due to a unit load at point jj equals the displacement at jj due to a unit load at ii: aij=ajia_{ij} = a_{ji} (so [a][a] and [k][k] are symmetric).

Proof: Apply load PiP_i first and then PjP_j. The work done (strain energy stored) is

U1=12Pi(aiiPi)+12Pj(ajjPj)+Pj(ajiPi)U_1 = \tfrac12P_i(a_{ii}P_i) + \tfrac12P_j(a_{jj}P_j) + P_j(a_{ji}P_i)

(the last term: PjP_j moves through the extra deflection at jj caused by PiP_i.) Now apply the loads in the reverse order:

U2=12Pj(ajjPj)+12Pi(aiiPi)+Pi(aijPj)U_2 = \tfrac12P_j(a_{jj}P_j) + \tfrac12P_i(a_{ii}P_i) + P_i(a_{ij}P_j)

The final state is the same and the energy stored in a linear elastic system does not depend on the order of loading, so U1=U2U_1 = U_2, giving aji=aija_{ji} = a_{ij}.

  • Practice · 8 marks

Three equal masses of 10 kg are connected in series to a fixed wall by three identical springs of stiffness 100 kN/m each (wall - spring - m1m_1 - spring - m2m_2 - spring - m3m_3, the last mass free). Set up the eigenvalue problem, find the natural frequencies and the mode shapes (normalised to X1=1X_1 = 1), and verify the orthogonality of the first two modes with respect to the mass matrix.

Answer

Data: m1=m2=m3=10m_1 = m_2 = m_3 = 10 kg, k=100 000k = 100\,000 N/m.

Mass and stiffness matrices

[M]=[100001000010],[K]=105[2−10−12−10−11] N/m[M] = \begin{bmatrix} 10 & 0 & 0 \\ 0 & 10 & 0 \\ 0 & 0 & 10 \end{bmatrix}, \qquad [K] = 10^5\begin{bmatrix} 2 & -1 & 0 \\ -1 & 2 & -1 \\ 0 & -1 & 1 \end{bmatrix}\ \text{N/m}

Eigenvalue problem

([K]−ω2[M]){X}={0}\left([K] - \omega^2[M]\right)\{X\} = \{0\}

Let λ=mω2k=10ω2105\lambda = \dfrac{m\omega^2}{k} = \dfrac{10\omega^2}{10^5}. The characteristic equation det⁡([K]−ω2[M])=0\det\left([K] - \omega^2[M]\right) = 0 becomes

∣2−λ−10−12−λ−10−11−λ∣=0\begin{vmatrix} 2-\lambda & -1 & 0 \\ -1 & 2-\lambda & -1 \\ 0 & -1 & 1-\lambda \end{vmatrix} = 0 (2−λ)[(2−λ)(1−λ)−1]−(1−λ)=0  ⇒  λ3−5λ2+6λ−1=0(2-\lambda)\left[(2-\lambda)(1-\lambda) - 1\right] - (1-\lambda) = 0 \;\Rightarrow\; \lambda^3 - 5\lambda^2 + 6\lambda - 1 = 0

Solving the cubic (Newton or trial):

λ1=0.1981,λ2=1.5550,λ3=3.2470\lambda_1 = 0.1981,\quad \lambda_2 = 1.5550,\quad \lambda_3 = 3.2470

Natural frequencies

ω=λ km=10000λ\omega = \sqrt{\lambda\,\frac{k}{m}} = \sqrt{10000\lambda}
Modeλ\lambdaω\omega (rad/s)ff (Hz)
10.198144.57.08
21.5550124.719.85
33.2470180.228.68

Mode shapes

From the first row, X2=(2−λ)X1X_2 = (2-\lambda)X_1; from the second row, X3=(2−λ)X2−X1X_3 = (2-\lambda)X_2 - X_1.

ModeX1X_1X2X_2X3X_3
111.8022.247
210.445-0.802
31-1.2470.555
 Mode 1: all in phase, no node
 Mode 2: one node (between m2 and m3)
 Mode 3: two nodes

Orthogonality check (modes 1 and 2)

{X}1T[M]{X}2=10[(1)(1)+(1.802)(0.445)+(2.247)(−0.802)]=0.000≈0  ✓\{X\}_1^T[M]\{X\}_2 = 10\left[(1)(1) + (1.802)(0.445) + (2.247)(-0.802)\right] = 0.000 \approx 0\ \ \checkmark

Answer: ω1=44.5\omega_1 = 44.5, ω2=124.7\omega_2 = 124.7, ω3=180.2\omega_3 = 180.2 rad/s; mode shapes as tabulated; orthogonality verified.

  • Practice · 6 marks

Prove that the natural modes of an undamped multi-degree-of-freedom system are orthogonal with respect to the mass and stiffness matrices. Explain how modal analysis uses this property to find the response to a general forcing function.

Answer

Orthogonality

Let ωr,{ϕ}r\omega_r, \{\phi\}_r and ωs,{ϕ}s\omega_s, \{\phi\}_s be two natural frequencies with their mode shapes, with ωr≠ωs\omega_r \ne \omega_s. Then

[K]{ϕ}r=ωr2[M]{ϕ}r(1),[K]{ϕ}s=ωs2[M]{ϕ}s(2)[K]\{\phi\}_r = \omega_r^2[M]\{\phi\}_r \quad (1), \qquad [K]\{\phi\}_s = \omega_s^2[M]\{\phi\}_s \quad (2)

Pre-multiply (1) by {ϕ}sT\{\phi\}_s^T and (2) by {ϕ}rT\{\phi\}_r^T:

{ϕ}sT[K]{ϕ}r=ωr2{ϕ}sT[M]{ϕ}r,{ϕ}rT[K]{ϕ}s=ωs2{ϕ}rT[M]{ϕ}s\{\phi\}_s^T[K]\{\phi\}_r = \omega_r^2\{\phi\}_s^T[M]\{\phi\}_r, \qquad \{\phi\}_r^T[K]\{\phi\}_s = \omega_s^2\{\phi\}_r^T[M]\{\phi\}_s

Since [K][K] and [M][M] are symmetric, the left-hand sides are equal (transpose of a scalar) and so are the mass products. Subtracting:

(ωr2−ωs2){ϕ}rT[M]{ϕ}s=0  ⇒  {ϕ}rT[M]{ϕ}s=0(r≠s)(\omega_r^2 - \omega_s^2)\{\phi\}_r^T[M]\{\phi\}_s = 0 \;\Rightarrow\; \{\phi\}_r^T[M]\{\phi\}_s = 0 \quad (r \ne s)

and hence also {ϕ}rT[K]{ϕ}s=0\{\phi\}_r^T[K]\{\phi\}_s = 0 for r≠sr \ne s.

For r=sr = s the products are the modal mass Mr={ϕ}rT[M]{ϕ}rM_r = \{\phi\}_r^T[M]\{\phi\}_r and modal stiffness Kr={ϕ}rT[K]{ϕ}r=ωr2MrK_r = \{\phi\}_r^T[K]\{\phi\}_r = \omega_r^2M_r. Normalising so that Mr=1M_r = 1 (mass-normalised) gives [Φ]T[M][Φ]=[I][\Phi]^T[M][\Phi] = [I] and [Φ]T[K][Φ]=diag(ωr2)[\Phi]^T[K][\Phi] = \text{diag}(\omega_r^2).

Modal analysis for forced response

For [M]{x¨}+[K]{x}={F(t)}[M]\{\ddot{x}\} + [K]\{x\} = \{F(t)\}:

  1. Find the natural frequencies ωr\omega_r and the mode-shape matrix [Φ][\Phi] (columns normalised to unit modal mass).
  2. Use the transformation {x}=[Φ]{q}\{x\} = [\Phi]\{q\}, where {q}\{q\} are the principal coordinates.
  3. Substitute and pre-multiply by [Φ]T[\Phi]^T. By orthogonality the equations uncouple:
q¨r+ωr2qr=Qr(t),Qr(t)={ϕ}rT{F(t)}\ddot{q}_r + \omega_r^2q_r = Q_r(t), \qquad Q_r(t) = \{\phi\}_r^T\{F(t)\}
  1. Solve each single-degree-of-freedom equation, for example with the Duhamel (convolution) integral:
qr(t)=qr(0)cos⁡ωrt+q˙r(0)ωrsin⁡ωrt+1ωr∫0tQr(τ)sin⁡ωr(t−τ) dτq_r(t) = q_r(0)\cos\omega_rt + \frac{\dot{q}_r(0)}{\omega_r}\sin\omega_rt + \frac{1}{\omega_r}\int_0^tQ_r(\tau)\sin\omega_r(t-\tau)\,d\tau
  1. Transform back: {x(t)}=[Φ]{q(t)}=∑r{ϕ}rqr(t)\{x(t)\} = [\Phi]\{q(t)\} = \sum_r\{\phi\}_rq_r(t).

For harmonic force Qrsin⁡ωtQ_r\sin\omega t, the steady-state result is qr=Qrωr2−ω2sin⁡ωtq_r = \dfrac{Q_r}{\omega_r^2 - \omega^2}\sin\omega t. Often only the first few modes are needed, which reduces the size of the problem greatly. With proportional damping ([C]=α[M]+β[K][C] = \alpha[M] + \beta[K]) the same method works with a damping ratio ζr\zeta_r for each mode.

Written from the official syllabus. Questions and answers are written for this site; check them against your class notes.

Chapter titles and hours from the IOE syllabus ↗