Open Educational Resources

Advanced Dynamics and Vibrations: Lagrange’s equations applied to dynamic systems

Analytical Mechanics – Lagrange’s Equations

Up to the present we have formulated problems using newton’s laws in which the main disadvantage of this approach is that we must consider individual rigid body components and as a result, we must deal with interaction forces that we really have no interest in. These forces come from constraints which one part of the system puts on another.

An alternate approach referred to as analytical mechanics, consider the system as a whole and formulate the equations of motion starting from scalar quantities – the kinetic and potential energies and an expression of the virtual work associated with non-conservative forces. If relies heavily on the concept of virtual displacements.

Any set of numbers that serve to specify the configuration of the system are examples of generalized coordinates. To these we can add constraints. The number of DOF is then the number of generalized coordinates minus the number of independent equations of constraint.

e.g a point moving on the surface of a sphere of radius R with center (x_{o},y_{0},\zeta _{0}) must obey:

    \[(x-x_{0})^{2} + (y-y_{0})^{2} + (\zeta- \zeta _{0})^{2} = R^{2}\]

There are 3 coordinates and 1 equation of constraint. Therefore it has 2 DOF. If we used two angles, say \theta & \phi then we would not need the constraint.

These constraints are classified as:

\textcircled{1} If the constraint can be written in finite form (as above) they are called holonomic if not (then in differential form) then they are non-holonomic.

\textcircled{2} If they are equalities then they are bilateral and unilateral if inequalities

\textcircled{3} If there is no explicit dependence on time then they are scleronomic and if there is explicit time dependence they are rheonomic.

Insert diagram

    \[x_{1}^{2} + y_{1}^{2} = a^{2}\]

    \[(x_{2} - x_{1})^{2} + (y_{2} - y_{1})^{2} = b^{2}\]

Therefore, 2 DOF holonomic, scleronomic bilateral system.

Insert diagram

x_{1}^{2} +y_{1}^{2} + \zeta _{1}^{2} \leq \ell^{2} (unilateral)

however if we allow the point of attachment to move (in the x-y plane)

x_{1}^{2} + (y_{1} - y_{o} \sin \omega t)^{2} = \ell^{2} (rheomic)

Vertical displacement & Virtual Work

The work done during the motion is

    \[\overline{d \underline{\omega}} = F \cdot d \underline{r}\]

where F & dr are vectors and \underline{d \omega} is a scalar. For a particle \underline{F} = m \ddot{r} where \underline{F} may be the sum of several forces. Therefore,

    \[\begin{split} d \overline{\omega} &= m \ddot{\underline{r}} \cdot dr \\&= m \frac{d \dot{\underline{r}}}{dt} \cdot ( \frac{d \underline{r}}{dt}dt) \\&= m \frac{d \dot{r}}{dt} \cdot (\dot{r} dt) \\&= m \dot{\underline{r}} \cdot d \overline{\underline{r}} \\&= d(\frac{1}{2} m \dot{\underline{r}} \cdot \dot{\underline{r}})\end{split}\]

where this is a true differential of the kinetic energy.

    \[T = \frac{1}{2} m \dot{\underline{r}} \cdot \dot{\underline{r}}\]

If the particle moves from positive \underline{r_{1}} to \underline{r_{2}} then

    \[\int_{\underline{r}_{1}}^{\underline{r}_{2}} \underline{F} \cdot d \underline{r} = \frac{1}{2} m [\underline{\dot{r}}_{2} \cdot \underline{\dot{r}}_{2}] - \frac{1}{2} m [\underline{\dot{r}}_{1} \cdot \underline{\dot{r}}_{1}]\]

if \underline{F} depends only on \underline{r}, \underline{F} = \underline{F}(\underline{r})

    \[dW = F \cdot d \underline{r} \equiv -dV(\underline{r})\]

where V(\underline{r}) is a scalar function that depends only on the position and not on \underline{\dot{r}} or time. V is term the potential energy.

    \[d(T+V) = 0\]

The total energy is constant.

When we have a conservative system we can use Rayleigh’s to estimate the natural frequencies.

Consider a simple problem and use visual work to solve.

Insert diagram

Two bars of length l. What is the equilibrium position

insert diagram

Therefore,

    \[P \delta x + 2mg \delta h = 0\]

    \[x = 2 \ell \sin \theta\]

    \[h = \frac{\ell}{2} \cos \theta\]

    \[\delta x = 2\ell \cos \theta \delta \theta\]

    \[\delta h = \frac{- \ell}{2} \sin \theta \delta \theta\]

    \[P(2 \ell \cos \theta \delta \theta) = 2mg(\frac{\ell}{2} \sin \theta \delta \theta)\]

    \[\frac{P}{mg} = \tan \theta\]

Consider a rigid beam (uniform) of mass M located by a uniform time dependent load f_{o}(t) and supported by a spring & dashpot as shown

Insert diagram

Using VW determine the equation of motion.

\sum Virtual work of force applied = \sum Virtual work of metric forces

    \[\frac{-kl \theta}{2}\Big( \frac{\ell \delta \theta}{2} \Big) - cl \dot{\theta}(l \delta \theta) + \int_{0}^{l} f_{o}(t)dx (x \delta \theta) = \frac{Ml^{2}}{12} \ddot{\theta} \delta \theta + M \frac{l \ddot{\theta}}{2} \Big( \frac{l}{2} \delta \theta \Big) \]

Therefore,

    \[M \ell^{2}\Big[ \frac{1}{12} + \frac{1}{4}\Big] \ddot{\theta} \delta \theta + \frac{k \ell^{2}}{4} \theta \delta \theta + c \ell^{2} \dot{\theta} \delta \theta = f_{o}(t) \frac{ \ell^{2}}{2} \delta \theta\]

    \[ \frac{M}{3} \ddot{\theta} + c \dot{\theta} + \frac{k}{4} \theta = \frac{f(t)}{2}\]

We can use virtual work to calculate the energy in a multi-degree of freedom system.

Insert diagram

For a displacement of \delta _1 only

    \[F_{1} = k_{11} \delta _{1} \hspace{1cm} F_{2} = k_{21} \delta _{1} \hspace{1cm} F_{3} = k_{31} \delta _{1}\]

and similarly for \delta _2, \delta _3 alone

    \[F_{1} = k_{12} \delta _{2} \hspace{1cm} F_{2} = k_{22} \delta _{2} \hspace{1cm} F_{3} = k_{32} \delta _{2}\]

    \[F_{1} = k_{13} \delta _{3} \hspace{1cm} F_{2} = k_{23} \delta _{3} \hspace{1cm} F_{3} = k_{33} \delta _{3}\]

So for the combination (assuming superposition)

    \[F_{i} = \sum _{j=1}^{n} k_{ij} \delta _{j}\]

and

    \[\delta _{i} = \sum _{j=1}^{n} a_{ij} F _{j}\]

The work done by force F_{i} in \delta _{i} is energy stored due to that force:

    \[V_{i} = \frac{1}{2} F_{i} \delta _{i}\]

If there are n forces and the displacements are \delta _{i} (i = 1,2, … n) then the total energy is:

    \[\begin{split} V &= \sum _{i=1}^{n} V_{i} \\&= \frac{1}{2} \sum _{i=1}^{n} F_{i} \delta _{i} \\&= \frac{1}{2} \sum _{i=1}^{n} \Bigg( \sum _{j=1}^{n} k_{ij} \delta _{j} \Bigg) \delta _{i} \end{split}\]

    \[\begin{split} V &= \frac{1}{2} \{\delta \}^{T} [k] \{ \delta\} \\&= \frac{1}{2} \{F \}^{T} [a] \{ F\} \end{split}\]

We can also show that the kinetic energy is

    \[T = \frac{1}{2} \{\dot{\delta}\}^{T} [m] \{\dot{\delta}\}\]

T & V are examples of so-called quadratic forms. Quadratic forms are said to be positive definite if they are never negative and only zero when the variables are zero (\{\delta\} = \{0\})

They are said to be positively semi-definite if they are never negative but can be zero for non-negative variables.

This is useful to know because if we have a stiffness matrix that is positive semi-definite it means there is a rigid body mode with zero natural frequency [ [m] is positive definite ]

For example a submarine:

insert diagram

has 2 rigid body modes

insert diagram

Principle of Virtual Work

We will consider a system of N particles moving in a 3D space and define the virtual displacements as \delta x_{i} ^{(1)} for particle 1 \delta x_{i} ^{(1)} for particle 2 etc.

These are not true displacements but small variations in the so-called generalized coordinates. These are imagined displacements (slight variations) which are at the same time and which satisfy the constraints given by

    \[g_{j} (x_{1}, y_{1}, \zeta _{1}, x_{2}, y_{2}, \zeta _{2} ..... t) = c \hspace{1cm} j = 1t m\]

Then for virtual changes in these coordinates:

    \[g_{j} (x_{1} + \delta x_{1} , y_{1} + \delta y_{1}, \zeta _{1} + \delta \zeta _{1} ..... t) = c \]

Expanding this in a Taylor series gives:

    \[g_{j}(x_{1}, y_{1}, \zeta _{1},..... t) + \sum _{i=1} ^{N} \frac{\partial g_{i}}{\partial x_{i}} \partial x_{i} + \frac{\partial g_{j}}{\partial xy_{i}} \partial y_{i} + \frac{\partial g_{j}}{\partial \zeta_{i}} \partial \zeta_{i} = C\]

Therefore,

    \[\sum _{i=1} ^{N} \frac{\partial g_{j}}{\partial x_{i}} \partial x_{i} + \frac{\partial g_{j}}{\partial y_{j}} \partial y_{j} + \frac{\partial g_{j}}{\partial \zeta _{j}} \partial \zeta _{j} = 0 \hspace{1cm} \forall _{j}\]

So that there are 3N - j arbitrary displacements

On each of these particles the forces acting are

    \[\underline{R}_{i} = \underline{F}_{i} + \underline{f}_{i} \hspace{1cm} i = 1,2, ... N\]

Where the resultant force \underline{R}_{i} is composed of the external forces \underline{F}_{i} and \underline{f}_{i} the constraint forces.

When all the particles are in static equilibrium

    \[\underline{R}_{i} = 0 \hspace{1cm} \forall _{i}\]

and

    \[\underline{R}_{i} \cdot \delta \underline{r}_{i} = 0\]

where this represents the virtual work performed by the resultant force on the i particle.

    \[\delta \overline{W} = \sum _{i=1} ^{N} \underline{R}_{i} \cdot \delta \underline{r}_{i}\]

    \[\delta \overline{W} = \sum _{i=1} ^{N} \underline{F}_{i} \cdot \delta \underline{r}_{i} + \underline{f}_{i} \cdot \delta \underline{r}_{i} = 0\]

If we rule out work by the constraint forces ( as the constraint force is perpendicular to the motion), therefore cannot handle friction.

Consider two particles held by a rigid massless rod.

Insert diagram

    \[|\underline{f} _{1}| = |\underline{f} _{2}|\]

    \[\delta W = \underline{f}_{1} \cdot \underline{\delta} r_{1} +  \underline{f}_{2} \cdot \underline{\delta} r_{2}\]

But because the rod is rigid

    \[\delta \underline{r}_{1} \cdot \underline{e}_{r} = \delta \underline{r}_{2} \cdot \underline{e}_{r}\]

    \[\delta W = f_{1} \underline{e}_{r} \cdot \delta \underline{r}_{1} - f_{2} \underline{e}_{r} \cdot \delta \underline{r}_{2}\]

Therefore for the constraint forces

    \[\begin{split} \delta W &= (f_{1} - f_{2}) \underline{e}_{r} \cdot \delta r_{2} \\&= 0 \end{split}\]

and the Principal of Virtual Work say

    \[\begin{split} \delta W &= \sum _{i=1} ^{N} \underline{F}_{i} \cdot \underline{\delta} r_{i}\\& = 0 \end{split}\]

that the work performed by the applied forces through infinitesimal virtual displacements compatible with the constraints is zero.

To make the virtual work equations more useful we write it in terms of the generalized coordinates

    \[\begin{split} \delta W &= \sum _{i=1} ^{N} \underline{F}_{i} \cdot \delta r_{i} \\&= \sum _{i=1} ^{N} \cdot \sum _{j=1} ^{N} \frac{\partial \underline{r}_{i}}{\partial q_{j}} \delta q_{j} \\&= \sum _{j=1} ^{N} \Bigg( \sum _{i=1} ^{N} \cdot \frac{\partial \underline{r}_{i}}{\partial q_{j}} \Bigg) \delta q_{j} \end{split}\]

We call

    \[Q_{j} =\sum _{i=1} ^{N} \underline{F}_{i} \cdot  \frac{\partial \underline{r}_{i}}{\partial q_{j}}\]

The generalized forces. These are not necessarily forces as they can be moments. It is only necessary that the quantity Q_{k} \delta q_{k} has the units of work.

For the situation in which we have a conservative system the work can be expressed differently.

    \[\begin{split} \delta W &= \sum _{i=1} ^{N} \underline{F}_{i} \cdot \delta \underline{r}_{i}  \\&= - \delta V \\&= - \sum _{i=1} ^{N} \Bigg( \frac{\partial V}{\partial x_{i}} \delta _{i} + \frac{\partial V}{\partial y_{i}} \delta y_{i} + \frac{\partial V}{\partial \zeta _{i}} \delta \zeta _{i} \Bigg) \\&= 0 \end{split}\]

as we can select the virtual displacement arbitrarily (i.e. only one non-zero at a time)

    \[\underline{F}_{x_{i}} = -\frac{\partial V}{\partial x_{i}} = 0, \hspace{0.5cm} \underline{F}_{y_{i}} = -\frac{\partial V}{\partial y_{i}} = 0, \hspace{0.5cm} \underline{F}_{\zeta _{i}} = -\frac{\partial V}{\partial \zeta _{i}} = 0\]

This states that V has a stationary value at the equilibrium position (can use this for stability consideration). To extend this to dynamical situations we use D’Alembert’s principle.

    \[\underline{F}_{i} +\underline{f}_{i} - m_{i} \underline{\ddot{r}}_{i} = 0\]

Where the m_{i} \underline{\ddot{r}}_{i} are the inertia forces. The virtual work for a system of particles is:

    \[\sum _{i=1} ^{N} (\underline{F}_{i} +\underline{f}_{i} - m_{i} \underline{\ddot{r}}_{i}) \cdot \delta \underline{r}_{i} = 0\]

and since we showed the virtual work of the interned forces is zero.

    \[\sum _{i=1} ^{N} (\underline{F}_{i}  - m_{i} \underline{\ddot{r}}_{i}) \cdot \delta \underline{r}_{i} = 0\]

This is called the generalized D ‘Alembert’s principle and is really not useful as it is still in vector form. To get it in a scalar form we rewrite the second term in terms of any generalized coordinates and this leads to Lagrange’s equations of motion.

Consider a coordinate transformation where radius vector is written in terms of q_{i} instead x_{i},y_{i}, \zeta _{i}.

    \[\underline{r}_{i} = \underline{r}_{i}(q_{1} ...... q_{n}) \hspace{1cm} i = 1,2 ... N\]

    \[\begin{split} \underline{\dot{r}}_{i} &= \frac{\partial \underline{r}_{i}}{\partial q_{1}} \dot{q}_{1} + \frac{\partial \underline{r}_{i}}{\partial q_{2}} \dot{q}_{2} + ... \\&= \sum _{k=1} ^{n}  \frac{\partial \underline{r}_{i}}{\partial q_{k}} \dot{q}_{k} \end{split}\]

Note: that since \frac{\partial \underline{r}_{i}}{\partial q_{k}} do not depend explicitly on time that.

    \[\frac{\partial \underline{\dot{r}}_{i}}{\partial \dot{q}_{k}} =  \frac{\partial \underline{r}_{i}}{\partial q_{k}} --- (+)\]

now consider a virtual change in \underline{r}_{i}

    \[\begin{split} \delta \underline{r}_{i} &= \frac{\partial \underline{r}_{i}}{\partial q_{1}} \delta q_{1} + ... \\&= \sum _{k=1} ^{n} \frac{\partial \underline{r}_{i}}{\partial q_{k}} \delta q_{k} \end{split}\]

and apply this to the second term of D ‘Alembert’s principle

    \[\begin{split} \sum _{i=1} ^{n} m_{i} \underline{\ddot{r}}_{i} \cdot \delta \underline{r}_{i} &= \sum _{i=1} ^{n} m_{i} \underline{\ddot{r}}_{i} \cdot \sum _{k=1} ^{n} \frac{\partial \underline{r}_{i}}{\partial q_{k}} \delta q_{k} \\&= \sum _{k=1} ^{n} \Bigg( \sum _{i=1} ^{n} m_{i} \underline{\ddot{r}}_{i} \cdot \frac{\partial \underline{r}_{i}}{\partial q_{k}} \Bigg) \delta q_{k} \end{split}\]

JASON SECTION STARTS HERE

Consider a typical term in the bracket:

    \[ m_i \ddot{\underline{r}_i} \cdot \frac{\partial \underline{r}_i}{\partial q_k} = \frac{\text{d}}{\text{dt}} \Big(m_i \dot{\underline{r}_i} \cdot \frac{\partial \underline{r}_i}{\partial q_k}\Big) - m_i \dot{\underline{r}_i} \cdot \frac{\text{d}}{\text{dt}} \Big( \frac{\partial \underline{r}_i}{\partial q_k}\Big) \]

Now interchange the order of differentiation in the last term.

    \[ \begin{split} m_i \ddot{\underline{r}_i} \cdot \frac{\partial \underline{r}_i}{\partial q_k} &= \frac{\text{d}}{\text{dt}} \Big( m_i \dot{\underline{r}}_i \cdot \frac{\partial \underline{\dot{r}}_i}{\partial q_k}\Big) - m_i \dot{\underline{r}_i} \cdot \frac{\partial \underline{\dot{r}}_i}{\partial q_k} \\&= \Big[\frac{\text{d}}{\text{dt}} \Big(\frac{\partial}{\partial \dot{q_k}}\Big) - \frac{\partial}{\partial q_k} \Big]\Big(\frac{1}{2}m_i \dot{\underline{r}}_i \cdot dot{\underline{r}}_i\Big) \end{split} \]

As a result, the total second term becomes:

    \[ \begin{split} \sum m_i \ddot{\underline{r}}_i \cdot \delta\underline{r}_i &= \sum_{k=i}^n \biggr\{ \Big[ \frac{\text{d}}{\text{dt}} \Big(\frac{\partial}{\partial \dot{q_k}}\Big) - \frac{\partial}{\partial q_k} \Big] \Big( \sum_{i=1}^N \frac{1}{2} m_i \dot{\underline{r}}_i \cdot \dot{\underline{r}}_i \Big) \biggr\} \delta q_k \\&= \sum_{k=1}^n \biggr[ \frac{\text{d}}{\text{dt}} \Big( \frac{\partial T}{\partial \dot{q_k}} \Big) - \frac{\partial T}{\partial q_k} \biggr] \delta q_k \end{split} \]

As T = \frac{1}{2} \sum_{i=1}^N m_i \dot{\underline{r}}_i \cdot \dot{\underline{r}}_i = T(q_1,q_2,...,q_n,\ \dot{q_1},\dot{q_2},...,\dot{q_n}) is the kinetic energy of the entire system. To complete the conversion, we must reunite the forces.

    \[\begin{split} \delta \tilde{W} = \sum_{i=1}^N \underline{F}_i \cdot \delta \underline{r}_i &= \sum_{i=1}^N \underline{F}_i - \sum_{k=i}^n \frac{\partial \underline{r}_i}{\partial q_k} \delta q_k \\&= \sum_{k=1}^n\Big(\sum_{i=1}^N \underline{F}_i \cdot \frac{\partial \underline{r}_i}{\partial q_k} \Big) \delta q_k \end{split} \]

And we call:

    \[ \sum_{i=1}^N \underline{F}_i \cdot \frac{\partial \underline{r}_i}{\partial q_k} \equiv Q_k \quad \text{(generalized forces)}\]

Therefore:

    \[ \delta \tilde{W} = \sum_{k=1}^n Q_k \delta q_k \]

Some of the forces will be conservative (derivable from a potential V = V(q_1,q_2,...,q_n) and some non-conservative).

    \[ \begin{split} \sum \tilde{W} &= \delta W_c + \delta \tilde{W_{N_c}} \\&= -\delta V + \sum_{k=1}^n Q_{k_{n_c}} \delta q_k \\&= -\sum_{k=1}^n \Big( \frac{\partial V}{\partial q_k} - Q_{k_{(n_c)}} \Big) \delta q_k \end{split} \]

So that the rewritten D’Alembert’s principle becomes:

    \[ \sum_{k=1} \Big[ \frac{\text{d}}{\text{dt}} \Big(\frac{\partial T}{\partial \dot{q_k}} \Big) - \frac{\partial T}{\partial q_k} + \frac{\partial V}{\partial q_k} - Q_{k_{n_c}} \Big] \delta q_k = 0\]

And since the displacements are arbitrary and independent.

    \[ \frac{\text{d}}{\text{dt}} \Big( \frac{\partial T}{\partial \dot{q_k}} \Big) - \frac{\partial}{\partial q} (T - V) = Q_{k_{n_c}} \quad k = 1,2,....,n \]

However, since the potential energy does not depend on the velocities, this can be written as:

    \[ \frac{\text{d}}{\text{dt}} \Big( \frac{\partial L}{\partial \dot{q_k}} \Big) - \frac{\partial L}{\partial q_k} = Q_k \quad k = 1,2,....,n \]

    \[ L \text{(Lagrangian)} = T - V \]

NOTES:

  • Does not have to be only for “small” oscillations.
  • Does not matter whether the conservative forces are included in the Q_k.
  • Sometimes the Q_k are able to be put into a “potential like” function.
  • Can include reaction if wanted using Lagrange multipliers.
  • Determine the Q_k using virtual work.

Example

Find the equations of motion:

    \[ x = r\theta \]

    \[\begin{split} T &= \frac{1}{2} m (\dot{x})^2 + \frac{1}{2}J(\dot{\theta})^2 \\&= \frac{1}{2} \Big(m + \frac{J}{r^2}\Big)\dot{x}^2 \end{split} \]

    \[V = \frac{1}{2}kx^2 \]

    \[\delta W\ \text{(damper)}\ = -c\dot{x} \delta x \]

    \[Q_x = -c\dot{x} \]

    \[\frac{\text{d}}{\text{dt}}\Big(\frac{\partial L}{\partial \dot{q}} \Big) - \frac{\partial L}{\partial q} = Q_{\text{NC}} \]

Thus:

    \[ \Big(m + \frac{J}{r^2} \Big) \ddot{x} + kx = -c\dot{x} \]

NOTE: Sometimes write the damping forces as a dissipation function.

    \[ D = \frac{1}{2}c\dot{q}^2,\quad Q = \frac{\partial D}{\partial q} = c\dot{x} \]

    \[\frac{\text{d}}{\text{dt}} \Big(\frac{\partial L}{\partial \dot{q_k}} \Big) - \frac{\partial L}{\partial \dot{q_k}} + \frac{\partial D}{\partial \dot{q_k}} = Q_{k_{\text{NC}}} \]

Example

Find the equations of motion:

    \[ J = \frac{1}{2} M r^2 \]

    \[T = \frac{1}{2} M \dot{x}^2 + \frac{1}{2}J\dot{\theta}^2 + \frac{1}{2}m\nu_A^2 \]

The potential energy is:

    \[ V = mgy + \frac{1}{2}kx^2 \]

And the damping force is the same. Note that:

    \[ x = r\theta,\quad y = \ell(1 - \cos \theta) \]

    \[\dot{x} = r \dot{\theta} \]

    \[ \nu_A = ?\]

    \[ \underline{\nu}_A = \underline{\nu}_o + \underline{\nu}_{A/o} \]

    \[ \ell \dot{\theta} = |\underline{\nu}_{A/o}|\]

    \[ |\underline{\nu}_o| = \dot{x} \]

    \[ \begin{split} |\underline{\nu}_A| &= |\underline{\nu}_o|^2 + (\ell \dot{\theta})^2 - 2\nu_o\ell\dot{\theta}\cos\theta \\&= \dot{x}^2 + \frac{\ell^2}{r^2}\dot{x}^2 - 2\dot{x}^2\frac{\ell}{r} \cos \frac{x}{r} \end{split} \]

Thus:

    \[ T = \frac{\dot{x}^2}{2}\Big[\frac{3}{2}M + m\Big(1 + \frac{\ell^2}{r^2} - \frac{2\ell}{r}\cos\frac{x}{r}\Big)\Big] \]

    \[ V = mg\ell\Big(1-\cos\frac{x}{r}\Big) + \frac{1}{2}kx^2 \]

    \[\frac{\partial L}{\partial \dot{x}} = \dot{x}\Big[\frac{3M}{2} + m \Big(1+\frac{\ell^2}{r^2} - \frac{2\ell}{r}\cos\frac{x}{r}\Big)\Big] \]

    \[\frac{\partial L}{\partial x} = \frac{\dot{x}^2}{2} \Big[ \frac{2\ell m}{r^2} \sin \frac{x}{r} \Big] -mg\frac{\ell}{r}\sin\frac{x}{r} - kx \]

    \[ \frac{\text{d}}{\text{dt}}\Big(\frac{\partial L}{\partial \dot x}\Big) = \ddot{x}\Big[\frac{3M}{2} + m \Big( 1 + \frac{\ell^2}{r^2} - \frac{2\ell}{r}\cos\frac{x}{r} \Big)\Big] + \dot{x}^2\frac{2m\ell}{r^2}\sin\frac{x}{r} \]

The equation of motion comes from:

    \[ \frac{\text{d}}{\text{dt}} \Big(\frac{\partial L}{\partial \dot{x}}\Big) - \frac{\partial L}{\partial x} = -c\dot{x} \]

And becomes:

    \[ \Big[ \frac{3M}{2} + m \Big(1+ \frac{\ell^2}{r^2} - \frac{2\ell}{r}\cos\frac{x}{r}\Big)\Big]\ddot{x} + \dot{x}^2\frac{m\ell}{r^2}\sin{\frac{x}{r}} + mg\frac{\ell}{r}\sin{\frac{x}{r}} + kx + c\dot{x} = 0\]

For small deflection \cos\frac{x}{r} \approx 1 and \sin \frac{x}{r} \approx \frac{x}{r}:

    \[ \Big[\frac{3}{2}M + m\Big(1+\frac{\ell^2}{r^2}-\frac{2\ell}{r}\Big)\Big] \ddot{x} + \frac{m\ell x}{r^2}\Big[\frac{\dot{x}^2}{r}+g\Big] + kx + c\dot{x} = 0 \]

2DOF Example

The uniform rod of length \ell and mass m carries a slider of mass m_o which is attached by a spring of stiffness k. The spring is unstretched at r = r_o.

    \[ T = \frac{1}{2}m_o\big[(r\dot{\theta})^2 + (\dot{r})^2\big] + \frac{1}{2}\Big(\frac{m\ell^2}{3}\dot{\theta}^2\Big) \]

    \[ V = \frac{1}{2}k(r-r_o)^2-m_ogr\cos\theta - mg\frac{\ell}{2}\cos\theta \]

    \[L = T - V \quad \text{(} \theta = \frac{\pi}{2} \text{ is the datum)} \]

    \[\frac{\partial L}{\partial \dot{\theta}} = m_or^2\dot{\theta} + \frac{m\ell^2}{3}\dot{\theta} \]

    \[\frac{\partial L}{\partial \theta} = -m_ogr\sin \theta - mg\frac{\ell}{2}\sin\theta \]

    \[ \frac{\text{d}}{\text{dt}} \Big(\frac{\partial L}{\partial \dot{\theta}}\Big) - \frac{\partial L}{\partial \theta} = 0 \]

Thus:

    \[m_o r [r\ddot{\theta} + 2\dot{r}\dot{\theta}] + \frac{m\ell^2}{3}\dot{\theta} +m_ogr\sin\theta + mg\frac{\ell}{2}\sin\theta = 0 \]

Also, note that:

    \[ \frac{\partial L}{\partial \dot{r}} = m_o\dot{r} \]

    \[ \frac{\partial L}{\partial r} = -k(r-r_o) + m_og\cos\theta + m_or\dot{\theta}^2 \]

    \[ \frac{\text{d}}{\text{dt}}\Big(\frac{\partial L}{\partial \dot{r}}\Big) - \frac{\partial L}{\partial r} = 0\]

Thus:

    \[ m_o\ddot{r} - m_or\dot{\theta}^2 + k(r-r_o) - m_og\cos\theta = 0\]

The equilibrium positions are:

    \[\sin\theta = 0 \Rightarrow \theta = 0,\ \pi \]

    \[ k(r_\text{ST} - r_o) = \pm m_og \]

We could now study the stability of these configurations.

Example with External Forces

    \[ T = \frac{1}{2}m_1\dot{x}^2 + \frac{1}{2}m_2\nu_G^2 + \frac{1}{2}I_G\dot{\theta}^2 \]

    \[ \nu_G^2 = \dot{x}^2 + \Big(\frac{\ell \dot{\theta}}{2}\Big)^2 - 2\dot{x}\Big(\frac{\ell\dot{\theta}}{2}\Big)\cos(180-\theta) \]

Using \cos(A-B) = \cos A \cos B + \sin A \sin B, we can say:

    \[ \begin{split} \cos(180-\theta) &= \cos 180 \cos \theta + \sin 180 \sin \theta \\& = -\cos \theta \end{split} \]

Thus:

    \[\nu_G^2 = \dot{x}^2 + \Big(\frac{\ell\dot{\theta}}{2}\Big)^2 + \dot{x}\ell\dot{\theta}\cos \theta \]

Therefore:

    \[T = \frac{1}{2}m_1\dot{x}^2 + \frac{1}{2}m_2\Big[\dot{x}^2 + \Big(\frac{\ell\dot{\theta}}{2}\Big)^2 + \dot{x}\dot{\theta}\ell\cos\theta \Big] + \frac{1}{2}I\dot{\theta}^2 \]

    \[V = \frac{1}{2}kx^2 + m_2g\frac{\ell}{2}(1-\cos\theta) \]

Where \theta = 0 is the datum. Now we must find the generalized forces. For a virtual displacement \delta x only:

    \[ \delta W = F(t)\delta x \Rightarrow Q_x = F(t) \]

For a virtual displacement \delta \theta only:

    \[ \delta W = \ell \delta \theta \cos\theta F(t) \]

Therefore:

    \[ Q_\theta = F(t)\ell\cos\theta \]

Then, note that:

    \[ \frac{\text{d}}{\text{dt}}\Big(\frac{\partial L}{\partial \dot{x}}\Big) - \frac{\partial L}{\partial x} = Q_x \quad (1) \]

    \[ \frac{\text{d}}{\text{dt}} \Big(\frac{\partial L}{\partial \dot{\theta}}\Big) - \frac{\partial L}{\partial \theta} = Q_\theta \quad (2) \]

    \[ \frac{\partial L}{\partial \dot{x}} = m_1\dot{x} + m_2\dot{x} + m_2\frac{\ell}{2}\dot{\theta}\cos\theta \]

    \[\frac{\partial L}{\partial x} = -kx\]

    \[ \frac{\partial L}{\partial \dot{\theta}} = m_2\frac{\ell^2}{4}\dot{\theta} + m_2\dot{x}\frac{\ell}{2}\cos\theta + I_G\dot{\theta} \]

    \[ \frac{\partial L}{\partial \theta} = -\frac{m_2}{2}\dot{x}\dot{\theta}\ell\sin\theta - m_2g\frac{\ell}{2}\sin\theta \]

Therefore, (1) becomes:

    \[ m_1\ddot{x} + m_2\ddot{x} + m_2\frac{\ell}{2}[\ddot{\theta}\cos\theta - \dot{\theta}^2\sin\theta]+kx = F(t) \]

And (2) becomes:

    \[m_2 \frac{\ell^2}{4}\ddot{\theta} + m_2\ddot{x}\frac{\ell}{2}\cos\theta - m_2\dot{x}\frac{\ell}{2}\dot{\theta}\sin\theta + \frac{1}{12}m\ell^2\ddot{\theta} + m_2\dot{x}\frac{\ell}{2}\dot{\theta}\sin\theta + m_2g\frac{\ell}{2}\sin\theta = F(t)\ell\cos\theta \]

    \[ m_2\frac{\ell^2}{3}\ddot{\theta} + \frac{m_2}{2}\ddot{x}\ell\cos\theta + m_2g\frac{\ell}{2}\sin\theta = F(t)\ell\cos\theta \]

For small angles these become:

    \[ (m_1+m_2)\ddot{x} + m_2\frac{\ell}{2}\ddot{\theta} + kx = F(t) \]

    \[ m_2\frac{\ell^2}{3}\ddot{\theta} + m_2\frac{\ell}{2}\ddot{x} + m_2g\frac{\ell}{2}\theta = F(t)\ell \]

If we have support motion instead of forcing,

then we cannot neccesairly use a potential, so now give the system a \delta x and \delta \theta. For a \delta x:

    \[ \delta W = -k(x-y)\delta x \]

Therefore:

    \[Q_x = -k(x-y) \]

    \[Q_o = 0 \]

Q_o = 0 as no work is done for \delta \theta alone. Thus, T is the same, and:

    \[ V = m_2g\frac{\ell}{2}(1-\cos\theta) \]

For small motions:

    \[ (m_1+m_2)\ddot{x} +m_2\frac{\ell}{2}\ddot{\theta} + kx = ky(t) \]

    \[ m_2\frac{\ell^2}{3} \ddot{\theta} + m_2\frac{\ell}{2}\ddot{x} + m_2g\frac{\ell}{2}\theta = 0\]

Thus:

    \[ \begin{bmatrix} m_1 + m_2 & m_2\frac{\ell}{2} \\ m_2\frac{\ell}{2} & m_2\frac{\ell^2}{3} \end{bmatrix} \begin{Bmatrix} \ddot{x} \\ \ddot{\theta} \end{Bmatrix} + \begin{bmatrix} k & 0 \\ 0 & \frac{m_2g\ell}{2} \end{bmatrix} \begin{Bmatrix} x \\ \theta \end{Bmatrix} = \begin{Bmatrix} ky(t) \\ 0 \end{Bmatrix} \]

Example

Consider the double pendulum:

    \[T = \frac{1}{2}m(\ell\dot{\theta_1})^2 + \frac{1}{2}m(\ell\dot{\theta_1} + \ell\dot{\theta_2})^2 \]

    \[V = mg\ell(1-\cos\theta_1) + mg\ell(1 - \cos\theta_1 + 1-\cos\theta_2) \]

Consider only small angles. Thus:

    \[ V \approx mg\ell\frac{\theta_1^2}{2} + mg\ell\Big(\frac{\theta_1^2}{2} + \frac{\theta_2^2}{2}\Big) \]

Then:

    \[ \frac{\partial L}{\partial \dot{\theta_1}} = m\ell^2\dot{\theta_1} + m\ell^2(\dot{\theta_1} + \dot{\theta_2}) \]

    \[ \frac{\partial L}{\partial \dot{\theta_2}} = m\ell^2(\dot{\theta_1} + \dot{\theta_2}) \]

    \[ \frac{\partial L}{\partial \theta_1} = -mg\ell\theta_1 - mg\ell\theta_1 \]

    \[ \frac{\partial L}{\partial \theta_2} = -mg\ell\theta_2 \]

Therefore:

    \[ \begin{bmatrix} 2 & 1 \\ 1 & 1 \end{bmatrix} \begin{Bmatrix} \ddot{\theta_1} \\ \ddot{\theta_2} \end{Bmatrix} + \frac{g}{\ell} \begin{bmatrix} 2 & 0 \\ 0 & 1 \end{bmatrix} \begin{Bmatrix} \theta_1 \\ \theta_2 \end{Bmatrix} = \begin{Bmatrix} 0 \\ 0 \end{Bmatrix} \]

    \[ [m]^{-1} = \begin{bmatrix} 1 & -1 \\ -1 & 2 \end{bmatrix} \]

    \[\begin{split} [m]^{-1}[m] &= \begin{bmatrix} 1 & -1 \\ -1 & 2 \end{bmatrix} \begin{bmatrix} 2 & 1 \\ 1 & 1 \end{bmatrix} \\&= \begin{bmatrix} 1 & 0 \\ 0 & 1 \end{bmatrix} \end{split}\]

page26

    \[ -p^2\begin{Bmatrix} X_1 \\ X_2 \end{Bmatrix} + \frac{g}{\ell}\begin{bmatrix} 1 & -1 \\ -1 & 2 \end{bmatrix} \begin{bmatrix} 2 & 0 \\ 0 & 1 \end{bmatrix} \begin{Bmatrix} X_1 \\ X_2 \end{Bmatrix} = 0 \]

    \[ \implies -\frac{p^2\ell}{g} \begin{Bmatrix} X_1 \\ X_2 \end{Bmatrix} + \begin{bmatrix}  2 & -1 \\ -2 & 2 \end{bmatrix} \begin{Bmatrix} X_1 \\ X_2\end{Bmatrix} = 0\]

Set \lambda^2 = \frac{p^2\ell}{g}:

    \[ \begin{vmatrix} 2 - \lambda^2 & -1 \\ -2 & 2 - \lambda^2 \end{vmatrix} = 0\]

    \[ (2 - \lambda ^2 ) ^2 - 2 = 0 \]

    \[ 4 - 4\lambda^2 + \lambda^4 - 2 = 0\]

    \[\lambda^4 - 4\lambda^2 + 2 = 0 \]

    \[ \lambda^2 = 2 \pm \sqrt{2} \]

    \[ (2-\lambda^2)X_1 - X_2 = 0\]

    \[ \frac{X_2}{X_1} = 2 - \lambda^2\]

    \[ \bigg(\frac{X_2}{X_1}\bigg)_1 = 2 - 2 + \sqrt{2} = \underline{\underline{\sqrt{2}}} \]

    \[ \bigg(\frac{X_2}{X_1}\bigg)_1 = 2 - 2 - \sqrt{2} = \underline{\underline{-\sqrt{2}}} \]

Example

Now try a different set of generalized coordinates for the same problem.

If we consider small angles:

    \[ T = \frac{1}{2} m(\ell \dot{\theta}_1)^2 + \frac{1}{2}m\big[\ell\dot{\theta}_1 + \ell(\dot{\theta}_1  \dot{\theta}_3)\big]^2 \]

    \[ V = mg\ell(1-\cos{\theta_1}) + mg\ell\big[ 1-\cos{\theta_1} + 1 - \cos{(\theta_1 + \theta_3)}\big] \]

Therefore:

    \[V = -mg\ell\frac{\theta_1^2}{2} - mg\ell\bigg[ \frac{\theta_1^2}{2} + \frac{(\theta_2 + \theta_3)^2}{2} \bigg] \]

    \[T = \frac{1}{2}m\ell^2(\dot{\theta}_1)^2 + \frac{1}{2}m\ell^2[2\dot{\theta}_1+\dot{\theta}_3]^2\]

    \[V = -mg\ell\theta_1^2-mg\frac{\ell}{2}(\theta_1 +\theta_3)^2 \]

    \[L = T-V\]

    \[\begin{split}\frac{\partial L}{\partial \dot\theta_1} &= m\ell^2\dot\theta_1 +\frac{1}{2}m\ell^22(2\dot\theta_1+\dot\theta_3)2 \\&= m\ell^2[5\dot\theta_1 + 2\dot\theta_3]\end{split}\]

    \[\begin{split}\frac{\partial L}{\partial \dot\theta_3} &= \frac{1}{2}m\ell^2(2)[2\dot\theta_1+\dot\theta_3] \\&= m\ell^2[2\dot\theta_1+\dot\theta_3]\end{split}\]

    \[\frac{\partial L}{\partial \theta_1 } = -2mg\ell\theta_1 - mg\ell(\theta_1+\theta_3)\]

    \[\frac{\partial L}{\partial \theta_3} = -mg\ell(\theta_1 +\theta_3)\]

    \[\frac{d}{dt}\bigg(\frac{\partial L}{\partial \dot\theta_1}\bigg) - \frac{\partial L}{\partial \theta_1} = 0\]

    \[ m\ell^2[5\ddot\theta_1 + 2 \ddot\theta_3] +3mg\ell\theta_1 + mg\ell\theta_3 = 0\]

    \[m\ell^2[2\ddot\theta_1 + \ddot\theta_3] + mg\ell(\theta_1 + \theta_3) = 0\]

Therefore:

    \[\begin{bmatrix} 5 & 2 \\ 2 & 1 \end{bmatrix}\begin{Bmatrix} \ddot\theta_1 \\ \ddot\theta_3 \end{Bmatrix} + \frac{g}{\ell}\begin{bmatrix} 3 & 1 \\ 1 & 1 \end{bmatrix}\begin{Bmatrix} \theta_1 \\ \theta_2 \end{Bmatrix} = \begin{Bmatrix} 0 \\ 0 \end{Bmatrix} \]

Therefore:

    \[ - \frac{p^2\ell}{g}\begin{bmatrix} 5 & 2 \\ 2 & 1 \end{bmatrix}\begin{Bmatrix} X_1 \\ X_2\end{Bmatrix} + \begin{bmatrix} 3 & 1 \\ 1 & 1 \end{bmatrix} \begin{Bmatrix} X _1 \\ X_2 \end{Bmatrix} = \begin{Bmatrix} 0 \\ 0 \end{Bmatrix} \]

    \[ [m]^{-1} = \begin{bmatrix} 1 & -2 \\ -2 & 5 \end{bmatrix} \begin{bmatrix} 5 & 2 \\ 2 & 1 \end{bmatrix} = \begin{bmatrix} 1 & 0 \\ 0 & 1 \end{bmatrix} \]

Set \lambda^2 = \frac{p^2\ell}{g}:

    \[ -\lambda\begin{Bmatrix} X_1 \\ X_3 \end{Bmatrix} + \begin{bmatrix} 1 & -2 \\ -2 & 5 \end{bmatrix} \begin{bmatrix} 3 & 1 \\ 1 & 1 \end{bmatrix} \begin{Bmatrix} X_1 \\ X_3 \end{Bmatrix} = \begin{Bmatrix} 0 \\ 0 \end{Bmatrix} \]

    \[ -\lambda\begin{Bmatrix} X_1 \\ X_3 \end{Bmatrix} + \begin{bmatrix} 1 & -1 \\ -1 & 3 \end{bmatrix}  \begin{Bmatrix} X-1 \\ X_3 \end{Bmatrix} = \begin{Bmatrix} 0 \\ 0 \end{Bmatrix} \]

Therefore:

    \[ (1 - \lambda)(3 - \lambda) - 1 = 0 \]

    \[\lambda^4 - 4\lambda^2 + 2 = 0 \quad \text{(same characteristic equation)}\]

Therefore:

    \[ \lambda^2 = 2 \pm \sqrt{2} \]

For mode shapes:

    \[ (1 - \lambda) X_1 - X_3 = 0 \]

    \[ \implies \frac{X_3}{X_1} = 1 - \lambda \]

    \[\begin{split} \bigg(\frac{X_3}{X_1}\bigg)_1 &= 1 - 2 + \sqrt{2} \\&= -1 + \sqrt{2} \end{split}\]

    \[\begin{split} \bigg(\frac{X_3}{X_1} \bigg) _2 &= 1 - 2 - \sqrt{2} \\&= -1 - \sqrt{2} \end{split}\]

These are different values of ratios than in the first case. However, when plotted:

Now use the newtonain approach to the same problem:

    \[ +\sum \circlearrowleft  M_0 = m\ell^2\ddot\theta_1 =  - mg\ell\theta_1 + T( \theta_2 - \theta_1) \ell \quad \text{(A)} \]

    \[\begin{split} + \downarrow \sum F_y &= m\ddot y \\&= -T\cos\theta_2 + mg \end{split}\]

    \[\begin{split} + \rightarrow \sum F_x &= m\ddot x \\&= - T\sin{\theta_2} \end{split}\]

    \[\ddot{x} \approx \ell ( \ddot\theta_1 + \ddot\theta_2) , \ \ddot y \approx 0 \]

Therefore:

    \[T \approx mg \]

    \[ m\ell(\ddot\theta_1 + \ddot\theta_2) + mg\theta_2 = 0 \quad \text{(B)}\]

    \[\begin{bmatrix} m\ell & 0 \\ m\ell & m\ell \end{bmatrix} \begin{Bmatrix} \ddot\theta_1 \\ \ddot\theta_2 \end{Bmatrix} + \begin{bmatrix} 2mg & -mg \\ 0 & mg \end{bmatrix} \begin{Bmatrix} \theta_1 \\ \theta_2 \end{Bmatrix} = \begin{Bmatrix} 0 \\ 0 \end{Bmatrix} \]

Note both mass and stiffness matrices are non-symmetric!

Calculating the eigenvalues gives the same result and the mode shapes are identical with the other 2 formulations.

NOTE: These sets(first and third formulation) are linearly dependent.

Lagrange’s Equation for Small Oscillations

Our main interest is in the motion of MDOF systems in the neighborhood of equilibrium positions. WLOG we assume the equilibrium position is q_1 = q_2 \dotsc = q_n = 0 and that the displacements are sufficiently small that the linear force- displacement and force-velocity relations hold. This means that the generalized coordinates are their time derivatives appear in the differential equations to only the first power.

Therefore:

    \[ \underline{\dot r} = \sum_{k = 1}^n \frac{\partial \underline{r}}{\partial q_k}\dot{q}_k \]

And \frac{\partial \underline r}{\partial q_k} are constants.

    \[\begin{split} T &= \frac{1}{2} \sum_{i = 1}^N m_i \underline{\dot r}_i \underline{\dot r}_i \\&= \frac{1}{2} \sum_{i = 1}^N m_i\bigg(\sum_{r = 1}^n \frac{\partial \underline{r}_i}{\partial q_r}\dot{q}_r\bigg)\bigg(\sum_{s = 1}^n \frac{\partial \underline{r}_i}{\partial q_s}\dot{q}_s\bigg) \\&= \frac{1}{2} \sum_{r = 1}^n \sum_{s = 1}^n\bigg(\sum_{i=1}^N m_i \frac{\partial \underline{r}_i}{\partial q_r}\frac{\partial \underline{r}_i}{\partial q_s}\bigg)\dot q_r\dot q_s \end{split}\]

Set m_{rs} = \sum_{i = 1}^N m_i \frac{\partial \underline{r}_i}{\partial q_r}\frac{\partial \underline{r}_i}{\partial q_s} = m_{sr} as the generalized masses.

Similarly we can write the potential energy as V = V(q_1, q_2 \dotsc q_n) and consider a Taylor expansion of V about the equilibrium configuration.

    \[V(q_n) = V_0  \sum_{n = 1} ^n \frac{\partial V}{\partial q_i} + \frac{1}{2}\sum_{i= 1}^n \sum_{j = 1}^n \frac{\partial^2V}{\partial q_i \partial q_j}q_iq_j+ \dotsc\]

Where the partial derivatives are all evaluated at q_i = 0.

As the equilibrium point is a condition in which:

    \[\frac{\partial V}{\partial q_i} = 0\]

V is approximately:

    \[\begin{split} V &= \frac{1}{2}\sum_{r = 1}^n \sum_{s = 1}^n \frac{\partial^2 V}{\partial q_r \partial q_s}q_rq_s \\&=\frac{1}{2}\sum_{r = 1}^n \sum_{s = 1}^n k_{rs}q_rq_s\end{split}\]

Where k_{rs} = k_{sr} are the stiffness coefficients. If we now use Lagrange’s equation the equations of motion become:

    \[\frac{\partial L}{\partial \dot{q}_j} = \sum_{s = 1}^n m_{js}\dot{q}_s \]

    \[\frac{\partial L}{\partial q_j} = \sum_{s = 1}^n k_{js}q_s \]

Therefore

    \[\sum_{s = 1}^n[m_{js} \ddot{q}_s(t) + k_{js}q(t)] = Q_j \]

    \[j = 1,2 \dotsc n\]

While this is very straightforward, we can use this result to obtain approximate formulations of the equations of motion as continuous systems it replaces them with a n degree of freedom system. It can also extend the technique to find approximate response to a specific forcing function. This is called the assumed modes method (AMM).

For discrete system it replaces them with a smaller number of DOF, while for continuous system it replaces them with an n degree of freedom system. We can also easily extend it to a forced system.

Then the method assumes:

    \[u(x,t) = \sum_{i = 1}^N\phi_iq_i(t) \]

Where \phi_i(x) are trial functions and q_i(t) are the generalized coordinates.

    \[ T = \frac{1}{2}\sum_{i=1}^N\sum_{j=1}^Nm_{ij}\dot{q}_i\dot{q}_j\]

    \[V = \frac{1}{2}\sum_{i = 1}^N\sum_{j = 1}^Nk_{ij}q_iq_j\]

\phi_i(x) needs only satisfy the geometric boundary conditions.

For continuous systems:

    \[\begin{split}T &= \frac{1}{2} \int dm\bigg(\frac{\partial y}{\partial t}\bigg)^2 \\&= \frac{1}{2} \sum_{i = 1}^N\sum_{j = 1}^N\dot{q}_i\dot{q}_j\int_m \phi_i(x)\phi_j(x)dm\end{split} \]

Therefore:

    \[m_{ij} = \int_m \phi_i(x)\phi_j(x)dm \]

While for discrete systems

    \[ m_{ij} = \sum_{p = 1}^N m_p\phi_i(x_p)\phi_j(x_p)\]

The potential energy can be determined is a similar manner.

Consider a generalized beam:

    \[ V = \frac{1}{2}\int_m EI \frac{\partial^2 y}{\partial x^2}dx \]

    \[\frac{\partial^2 y}{\partial x^2} = \sum_{i=1}^N \phi''(x)(t)\]

Therefore:

    \[ V = \frac{1}{2}\sum_{i = 1}^N\sum_{j=1}^Nq_iq_j\int_LEI\phi''_i\phi''_jdx\]

    \[k_{ij} = \int EI\phi''_i\phi''_jdx\]

This will allow calculation of a consistent mass and stiffness matrices.

Example: Discrete System

    \[m_1 = m_2 = m_3 = m\]

    \[ [k] = \begin{bmatrix} k & \frac{k}{3} & 0 \\ \frac{k}{3}& \frac{11k}{9} & -k \\ 0 & -k & k\end{bmatrix} \]

We want to estimate the lowest natural frequency using assumed modes method.

Try:

    \[ \phi_1 = \begin{bmatrix} 1 \\ -1 \\ -\frac{3}{2} \end{bmatrix} \]

Now calculate the consistent mass and stiffness using this assumed shape:

    \[ m_{11} = m\begin{bmatrix} 1 & -1 & -\frac{3}{2} \end{bmatrix}\begin{bmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1\end{bmatrix}\begin{bmatrix} 1 \\ -1 \\ -\frac{3}{2} \end{bmatrix} = \frac{17}{4}m\]

    \[ k_{11} = k\begin{bmatrix} 1 & -1 & -\frac{3}{2} \end{bmatrix}\begin{bmatrix} 1 & \frac{1}{3} & 0 \\ \frac{1}{3} & \frac{11}{9} \ -1 \\  0  & -1 & 1\end{bmatrix}\begin{bmatrix} 1 \\ -1 \\ -\frac{3}{2} \end{bmatrix} = \frac{29}{36}k \]

    \[m_{11}\ddot{q}_1 + k_{11}q_1 = 0\]

    \[p = \sqrt{\frac{29}{36}\frac{4}{17}\frac{k}{m}} = 0.435\sqrt{\frac{k}{m}}\]

What if we choose \phi_1 = \begin{bmatrix} 1 \\ -1 \\ -\frac{3}{2} \end{bmatrix}, \ \phi_2 = \begin{bmatrix} 1 \\ -2 \\-3 \end{bmatrix}.

Then m_{11} and k_{11} are the same.

    \[m_{22} = m(1)(1) + m(-2)(-2) +m (-3)(-3) = 14m\]

    \[ k_{22} = k\begin{bmatrix} 1 & -2 & -3 \end{bmatrix}\begin{bmatrix} 1 &  \frac{1}{3} & 0 \\ \frac{1}{3} & \frac{11}{9} & -1 \\  0  & -1 & 1\end{bmatrix}\begin{bmatrix} 1 \\ -2 \\ -3 \end{bmatrix} = \frac{14}{9}k\]

    \[  k_{12} = k_{21} = k\begin{bmatrix} 1 & -1 & -\frac{3}{2} \end{bmatrix}\begin{bmatrix} 1 & \frac{1}{3} & 0 \\ \frac{1}{3} & \frac{11}{9} & -1 \\  0  & -1 & 1\end{bmatrix}\begin{bmatrix} 1 \\ -2 \\ -3 \end{bmatrix} = \frac{17}{18}k\]

    \[ m_{12} = m\begin{bmatrix} 1 & -2 & -3 \end{bmatrix}\begin{bmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1\end{bmatrix}\begin{bmatrix} 1 \\ -1 \\ -\frac{3}{2} \end{bmatrix} = \frac{15}{2}m\]

So we have replace the system by:

    \[ m\begin{bmatrix} \frac{17}{4} & \frac{15}{2} \\ \frac{15}{2} & 14 \end{bmatrix} \begin{bmatrix} \ddot{q}_1 \\ \ddot{q}_2\end{bmatrix} + \begin{bmatrix} \frac{29}{36} & \frac{17}{18} \\ \frac{17}{18} & \frac{14}{9}\end{bmatrix} \begin{bmatrix} q_1 \\ q_2 \end{bmatrix} \]

Now calculate the p‘s.

    \[ \begin{bmatrix} \frac{29}{36}k - \frac{17}{4}mp^2 & \frac{17}{18} k - \frac{15}{2}mp^2 \\ \frac{17}{18}k - \frac{15}{2}mp^2 & \frac{14}{9} k - 14mp^2 \end{bmatrix} \begin{Bmatrix} Q_1 \\ Q_2 \end{Bmatrix} = 0\]

    \[ \bigg( \frac{29}{36}k - \frac{17}{4}mp^2\bigg)\bigg(\frac{14}{9}k - 14 mp^2 \bigg) - \bigg(\frac{17}{18}k - \frac{15}{2}mp^2\bigg)^2 = 0\]

    \[ \frac{29\cdot 14}{36\cdot 9}k^2 - kmp^2\bigg(\frac{17\cdot 14}{36} + \frac{14\cdot 29}{36}\bigg) + m^2p^4\bigg(\frac{17\cdot 14}{4}\bigg)\]

    \[-\frac{17^2}{18^2}k^2 + 2\bigg(\frac{17\cdot 15}{18\cdot 2}\bigg)kmp^2 - \frac{225}{4} m^2p^4 = 0 \]

    \[p^4\bigg[ \frac{238}{4} - \frac{225}{4} \bigg] - \frac{k}{m} p^2 \bigg[ \frac{17\cdot 14}{36} + \frac{14\cdot 29}{36} - \frac{17\cdot 15}{18}\bigg] + \bigg[\frac{29\cdot 14}{36\cdot 9} - \frac{17^2}{18^2}\bigg]\bigg(\frac{k}{m}\bigg)^2 = 0\]

    \[ \frac{13}{4}p^4 - p^2\frac{k}{m}\bigg[ \frac{17\cdot 14 + 14\cdot 29 - 2\cdot 17\cdot 15}{36}\bigg] + \frac{406-289}{324}\frac{k^2}{m^2} = 0\]

    \[\frac{13}{4}p^4 - \frac{134}{36}p^2\frac{k}{m} + \frac{117}{324} = 0  \]

    \[\frac{13}{4}p^2 - \frac{67}{18}p^2\frac{k}{m} + \frac{39}{108}\frac{k^2}{m^2} = 0\]

    \[p^2 = \frac{\frac{67}{18} \pm \sqrt{(\frac{67}{18})^2 - 4\frac{13}{4}\cdot \frac{39}{108}}}{\frac{13}{2}}\frac{k}{m} = \frac{3.722 \pm \sqrt{3.722^2 - 4.694}}{6.5}\frac{k}{m}\]

    \[ p^2 = \frac{3.722 \pm 3.024}{6.5}\frac{k}{m} \]

    \[p_1^2 \leq 0.1074 \frac{k}{m}\]

    \[\implies p_1 \leq 0.3277\sqrt{\frac{k}{m}}\]

The answer is dependent on the assumption of mode shape

EXAMPLE: Use static deflection as mode.

Add masses to the diagram

    \[F = 5mg\]

    \[\Delta_1 = \frac{mg}k-5\frac{mg}k = -4\frac{mg}k\]

    \[\Delta_2 = 15\frac{mg}k\]

    \[\Delta_3 = 16\frac{mg}k\]

Therefore choose \phi_1 = \begin{bmatrix}4 \\ -15 \\ -16 \end{bmatrix}

    \[ m_{11} = m(4)(4) + m(-15)(-15) + m(-16)(-16) = 497m \]

    \[\begin{split} k_{11} &= k\begin{bmatrix} 4 && -15 && -16 \end{bmatrix} \begin{bmatrix} 1 && \frac{1}{3} && 0 \\ \frac{1}{3} && \frac{11}{9}  && -1 \\ 0 && -1 && 1 \end{bmatrix} \begin{bmatrix} 4 \\ 15 \\ 16 \end{bmatrix} \\ &= 27k \end{split} \]

Therefore:

    \[ \begin{split} p_1 &= \sqrt{\frac{27}{497}}\sqrt{\frac{k}{m}} \\ &\leq 0.23308\sqrt{\frac{k}{m}} \end{split} \]

It is sometimes useful to consider these approximate solutions to finding the eigenvalues and eigenvectors. The UAMM gives upper bounds as do most techniques. We can get lower bounds as well.

Consider again the equations and motion.

    \[ \begin{bmatrix} m \end{bmatrix} \begin{Bmatrix} \Ddot{q} \end{Bmatrix} + \begin{bmatrix} k \end{bmatrix} \begin{Bmatrix} q \end{Bmatrix} = \begin{Bmatrix} 0 \end{Bmatrix} \]

To get to an eigenvalue problem we can do it in two ways.

  1. Premultiply by \begin{bmatrix} k \end{bmatrix}^{-1} = \begin{bmatrix} a \end{bmatrix}

        \[ \begin{split} \begin{Bmatrix} \begin{bmatrix} a \end{bmatrix} \begin{bmatrix} m \end{bmatrix} && - \frac{1}{p^2} \begin{bmatrix} I \end{bmatrix} \end{Bmatrix} \begin{Bmatrix} Q \end{Bmatrix} &= 0 \\ \begin{bmatrix} a \end{bmatrix} \begin{bmatrix} m \end{bmatrix} &:= \begin{bmatrix} D \end{bmatrix} -\text{dynamical matrix} \end{split} \]

  2. Premultiply by \begin{bmatrix} m \end{bmatrix}^{-1}

        \[ \begin{split} \begin{Bmatrix} \begin{bmatrix} m \end{bmatrix}^{-1} \begin{bmatrix} k \end{bmatrix} && - p^2 \begin{bmatrix} I \end{bmatrix} \end{Bmatrix} \begin{Bmatrix} Q \end{Bmatrix} &= 0 \\ \begin{bmatrix} m \end{bmatrix}^{-1} \begin{bmatrix} k \end{bmatrix} &:= \begin{bmatrix} \Delta \end{bmatrix} -\text{system matrix} \end{split} \]

either gives the same eigenvalues and equivalent eigenvectors.

One can use these to find the eigenvalues/vectors through matrix iteration.

One clever idea that gives a lower bound is Dunkerleys formula. Consider the Dynamic Matrix formulation

    \[ \begin{Bmatrix} \begin{bmatrix} a \end{bmatrix} \begin{bmatrix} m \end{bmatrix} - \frac{1}{p^2}\begin{bmatrix} I \end{bmatrix} \end{Bmatrix} \begin{Bmatrix} Q \end{Bmatrix} = 0 \]

and expand the determinant of the coefficients

    \[ \begin{vmatrix} D_{11}-\frac{1}{p^2} && D_{12} && ... && D_{1n} \\ D_{21} && (D_{22}-\frac{1}{p^2} && ... && D_{2n} \\ \begin{matrix} . \\ . \\ . \end{matrix} && && && \\ D_{n1} && && && (D_{nn}-\frac{1}{p^2} \end{vmatrix} = 0 \]

    \[ \biggr( \frac{1}{p^2} \biggr)^n - \big( D_{11} + D_{22} + D_{33} + ... + D_{nn} \big) \biggr( \frac{1}{p^2} \biggr)^{n-1} + ... = 0 \]

Now consider that we have the roots of this polynomial p_1^2, p_2^2, ... p_n^2. This equation is equivalent to (factored)

    \[ \biggr( \frac{1}{p^2}-\frac{1}{p_1^2} \biggr) \biggr( \frac{1}{p^2}-\frac{1}{p_2^2} \biggr) \biggr( \frac{1}{p^2}-\frac{1}{p_3^2} \biggr)  \biggr( \qquad \biggr) ... = 0 \]

and expand this to get

    \[ \biggr( \frac{1}{p^2} \biggr)^n - \biggr( \frac{1}{p_1^2} + \frac{1}{p_2^2} + \frac{1}{p_3^2} + ... + \frac{1}{p_n^2} \biggr) \biggr( \frac{1}{p^2} \biggr)^{n-1} + ... = 0 \]

Now equate the coefficients

    \[ tr D = \biggr( \frac{1}{p_1^2} + \frac{1}{p_2^2} + ... + \frac{1}{p_n^2} \biggr) \]

However the largest term is the first one and often dominates

Therefore:

    \[ \begin{split} \frac{1}{p_1^2} &< tr D \\ p_1^2 &> \frac{1}{tr D} \end{split} \]

A special case occurs when the mass matrix is diagonal. Then tr D is

    \[ tr D = a_{11}m_1 + a_{22}m_2 + ... + a_{nn}m_{nn} \]

and we need calculate only a_{ii}, i=1 \text{ton}

In this case there is a physical interpretation of the expressions

    \[ \frac{1}{p_1^2} \leq a_{11}m_1 + a_{22}m_2 + a_{33}m_3... + a_{nn}m_{nn} \]

namely that a_{11}m_1 is the natural frequency of the system with only m_1 as the mass (all other masses set to zero).

Therefore:

    \[ \frac{1}{p_1^2} \leq \frac{1}{p^2_{11}} + \frac{1}{p^2_{22}} + ... + \frac{1}{p^2_{nn}} \]

This is also sometimes shown as Dunkerley’s formula.

Consider previous example

Calculate a_{ii}

  1. Apply unit load to m_1

        \[ \begin{split} \Delta_1 &= \frac{1}{k} + \frac{1}{k} = \frac{2}{k} = a_{11} \\ \Delta_2 &= -\frac{3}{k} = \Delta_3 \end{split} \]

Apply unit load to m_2

    \[ \begin{split} F(L) &= 1(3L) \\ F &= 3 \\ \hat{\Delta} &= \frac{3}{k} \\ a_{22} &= \frac{g}{k} \end{split} \]

Apply unit load to m_3

    \[ \begin{split} \hat{\Delta} &= \frac{g}{k} \\ a_{33} &= \frac{9}{k} + \frac{1}{k} = \frac{10}{k} \end{split} \]

Therefore:

    \[ \begin{split} \frac{1}{p_1^2} &< m(\frac{2}{k}) + m(\frac{9}{k}) + m(\frac{10}{k}) = \frac{21m}{k} \\ p_1^2 &> \sqrt{\frac{1}{21}}\sqrt{\frac{k}{m}} = \underline{0.2182\sqrt{\frac{k}{m}}} \end{split} \]

Consider an example of a continuous system

    \[ \begin{split} \phi_1 &= x^2 \\ \Ddot{\phi_1} &= 2 \\ \phi_2 &= x^3 \\ \Ddot{phi_2} &= 6x \end{split} \]

    \[ \begin{split} m_{ij} &= \int_m \phi_i\phi_j dm \\ &= \int_0^L \phi_i\phi_J mdx \qquad \text{where } m= \text{ mass/length} \end{split} \]

    \[ \begin{split} m_{11} &= m\int_0^L x^4 dx = \frac{mL^5}{5} \\ m_{12} &= m\int_0^L x^5dx = \frac{mL^6}{6} = m_{21} \\ m_{22} &= \frac{mL^7}{7} \end{split} \]

    \[ \begin{split} k_{ij} &= \int EI\phi_i''\phi_j'' dx \\ k_{11} &= EI \int_0^L 4x dx\\ &= EI(4L) \\ k_{12} &= EI\int_0^L 12x dx \\ &= 6EIL^2 \\ k_{22} &= 12L^3EI \end{split} \]

    \[ mL^5 \begin{bmatrix} 1/5 && L/6 \\ L/6 && L^2/7 \end{bmatrix} \begin{Bmatrix} \Ddot{q}_1 \\ \Ddot{q}_2 \end{Bmatrix} + EIL \begin{bmatrix} 4 && 6L \\ 6L && 12L^2 \end{bmatrix} \begin{Bmatrix} q_1 \\ q_2 \end{Bmatrix} = \begin{Bmatrix} 0 \\ 0 \end{Bmatrix} \]

Therefore the eigenvalue problem is

    \[ \begin{bmatrix} 4 - \frac{\lambda}{5} && 6L - \frac{\lambda L}{6} \\ 6L - \frac{\lambda L}{6} && 12L^2 - L^2 \frac{\lambda}{7} \end{bmatrix} \begin{Bmatrix} Q_1 \\ Q_2 \end{Bmatrix} = \begin{Bmatrix} 0 \\ 0 \end{Bmatrix} \]

    \[ \begin{split} (4-\frac{\lambda}{5})L^2(12-\frac{\lambda}{7}) - L^2(6-\frac{\lambda}{6})^2 &= 0 \\ 48 - \frac{12}{5}\lambda - \frac{4}{7}\lambda + \frac{\lambda^2}{35} - (36-2\lambda + \frac{\lambda^2}{36}) &= 0 \\ 12 - \frac{(84+20-70)}{35}\lambda + \lambda^2(\frac{1}{35}-\frac{1}{36}) &= 0 \\ 12(35)(36) - (34)(36)\lambda + \lambda^2 &= 0 \end{split} \]

    \[ \lambda^2 - (34)(36)\lambda - 12(35)(36) \]

    \[ \begin{split} \lambda &= (34)(18) \pm \sqrt{(34)^2(18)^2 - 12(35)(36)} \\ &= 612 \pm \sqrt{359424} \\ &= 612 \pm 599.5 \\ &= 12.48, 1211.5 \end{split} \]

Therefore

    \[ \begin{split} p_1^2 &= \sqrt{12.48} \sqrt{\frac{EI}{mL^4}} \\ &= 3.53\sqrt{\frac{EI}{mL^4}} \\ p_2^2 &= 34.8 \sqrt{\frac{EI}{mL^4}} \end{split} \]

    \[ (4-\frac{\lambda}{5})q_1 + (6-\frac{\lambda}{6})Lq_2 = 0 \]

    \[ \begin{split} \frac{Lq_2}{q_1} &= \frac{-(4-\frac{\lambda}{5})}{(6-\frac{\lambda}{6})} \\ &= \frac{-(4-\frac{12.48}{5})}{6-\frac{12.48}{6}} \\ &= -0.384 \end{split} \]

The mode shape is

    \[ \phi = x^2 - 0.384x^3 \]

What if we apply some dynamic loading to the structure? Then we must calculate the generalized form for our model. Consider a virtual displacement

    \[ \delta_y = \sum_{i=1}^n \phi_i(x)\delta q_n \]

then

    \[ \begin{split} \delta W &= \sum_{j=1}^m F_j \biggr( \sum_{i=1}^n \phi_i (x_j) \delta_{q_i} \biggr) \\  &= \sum_{i=1}^n \delta_{q_i} \biggr( \sum_{j=1}^m F_j \phi_i (x_j) \biggr) \\ &= \sum_{i=1}^n Q_i \delta q_i \end{split} \]

Therefore

    \[ Q_i = \sum_{j=1}^m F_j \phi_i (x_j) \]

Consider the previous example

    \[ \begin{split} Q_1 &= F(t) \phi_1 (L) = F(t) L^2 \\ Q_2 &= F(t) \phi_2 (L) = F(t) L^3\end{split} \]

    \[ \begin{bmatrix} m \end{bmatrix} \begin{Bmatrix} \Ddot{q} \end{Bmatrix} + \begin{bmatrix} k \end{bmatrix} \begin{Bmatrix} q \end{Bmatrix} = F(t) \begin{Bmatrix} L^2 \\ L^3 \end{Bmatrix} \]

Whirling of Rotating Shafts

Rotating shafts tend to bow out at certain speeds and whirl in a complicated manner. Whirling is defined as the rotation of the bent shaft and the line of centers of the bearing. This phenomenon results from various causes as mass unbalance, hysteresis dangers, gyroscopic effects, fluid friction, etc. This whirling can take place in the same or opposite direction as that of the rotation of the shaft. The whirling speed may or may not be equal to the rotation speed.

Let us assume a simple system

k is the stiffing of the system to lateral of flection.

We will assume the shaft to be rotating at an instant angular velocity \omega and OS to be whirling at angular speed \theta

    \[ \underline{v}_G = \underline{v}_O + v_{G/s} \]

Where \phi = \omega t - \theta

    \[ \begin{split} v_r^G &= \dot{r} - e\omega\sin(\omega t - \theta) \\ v_\theta^G &= r\dot{\theta} + \omega w\cos(\omega t - \theta) \end{split} \]

Therefore:

    \[ \begin{split} \Big( v^G \big)^2 &= \Big[ \dot{r} - e\omega \sin(\omega t - \theta)\Big]^2 + \Big[ r\dot{\theta} - e\omega \cos(\omega t - \theta)\Big]^2 \\ &= (\dot{r})^2 - 2\dot{r}e\omega\sin(\omega t - \theta) + (r\dot{\theta})^2 + 2r\dot{\theta}e\omega\cos(\omega t - \theta) + e^2\omega^2 \\ (v_G)^2 &= (\dot{r})^2 + (r\dot{\theta})^2 + e^2\omega^2 - 2\dot{r}e\omega\sin(\omega t -\theta) + 2r\dot{\theta}e\omega\cos(\omega t -\theta) \end{split} \]

Therefore:

    \[ \begin{split} T &= \frac{1}{2} I_G \omega^2 + \frac{1}{2}mv^2_G \\ V &= \frac{1}{2}kr^2 \\ L &= T - V \end{split} \]

Therefore:

    \[ \begin{split} \frac{d}{dt}\Big(\frac{\partial L}{\partial\dot{r}}\Big) - \frac{\partial L}{\partial r} &= Q_r \\ \frac{d}{dt}\Big(\frac{\partial L}{\partial\dot{\theta}}\Big) - \frac{\partial L}{\partial\theta} &= Q_\theta \end{split} \]

    \[ \begin{split} \frac{\partial (V_G)^2}{\partial \dot{\theta}} &= 2r^2\dot{\theta} + 2re\omega\cos(\omega t - \theta) \\ \frac{\partial (V_G)^2}{\partial \dot{r}} &= 2\dot{r} - 2e\omega\sin(\omega t - \theta) \\ \frac{\partial (V_G)^2}{\partial \theta} &= +2\dot{r}e\omega\cos(\omega t -\theta) + 2r\theta e\omega\sin(\omega t - \theta) \\ \frac{\partial (V_G)^2}{\partial r} &= 2r(\dot{\theta})^2 + 2\dot{\theta}e\omega\cos(\omega t - \theta) \end{split} \]

Therefore:

    \[ \begin{split} \frac{md}{2dt}\big[2\dot{r}-2e\omega\sin(\omega t-\theta)\big] - \frac{m}{2}\big[2r(\dot{\theta}^2)+2\dot{\theta} e\omega\cos(\omega t -\theta) + kr &= -c\dot{r} \\ m\Ddot{r} - me\omega\cos(\omega t - \theta)(\omega - \dot{\theta}) - mr(\dot{\theta})^2 - m\dot{\theta}e\omega\cos(\omega t - \theta) + kr &= 0 \\ m\Ddot{r} - mr(\dot{\theta})^2 - me\omega^2\cos(\omega t -\theta) + kr &= -c\dot{r} \end{split} \]

Therefore:

    \[ \Ddot{r} + \frac{k}{m}r + \frac{c}{m}\dot{r}-r(\dot{\theta})^2 = e\omega^2\cos(\omega t - \theta) \]

    \[ \frac{d}{dt}\Big(\frac{\partial L}{\partial \dot{\theta}}\Big) - \frac{\partial L}{\partial \theta} \]

    \[ \begin{split} &\frac{m}{2}\frac{d}{dt}\big(2r^2\dot{\theta} + 2re\omega\cos(\omega t -\theta)\big) - \frac{m}{2}\big[2\dot{r}e\omega\cos(\omega t - \theta) + 2r\dot{\theta}e\omega\sin(\omega t -\omega)\big] \\ &= m\big[ 2r\dot{r}\dot{\theta} + r^2\Ddot{\theta} + \dot{r}e\omega\cos(\omega t -\theta) - re\omega(\omega - \dot{\theta})\sin(\omega t - \theta)\big] \\ &- m\big[\dot{r}e\omega\cos(\omega t - \theta) + r\dot{\theta}e\omega\sin(\omega t - \theta)\big] \end{split} \]

Therefore:

    \[ mr\Ddot{\theta} + 2m\dot{r}\dot{\theta} - me\omega^2\sin(\omega t - \theta) = -cr\Ddot{\theta} \]

Therefore:

    \[ r\Ddot{\theta} + \frac{c}{m}r\dot{\theta} + 2\dot{r}\dot{\theta} = e\omega^2\sin(\omega t - \theta) \]

These are nonlinear equations and there are more than one solution.

Consider the simplest case of synchronous whirl i.e. try

    \[ \begin{split} \dot{\theta} &= \omega \qquad \theta = \omega t - \phi \\ \Ddot{\theta} &= 0 \qquad \Ddot{r} = \dot{r} = 0 \end{split} \]

Therefore:

    \[ \begin{split} \frac{k}{m}r - r\omega^2 &= e\omega^2\cos\phi \\ \frac{c}{m}r\omega &= e\omega^2\sin\phi \\ r\Big(\frac{k}{m} - \omega^2\Big) &= e\omega^2\cos\phi \end{split} \]

    \[ \begin{split} \tan\phi &= \frac{\frac{c}{m}r\omega}{r\big(\frac{k}{m}-\omega^2\big)} \\ p &= \frac{k}{m} \\ \zeta &= 2mp \end{split} \]

    \[ \begin{split} r &= \frac{e\omega^2}{\big(\frac{k}{m}-\omega^2\big)}\cos\phi \\ &= \frac{e\omega^2}{\big(\frac{k}{m}-\omega^2\big)} \frac{\big(\frac{k}{m}-\omega^2\big)}{\sqrt{\big(\frac{k}{m}-\omega^2\big)^2+\big(\frac{c}{m}\omega\big)^2}} \end{split} \]

    \[ r = \frac{e\big(\frac{\omega}{p}\big)^2}{\sqrt{\big(1-\big(\frac{\omega}{p}\big)^2\big)^2 + \big( Z\zeta\frac{\omega}{p}\big)^2 }} \]

This looks exactly the same as rotating inbalance excitation

Therefore the

So that synchronous whirl occurs when \frac{\omega}{p} \approx 1

Self Excited Vibrations

Almost all of the vibrations considered so far were either free or forced vibrations. A fundamentally different class are the so-called self-excited vibrations.

“In a self-excited vibration the alternating force that sustains the motion is created or controlled by the motion itself ; when the motion stops the alternating force disappears”

In a forced vibration the sustaining alternating force exists independently of the motion and persists even when the vibratory motion is stopped.

Insert diagram

Insert diagram

Engineered art

  • Use vibration phenomena for the display of classic ideas that ask the question “Why does that occur”
  • Search kinetic art
  • Naum Gabo Stationary Wave (1919)