DEVELOPING AN UNDERSTANDING OF

WASHING MACHINE DYNAMICS

Clive Marsh∗ , Steve Taylor† , Paul Milliken‡ and Galkadowite Senaratne

(Sena)§

Abstract

This report summarises the findings of one of the problems presented at the 2006
MISG. The problem, presented by Fisher and Paykel, was to develop a better un-
derstanding of the dynamics of modern washing machines that use ‘balance rings’.
Balance rings are toroidal chambers that are partially filled with fluid. The wash-
ing machine is a complicated system so several simplified systems were examined
instead of attempting to model the complete system. These approaches included
1 experimental work using a washing machine with transparent balance rings
and a strobe-light to observe the fluid-flow in the balance rings,
2 developing a 2D model of a balance ring and examining the system behaviour,
3 developing a simplified 3D model of a washing machine, without balance rings
or an out-of-balance mass, and examining the system behaviour,
4 developing a simplified model to describe waves that might occur in the fluid
within the balance rings.

1. Introduction
1.1. The product
The industry partner manufactures washing machines which have a
cylindrical drum (with a nominally vertical axis) suspended within a
cuboid outer box (called the envelope). The drum has a non-rotating
outer part and a rotating inner bowl where the clothes (load) are placed.

∗ Canesis Network Ltd, Christchurch, New Zealand. Email: marsh@canesis.com

† Dept of Mathematics, University of Auckland, Auckland, New Zealand. Email:
s.taylor@math.auckland.ac.nz
‡ Nicolaou Consulting Ltd, Rotorua, New Zealand. Email: paul.milliken@gmail.com
§ Massey University, Auckland, New Zealand. Email: G.Senaratne@massey.ac.nz

133
134

The centre of mass of the load will not usually lie on the axis of symmetry
of the inner drum and thus there will be an out of balance load (OOBL)
when the drum rotates. This causes the drum to rotate about an axis
other than its axis of symmetry and causes the motion to be eccentric.
The industry partner seeks to maximise the size of the inner drum
(to maximise capacity) whilst minimising the size of the envelope (space
required in laundry) and thus the clearance between the bowl and the
envelope needs to be minimised (it is typically 25mm). This constrains
the amount of eccentricity that can be tolerated before the drum collides
with the envelope. Eccentricity is reduced by the use of balancing rings
on the inner drum and by the design of the suspension system of the
drum.

1.2. Objective
The industry partner seeks a more comprehensive understanding of
the dynamics of the system and in particular what causes the motion to
be eccentric. They have always been able to successfully design balance
rings and suspension rods which keep the eccentricity to acceptable levels
for all reasonable OOBLs but the design approach has been somewhat
“trial and error”.

1.3. The approach

A modern washing machine with balance rings is a complicated system
so, rather than modelling the complete system, simplified models were
developed for various parts of the system. Thus, several approaches
were taken to gain a better understanding of the dynamics of a washing
machine. The four main approaches were
Experimental work was undertaken to observe the fluid flow in
the balance rings of a washing machine at a range of speeds and
with a range of out-of-balance loads. Fisher and Paykel supplied a
washing machine with transparent balance rings to facilitate this
experimentation.

A simplified 2D1 model of a washing machine, including balance

rings and an out-of-balance load, was developed and examined.

A simplified 3D2 model of a washing machine, without balance

rings or an out-of-balance load was developed.

1 2D refers to the fact that the model is assumed to lie completely in the horizontal plane.
2 3D refers to 3 spatial dimensions.
DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 135

A simplified model of the fluid in the balance rings was developed

which included an estimate of the speed at which surface waves
would travel in the balance rings.

These approaches are considered in the remainder of this report.

2. Experimental work
2.1. Introduction
Fisher and Paykel provided a washing machine with transparent bal-
ance rings and a strobe-light so that the group could observe the dy-
namics of the fluid within the rings.

2.2. Method and Main results

2.2.1 Background. Observations and the industry partner’s
experience suggest that eccentricity is relatively small during the low
speeds of the wash cycle (say below 30rpm) and during the high speeds
reached towards the end of the spin cycle (above 300rpm and up to
approximately 1,000rpm) but that there are two speed ranges (traversed
during the spin cycle) where the eccentricity can become large. The
first occurs typically between 30 and 60rpm and across a narrow speed
range (for any given configuration and load). At such low speeds the
drive motor has abundant torque and the undesirable eccentric motion
can be avoided by rapidly accelerating through the speed range where it
would occur. The second speed range occurs typically between 150 and
300rpm and (for any given configuration and load) prevails over a wider
speed range. This wide speed range combined with reduced motor drive
torque available at these higher speeds mean that it is not possible to
simply rapidly accelerate through the speed range and careful attention
needs to be paid to the design of the suspension and the balance rings
to achieve acceptable behaviour.

2.2.2 Observations at low speed. The low speed mode corre-

sponds to the natural frequency of the swing mode of the drum hanging
on the suspension rods (with or without drum rotation, the drum can
simply swing from the suspension rods). This natural frequency was
measured to be slightly below 1 Hz and the motion was excited at a
rotational speed of approximately 50rpm at which the out of balance
centrifugal force would excite the mode. The amplitude (and natural
frequency) of the motion would change with suspension characteristics
(stiffness and damping). However, the design of the suspension system
is a compromise between minimising the amplitude of the motion of
136

the drum relative to the envelope and minimising the transmission of

high frequency vibrations (hence noise) from the drum to the envelope.
Thus, though it would be possible to reduce the amplitude of the swing
by stiffening the suspension or increasing the damping, both would tend
to increase the transmission of high frequency vibrations to the envelope.
It should be noted that, at the speeds where the low-speed mode is
observed, the effect of gravity is greater than the centripetal acceleration
at the edge of the balance ring (due to the fact that the radius of the
balance ring was 250mm). The simple model in the “Simple Theory”
section of [1], though for a system with a massive base plate on the drum
and with horizontal suspension connections to the envelope, illustrates
this mode. A model to describe this behaviour for the modern designs
(no massive base plate) could be developed relatively easily. Overall,
the team suggest that the current suspension system design achieves
a reasonable compromise between suppressing eccentric motion in this
mode and minimising vibrations and noise at higher frequencies and
that the solution of rapidly accelerating through the critical speed is a
sensible one.

2.2.3 Observations at higher speed. Once higher speeds are

reached, centrifugal forces dominate the system. This is because the
radius of the balance ring is 250mm so that, at 135rpm, the centripetal
acceleration is 5g. Like a spinning top, the inner drum will tend to ro-
tate about its principal axis of inertia (axis about which the moment of
inertia is minimal). Due to the OOBL, the principal axis is dissimilar
to the axis of symmetry and hence the motion is eccentric. The balance
rings are able to partially counterbalance the OOBL and bring the prin-
cipal axis closer to the axis of symmetry (thus reducing eccentricity)
but it can be shown that they are not able to completely remove the
eccentricity. The model in MISG 2000 [2] provides a good description
of the operation of the balance rings and the resulting, reduced eccen-
tricity. A key assumption of this 2-D model is that the axis of rotation
will be vertical and stationary, thus the motion of the drum is a simple
1-degree of freedom (dof) rotation (albeit about an axis inclined to the
axis of symmetry). Key characteristics of such a motion are that the gap
(vertical and horizontal) between a fixed point on the envelope and the
upper rim of the drum is sinusoidal through time with fixed amplitude
and frequency equal to the rotational frequency of the inner drum and
the length of the suspension rods is also sinusoidal through time with
fixed amplitude and similar frequency. Further, for this motion, the wa-
ter in the balance rings will be stationary with respect to the ring and
its position and profile will not be speed dependant.
DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 137

Observations of the motion suggest that these displacements are sinu-

soidal, that the balancing water is stationary with respect to the balance
rings, and that the motion is the simple 1-dof motion predicted for both
high speeds (say above 300rpm) and for speeds below a critical thresh-
old (approximately 180rpm for the configuration of the experimental rig
available during the workshop). However, at the speeds where the eccen-
tricity is high (approximately 180rpm to 250rpm), the motion is not the
simple 1-dof rotation described by the MISG 2000 model. This is evident
as the gap between the drum and envelope does not vary sinusoidally
and the water moves within the balance rings.
Thus, it appears that the assumption made in the MISG 2000 model
about the axis of rotation being vertical and stationary is not valid at
the speeds where the problem occurs and a richer model is required to
describe the motion and understand the cause of the eccentricity. The
two degrees of rotational freedom, suppressed in the MISG 2000 model,
can be described as precession and nutation, in these degrees of freedom
the axis of rotation of the spin will itself pitch and roll. When precession
and nutation1 occur, the eccentricity (of the drum’s rotation viewed
in a horizontal plane) is likely to increase. There is no fundamental
reason why precession and nutation will not occur. A brief analysis
of a simple model developed to describe the dynamics and permitting
precession and nutation suggest that there are spin-speed dependant
terms which significantly affect the precession and nutation behaviour at
higher speeds. However, much further modelling and analysis is required
to provide a complete explanation.

2.2.4 Effects of suspension, fluid motion and motor-drive.

The suspension system will play a vital role isolating and damping the
resulting motion and it is possible that baffles which damp the motion
of the water within the balance rings will also affect the motion. This
is consistent with the industry partner’s experience of significant design
parameters which influence the eccentricity.
Further phenomena which were observed during the non-simple rota-
tional motion were that there appeared to be standing waves between
the baffles within the balance rings and that it appeared that the spin
speed of the drum was not constant.
The spin speed is regulated by feedback control which manipulates the
torque applied by the motor to drive the spin. Ideally, the spin speed
will be tightly controlled so that variations in spin speed do not create

1 Precession is when the axis of spin rotates so that it describes a cone shape. Nutation is

when the angle that the cone of precession describes changes periodically.
138

undesirable motions, and potential disturbances to spin speed, caused

by undesirable motions, are rejected by the control system.
It is not certain whether the standing waves in the balance ring sectors
are an effect or a cause of the non-simple rotational motion. However,
since the water is stationary within the balance rings when the rotational
motion is simple (outside the 180 to 250rpm range), there would have
to be some self-generating effect for the waves to be the cause of the
non-simple motion as the spin speed entered the 180 to 250rpm range.

2.3. Summary
At high speed (say greater than 250rpm) and below a threshold (typi-
cally 180rpm), the motion is predominantly a simple 1dof rotation with
the axis of rotation slightly tilted with respect to the axis of symmetry
and slight (but acceptable) eccentricity. The balance rings operate well
with the water stationary within them and the MISG 2000 model ade-
quately describes the behaviour of the system. At intermediate speeds
(say 180 to 250rpm) the motion is non-simple with excessive (potentially
unacceptable) eccentricity. Three candidate causes have been identified:

1 The motion is naturally non-simple and precession and nutation

naturally occur at these speeds.

2 Waves occur in the balance ring sectors and these force the rota-
tional motion to be non-simple.

3 The spin speed control is ineffective and spin speed fluctuations

are the cause.

The third potential cause is essentially a stand-alone issue and can be

analysed and resolved independently. A richer model could be devel-
oped which describes the non-simple motion and how it is influenced by
suspension system and balance ring design. This would be potentially
useful to the industry partner as a design aid.

3. 2D model of a rigid-body system

3.1. Introduction
An analysis of a simplified 2D model of a balance ring was considered.
This analysis illustrated the benefits of using balance rings to reduce
vibration in a washing machine.
DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 139

3.2. Main result

The analysis conducted by the group produced almost identical results
to [2], so the reader is referred to that paper for a good summary of the
2D case.

4. 3D model of a rigid-body system

4.1. Introduction
A model of a 3D rigid-body washing machine was considered. Thus,
the model was simplified by removing balance rings and out-of-balance
loads. This approach was useful because it helped the group to determine
if certain effects were likely to be due to rigid-body motion of the washing
machine or due to dynamic effects in the balance rings. Although a
simplified model was considered, the equations of motion were found to
be non-linear and highly coupled.

Figure 1. Sets of unit vectors used in the model

4.2. Preliminaries
First, some notation must be defined. Figure 1 shows a set of unit
vectors E = {e1 , e2 , e3 } fixed in the earth and a set of unit vectors
B = {b1 , b2 , b3 } such that b3 is in the direction of the axis of symmetry of
the bowl and b2 is perpendicular to e3 . The steps in the transformation
required to get from E to B are:

1 Start with the axis set defined by E and rotate the axes by an angle
φ about the e3 axis. Call this intermediate axis set C = {c1 , c2 , c3 }.
2 Then rotate the axes by an angle θ about c2 . This axis set is B.

Thus, the relationship between B and E is given by (1).

140
 
cos φ sin φ 0
b =  − sin φ cos θ cos φ cos θ sin θ  e. (1)
sin θ sin φ − sin θ cos φ cos θ

Figure 2. Sketch of the bowl being supported by springs

Now, suppose the bowl is spinning with an angular speed of ψ̇ about

the b3 axis. Therefore, it can be seen that φ̇ is the precession speed and
θ̇ is the nutation speed of the motion.
Note that, in Figure 1, the vector b2 is in the plane defined by e1 and
e2 .
Figure 2 shows that the bowl is supported by four springs which coun-
teract the weight of the bowl. There are seven assumptions that are
used for the model. They are:

1 The springs that support the washing machine bowl are oriented
vertically and are assumed to be long so that they remain vertical
at all times. See Figure 2.

2 There is no damping or friction in the system.

3 The bowl is allowed to spin freely and there are no external mo-
ments such as motor torque acting on the bowl.

4 The vertical bounce mode of the bowl and any horizontal transla-
tions of the bowl are ignored.

5 The height of the centre of gravity of the washing machine bowl is

the same as the height where the springs attach to the bowl when
the bowl is at rest in its equilibrium position.
DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 141

6 There is no out-of balance mass so that the moment of inertia

about the b3 axis is I3 and the moments of inertia about the b1
and b2 axes are both I⊥ . Furthermore, I3 < I⊥ . See Figure 1.
7 The washing machine bowl is a rigid body so balance-rings and
fluid in the bowl are not included in the model.
Although the assumptions make the model considerably simpler than
a real washing machine, the simplified model does contain enough detail
to provide a good amount of insight into the likely behaviour of a real
machine. In particular, precession and nutation are permitted in this
model.

Figure 3. Pictorial representation of the radius of the bowl and stiffness of the
springs

4.3. Main Result

This section presents the equations of motion for the washing machine
(cf. Figure 3).
Theorem 1 The equations of motion are given by the set of non-linear
differential equations (2), (3), (4).
sin θ  
θ̈ = (I⊥ − I3 ) cos θ φ̇2 − I3 ψ̇ φ̇ − 2Rk cos θ , (2)
I⊥
(2I⊥ + I3 ) cos θ
φ̈ = − φ̇θ̇, (3)
I⊥ sin θ
 
(2I⊥ + I3 ) cos θ
ψ̈ = + sin θ φ̇θ̇. (4)
I⊥ sin θ
142

Proof: The potential energy U stored in the springs is given by

U = Rk sin2 θ (5)

The kinetic energy T is given by summing the kinetic energy of rota-

tion in the direction of each of the principal moments of inertia. These
principal moments of inertia are in the directions of the three unit vectors
b1 , b2 and b3 .
To calculate the kinetic energy, first consider the angular velocity ω
where

ω = θ̇b1 + φ̇e3 + ψ̇b3 (6)

Rearranging (1) and substituting for e3 in (6) gives

 
ω = θ̇b1 + φ̇ sin θb2 + ψ̇ + φ̇ cos θ b3 . (7)

Thus, the kinetic energy is

1 1 1  2
T = I⊥ θ̇ 2 + I⊥ φ̇2 sin2 θ + I3 ψ̇ + φ̇ cos θ . (8)
2 2 2
Equations (5) and (8) may be used to calculate the Lagrangian,

L = T −U (9)
1 1 1  2
= I⊥ θ̇ 2 + I⊥ φ̇2 sin2 θ + I3 ψ̇ + φ̇ cos θ − Rk sin2 θ. (10)
2 2 2

Hamilton’s principle of least action gives the Euler-Lagrange equations

for the system as (11) to (13).

∂L d ∂L
0 = − , (11)
∂θ dt ∂ θ̇
∂L d ∂L
0 = − , (12)
∂φ dt ∂ φ̇
∂L d ∂L
0 = − . (13)
∂ψ dt ∂ ψ̇

The following partial derivatives are useful in the sequel:

DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 143

∂L  
= sin θ (I⊥ − I3 ) φ̇2 cos θ − ψ̇ φ̇I3 − 2Rk cos θ , (14)
∂θ
∂L
= 0, (15)
∂φ
∂L
= 0, (16)
∂ψ
∂L
= I⊥ θ̇, (17)
∂ θ̇
∂L  
= I⊥ φ̇ sin2 θ + I3 ψ̇ + φ̇ cos θ cos θ, (18)
∂ φ̇
∂L  
= I3 ψ̇ + φ̇ cos θ . (19)
∂ ψ̇

Differentiating (17), (18) and (19) with respect to time gives:

d ∂L
= I⊥ θ̈, (20)
dt ∂ θ̇
d ∂L
I⊥ sin2 θ + I3 cos2 θ φ̈ + I3 cos θ ψ̈


=
dt ∂ φ̇
+2 (I⊥ − I3 ) sin θ cos θ θ̇φ̇, (21)
d ∂L  
= I3 ψ̈ + φ̈ cos θ − φ̇θ̇ sin θ . (22)
dt ∂ ψ̇

Evaluating (11), (12) and (13), some algebraic manipulations yield

(2), (3) and (4) as required. ⊓
⊔

4.4. Discussion
Examining the equations of motion for the 3D rigid-body model (2)
to (4), it can be seen that there is a spin motion, precession and nuta-
tion. These equations are non-linear and coupled so characterising the
motion is difficult. It is believed that computer simulations would be
the best way to examine this system further. Unfortunately, there was
not sufficient time to complete computer simulations during the MISG
meeting. Once an understanding of the equations of motion have been
gained, additional complexity, such as an out-of-balance load, could be
added to the model. This method would allow the effects of changes to
the model to be assessed.
144

5. Fluid (wave) analysis

5.1. Introduction
The experimental work showed that the fluid was not stationary rel-
ative to the washing machine bowl when excessive vibrations were oc-
curring. This suggested that there was some interaction between the
washing machine system (excluding the balance rings) and the fluid in
the balance rings such that waves were occurring in the fluid. Equations
for the water wave frequency were developed for a simplified system.

5.2. Method
We follow [4]. We wish to study the resonances of the fluid motion
under the assumption that the drum is rotating with constant angular
velocity about a fixed axis. If such resonance modes exist, it is possible
that they could be excited by perturbations to the rotational motion of
the drum. If this were to happen, the resulting oscillations could grow in
amplitude to those observed for rotational frequencies between 180rpm
and 250rpm. Thus our goal is to see if there are any such resonances for
rotational frequencies in this range.

5.3. Analysis
Let q be the velocity of water in the balance rings measured in a
co-ordinate system which is fixed relative to the motion of the drum.
We assume that the drum is rotating with constant angular velocity Ω.
Thus the vector Ω points in the direction of the axis of rotation (which
is vertical) and its length ω = |Ω| is the angular speed.
The equations of motion of the fluid in this co-ordinate system are for
conservation of mass (water is assumed to be incompressible)
∇·q =0 (23)
and conservation of momentum
∂q 1
+ q · ∇q + 2Ω × q = − ∇p − ν∇ × (∇ × q). (24)
∂t ρ
Here ν and ρ are respectively the kinematic viscosity and density of
water. The scalar p is the reduced pressure which is related to the actual
pressure P by the equation
1
p = P + ρgz − ρω 2 r 2 , (25)
2
in which r and z are the radial and axial co-ordinates in a cylindrical
co-ordinate system with axis Ω.
DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 145

5.3.1 Steady State Motion. The ideal state of the system is

for the fluid to be stationary relative to the rotating reference frame.
In this case, q = 0 and (24) shows that ∇p = 0 so that p = constant.
Equation (25) gives the actual pressure within the rotating liquid.
On the air-liquid interface within the balance rings, the air pressure P0
must be the same as the liquid pressure. Thus (25) gives us an equation
for the shape of this interface:

ω2 2
z= r + constant, (26)
2g
showing that the shape is a paraboloid.
When ω = 0 the paraboloid of course degenerates to a plane horizontal
surface. But we are more interested in what happens near 200rpm, for
which ω ≈ 21. The balance rings have a radius of approximately 220mm,
so the slope of the interface,
dz ω2r
= ≈ 10,
dr g
is quite steep at such frequencies.

5.3.2 Waves: Equations and boundary conditions. As-

suming that the nonlinear and viscous effects of the motion may be
ignored, equation (24) becomes
∂q 1
+ 2Ω × q = − ∇p. (27)
∂t ρ
This is a reasonable approximation provided that the two dimensionless
parameters

ν U
E= (Ekman number), ǫ= (Rossby number),
ωL2 ωL
are small. Here, U and L represent typical velocity and length scales for
the motion of the liquid.
We seek oscillatory solutions

q = Qeiλt ,
p = constant + Φeiλt ,

where Q and Φ are complex-valued functions of the spacial coordinates

and it is understood that the physical velocity and reduced pressure are
the real parts of q and p.
146

Substituting these expressions into (23) and (27) gives the equations
that must be satisfied by Q and Φ:

∇ · Q = 0, (28)
1
iλQ + 2Ω × Q = − ∇Φ. (29)
ρ
The normal flux of water at solid boundaries must be zero. Thus if Γ0
denotes the portion of the inner surface of the balance rings in contact
with the water, and if n denotes the outward unit normal vector on Γ0
then we must have
n.Q = 0 on Γ0 . (30)
Let Γ1 denote the water surface in contact with air inside the balance
rings. Fluid particles on this surface will stay on this surface so they
experience constant pressure P = P0 . Thus dP/dt = 0 on Γ1 , where
d/dt denotes the time derivative following the water,
d ∂
= + q · ∇.
dt ∂t
Applying this operator to (25), we find

dp 1
= q · ∇(ρgz − ρω 2 r 2 ) on Γ1 .
dt 2
But the unit normal n to the steady state surface (26) is given by

∇(ρgz − 21 ρω 2 r 2 )
n= .
|∇(ρgz − 12 ρω 2 r 2 )|
Further,
dp ∂p ∂p
= + q · ∇p ≈ .
dt ∂t ∂t
Thus the free surface boundary condition for our linear waves is given
by
iλΦ = Θn · Q on Γ1 , (31)
where Γ1 is given approximately by (26) and
1 p
Θ = |∇(ρgz − ρω 2 r 2 )| = ρ g2 + ω 4 r 2 .
2
We see then that the water waves must satisfy equations (28), (29)
and the two boundary conditions (30), (31).
DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 147

5.3.3 Waves: Rough calculation. Figure 4 shows schemati-

cally how the two concentric balance rings are divided into cells. Narrow
slots connecting adjacent cells in each ring allow the water to flow around
each ring. The slots provide a simple damping mechanism for large scale
movement of the fluid in the rings.

Figure 4. The shape of the cells within balance rings

The videos provided by Fisher & Paykel and the experiments per-
formed at MISG showed strong fluid oscillations within the cells during
the resonant motion of the washing machine. The fluid flow for these
oscillations appeared to be two dimensional, the direction of motion be-
ing perpendicular to the axis of rotation of the drum. The fluid flow
from one cell to another seemed to be on a slower time scale (period ≈ 4
seconds) than the oscillations within each cell.
These observations partly justify some simplifying assumptions:
The flow is two dimensional.
The oscillations of fluid within each cell are not significantly af-
fected by the flow of fluid through the slots connecting adjacent
cells. Hence we ignore the slots in this analysis.
The shape of the cells themselves will affect the oscillation frequency.
However, in order to get a rough idea of the nature of these oscillations,
we assume a simplified geometry for the cells. This is shown in Figure
5, which also illustrates a local co-ordinate system for a cell.
With respect to this co-ordinate system, equations (28), (29) for Q =
(Q1 , Q2 , 0) and Φ take the form
∂Q1 ∂Q2
+ = 0, (32)
∂x ∂y
1 ∂Φ
iλQ1 + 2ωQ2 = − , (33)
ρ ∂x
1 ∂Φ
iλQ2 − 2ωQ1 = − . (34)
ρ ∂y
148

y direction of rotation

free surface

0 L x
Figure 5. Simplified cell geometry and local co-ordinate system

We noted in Section 5.3.1 that the free surface is almost cylindrical

for the large values of ω that interest us. Given that the cells are small,
we take this approximation one step further and approximate the steady
free surface within a cell by a plane parallel to the axis of rotation. The
approximate version of Equation (31) takes the form
iλΦ = ρω 2 RQ2 , (35)
where R is the balance ring radius. On the solid boundaries x = 0,
x = L we must have Q1 = 0 and on y = 0 we must have Q2 = 0.
In terms of the dimensionless variables
x y Q1 Q2 Φ λ
ξ= , η= , u1 = , u2 = , φ= , σ= ,
L L U U ρωU L ω
our equations take the form
∂u1 ∂u2
+ = 0, (36)
∂ξ ∂η
∂φ
iσu1 + 2u2 = − , (37)
∂ξ
∂φ
iσu2 − 2u1 = − . (38)
∂η
The boundary conditions are
u1 (0, η) = u1 (1, η) = 0, 0 < η < h, (39)
u2 (ξ, 0) = 0, 0 < ξ < 1, (40)
σL
u2 (ξ, h) = i φ(ξ, h), 0 < ξ < 1, (41)
R
where η = h ≈ constant is the dimensionless equation for the free sur-
face. We may think of h as being the radial depth of the water in a cell
divided by the cell’s length L.
DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 149

The flow is two dimensional so the equation of continuity (36) implies

the existence of a stream function ψ such that
∂ψ ∂ψ
u1 = − , u2 = .
∂η ∂ξ
This allows us to write (37), (38) as
∂ ∂ψ
(φ + 2ψ) = iσ ,
∂ξ ∂η
∂ ∂ψ
(φ + 2ψ) = −iσ .
∂η ∂ξ
These equations imply that i(φ + 2ψ)/σ is a complex velocity potential,
i.e.
i ∂ i ∂
u1 = (φ + 2ψ), u2 = (φ + 2ψ).
σ ∂ξ σ ∂η
The equations also show that φ and ψ satisfy Laplace’s equation
∂2φ ∂2φ ∂2ψ ∂2ψ
+ 2 = 0, + = 0.
∂ξ 2 ∂η ∂ξ 2 ∂η 2
Remark 1 One could make use of this potential flow. It allows the
use of the conformal mapping technique to treat cell geometries that are
more realistic than that shown in Figure 5. This technique is discussed
in many fluid mechanics and complex analysis texts, such as [3].
The solid boundary conditions (39) and (40) are satisfied if we take
ψ to be zero on the solid boundaries. The free boundary condition (41)
takes the form
∂ψ iσL
= φ if η = h. (42)
∂ξ R
The technique of separation of variables furnishes an infinite sequence
of solutions that satisfy the solid boundary conditions:
ψn (ξ, η) = sin(nπξ) sinh(nπη)
φn (ξ, η) = −iσ cos(nπξ) cosh(nπη) − 2 sin(nπξ) sinh(nπη),
for n = 1, 2, . . . , ∞. We must take an infinite linear combination of these
functions to find solutions that satisfy the free boundary condition
∞ ∞
X bn X bn
ψ(ξ, η) = ψn (ξ, η), φ(ξ, η) = φn (ξ, η),
sinh(nπh) sinh(nπh)
n=1 n=1

where the sinh(nπh) terms in the denominators are introduced to sim-

plify the following calculations.
150

The free boundary condition (42) implies that

∞ ∞
X X σ2 L
nπbn cos(nπξ) = bn coth(nπh) cos(nπξ) (43)
n=1 n=1
R
2iσL
− bn sin(nπξ), 0 < ξ < 1.
R
We are interested in non-trivial solutions of this equation and such so-
lutions will only exist for certain values of the eigenvalue parameter σ.
It is more convenient to work with an equivalent equation obtained by
multiplying (43) by cos(mπξ) and integrating the resulting equation over
[0, 1]:
∞
σ2 L
 
−8iσL X
mπ − coth(mπh) bm = bn Γmn , m = 1, 2, . . . , ∞,
R R n=1
(44)
where (
0, if n + m is even,
Γmn = n
π(n2 −m2 )
, if n + m is odd.
For computational purposes it is sufficient to take 1 ≤ m, n ≤ N
for some sufficiently large value of N . If this is done then the non-
dimensional eigenfrequencies σ are simply the zeros of the determinant of
the coefficient matrix for the equations giving the constants bm . We used
this procedure to plot the first three eigenfrequencies against relative
radial depth h in Figure 6.
Figure 6 indicates that for certain values of h, we can expect relative
eigenfrequencies close to 1. This corresponds to the liquid oscillating
with the rotational frequency ω, which is the predominant forcing fre-
quency of the system. Thus we cannot rule out the possibility that
there is a resonance between the cellular oscillations of the liquid and
the forced motion of the drum which leads to the unwanted motion. We
note that we have not shown that this is necessarily the cause of the
unwanted motion.

6. Conclusions
Modern washing machines with balance rings are complicated sys-
tems and it was not possible to build a 3D model including balance
rings and an out-of-balance load in the time available.

[2] showed that balance rings can reduce eccentricity due to an

out-of-balance-load but cannot eliminate it.
DEVELOPING AN UNDERSTANDING OF WASHING MACHINE DYNAMICS 151

The first three eigenfrequencies

6

5
nondimensional eigenfrequency σ

σ
2 1
σ2
σ
3

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
relative depth h

Figure 6. The first 3 eigenfrequencies as a function of relative radial depth h. We

used L/R = 2π/24 which corresponds to an average relative cell length when there
are 24 cells.

A 3D model of a washing machine without balance rings, and

without an out-of-balance-load showed that the motion can be de-
scribed as a combination of spin, precession and nutation.

The equations of motion for the simplified 3D model were non-

linear. Furthermore, linearising the equations removed the motion
of interest (precession and nutation).

It is clear that the motion of the water in the balance rings exac-
erbates the resonant behaviour of the washing machine. It is also
clear that this is because the water moves away from its balancing
state, possibly even adding to the effect of the out of balance load.
Our analysis of water oscillations within the balance ring does not
rule out the possibility that the oscillations themselves start the
unwanted motion. However, further work is needed to see if this
is the case.

A good starting point for future research on the effect of the bal-
ance rings is Equation 33, which for low frequency (λ ≈ 0) oscilla-
152

tions becomes
1 ∂Φ
2ωQ2 = − .
ρ ∂x
This shows that motion of the balance ring in the y-direction in-
duces a pressure gradient in the x-direction. It is conceivable that
such pressure gradients cause pressure jumps between adjacent
cells which force the motion of water from one cell to another.
It would be interesting to analyse this effect more closely to see if
it is the mechanism for the onset of the resonance.

Acknowledgements

We gratefully acknowledge the support of the following MISG partic-

ipants in this work: Andrew Bassom, Philip Broadbridge, Ray Hoare,
Nigel Holland, Ian Howells, Peter Howell, Alex McNabb, Dion O’Neale,
Brett Ryland, Peter Sale and Shixiao Wang.

We are especially grateful to our Fisher and Paykel representatives,

Bruce Otte and Philip Renwick, for bringing this interesting problem to
MISG. Their video recordings, experiments and informative discussions
were excellent.

References
[1] VanKirk J. and Burmeister L., “An automatic balancer design for a vertical-axis
clothes washing machine”, Proceedings of ASME Design Engineering Conference
& Show, Chicago, Ill., (1994).
[2] Whiten W. J. and Broadbridge P., “A washing machine balancing problem”,
Proceedings of MISG 2000, Adelaide, (2000).
[3] Churchill R. V. and Brown J. W., Complex variables and applications, 4th Ed.,
McGraw-Hill Book Co., New York, (1984).
[4] Greenspan H.P., The theory of rotating fluids, Reprint of the 1968 original,
Cambridge Monographs on Mechanics and Applied Mathematics, Cambridge
University Press, (1980).