Applied Ergonomics 30 (1999) 235—245

Factors affecting minimum push and pull forces of manual carts

Khaled W. Al-Eisawi *, Carter J. Kerk
, Jerome J. Congleton, Alfred A. Amendola,
Omer C. Jenkins, Will Gaines
SABRE Technology Solutions, 1 East Kirkwood Blvd. MD 7385, Southlake, TX 76092, USA

South Dakota School of Mines and Technology, Industrial Engineering Program, Rapid City, SD 57701-3995, USA
Safety Engineering Program, Department of Nuclear Engineering, Texas A&M University, College Station, TX 77843-3133, USA
Department of Statistics, Texas A&M University, College Station, TX 77843-3143, USA
Scott & White, Department of Occupational Medicine, 1600 University Drive East, College Station, TX 77840, USA
Received 18 February 1997; accepted 2 March 1998

Abstract

The minimum forces needed to manually push or pull a 4-wheel cart of differing weights with similar wheel sizes from a stationary
state were measured on four floor materials under different conditions of wheel width, diameter, and orientation. Cart load was
increased from 0 to 181.4 kg in increments of 36.3 kg. The floor materials were smooth concrete, tile, asphalt, and industrial carpet.
Two wheel widths were tested: 25 and 38 mm. Wheel diameters were 51, 102, and 153 mm. Wheel orientation was tested at four levels:
F0R0 (all four wheels aligned in the forward direction), F0R90 (the two front wheels, the wheels furthest from the cart handle, aligned
in the forward direction and the two rear wheels, the wheels closest to the cart handle, aligned at 90° to the forward direction), F90R0
(the two front wheels aligned at 90° to the forward direction and the two rear wheels aligned in the forward direction), and F90R90 (all
four wheels aligned at 90° to the forward direction). Wheel width did not have a significant effect on the minimum push/pull forces.
The minimum push/pull forces were linearly proportional to cart weight, and inversely proportional to wheel diameter. The
coefficients of rolling friction were estimated as 2.2, 2.4, 3.3, and 4.5 mm for hard rubber wheels rolling on smooth concrete, tile,
asphalt, and industrial carpet floors, respectively. The effect of wheel orientation was not consistent over the tested conditions, but, in
general, the smallest minimum push/pull forces were measured with all four wheels aligned in the forward direction, whereas the
largest minimum push/pull forces were measured when all four wheels were aligned at 90° to the forward direction. There was no
significant difference between the push and pull forces when all four wheels were aligned in the forward direction.  1999 Elsevier
Science Ltd. All rights reserved.

Keywords: Push and pull forces; Carts; Rolling friction

1. Introduction
Most of the ergonomics studies that have been per-
formed on pushing and pulling, including the psycho-
physical guidelines developed by Snook and Ciriello
(1991) for the design of manual material handling tasks,
report results in terms of hand force exertions. When
designing a cart pushing or pulling task, a job designer
must determine the maximum weight with which to load
the cart so that the hand forces needed to push or pull the
cart do not exceed safe limits. In designing or selecting
* Corresponding author. Tel.: 001 817 264 3590; fax: 001 817 264 3802.
a cart, one should select a wheel diameter that will keep
the hand forces exerted in pushing or pulling the fully
loaded cart below safe limits. This is not an easy task
because the hand forces required to push or pull a cart
depend on many factors such as wheel diameter, wheel
orientation, floor material, and wheel material. This
study attempted to facilitate the determination of hand
forces from factors that are easier to measure including
cart weight, wheel diameter, floor material, and wheel
orientation.
The resistance between rolling wheels and the floor
surface is termed ‘rolling friction’. Its laws are not yet
definitely established. A widely acceptable relationship is
that if ¼ is the total weight, r the wheel radius, and o
the coefficient of rolling friction, then F is equal to the

0003-6870/99/$ — see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 3 - 6 8 7 0 ( 9 8 ) 0 0 0 1 9 - 2
236 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245
resistance applied at the circumference of the wheel and is
given by
¼ ings. Wheel bearings are typically well lubricated and
F"o (1)
r provide very little resistance to rolling as compared to
To keep the wheel rolling in uniform motion without
slipping, an external force equal to F must be applied.
Notice that o has units of length while the coefficient of
static friction and the coefficient of sliding friction are
dimensionless quantities. o/r is called the coefficient of
resistance to rolling (Kragesky et al, 1982). The deriva-
tion of the above relationship can be found in Shames
(1966). It is important to note that the above relationship
has been derived analytically and not empirically, which
means that it is theoretically justified.
Kragelsky et al. (1982) provided a summary of the
earlier research on rolling friction. Rolling friction is
typically much smaller than sliding or static friction. This
is why vehicles have wheels instead of runners, and ball
bearings are used in some machines instead of sleeve-
bearings. Reimann (1971) reported that rolling friction is
almost unaffected by the presence of lubricants and ex-
plained rolling friction as the result of the inability of an
elastically deformed body to retain its original shape
immediately on release of the stress. The harder the
rolling wheel or ball, and the harder the surface over
which it rolls, the less the force of rolling friction (White,
1969).
Very little information is available about the values of
o for different materials. For wood on wood, the coeffic-
ient of rolling friction is 0.152 mm while the coefficient of
static friction is 0.25—0.5 (Oberg et al, 1987). The coeffic-
ient of rolling friction may be assumed to be 0.051 mm
for iron on iron, 0.216 mm for iron on granite, 0.381 mm
for iron on asphalt, and 0.559 mm for iron on wood
(Oberg et al, 1987). Marks and Baumeister (1978) in-
cluded a table for the coefficients of rolling friction for
steel and pneumatic tires on different surfaces. Car-
michael (1950) provided a small table of the coefficient of
rolling friction for different materials, mainly wood and
steel.
Eq. (1) describes the force needed to keep a wheel
rolling without slipping and not the initial force required
to start a wheel rolling. One would hypothesize that the
initial force has to be higher than the force needed to
maintain uniform rolling. What is not known is whether
the initial force has the same kind of relationship with
total weight and wheel radius as the force needed to
sustain rolling. If that is the case, the only difference
would be that the coefficient of rolling friction (o) would
be higher for the initial force than that for the sustaining
force. Since ergonomists focus on the causes of higher
stresses, this study investigated only the initial force and
its relationship to total weight and wheel radius under
different conditions.
When a cart is pushed or pulled from a stationary
position, the applied force must overcome the rolling
friction at the wheels and the friction at the wheel bear-
the friction between the wheels and the floor. This study
focused on rolling friction between the wheels and the
floor and investigated the relationship between rolling
friction, cart weight, and wheel diameter for different
floor materials.
2. Objectives
This study was divided into four experiments to ad-
dress the following questions:
1. Does wheel width have an effect on the minimum
initial cart pull forces? The relationship described in
Eq. (1) does not include wheel width, which implies
that wheel width should not have an effect on the
minimum force required to move a cart.
2. What is the effect of floor material? Floor material
affects the coefficient of rolling friction, which is the
constant of proportionality o in Eq. (1). According to
White (1969), harder surfaces are expected to yield
lower cart push/pull forces. The coefficients of rolling
friction were determined for a number of floor surfaces
and one wheel material.
3. What are the effects of wheel diameter, wheel orienta-
tion and their interaction? Eq. (1) has been derived for
the case where the force applied at the circumference
of the wheels acts in the same direction in which the
wheel rolls. Therefore, it is expected that the relation-
ship in Equation 1 will hold when all four wheels are
aligned in the forward direction (F0R0). It is expected
that when one or more of the wheels is not oriented in
the forward direction, higher forces are required to
move the cart forward.
4. Is there a difference between cart push and cart pull
forces for different wheel orientations? The basic laws
of statics necessitate that cart push and pull forces are
the same when all wheels are oriented in the forward
direction. What remains unknown is how the relation-
ship between cart push and pull forces changes for
other wheel orientations.
3. Methods
A Wagner push—pull force gauge (Model FDL 100)
with a maximum reading of 100 lb and an accuracy of
$0.3% of full scale was used to measure the minimum
cart push/pull forces required to initiate the movement of
a cart under the different conditions. The cart started
from a stationary position for each measurement. The
experimenter attached the force gauge to the cart at
 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245
Fig. 1. Illustration of the cart used in this study.
a point midway along its width and pushed or pulled it,
increasing the force slowly and steadily without jerking
until the cart moved. The force gauge was perpendicular
to the cart handle. For measuring pull forces, the force
gauge contacted the cart at the end of a fork-type hook
attachment. For measuring push forces, the attachment
was cylindrical with a diameter of 5 mm. The dial on the
force gauge recorded the force and stored it after the
release of the force. The force gauge was reset before the
next measurement. Each measurement was repeated six
times, each trial being at a different random location. The
floor was level and of consistent quality.
The cart was a 4-wheel Rubbermaid plastic utility cart.
Its external dimensions were: 610 mm wide, 1020 mm
long, and 820 mm high. The cart is illustrated in Fig. 1.
The cart weighed 15.3 kg without the wheels. It had
a handle along its width. Cinder blocks were used to load
the cart. Each block weighed 9.1 kg. The cinder blocks
were distributed evenly with respect to the center of the
cart.
3.1. Effect of wheel width
Two types of wheels were tested: 25 mm wide, and
38 mm wide. This was the widest range of wheel widths
available for the same wheel material, wheel diameter,
type of bearing, and load capacity. In other words, wheels
are not available in all materials, diameters, bearings, and
widths. This is not a surprise because smaller diameter
wheels with relatively small widths and soft materials are
designed for light duty while the larger diameter, larger
width, and harder material wheels are designed for heavy
duty. Obviously, the cost also increases as one starts
considering the heavier duty wheels. Therefore,
a 25—38 mm wheel width range was the range that a cart
237
designer would most likely select for the type of cart used
in this experiment.
The two types of wheels were made of hard rubber and
had a diameter of 102 mm. The experiment was per-
formed on smooth concrete and industrial carpet floors.
Cart load was increased from 0 to 181.4 kg in in-
crements of 36.3 kg. All four wheels were maintained in
the forward direction (F0R0) throughout this part of the
experiment. All the measured forces were pull forces.
3.2. Effect of floor material
Minimum cart pull forces were measured using the
102 mm diameter hard rubber wheels on smooth con-
crete, tile, asphalt, and industrial carpet. These are sur-
faces commonly seen in different industries. The asphalt
floor was free of cracks, bumps, or gravel. The carpet
floor was made of carpet tiles (heuga Flor S, 55% nylon,
27.5% animal hair, and 17.5% rayon). The latter type of
floor is common in large, open floor areas in ‘electronic’
and modular design offices. All floors were level and of
consistent quality. The orientation of the wheels was
maintained in the forward direction. Cart load was
changed in increments of 36.3 kg from 0 to 217.7 kg. Only
pull forces were measured.
3.3. Effect of wheel diameter and wheel orientation
In this part of the experiment, four factors and all of
their combinations were tested. The factors were: wheel
diameter, wheel orientation, floor surface, and cart load.
Three wheel diameters were tested: 51, 102, and 152 mm.
All wheels had the same width, 25 mm, were made
of the same material (hard rubber), and had the same
type of bearing. Only pull forces were measured. Wheel
238 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245

orientation was tested at four levels: F0R0 (all four Table 1

wheels aligned in the forward direction), F0R90 (the two
front wheels, the wheels furthest from the cart handle,
aligned in the forward direction and the two rear wheels,
the wheels closest to the cart handle, aligned at 90° to the
forward direction), F90R0 (the two front wheels aligned
at 90° to the forward direction and the two rear wheels
aligned in the forward direction), and F90R90 (all four
wheels aligned at 90° to the forward direction). All
swiveling wheels had an offset swivel base. When aligned
at 90° to the forward direction, the swiveling wheels faced
the same direction. Measurements were taken on both
smooth concrete and carpet floors. Cart load was varied
from 0 to 181.4 kg in increments of 36.3 kg.
3.4. Comparison of minimum push and pull forces
In this part of the experiment, both the minimum push
and pull forces required to initiate cart movement of the
cart were tested. The cart was equipped with 102 mm
diameter hard rubber wheels. Three wheel orientations
were tested: F0R0, F90R0, and F0R90. Testing was con-
ducted on a carpet floor.
4. Results
4.1. Effect of wheel width
Fig. 2 illustrates the results of this part of the study.
Each point on the graph represents the average of six
data points. Cart weight is the total weight including the
Least squares regression lines attributes for wheel width data
R Slope 95% confidence
interval on slope
1 wide, concrete 0.975 0.0419 0.0407—0.0430
1.5 wide, concrete 0.981 0.0429 0.0420—0.0439
1 wide, carpet 0.977 0.0894 0.0874—0.0915
1.5 wide, carpet 0.992 0.0894 0.0880—0.0908
weight of the empty cart, the cinder blocks, and the
wheels. The graph shows that there was no significant
effect of wheel width on the minimum cart pull forces on
either floor. It is obvious that all four curves exhibit
a strong linear trend. Table 1 summarizes the R values,
slopes, and 95% confidence intervals on the slopes of the
least squares regression lines fit to the data sets in Fig. 2.
The analysis was done on the individual observations. All
regression lines were forced to pass through the origin
since a y-intercept other than zero would mean that it
takes a force equal to the y-intercept value to move a cart
of zero weight. This does not make sense from a physical
point of view. Conceptually, zero force is required to
move an object of zero weight.
Analysis of covariance to compare the two wheel
widths at the different loads was conducted on the min-
imum pull forces on concrete and on carpet separately,
with cart weight as the covariate. Table 2 summarizes the
p-values of wheel width, cart weight, and their interac-
tions. Although the effect on pull forces of wheel width on
concrete was statistically significant (p"0.0083), it was

Fig. 2. Effect of wheel width on minimum cart pull forces.

 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245 239

Table 2 Table 3
p-values from ANCOVA to compare wheel widths Least-squares regression lines attributes for the different floor materials

Concrete Carpet R Slope 95% confidence

interval on slope
Wheel width (¼) 0.0083 0.2367
Cart weight (¸) 0.0001 0.0001 Concrete 0.975 0.0434 0.0422—0.0446
¼;¸ 0.8867 0.5400 Tile 0.990 0.0465 0.0458—0.0473
Asphalt 0.956 0.0642 0.0618—0.0666
Carpet 0.977 0.0894 0.0874—0.0915

obvious from the graph in Fig. 2 that it had no practical

significant effect on the minimum cart pull forces. In fact, Table 4
the 95% confidence interval on the difference between Estimated coefficients of rolling friction for the different floor materials
the means of the minimum pull forces for both wheel (mm)
widths was (0.07—0.42 kg). Floor Coefficient 95% confidence Comparison with
of rolling interval on concrete
friction (mm) coefficient of
4.2. Effect of floor material rolling
friction (mm)
Fig. 3 shows the means and the standard deviations of Concrete 2.205 2.144—2.266 —
the minimum cart pull forces for the four surfaces tested Tile 2.362 2.327—2.403 7% higher
in this experiment: concrete, tile, asphalt, and carpet. Asphalt 3.261 3.139—3.383 48% higher
Least-squares regression lines were fitted to the data of Carpet 4.541 4.440—4.648 106% higher
each floor surface. The lines were forced to pass through
the origin. Table 3 provides the R values, slopes, and
confidence intervals for these slopes. Multiplying these 4.3. Effect of wheel diameter and wheel orientation
slopes by the radius of the wheels (51 mm) gave estimates
of the coefficients of rolling friction for hard rubber Figs. 4 and 5 show that increasing wheel diameter
wheels on the different floor materials. This information caused the minimum cart pull forces to decrease on both
is presented in Table 4, which also shows how the differ- the concrete and carpet floors when all four wheels were
ent floor materials compared to concrete. aligned in the forward direction (F0R0). Fig. 6 is a plot of

Fig. 3. Effect of floor material on minimum cart pull forces.

240 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245

Fig. 4. Effect of wheel diameter on minimum cart pull forces on concrete floor with all wheels in forward direction.

Fig. 5. Effect of wheel diameter on minimum cart pull forces on carpet floor with all wheels in forward direction.

the minimum cart pull forces against the ratio of cart
weight to wheel diameter for the concrete and carpet
floors with the wheels at the F0R0 orientation. The graph
also shows the least-squares regression lines fitted to the
concrete and carpet data. The least-squares regression
lines were forced through the origin. Table 5 summarizes
the attributes of the regression lines.
Least-squares regression lines were fitted to the min-
imum cart pull forces as a function of cart weight for each
of the wheel diameters, wheel orientations, and floor
combinations. The lines were forced through the origin.
All lines showed strong linear trends (R'0.95). Estim-
ates of the slopes are presented in Table 6. Table 7 shows
how the different wheel orientations compare to the
 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245 241

Fig. 6. Minimum cart pull force versus ratio of cart weight to wheel diameter.

Table 5 Table 7
Attributes of least-squares regression lines describing the minimum cart Comparison of slopes for F90R0, F0R90, and F90R90 wheel orienta-
pull forces as a function of the ratio of cart weight to wheel diameter tions to slopes for F0R0 wheel orientation (within same wheel diameter)

Floor R Slope 95% confidence Concrete Carpet

(mm) interval on slope
Orientation 51 mm 102 mm 153 mm 51 mm 102 mm 153 mm
Concrete 0.953 3.73 3.64—3.82
Carpet 0.957 7.98 7.82—8.13 F0R0 1.00 1.00 1.00 1.00 1.00 1.00
F90R0 0.84 1.18 1.10 0.82 1.24 1.32
F0R90 0.77 1.35 0.90 0.92 1.53 1.44
F90R90 0.91 1.81 1.22 1.01 1.53 1.90
Table 6
Estimated slopes of least-squares regression lines modeling minimum
cart pull forces as a function of cart weight Table 8
Comparison of slopes for 102 mm and 153 mm wheel diameters with
Concrete Carpet
slopes for 51 mm wheel diameter (within same wheel orientation)
Orientation 51 mm 102 mm 153 mm 51 mm 102 mm 153 mm Concrete Carpet
F0R0 0.0685 0.0434 0.0237 0.145 0.0894 0.0584 Orientation 51 mm 102 mm 153 mm 51 mm 102 mm 153 mm
F90R0 0.0578 0.0514 0.0261 0.119 0.110 0.0772
F0R90 0.0530 0.0586 0.0214 0.133 0.136 0.0843
F0R0 1.00 0.63 0.35 1.00 0.62 0.40
F90R90 0.0624 0.0787 0.0290 0.147 0.136 0.110 F90R0 1.00 0.90 0.45 1.00 0.92 0.65
F0R90 1.00 1.11 0.40 1.00 1.02 0.63
F90R90 1.00 1.26 0.46 1.00 0.93 0.75

F0R0 wheel orientation for each wheel diameter on the

two floor surfaces. For example, on the concrete floor,
the minimum pull forces with the wheels at the F90R0
were 1.18 times the minimum pull forces under F0R0
wheel orientation and with all the other conditions being
the same. Table 8 shows how the wheel diameters com-
pare to the 51 mm diameter for each wheel orientation.
As the effect of wheel orientation seems to be inconsist-
ent between the different wheel diameters, the minimum
cart pull forces for all conditions within each of the wheel
orientations tested were averaged and compared to the
average force for the F0R0 wheel orientation. For
242 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245

Fig. 7. Minimum cart pull and push forces for the F0R0 wheel orientation.

Fig. 8. Minimum cart pull and push forces for the F90R0 and F0R90 wheel orientations.

F90R0, the average force was 2.7% higher than that for 4.4. Comparison of minimum push and pull forces
F0R0. For F0R90, the average force was 13.1% higher;
and for F90R90, the average force was 30.7% higher than Figs. 7 and 8 show the minimum cart pull and push
that for F0R0. forces for the three wheel orientations: F0R0, F90R0, and
 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245
Table 9
Slopes of least-squares regression lines for push and pull data
F0R0 F90R0 F0R90
Pull 0.0894 0.110 0.136
Push 0.100 0.153 0.114
F0R90. The slopes of the least-squares regression lines
fitted to each data set are shown in Table 9. The least-
squares regression lines were forced through the origin.
Analysis of variance was conducted on the minimum
force to test for the significance of Task (pull versus push),
Load, and their interaction for the F0R0 wheel orienta-
tion. The p-values were 0.0003, 0.0001, and 0.0655 for
Task, Cart Load, and Task;Cart Load, respectively.
Since a statistically significant difference existed between
push and pull forces, a 95% confidence interval was
constructed on the difference between the average push
force and the average pull force for the F0R0 wheel
orientation to test the practicality of the statistical signi-
ficance. The confidence interval was (0.44!1.41 kg), the
average push force being higher than the average pull
force. Thus, the maximum difference between push and
pull forces could be as high as 1.4 kg, which attests that it
is not a practically significant difference for ergonomics
design and evaluation purposes.
Analysis of variance was also conducted to compare
the pull forces with the F90R0 wheel orientation to the
push forces with the F0R90 wheel orientation. The p-
values were 0.0659, 0.0001, and 0.0092 for Task, Cart
Load, and Task;Cart Load, respectively, with Task
being pull with F90R0 versus push with F0R90 wheel
orientation.
Similarly, analysis of variance was conducted to com-
pare the pull forces with the F0R90 wheel orientation to
the push forces with the F90R0 wheel orientation. The
p-values were 0.0260, 0.0001, and 0.0041 for Task, Cart
Load, and Task;Cart Load, respectively. Since Task
was statistically significant, a 95% confidence interval
was constructed for the difference between the average
pull force with F0R90 and average push force with
F90R0 to investigate practical significance. The confid-
ence interval was (0.10—1.22 kg) with the average push
force being higher. Again, the confidence interval was not
wide enough to conclude that the push-F90R0 and pull-
F0R90 forces were significantly different for ergonomics
design and evaluation purposes.
5. Discussion
A significant finding of all four experiments, as evident
in the graphs presented in the previous section, is the
strong linear relationship between the minimum cart
push/pull forces and cart weight. This was evidenced in
243
the large R values (higher than 0.95) for all regression
lines described in the previous section. This strong linear
relationship existed for all floor surfaces, wheel dia-
meters, and wheel orientations. Therefore, a linear rela-
tionship exists between cart weight and the force to
initiate cart movement, similar to the linear relationship
between cart weight and the force to sustain wheel roll-
ing, as described by Eq. (1). A similar linear relationship
between load weight and force was found by Strindberg
and Petersson [9].
5.1. Effect of wheel width
It was expected that a 50% increase in wheel width
would not have a significant effect on the force needed to
initiate cart movement. The tables and graphs in the
previous section confirmed this hypothesis. Since wheel
width does not appear in Eq. (1), it is known that it does
not affect the force needed to sustain cart movement.
Therefore, whether it is the force to initiate or sustain
cart movement, wheel width does not have a signifi-
cant effect. In fact, a result other than that would be
hard to explain physically. An analogy can be made
here with the well-known fact that area of contact
between two objects does not affect translational friction
between them. It is surprising that Eastman Kodak
(1986) stated that a wider tread increases the wheel’s
rolling resistance.
It is worth noting that since the 38 mm wide wheel was,
as expected, heavier than the 25 mm wide wheel, total
cart weight at each cart load was not the same for the two
types of wheels. This is why analysis of covariance and
not an analysis of variance had to be conducted with
total cart weight as the covariate so as to filter out
differences in cart pull forces that were merely due to cart
weight difference for the two wheel types.
5.2. Effect of floor material
Cart pull forces were the lowest for concrete, then
increased for tile, asphalt, and carpet surfaces. Drury et
al. (1969) found a similar relationship between the forces
to push vehicles on hospital carpeted and hard floors.
White (1969) stated that the harder the wheels and the
harder the floor material, the less the force needed to
sustain rolling. Among the tested floor surfaces, concrete
is the hardest and carpet is the softest. Although asphalt
might be generally harder than tile, the rougher surface of
asphalt may be the cause for the higher forces measured
on asphalt as compared to tile.
Also, this experiment produced estimates of the coeffi-
cients of rolling friction for initiating movement of hard
rubber wheels on the four floor surfaces. Other wheel
materials will certainly have different values for the coef-
ficients of rolling friction. The reason the coefficients of
rolling friction reported in this study differ significantly
244 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245
from those reported by previous researchers such as
Oberg et al. (1987) is that the wheel and surface materials
used in this study were different from those used in
previous studies. This attests to the significant impact of
wheel and surface materials on the coefficient of rolling
friction. Future research should establish a database of
the coefficients of rolling friction for various wheel mater-
ials, tires, and floor surfaces that exist in industry. In
order to build this database, it would be imperative to
devise a standard protocol for measuring the minimum
force required to initiate wheel rolling. It is recommended
that the force exertion by the human experimenter be
replaced by a more automated technique to improve
cross-study comparisons.
5.3. Effect of wheel diameter and wheel orientation
Figs. 4 and 5 and Tables 6 and 8 show that there was
a consistent drop in the minimum pull forces for the
F0R0 wheel orientation as wheel diameter increased.
Eq. (1) states that the force to sustain rolling is inversely
related to wheel radius. In order to test whether the same
relationship applies to the forces to initiate cart move-
ment, the minimum cart pull forces measured in this
experiment were regressed on the ratio of cart weight to
wheel diameter as illustrated in Fig. 6. The strong linear
relationship, evidenced by the high R values as shown
in Table 5, suggest that the force needed to initiate
cart movement is inversely proportional to wheel
diameter, or, in other words, directly proportional
to the ratio of cart weight to wheel diameter. Therefore,
the relationship of cart weight and wheel diameter to the
force needed to sustain movement is the same as that
between those factors and the force needed to initiate cart
movement.
Drury et al. (1975) measured speed and pulse rate for
pushing hospital vehicles of different wheel sizes. Pushing
speed was higher and pulse rate was lower for the larger
wheel size. This result is congruent with the relationship
between wheel diameter and force that our study demon-
strated.
It was expected that when the wheels are oriented
away from the forward direction, higher forces would be
required to move the cart. The results of this experiment,
as shown in Tables 6 and 7, suggest that this hypothesis
was not supported consistently. The only cases that fol-
lowed the expected trend were with the 102 mm diameter
wheels, and the 153 mm diameter wheels on carpet. The
51 mm diameter wheels showed obvious contradiction to
the expectations. This suggests a strong interaction be-
tween wheel orientation and wheel diameter and between
wheel orientation and floor surface. When such interac-
tions exist, precise conclusions are typically difficult to
make.
It was known from the beginning that Eq. (1) does not
apply except for the case when the wheels are all in the
forward direction because it was derived for the case
when the direction of the applied force is the same as that
in which the wheel rolls. When the wheels are not aligned
in the forward direction and the cart is pulled straight,
factors other than the simple rolling friction come into
play. Ideally, the wheels might be expected to align them-
selves first in the forward direction and then start to roll.
However, this did not occur consistently. What happened
was that the cart started to turn with the wheels aligning
themselves as the cart turned. This might have con-
taminated the measurement since the cart was always
pulled straight backwards with the force gauge and when
the cart turned the exerted force was no longer parallel to
the direction of the cart. The somewhat inconsistent
effect of wheel orientation warrants more research to
develop theories about what happens when the wheels
are not all aligned in the forward direction and to test
them in the lab.
It is worth noting that when all four wheels were
aligned at 90° to the forward direction (F90R90) the pull
forces were consistently higher than when only two
wheels were aligned at 90° to the forward direction
(F0R90 and F90R0).
5.4. Comparison of minimum push and pull forces
It was expected that the minimum push and pull forces
would be the same for the F0R0 wheel orientation since
the ultimate direction of the applied force was the same
whether it was a push or pull force. Although the results
tended to agree with the expectations, the small
differences detected might be attributed to the difficulty
experienced in applying pure push and pull forces.
Applying a pure force that does not create a turning
couple necessitates applying a force strictly perpendicu-
lar to the width of the cart at a point midway across the
cart width. Applying a force that does not conform to
these rules will cause a force higher than the correct
one to be measured and may cause the cart to turn.
Applying pure push forces was slightly more difficult
than applying pure pull forces because in the case of push
forces, the force gauge contacted the cart at the end of
a small attachment that was 5 mm in diameter. In the
case of pull forces, the attachment was a fork-type hook.
The minimum pull forces at the F90R0 wheel orienta-
tion were, on average, 28% less than the minimum push
forces. On the other hand, the minimum pull forces at
the F0R90 wheel orientation were, on average, 19%
higher than minimum push forces. This suggests that
swiveling wheels are recommended in the front if the
cart is primarily pulled, while the swiveling wheels are
recommended in the rear if the cart is primarily pushed.
This recommendation is based entirely on the minim-
um forces required to push or pull the cart forward and
does not take into account the forces required to turn
the cart or maneuver it.
 K.W. Al-Eisawi et al. / Applied Ergonomics 30 (1999) 235—245
This experiment showed also that the pull-F90R0 for-
ces were not significantly different from the push-F0R90
forces. Similarly, the pull-F0R90 forces were not signifi-
cantly different from the push-F90R0. This results sug-
gests that pulling a cart with swiveling front wheels
would be the same as pushing a cart with swiveling rear
wheels. Similarly, pulling a cart with swiveling rear
wheels would be the same as pushing a cart with swivel-
ing front wheels.
In more than one case, it appeared that the measured
force values and the graphs suggested no practical signifi-
cant difference, while statistical analysis produced a sig-
nificant effect. One should be cautious not to conclude
hastily that a significant effect exists because small
variances and large sample sizes make small differences
appear statistically significant. In such a case, a 95%
confidence interval on the difference between the two
treatment means will help determine the practicality of
the significant difference.
Concerning the question of whether the front or rear
wheels should swivel, this study did not provide a definite
answer since it did not investigate turning carts, and
wheel orientation was found to have strong interactions
with wheel diameter and floor surface. This question
remains to be answered in future research. However, it is
generally easier to turn a cart when the wheels closer to
the handles of the cart (rear wheels) are swiveling as
opposed to when the front wheels swivel. In fact, turning
a cart with swiveling front wheels requires the applica-
tion of a couple at the cart handles; while turning a cart
with swiveling rear wheels requires the application of
a transverse force. However, controlling a cart with
245
swiveling rear wheels is generally harder. Therefore, we
would generally recommend having the wheels closer to
the cart handles swivel if the cart will have to go around
corners more often than be pushed or pulled straight.
Based on this limited study, having all four wheels swivel
is generally discouraged unless the cart is intended for use
in very tight spaces and is not pushed or pulled for long
distances.
References
Carmichael, C., 1950. Kent’s Mechanical Engineers’ Handbook: Design
and Production. 12th ed. Wiley, New York.
Drury, C. G., Barnes, R. E., Daniels, E. B., Pedestrian operated vehicles
in hospitals. Proceedings of the American Institute of Industrial
Engineers, 26th Annual Conference and Convention, pp. 184—191.
Eastman Kodak, 1986. Ergonomic Design for People at Work. vol. 2.
Van Nostrand Reinhold, New York.
Kragelsky, I. V., Dobychin, M. N., Kombalov, V. S., 1982. Friction and
Wear Calculation Methods. Pergamon Press, Oxford.
Marks, L. S., Baumeister, T., 1978. Mechanical Engineers’ Handbook.
6th edn. McGraw-Hill, New York.
Oberg, E., Jones, F. D., Horton, H. L., 1987. Machinery’s Handbook.
22nd ed. Industrial Press, New York.
Reimann, A. L., 1971. Physics, vol. I: Mechanics and Heat. Barnes and
Noble, New York.
Shames, I. H. 1966. Engineering mechanics: Statics. 2nd ed. Prentice-
Hall, Englewood Cliffs.
Snook, S. H., Ciriello, V. M., 1991. The design of manual handling tasks:
revised tables of maximum acceptable weights and forces.
Ergonomics 34, 1197—1213.
Strindberg, L., Petersson, N. F., 1972. Measurement of force perception
in pushing trolleys. Ergonomics 15, 435—438.
White, H. E., 1969. Introduction to College Physics. Van Nostrand,
New York.