Development of a Biomechanical Model and Validation of Assessment Tools for Personal Load Carriage Systems BY Wm. Alan H. Rigby A thesis submitted to the School of Physicai and Health Education in confonnity with the requirernents for the degree of Master of Science Queen's University at Kingston Kingston, Ontario, Canada September, 1999 copyright 8 Wm. Alan H. Rigby, 1999
137
Embed
Development Biomechanical Model Validation Assessment · predict pack-person interface variables. The first technique used the principles of static equilibrium of the pack-person
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Development of a Biomechanical Model and Validation of Assessment Tools for Personal Load Carriage Systems
BY
Wm. Alan H. Rigby
A thesis submitted to the School of Physicai and Health Education in confonnity with the requirernents
for the degree of Master of Science
Queen's University at Kingston Kingston, Ontario, Canada
September, 1999
copyright 8 Wm. Alan H. Rigby, 1999
National Library Bibliothèque nationale du Canada
Acquisitions and Acquisitions et Bibliogiaphic Services services bibliographiques
The author has granted a non- exclusive licence allowing the National Library of Canada to reproduce, loan, distribute or sel1 copies of this thesis in rnicroform, paper or electronic formats.
The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts ffom it may be printed or otherwise reproduced without the author's permission.
L'auteur a accorde une licence non exclusive permettant à la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la forme de microfiche/nlm, de reproduction sur papier ou sur format électronique.
L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.
Abstract
This study was part of a larger military project to improve personal load carriage
systems for soldiers. The goal of this study was to develop and validate a personal ioad
carriage system biomechanical model. The mode1 would serve as the basis for a personal
load camage system design tool, which would provide a better understanding the pack-
person interface and in tum help development of new systems for soldiers. A sub-
problem of this study was to develop and validate an improved pack testing system for
evaluation of the biomechanical modei and for fbture scientific and field studies.
A load distribution mannequin and force platform were part of the comprehensive
testing system designed to provide the necessary measurements to validate past and
current personal load carriage system biomechanical models. In addition, hkro new
devices, a strap tension probe and an instmmented test pack, capable of measuring pack
strap tensions and lumbar contact forces respectively, were created. These new
measurement tools were validated within a 5% average error and suggestions were made
for funher irnprovements.
The current biomechanical model was the third iteration in a series of persona1
load carriage system models. The current model employed two different techniques to
predict pack-person interface variables. The first technique used the principles of static
equilibrium of the pack-person interface to determine unknown variables and preûict
contact forces between the pack and the human form. Tension in the upper shoulder,
lower shoulder, and load lifter straps supported much of the pack mass. Through fiction
and anatomical geometry the waist belt and lumbar region provided vertical lie. Al1
unknown elements of the equilibrium needed to be solved. The second technique used
logical relationships between different elements of the pack-person geometry, interna1
forces and moments, and contact forces to predict unknown variables and pack-person
contact forces. Again, tension in the upper shoulder, lower shoulder, and load lifter
straps supponed much of the pack mass. The relationship between the shouider straps
was modeled using a modified pulley equation. The waist belt and lumbar pad lie forces
were predicted bas4 on fiictional contributions and vertical components of reaction
forces.
The current model could not be used as a robust xientific tool. Equilibrium
predictions of pack-person interfiice variables were quite poor compared to measured
values (average p-values less than 1 . 0 * 1 ~ ~ ) . The large coefficients of fiction at the
shoulder, lumbar region, and waist caused the predicted ranges of the regional models to
be so large that their ability to contain the associated validation measurements were
suspect, despite the predicted ranges encompassing an average of 7 1.3% of the rneasured
values. On the other hand, the geometric components of the model were valid as the
model predictions and measured values were statistically correiated (average p-values of
0.99). The sensitivity analysis proved that the equilibrium expression that predicted the
outputs were highly sensitive to input variables, implicating the load lifter strap model as
a potential cause of the "ill-conditioned" system. Regional model and geometric outputs
were less sensitive to changes in input variables.
In general the model, modeling process, and sensitivity analysis provided insight
into and qualitative information about the pack-person interface. In addition, two new
measurement tools were validated and can be added to the personal load carriage system
battery of test. Future directions outlined in the thesis showed how the current rnodel and
validation techniques could be used for the next iteration of the model, other scientific
studies, and field tests.
Acknowledgements
I would like to acknowledge the Defense and Civil Institute of Environmental
Medicine for their financial and materials contributions to this thesis. 1 found
this work very stimulating and enjoyable, and 1 appreciate DCiEM's and Major Linda's
Bossi's help in making this research possible.
1 would also like to take this opportunity to thank a number of individuals,
without whom this thesis would no€ have been possible: Dr. Joan Stevenson, for ail your
help, guidance, and encouragement over the past 4 years; Dr. Ron Pelot and Dr. Tim
Bryant for your invaluable contributions and advice; Gerry Saunders of the Clinical
Mechanics Group for your precision and tireless efforts in construction of the strap
tension probe; The gang in the lab: Jon, Wayne, Pat, Derek, and Sue, for making
yourselves so available for al1 the day to day help.
1 would also like to recognize my many friends and family, who's support has
rneant so much. A special thank you to my mother, Sandy Rigby, for everything you
have done, your credits are too long to list.
Finally, 1 would like to acknowledge Jenn Ellis. Your motivation, support, love,
and understanding throughout this thesis and in life continue to be my inspiration. Thank
Lumbar Pad Mode1 .........................................................................
Waist Belt Model ............................................................................
List of Figures
Figure 1-1 : Phase 1 shoulder based pack-person interface mode1 ............................ ..................................................................... Figure 1-2: Phase II waist belt model
................................................................... Figure 1-3 : Phase II lumbar pad modd
....................................... Figure 1-4: Phase II personal load carriage system model
.......................................................... Figure 2- 1 : The load distribution mannequin
............................................................................... Figure 2-2: Modified test pack
Figure 3-1 : Elements of a standard commercial pack and the biomechanical mode1 .
...................... Figure 3-2: Biomechanical model of a personal load carnage system
Figure 3-3: Determination of strap and wrap angles ...............................................
................................................................... Figure 3-4: Net shoulder contact force
......................... Figure 3-5 : Anatomical simplification of a transverse waist section
................................ Figure 3-6: Wedge shape of hips in Frontal and sagittal planes
............................ Figure 3-7: Complete anatomical simplification of the hip region
Figure 3-8: Liff capability of the waist belt .............................................................
... Figure 3-9: Transverse view of a typical quarter section of a waist belt in tension
.............. Figure 3 - 1 0: Anatomical simplification of the lumbar region, sagittal Mew
.................................................... Figure 3- 1 1 : Lumbar region - lumbar pad model . . ............................................................................. Figure 3 - 1 2: The role of fhction
vii
List of Tables
th Table 2- 1 : 50 percentile male human fonn mannequin .......................................... 18
Table 2-2: Calibration coefficients and error measures ............................................ 31
Table 2-3: Accuracy and reliability results with respect to tension .......................... 32
Table 2-4: Accuracy and reliability results with respect to stifiess ......................... 32
Table 5-8 shows the measured lumbar contact force (Fx) compared to the
equilibrium expression prediction of that force. A paired t-test revealed a p-value of
3.20*10". Table 5-8 also shows the range of Fx predicted by the regional model
(Equations 3-24 through 3-25). Seventeen of the 48 measures fall within the predicted
range. The complete data set can be seen in Appendix 2.
Table 5-9 shows the measured lifi force of the lumbar pad ( F Z ~ ) compared to the
range predicted by the regional model (Equation 3-24.2). Twenty-four of the 48 measures
fa11 within the predicted range. The complete data set can be seen in Appendix 2.
Remember that there is no equilibrium expression representation of the lumbar pad lie
force ( F ~ ~ ) . Hence only a range predicted by the regional model exists.
Table 5-8: Lumbar contact force (Fx) mode1 validation results (in Newtons).
Test Condition Measu red Equilibrium Regional Mode1 Range Ptediction Ptediction Minimum Maximum
1-1 122.33 1 13.71 88.82 128.03
Table 5-9: Lumbar pad lift force ( F ~ ~ ) mode1 validation results (in Newtons).
Test Condition Measured Regional Mode1 Range Prediction Minimum Ma-ximum
1-1 40.33 10.91 59.58
Waist Belt Mode1
Table 5-10 shows the measured waist belt lia force (FzB) cornparecl to the range
predicted by the regional model (Equations 3- 16 through 3-23). Forty-six of the 48
measures faIl within the predicted range. The complete data set can be seen in Appendix
2. Remember that there is no equilibrium expression representation of the waist belt lie
force (IJzB). Hence only a range predicted by the regional model exists.
Table 5-1 1 shows the measured lift force of the waist belt and lumbar complex
(Fz) compared to the equilibrium expression prediction. A paired t-test produced a p-
value of 3.99* l O-'. Table 5-1 l a l s ~ shows the range of Fz predicted by the regional model
(Equations 3-16 through 3-23). Forty-eight of the 48 measures fa11 within the predicted
range. The complete data set can be seen in Appendix 2.
Table 5- 10: Waist belt lifi force ('FZB) model validation results (in Newtons).
Test Condition Measuted Regional Mode1 Range Prediction Minimum &ximwn
1-1 163.36 39.89 200.1 5
Table 5-1 1 : Net lifi force (Fz) mode1 validation results (in Newtons).
Tesi Condition Measured Equilibrium Regional Mode1 Range Prediction Prediction Minimum Maximum
1-1 21 2.69 264.82 50.80 258.73
Sensitivity Amlysis
TaMe 5-12, Table 5-13, and Table 5-14 summanze the results of the sensitivity
analysis for the geometric outputs, regionai model outputs, and equilibrium expression
outputs respectively. The detailed sensitivity analysis results can be seen in Appendix 3.
The table shows the fraction of change in the output variables (rows) as the input variables
(columns) were systernatically and independently increased by 10%. Cells only contain
values if the change in output variables were greater than or equal io the 50' decile. The
most sensitive model output was clearly the equilibrium expression prediction of the upper
shoulder strap tension (Ti). One can see from Table 5-14 that small changes in model
inputs (lm) led to consistently large changes in the equilibrium predictions of the upper
shoulder strap tension (Ti). It might appear as though the shoulder reaction forces (!SN,
s\, !SNZ), and the force of friction around the shoulder (FR) were also highly sensitive.
However, the upper shoulder strap tension (Ti) was used to calculate these variables
directly, so sensitivity of shoulder reaction forces (sN, s\, and sNz) was a product of the
sensitivity of Ta. Conversely, the lumbar reaction force (Fx) and vertical lifi of the waist
belt - lumbar pad complex (FZ) were relatively insensitive to changes in the inputs.
Table 5- 12: Summary of sensitivity analysis for geometric outputs, expressed in percent
change.
Model
outputs
01
02
0 4
a1
a
Model inputs increased by 10 %
ds (mm) ds (mm) r (mm) dl (mm) 6 (mm) O. 12 0.08 0.05 0.20
0.07 0.04
0.32 0.09 O, 16
0.05
0.1 1 0.08
Table 5- 14: Summary of sensitivity analysis for equilibrium expression outputs, expresseci
in percent change.
Mode1 inputs increased by 10 %
Cbapter 6
Discussion
Coefficients of Friction Determination of the coefficients of friction about the thtee anatomical regions
(shoulder, waist, and lumbar region) revealed values of 0.39, 0.32, and 0.35 respectively.
Of note was the dramatic increase in these values fiom MacNeil's (19%) measurements.
MacNei1 (1996) used a rnilitary combat shirt, over Tekscanm pressure sensing devices,
which were mounted on the mannequin. The current work used only the mannequin.
These differences accounted for the variation in coefficients of fiction.
Geometric Predictions
Table 5-2 clearly illustrates the ability of the mode1 to accurately predict the test
pack's geometry. The predicted values and the measured values of strap geometry (el, Oz, 04, ai, and a) were not significantly different (average p-value = 0.94) and thus Rom
the same sample.
The ability of the model to accurately predict the test pack's gecmetry cannot
only be attributed to the model itself, but also the repeatability of the validation protocol.
The mannequin was marked and labeled so pack elements could be placed and replaced
in an accurate manner. Furthermore, input data corn the test pack and mannequin
geometry were accurately measured and remained unchanged between trials.
While predictions of al1 the geometric masures were very accurate, the validation
protocol revealed that the vertical stays supponing the attachment points for the load
lifter straps were quite compliant. When the load lifter straps were tightened the vertical
stays flexed, altering the geometnc relationship between the straps and the stays and the
stays and the pack body. However, angles of the load lifter straps (Or) were measured
relative to the main body of the pack itself and the compliance of the stays did not
significantly effect the geometric reiationship between the straps and the test pack body.
While this may have had a signiticant effect on other elements of the pack-person
interface, such as strap tension, it is of littie importance to the geometric predictions.
Since the relative angle of the strap to the pack was not significantly affected the factor
can be considered inconsequential. However, the cornpliance of the vertical stays is
addressed fbrt her below.
Comparison of the geometric measures on the right and left-hand-sides of the test
pack indicated that the right and lefl shoulder strap angles (81, &, and Br) and shoulder
wrap angles (al and a) (p = 0.67, 0.85. 0.92, 0.71, and 0.80 respectively) were not
significantly different. This indicated that the right-lefi symmetry assumption was
accurate for pack geometry.
Equilibrium Predictions
In general the equilibrium predictions of upper shoulder strap tensions (Tl),
lumbar reaction forces (Fs), waist belt - lumbar pad complex lift forces (Fz), shoulder
reaction forces (sN, s\, sNZ), and the forces of friction around the shoulder (FB) were
poor as evidenced by the lack of statistical relationship (p < 0.001) for al1 seven outcome
variables. However, because upper shoulder strap tensions Tl, which were poorly
predicted, were used to calculate the shoulder reaction forces (sN, sNx, and sNz) and the
forces of friction around the shoulder (FB) it was not surprising that the later predictions
were also poor. It is not entirely understood why the equilibrium expressions were so
poor in predicting upper shoulder strap tensions (TI), lumbar reaction forces (Fx), and
waist belt - lurnbar pad complex lift forces (Fz). The author hypothesizes that the
sensitivity of the equilibrium system and the accuracy of the equilibrium expressions
themselves may both have had a detrimental effect.
This hypothesis was examined by conducting a sensitivity analysis to determine
the effect that each input variable had on the biomechanical model output variables. It
was apparent from the sensitivity analysis that upper shoulder strap tensions (Ti) were
highly sensitive to changes in the model inputs, whereas lumbar reaction forces (Fx) and
waist belt - lumbar pad complex lie forces (Fz) were noticeably less sensitive. This
discrepancy in sensitivity created an "ill-conditioneâ" system. The equilibnum
expressions were not capable of depicting the pack-person interface. The details of this
hypothesis are outlined below during discussion of the biomechanical rnodel's sensitivity
anal ysis.
The second possible explanation for poor predictive power of interface variables
could be incorrect modeling. This would occur if a kinetic element that existed in the
physical equilibrium was not included in the rnodel or kinetic elements that were both a
part of the physical system and the model were misrepresented. Since previous, simpler
phases of the model were more accurate predictors (MacNeil, 1996; Rigby, 1997), the
author feels that elements added to this phase of the model were rnisrepresented. One
such element was the load lifter strap tension (T4). Inspection of the model output
surnmary table in Appendix 2 revealed that when load lifter strap tensions (T4) were
large, the prediction of upper shoulder strap tensions (Ti) became less accurate and often
becarne negative, a physical impossibility for tension. Furthermore, when load lifter
strap tensions (T4) were eliminated from the equilibrium expressions (a variable value of
O N) the equilibrium expression's predictions of upper shoulder strap tensions (TI) were
much more accurate. For example, the two closest predictions occurred on test setup 15-
3 and 16-2, both of which had T4 inputs of zero.
The author postulates that the vertical component of the load lifter strap was given
too much weight in the equilibrium expressions. The current model suggested that the
entire vertical component of the load lifter strap, the cosine of the strap tensions (T,),
acted on the pack-body by pulling the pack down. This condition would suggest that, as
the load lifter strap was tightened, the pack would effectively have a greater downward
force and thus lessen the vertical lift created by the increase in lower shoulder strap
tensions (TI). Practically, however, a wearer of the pack would not feel an increase in
"weight" of the pack when the load lifter straps were tightened. m i l e it was clear that
load lifter strap tensions (T4) were an important element of the load carriage system that
exerted some force on the pack and user, it was not entirely understwd how this
occurred.
Future work is needed to fùrther explore how the load lifter straps transfer forces
between the pack, trunk, and shoulders. One possibility is to assume that the load lifter
straps contributed only anterior-posterior force to the system and not downward force.
The downward forces that would be generated by the load lifter strap geometry may not
be transferred to the pack because the shoulders are unable to transfer the reaction force
to the load lifter straps as the shoulders do to the upper and lower shoulder straps. The
upper and lower shoulder straps wrap around the shoulders at approximately 180 degrees
and the reaction force of the shoulders can oppose the tension in the straps directly,
allowing the straps to act on the pack itself Whereas the load lifter straps wrap angle
was much smaller and the contact with the shoulder was almost entirely on the anterior
surface of the shoulder thus the reaction forces of the load lifter straps may not act in the
superior-inferior direction.
If a system of equations is "ill-conditioned" by an incorrectly represented
element, such as load lifter strap tension (T4), solutions for the other variables will be
incorrect also. The author feels that this is the case for equilibrium predictions of lumbar
reaction force (Fx) and waist belt - lumbar pad complex lifi force (Fz). Changing the
shoulder suspension model will have a significant effect on the remainder of the
equilibrium expressions. Once the representation of load lifter strap tension (TI) is
improved, the remainder of the equilibrium expression predictions will improve also.
Furthemore, because the shoulder reaction forces (sN, sNx, and sNz) and the force of
improvements in the shoulder suspension model will improve predictions of the former
variables as well.
Comparison of the upper shoulder strap tensions on the tight and lefi-hand-sides
of the test pack indicated that the right and left upper shoulder strap tensions (Ti) were
not significantly different (p = 0.93). This indicated that the assumption of right-left
symmetry was accurate for strap tensions.
Regional Mode1 Predictions
The most notable outcome of the regional model predictions was the size of the
predictive ranges. Chapter 3 outlined the variable contribution of fiction throughout the
model and how these variations led to predictive ranges rather than predictive values.
Inspection of Tables 5-2 through 5- 1 1 illustrated that these ranges were quite large; on
average 152% of the minimum prediction and 56% of the maximum prediction.
Furthermore, large coefficients of fiction at the shoulder (b), the waist De), and lumbar
region (pL) the sudaces in question exacerbated this situation. Most imponantly, while
the regional model predictions encompassed most of the validation outputs, suggesting
the regional models were able to accuratel y predict pack-person interface variables, the
size of the ranges made this statement questionable.
A better understanding of key factors affecting the model could be gleaned by
eliminating the fictional contribution of the lumbar pad, waist belt, and shoulder. Since
the precision of the model was so pwr, it was dificult to access its accuracy. While
most predictions fell within the regional model ranges, the ranges resulting from
frictional components were too large to be accepted as an effective model. It is
recommended that future validations alter the test mannequin andlor test pack to ensure
that smaller ranges for the fictional forces are determined. One suggestion is that the
mannequin be covered with TenonTM sheeting, reducing the coefficients of friction and
thus the ranges of the frictional forces. While this does not provide much information
about the contribution of frictioii, it will help researchers better understand the pack-
person reaction forces.
Inspection of Tables 5-5 and 5-8 reveals that the regional model predictions of
lurnbar contact forces (Fs) were consistently high, while predictions of shoulder reaction
forces (sNX) were more accurate. Since lumbar reaction force (Fx) and shoulder reaction
forces (sNX) were in direct opposition and were modeled as counter forces to each other
they should be equivalent. However, consistently high lumbar reaction force (Fs) values
led the author to believe that, while it was not apparent during testing, the waist belt
contributed some anterior-posterior force to the pack. This would challenge the
assumption that the waist belt did not contribute to anterior-posterior forces.
Future work should attempt to either eli minate any anterior-posterior force
generated by the waist belt or include these forces into the regional model. The . attachment of the waist belt to the pack could be changed to one that allowed free
movement in the antenor-posterior plane. If the waist belt was comected to a rod that
was permitted to slide within a chamber attached to the pack-person and oriented in the
anterior-posterior plane, 2-axis, and Y-axis forces would still be transmitted and X-axis
forces would be eliminated. The other option would be to account for the anterior-
posterior force in the model by calculating it as twice the product of the belt tension and
the cosine of the transverse angle of the belt attachent point.
Despite the two factors noted above, the regional rnodel predictions encompassed
71.3% of the measured values. The most accurate predictions were of waist belt - lumbar pad complex lift force (Fz) whose predictions encompassed 10W of the
measured values while the lumbar reaction force (Fx) predictions were the least accurate,
only encompassing 35.4% of the measured values. The relatively low accuracy of the
latter was best explained by the fact that the waist belt generated some anterior-posterior
force. Other poor predictions included shoulder reaction forces (sN) and lumbar lift
forces ( F ~ ~ ) . Because net shoulder reaction forces (sN) were calculated fiom the root of
the sum of the squares of horizontal (sNx) and vertical (sNz) shoulder reaction forces, the
inherent error in the later two cornbineci to increase the enor in the former. Similarly,
lumbar lift forces ( F ~ ~ ) were derived from lumbar reaction force (Fx) predictions, which
as noted above, were relatively poor.
Sensi tivity Analysis
The sensitivity of upper shoulder strap tensions (Ti) suggested that the
mathematical relationships generated by the equilibrium expressions were "ill-
conditioned" toward upper shoulder strap tensions (Ti). Because the mathematical
relationship disproportionately weighted upper shoulder strap tensions (TI), the physical
system could not be well represented without an improved method of assessrnent for
upper shoulder strap tensions (Ti). Small errors in input values significantly altered the
output value of upper shoulder strap tensions (TI) thus leading to poor predictability.
Alternatively, insensitivity of the lumbar contact forces and waist belt - lumbar pad
complex lie forces (Fs and Fz) led to similar problems for opposite reasons. To register
change in lumbar reaction forces (Fs) or waist belt - lumbar pad complex lifi forces (Fz),
the input values must be changed drastically. It would seem that lumbar reaction forces
(Fx) and waist belt - lumbar pad complex lia forces (Fz) were more responsive to load
factors than pack geometry. The physicai system required that model predictions be
more accurately measured because of the high sensitivity of some model outputs and
relatively low sensitivity of others.
The sensitivity analysis provided support for logical model relationships:
Geometric outputs were sensitive to changes in geometric inputs. Upper shoulder swap
angles (el), lower shoulder strap angles (&), and load lifter strap angles (e4) were al1
sensitive to changes in: the lumbar contact point to shoulder center (d5), pack body to
shoulder center (&), shoulder radius (r), lumbar center to load lifter strap attachment
points (&), and lumbar center to upper shoulder strap attachment points (di) dimensions.
The shoulder wrap angles (al and a4) were also sensitive to changes in similar inputs.
Considering that upper shoulder strap tensions (Tl) were related to lower shoulder strap
tensions (TI) by the modified pulley equation, it was not surprising that upper shoulder
strap tensions (Ti) were sensitive to changes in lower shoulder strap tensions (4). The
same relationship dictated that upper shoulder strap tensions (Ti) and frictional forces
around the shoulder (FR) were sensitive to changes in the coefficient of friction around
the shoulder ($1. It was also logical that shoulder strap tensions (Ti, Tt, and T4) had
significant effects on the shoulder reaction forces (sN, sNX and sNZ). Because forces of
fiction around the shoulder (FR) were derived fiom the difference between the lower
shoulder strap tensions (TI) and the combination of the upper shoulder strap tensions (Tl)
and the load lifter strap tensions (TJ), it was not surprising that the forces of friction
around the shoulder (Fa) were sensitive to shoulder strap tensions. The lifl forces of the
waist belt (F~z) and lumbar pad ( F ~ z ) were sensitive to changes in anatomical angles (ye
and yL) or the hip and lumbar region. Waist belt forces, such as the compressive forces
(Tx and TX~) , were sensitive to increases in waist belt tension (T3) And finally, the lie forces due to fiction at the waist (Tm) and lumbar regions (Fsr) were sensitive to
changes in the coefficients of fiction in these regions (pe and pt).
Some input-output relationships were less obvious. Increasing the rnass of the
pack (W) while not increasing the shoulder strap tensions (Ti, T2, and T4), which
occuned in the sensitivity analysiq required that the extra mass be bom elsewhere.
Therefore, waist belt - lurnbar pad complex lie forces (Fz), which was essentially the
difference between the vertical shoulder suspension forces (sNX) and the mass of the pack N N (W), was sensitive to changes in pack mass. Shoulder reaction forces (S , S X, and sNz)
and lumbar contact forces (Fx) were highly sensitive to changes in the lumbar contact to
shoulder center (d5) and pack body to shoulder center (&) dimensions. More accurately, N N however, the shoulder reaction forces (S . S X, and sNz) were sensitive to the geometric
outputs, which were in tum sensitive to the geometric inputs. Changes in the angle of
pull of the shoulder straps (01, O*, and 04) would certainly e f f ~ t the distribution of force
about the shoulder. As well, because the lumbar contact forces (Fx) counteracted the
shoulder reaction forces (sNx) in the anterior-posterior plane, the lum bar contact forces
(Fx) were highly sensitive to shoulder reaction forces (sNX). Altering the tensions in the
lower shoulder strap (T2) also created the same chain reaction effect of sensitivity. The
lumbar contact forces (Fx and Fz) were highly sensitive to lower shoulder strap tensions
(Ta) because lower shoulder strap tension (T2) significantly effected the shoulder reaction
forces (sN, sNx, and s'z), which in tum effected the lumbar contact forces (FA and Fz).
The shoulder Contact forces (sN, sNy, and sN2) and lumbar contact forces (Fx and Fz)
were also sensitive to variations in the lumbar centre to upper shoulder straps (di) and
lumbar centre to load Mer stnps (6) dimensions as a result of a similar chain reaction.
Methods and Materials In validating the system, it is important to have an accurate benchmark pack as a
reliable measurement tooi. The test pack was a simplified extemal frame system with
most features of the pack-person interface sirnilar to a standard pack. It is clear that the
pack was representative of a stiff, extemal frame style pack. The results of the study
cannot currently be extended to sofier interna1 frame style packs. Further investigation
into how the mode1 represents these packs would be necessary. The main body of the test
pack was relatively rigid. However, as mentioned earlier, the vertical stays chat served as
the load lifter anachment points were much less rigid than the remainder of the pack.
While this did not effect the model's ability to predict geometry, it may have had some
effect on the force transmitted between the load lifter straps and the pack itself Under
normal static conditions a more compliant connection would not Iead to lower force
transmission. However, the relative sensitivity of the system to extraneous factors
suggested that alteration in any element of the pack-person interface test setup may have
had significant effects on output masures. Furthemore, a dynamic mode1 has been
suggested as a future goal of the research team thus making test pack cornpliance a much
more critical variable.
It is recommended that the vertical stays on the test pack be changed. Altering the
stays so that a uniform stifiess is achieved throughout the pack will reduce the
extraneous factors that may tend to effect measurements. Specifically, because the effect
of varied cornpliance is not well understood, future work should strive to eliminate such
concems. One recommendation is to mn aluminum stays along the entire height of the
pack so that al1 shoulder straps are attached to the pack through the same medium.
Another concem with respect to pack stiffness was the assumption that the waist
belt transmitted forces to the pack body by a pin joint. While this assumption was made
it was known that the connection was not actually a pin joint and a small error was
expected. However, since it has become apparent that the pack-person interface outputs
are quite sensitive to extraneous factors this assumption has become suspect. It is not
ent ire1 y understood what effects t his assumpt ion may have had on the pack-person
interface measures. However, it is believed that the assumption may have had more of an
effect than previously anticipated and moments at the waist k l t may have been
transfened to the pack. The actual rotational equilibrium of the pack would be affected
and thus the test pack may not be indicative of a standard pack in this respect.
Future endeavors should strive to consolidate the pin joint concem. Either by
gaining an understanding of how the stifhess of the waist belt effects the system or
changing the connection of the waist belt to the pack to an actual pin, this concern can be
eliminated. The most effective solution may be to reattach the waist belt to the pack with
a simple bal1 and socket style connection. This would ensure that a pin joint connection
exists and no assumptions need be made.
As was mentioned in Chapter 3, the order of tightening pack straps affected the
direction of the force of friction around the shoulder (FR). It was also discovered during
the initial stages of a broader military pack investigation that the method of donning and
dofing the pack had a significant effect on pack-person interface variables. For example,
doming the pack by tightening the shoulder straps and then the waist belt would result in
a different set of measured outputs than donning the pack by tightening the waist belt and
then the shoulder straps. The doming factor was accounted for during the validation
study by employing the same doming technique throughout the study (taken fiom
protocol employed by experience pack users). However, this result indicates the
sensitivity of the system to extraneous factors and suggests that Gare must be taken during
ftture validation studies or interpretation of results.
As was discussed above, the method used to don the pack significantly affected
the outcome measures. Whiie the same method could be used consistently throughout
future validation studies it may be more important to gain an understanding of how
donning methods affect outcornes. The validation protocol could be repeated using the
various donning procedures outlined in the literature. The outcome measures could then
be compared using a paired analysis and thus gain an understanding of the effects of
doming techniques. Not only can this information be used to better validate the
biomechanical model, but it rnay also provide more insight into the pack-person interface.
The assumption that contact between the pack body and the users' back only
occurs in the lumbar region rnay not be entirely accurate. During two test setups (3-2 and
7-3), the pack-body came into contact with the upper back. The author believes that the
contact was the result of extreme shoulder strap tensions. These two test condition setups
(Table 4-2) required strap tensions that were higher than the predicted equilibrium could
provide. The extra shoulder force pulled the pack against the upper back of the test
mannequin creating another counter force to the shoulder force. The sum of the lumbar
contact forces (Fx) and the new upper back contact force counteracted the excessive
anterior-posterior tensions of the shoulder straps (TI, TI, and T4), thereby permitting the
required tensions.
Previous literature reviews and investigations at Queen's University suggested
that experienced users oflen tighten the shoulder straps beyond the minimum necessary
tension (Pelot et al., 1995). This tension draws the pack into contact with not only the
lumbar region, but also the upper back as well. It is not surprising that this occurred
during two test setups. Therefore, it is suggested that future work attempts to understand
what conditions produce contact with the upper back and how this contact affects
equilibrium conditions. With intemal fkame packs, this issue would be exacerbated.
Some concem existed with the strap tension probe. The force transducer of the
strap tension probe was louited between the pivot and the handles. In this configuration,
the strap tension probe was sensitive to the location of the force applied to the handles. If
the force wes not applied directly over the stop rod, a bending moment was generated
about the stop rod thus altering the output of the force transducer. In this controlled
laboratory setting, a grip that concentrateci the force directly over the stop rod was
employed and was considered reliable and repeatable. However, for future field studies,
the probe should be modified so variable grips on the pliers do not affect the probe's
readings. It is suggested that the force transducer be move to the shaft of the pliers
between the pivot and the strap prongs and the pliers be re-calibrated and re-validated.
With the transducer located between the pivot and prongs no extemal bending moment
would be applied. This would ensure that any gripping technique employed would
provide accurate outputs.
Chapter 7
Future Directions and Conclusions
Future Directions Many concems were expressed and recommendations made about the model and
testing materials in the previous chapter. The concems specitic to this wotk should be
examined and decisions made to consolidate this model before significant model
applications or funher model advances are attempted. Specifically, the following
concerns and recommendations should ail be investigated fbrther:
The stiffness of the vertical stays for anachment of the load lifter straps to the pack
should be altered so that the connection of al1 the shwlder straps provide consistent
cornpliance;
Re-evaluate certain elements of the model, specifically the way in which the load
lifter strap tensions (T4) transfers forces between the shoulder and the pack body. In
fact, the waist-lumbar region should be evaluated in isolation fiom the shoulder
model. Once both elements are accurately represented independently, a combined
comprehensive model, such as the curent attempt can be realized.
The size of the predictive ranges of the regional models should be reduced and the
mode1 revalidated to better understand the pack-person reaction forces;
Eliminate the antenor-posterior force generated by the waist belt on the test pack or
include this force in the model;
Re-evaluate contact between the pack and the user's (or test mannequin's) upper
back;
Detennine the efTects of different donning methods for the test pack;
Improve the design of the strap tension probe so that it may be used in various
situations and be considered more reliable;
Improve the swap tension probe to be insensitive to gripping techniques.
Once this work is improved through consolidation of the model and associateci
validation, researchers would be better able to undertake the following two related
94
investigations: further development of the current optimization routine and development
of a dynamic personal load carriage system biomechanical model.
Previous research (Pelot et ai., 1998) outlined an optimization routine that used a
biomechanical model as its basis. The routine was designed to optimize the pack
geometry, pack materials, kit selection, and kit placement in a pack to minimize negative
effects in the pack-person interface. Currently, the optimization routine uses previous
phases of this model as a base. Once this phase of the model is improved and
revalidated, it can be incorporated in an attempt to funher the optimization routine. The
routine can only be advanced as the biomechanical mode1 is advanced.
The ultimate goal of any model is to provide a perfect representation of a system.
Since the personal load carriage system model represents a dynamic situation, eventually
a dynamic load carriage model needs to be developed. Furthemore, because this phase
of the model has begun to deal with dynamic elements, such as pack stiffness, it seems
that the next phase should attempt such an undertaking. The author suggests that the
biomechanical model be investigated in a cyclical motion, such as the vertical oscillations
associated with gait. Under these conditions, peak forces, moments and pressures could
be studied and the stiffness of certain pack elements evaluated.
Conclusions The main objective of this work was to develop and validate the next phase in a
series of pack-person interface biornechanical models. It was also expected that four sub-
objectives would also be met:
1. The biomechanical model would be the basis for the persona1 load camage system
design tool outlined in Chapter 1;
2. The equilibrium expressions and the regional models contained within the personal
load camage system model would provide researchers and designers alike with more
information about the pack-person interface;
3. The process of developing and validating the biomechanical model would provide
more insight into the pack-person interface;
4. And the load cadage system measumnent tools would be improved through the
addition of a strap tension probe and test pack.
The current phase of the biomechanicai model is not as representative of the
physical pack-person interface as the author had hypothesized. The equilibrium
predictions of the pack-person interface variables were quite poor. A number of
approaches should be taken to improve this phase of the model before it is considered
accurate and appropriate for xientific investigations of design work. The regional
models did a reasonable job at predicting the range in which the physical variables will
reside. However, the ranges were too large to be conclusive about iheir accuracy. The
author also believes that the current representation of the load IiAer strap is quite poor
and it had major impact on many other output variables in the model.
Despite the poor ability of the current model to predict pack-person contact
forces, the model proved to be an excellent predictor of pack geometry. The relative
angles of the three shoulder draps to the pack and the wrap angles of these straps were al1
accurately predicted. This aspect of the rnodel can be used in future investigations.
The biomechanical model could not considered a scientifically valid
representation of the physical pack-person system and should not yet be used to conduct
pack performance evaluations of current or prototype pack designs. However, despite the
fact that the model did not stand up to a rigorous scientific validation, many elements of
the model can provide qualitative information about pack-person interface variables.
Designers may be able to gain some insight into trends or descriptive relationships
between variables, helping the design process.
Development and validation of the current biomechanical model has improved the
understanding of the pack-person interface. Results of this study has illustrated the
following points:
The method of donning a pack can have a significant effect on the pack-person
interface variables;
The direction of the force of fiction is determined by the order in which the shoulder
straps are tightened;
The cornpliance of the pack may have a significant effect on pack-person contact
forces;
The sensitivity analysis revealed that the pack-person intertace might be quite
sensitive to design features;
5. The sensitivity analysis also illustrated that changes in most model outputs were the
logical result of changes to model inputs, illustrating some legitimacy within the
current biomechanical model.
6. Friction within the pack-person interface is a very important characteristic that must
be fiilly understood for future designs.
It is interesting to note that the complex pack-person interface may not be best
represented with an equally complex model, since previous versions of the model seemed
to better represent the system than the current model.
The strap tension probe and the validation test pack make excellent additions to
the Queen's Ergonomics Research Grou p' s battery of persona1 load carriage system
measurement tools. Validation of these tools illustrated that the devices were an effective
means of measuring strap tensions and numerous pack-person contact forces. The
current validation test pack could be used to conduct future pack-person interface
analyses. However, if the stiffness of the pack is required to be uniform, the vertical
stays that serve as the attachment point for the load lifter straps must be improved. Also,
if the waist belt connection is to be assumed a pin joint, the rigidity of that connection
must be reduced. The strap tension probe also requires some fiirther investigation before
it can be used in fùture analyses. Despite the fact that the probe was considered valid for
this study, it should be modified to be less sensitive to grip location and thus be more
versatile for field use.
In general the author feels that this work has contnbuted to the larger military
load carriage project. While the current model cannot be used as a robust scientific tool,
much information has be learned and added to the load carriage system knowledge base.
Furthermore, this work provides an excellent beginning for improvements to the current
model and the next iteration of a dynamic biomechanical model. Although the model
cannot be used as the basis for a valid design tool, qualitative information can be gleaned
€rom the model for design purposes. In addition, two new measurement tools have been
added to the load carriage system measurement battery and can be used for both fùture
validation studies as well as investigations in a larger scientific scope. Finally, the tiiture
directions outlined above detail the strategy by which the current model and measurement
tools can be used as a stepping stone to tiiture scientific work.
References
AMTI. (1 989). Biomechanics Platfonn Set Instructions Manual. Boston, Mass.
............................................................ Validation measurements: Group averages 1 17
..................................................................... Sample data set of model predictions 1 18
Minhum Mpdmum MWmun Markmim Minhum Mixlmum Minimum W m u m Minimum 27.10 37.24 1934 23.1 7
13.03
31.6s
20.14 46.50
27.51
40.60
21 .?O
23.03
4-43
21 .O4
13.05
27.62 15.29
4.43
1 7.35
22.28
49.65
32.23
2 7.29 33.50
12.74
40.06
26.m 20.86 25.S 34.87
16.07
35.66 18.1 7
31.0t
26.20 39.8s 46.06 38.06
nat 22.44 17.57
13.66
20.1 1
40.22
29.72 18.73
n.n
X s*, f, Pt Pz Modmum Mlnimum Msximum Minimum Mprknum Mnknum W m w n Minimum Msximurn
irwu of pck input, Symbd M a Unib W 33.00 kg 1 up9ll8h. Stnp q k from n m .
V W t l c r l p o J t k n o f ~ o f m r u h o i f i o r i t o l ~ o f ~ o f m r u wrtkrl dimemion of pack horimtal dl- of prck lumbr contrct to di. C m pack to rti. Center dius of rh0Uw.r fadiw of h i p body ban angb low b c k angk hip angle bwer stl. Stmp tsmkn laid lifter tendon lumbar Mer to upper sh. Stmp lumtmr Wdl to kwr sh. Strsp lumôar Ebntw to kad iiitff drap eoelflcknt of friction iround 8h. Waia belt tam&n lumbar psd mter to bottom of prck coefficht of fricth of wrht bsn coeflicknt of frktbn of lumôar pnâ
force of fricth wwnd diouidar
c o m p r ~ v e r01C-e of WaM belt normal fana componsnt of T3C force of friction due to f X
Fx 115.09 N Fz 236.63 N TI 26.11 N lift on pack from lumbar cornpiex
Statk Equilikium Output8 [~meripcim ~yrnimi ~ a t a Units
force of friction due Co fXN
-ci0 min )LS
~ ) i ô
min Ps Ps
min Cis lfux Ci6 min )rs - PB min PB
lfuxcie min W c i e min )y
m)ie
min )is
WIis min Ps
Cie min )is mmt Cis min Ci8
m)is
min Ps Ci6
min
-)is
min r s
min Mi -Cis -)ib
M C i e fnin P?l
inhl Pa
Appendix 3
Cornpiete Model Sensitivity Analysis Results
Unique variable outputs as input variables increase by 1 û% ................................... 121
N o m l i z e d increase in output variables as input variables increase by 10% .......... 123