Professional Documents
Culture Documents
Fall 2017
Maria Hudock BME 303 Fall 2017 Honors Project
I. Abstract
From a macro-scale biology view, bone might be considered to be a solid. However, its composition and
organization on meso- and microscales mean that bone, like any biomaterial, is along the spectrum of
viscoelasticity, and a good deal of fluid must also be taken into account. Here, the principles of BME 303
are applied to understand the non-intuitive mechanical situations that arise due to 1) essential
interstitial fluid movement for mechanotransduction, and 2) the viscoelastic-viscoplastic properties of
bone. An equation and flow profiles to describe the pressure-driven movement of interstitial fluid in
canaliculi are derived, and the viscoelastic component of bone is used as a model to understand stress
and strain rate responses to varied inputs.
II. Introduction
Bone is a mechanically remarkable biomaterial: it sustains cyclic and impact loading at various angles and
load rate without easily fracturing, yet is porous and interconnected enough such that it provides proper
circulation of nutrients and waste to and from the cells that dynamically sustain it. It is extremely well-
ordered on a hierarchal scale: from the alignment within collagen fibers themselves, to the staggered
alignment and crystal pattern of mineralized fibrils, to precise “twisted plywood” arrangement between
layers in a lamellar wall, to the concentric lamellae, to the organization of osteons and the vascular
canals.1,2 This hierarchy, summarized in Figure 1,2,3 imparts the properties that make bone the light yet
remarkably durable material that it is.
a
b
Researchers have long been occupied with understanding the mechanics of bone from various
perspectives. Solid mechanics might be the most obvious route of analysis: after all, bone is often
characterized as the solid support frame that keeps our otherwise sac-like organs in their proper locations.
However, bone is not a simple elastic solid. Like most any tissue biomaterial, it is filled with both
vasculature and interstitial fluid – thus, fluid mechanics plays a role in bone. Bone is also hydrated and
rests along the spectrum of viscoelastic materials. As will be shown, two important applications of fluid
and viscoelastic material analysis in bone are to understand mechanotransduction and stress dissipation
in bone, respectively.
In this project, the fundamental approaches developed in BME 303 will be applied to the abovementioned
biological problems related to bone. Fluid mechanics analysis will be used to understand the interstitial
Maria Hudock BME 303 Fall 2017 Honors Project
fluid flow profiles in canaliculi that mediate the transduction of mechanical stress to bone cells to
stimulate growth or atrophy. Then, bone will be examined as a viscoelastic biomaterial to understand its
deformation response to loading.
Bone, like most any other tissue, is vascularized and contains ample interstitial fluid to transport
nutrients and waste, so fluid mechanics for transport plays a vital role within this hard-tissue organ. This
section will focus on flow of interstitial fluid; the interstitial fluid compartments of bone will be
introduced briefly before our fluid mechanics problem is introduced.
below 0.5% strain, yet physiological tissue strain seldom exceeds 0.2%.4,5 However, there has been
convincing evidence for the second hypothesis: in 1994, Weinbaum et. al. developed a model to show
that physiologically relevant tissue strain could cause pressure changes within bone canaliculi that were
large enough to drive interstitial fluid flow, causing a shear stress on osteocyte processes within the
canaliculi. This fluid shear was, surprisingly, predicted to be of the same order of magnitude as that
experienced by vascular endothelial cells.4,6 Later, when it was noted that even this shear would
probably not significantly strain the tough walls of the cell processes, others built on this model by
including the influence on drag on the proteoglycan matrix “tethers” between the cell process and the
canalicular walls.4 If this matrix is considered to be flexible relative to the cell process, which is
additionally anchored in place by rigid integrins, the tension developed in the proteoglycan matrix under
flow translates to additional hoop and shear stress on the cell process, significantly amplifying the stress
effect on the cell during mechanical loading of the bone. This is illustrated in Figure 3, an adaption of a
figure from Han et al. (2004) taken from Fritten and Weinbaum’s 2004 review.4
In this section, an equation to describe the flow of the interstitial fluid under the time-variant pressure
gradient developed during loading will be derived, starting with
the general Navier-Stokes equation for a Newtonian viscous fluid
(considering the interstitium to be a Newtonian fluid). The steps to
solving it will be put forth, although the mathematics for the
actual solution are beyond the scope of usual differential
equations methods, so an exact solution was not found. The
structure of the solution for different cases will be compared to
the solutions in Weinbaum et al.’s 1994 derivation.6
To model this flow, we will consider the canaliculi with cell process
extending down the center as two concentric cylinders, with the
proteoglycan matrix and integrins partially occupying the space Figure 4: Cylindrical model of a canaliculus.
between the two cylinders (Figure 4).
Starting with the general Navier-Stokes to time because the pressure gradient varies
equation for a Newtonian viscous fluid, with time, so we have:
the velocity is not necessarily the same at the Where each term goes to 0, either because the
canaliculus was as it is at the surface of the cell corresponding velocity or spatial acceleration is
process (the cell process can move to some 𝛿 2 𝑣1
zero as indicated above. Expanding 𝜇 𝛿𝑥 , the
extent), v1 will vary as a function of x2 and x3. 𝑘 𝛿𝑥𝑘
There is no reason it should vary along the axial first term can be dropped because the change
direction, x1. Velocity will change with respect in velocity along x1 is zero:
Maria Hudock BME 303 Fall 2017 Honors Project
This is a linear, non-homogenous second-order PDE. To approach the solution, we start by assuming an
𝛿𝑝(𝑡)
oscillatory pressure function 𝛿𝑧
= 𝑃1 𝑐𝑜𝑠𝑡(𝜔𝑡) = 𝑅𝑒[𝑃1 𝑒 𝑖𝜔𝑡 ]. , so that the velocity can be assumed
to take the form 𝑣𝑧 (𝑟, 𝑡) = 𝐴(𝑟) cos(𝜔𝑡) = 𝑅𝑒[𝐴(𝑟)𝑒 𝑖𝜔𝑡 ], where A(r) is the amplitude of the velocity
as a function of r, which we can solve for in the complex plane in the following.
𝛿 𝛿𝑣𝑧 𝛿
(𝑟 ) = 𝑅𝑒[𝑟𝐴′(𝑟)𝑒 𝑖𝜔𝑡 ] = 𝑅𝑒[𝑟𝐴′′ (𝑟)𝑒 𝑖𝜔𝑡 + 𝐴′ (𝑟)𝑒 𝑖𝜔𝑡 ]
𝛿𝑟 𝛿𝑟 𝛿𝑟
So the differential equation in cylindrical coordinates, transformed to the complex plane, becomes:
𝜗 1 𝜗
𝑅𝑒[𝑖𝜔𝐴(𝑟)𝑒 𝑖𝜔𝑡 ] = − 𝑅𝑒[𝐴(𝑟)𝑒 𝑖𝜔𝑡 ] − 𝑅𝑒[𝑃1 𝑒 𝑖𝜔𝑡 ] + 𝑅𝑒[𝑟𝐴′′ (𝑟)𝑒 𝑖𝜔𝑡 + 𝐴′ (𝑟)𝑒 𝑖𝜔𝑡 ]
𝑘 𝜌 𝑟
Cancelling out the 𝑒 𝑖𝜔𝑡 terms and dropping the Re operator, with the understanding that the solution
will consist of only the real part, we have:
𝜗 𝑃1 𝜗
𝑖𝜔𝐴(𝑟) = − 𝐴(𝑟) − + 𝜗 𝐴′′ (𝑟) + 𝐴′(𝑟)
𝑘 𝜌 𝑟
Rearranging, and dividing by 𝜗, this is:
Maria Hudock BME 303 Fall 2017 Honors Project
𝑃1 𝑖𝜔 1 1
= (− − ) 𝐴(𝑟) + 𝐴′ (𝑟) + 𝐴′′ (𝑟)
𝜇 𝜗 𝑘 𝑟
We note that the general form of this equation is a second-order linear ODE for A(r), but it does not
have constant coefficients. Therefore, we cannot solve it using eigenfunctions. The general form of this
function is:
𝐴′′ + 𝑝(𝑟)𝐴′ + 𝑞(𝑟)𝐴 = 𝑓(𝑟), where q and f are constant functions, but p(r) = 1/r.
We can assume that the solution is the sum of a particular and a homogenous solution. Assume the
simplest particular solution, that is, Ap(r) = constant, so that its first and second derivatives are zero:
𝑃1 𝑖𝜔 1
= (− − ) 𝐴𝑝 (𝑟) + 0 + 0
𝜇 𝜗 𝑘
𝑃1 −𝑃1 𝜗𝑘
𝐴𝑝 (𝑟) = =
𝑖𝜔 1 𝜇(𝑖𝜔𝑘 + 𝜗)
𝜇 (− − )
𝜗 𝑘
With the help of mathematics software8, the homogenous equation can be recognized as of the form:
1 𝑑 𝑖𝜔 1
𝐴′′ + 𝐴′ + 𝛼𝑟𝐴 = (𝑟𝐴′ (𝑟)) + 𝛼𝑟𝐴(𝑟) = 0, 𝛼 = (− − )
𝑟 𝑑𝑟 𝜗 𝑘
This is a Sturm-Liouville equation, and the homogenous solution takes the form:
Where c1 and c2 are constants, and J0 and Y0 are Bessel functions of the first and second kind,
respectively, defined as follows with m= 0:
∞
(−1)𝑙 𝑥
𝐽𝑚 (𝑥) = ∑ ( )2𝑙+𝑚 , 𝑌𝑚 (𝑥) = lim (𝐽𝑛 (𝑥) cos(𝑛𝜋) − 𝐽−𝑛 )/sin(𝑛𝜋)
𝑙! Γ(𝑙 + 𝑚 + 1) 2 𝑛→𝑚
𝑙= 0
Therefore, substituting the particular and homogenous solutions for A(r) into our general solution for vz:
𝑖𝜔 1 𝑖𝜔 1 −𝑃1 𝜗𝑘
𝑣𝑧 (𝑟, 𝑡) = 𝑅𝑒 [(𝑐1 𝐽𝑜 (𝑟√− − ) + 𝑐2 𝑌𝑜 (𝑟√− − ) + ) 𝑒 𝑖𝜔𝑡 ]
𝜗 𝑘 𝜗 𝑘 𝜇(𝑖𝜔𝑘 + 𝜗)
Because the arguments of the Bessel functions are complex, the true operators should be modified
Bessel functions, I0 and K0 for J0 and Y0. The coefficients c1 and c2 would be determined from the initial
conditions, which are as follows:
1) At the canalicular wall, defined as r = b, under no-slip conditions, v(b, t) = 0 because the wall
does not move.
2) At the surface of the cell process, defined as r = a, the wall moves slightly as it is pulled by the
flexible proteoglycan matrix. If we consider this matrix as deforming with spring constant c,
moving due to the pressure exerted over the surface area of the matrix fibers, then we can say:
𝐹𝑝𝑟𝑒𝑠𝑠𝑢𝑟𝑒 − 𝐹𝑠𝑝𝑟𝑖𝑛𝑔 = 𝑚𝑎𝑤𝑎𝑙𝑙
𝑝(𝑡) ∗ 𝐴𝑓𝑖𝑏𝑒𝑟𝑠 − 𝑐Δz = 𝑚𝑎𝑧 (𝑎, 𝑡), Δz = fiber displacement (bending)
Maria Hudock BME 303 Fall 2017 Honors Project
1
𝑣𝑧 (𝑎, 𝑡) = ∫(𝑝(𝑡) ∗ 𝐴𝑓𝑖𝑏𝑒𝑟𝑠 − 𝑐Δz)dt
𝑚
Since this spring constant c is unknown, and with the complexity of the modified Bessel functions, it is
difficult to evaluate the actual values of c1 and c2. It is sufficient to say that the amplitude at the
canalicular wall will be 0 at all time, and it will be a function of time at the cell process membrane. With
this information, and using general graphs for these Bessel functions,* we can predict that the velocity
profile for an instant in time will be the shape of J0 + Y0 (stretched by the constants, scaled by their
arguments, and shifted by the particular solution) from some nonzero value at r = a to zero at r = b. That
is, we can truncate the graph based on knowledge of the boundary conditions to get an approximation
to the velocity profile.
The graph of J0 + Y0 at t such that Re[eiwt] = cos(wt) = 1 and the resulting velocity profile in the
canaliculus are demonstrated in Figure 5. Since this is the maximum value of cosine, this is the maximum
scaling of the graph of J0 + Y0 and hence the “maximum profile” (most positive velocities).
Interstitial fluid +
proteoglycan matrix
cell process
We can compare this work to the profile derived by Weinbaum et al. First, it is necessary to compare the
initial assumptions to understand any differences in how the velocity profile is described. The
𝜇
Weinbaum group started the flow profile derivation with Brinkman’s equation, ∇𝑝 = − 𝑘 𝑣𝑖 + 𝜇∇2 𝑣𝑖 .
Rearranging to compare to the Navier-Stokes equation used in this work, we have
𝜇
0 = − ∇𝑝 − 𝑘 𝑣𝑖 + 𝜇∇2 𝑣𝑖 , which assumes that velocity does not vary in time. This is the only key
difference between the derivation in the present work and the work of Weinbaum et al.
The solution put forth by this group (modified to match the coordinate axes in the present work) also
takes the form of the sum of two (modified) Bessel functions minus a constant:
𝑘𝑝 𝛿𝑝 𝑟 𝑟
𝑣𝑧 = [𝐴1 𝐼0 (𝛾 ) + 𝐵1 𝐾0 (𝛾 ) − 1],
𝜇 𝛿𝑧 𝑏 𝑏
𝛾 𝛾
𝐾0 (𝛾) − 𝐾0 (𝑞 ) 𝐼0 (𝑞 ) − 𝐼0 (𝛾)
𝐴1 = 𝛾 𝛾 , 𝐵1 = 𝛾 𝛾 , 𝛾 = 𝑏/√𝑘𝑝
𝐼0 (𝑞 ) 𝐾0 (𝛾) − 𝐼0 (𝛾)𝐾0 (𝑞 ) 𝐼0 (𝑞 ) 𝐾0 (𝛾) − 𝐼0 (𝛾)𝐾0 (𝑞 )
Bone might be generally considered solid in everyday life, but it is indeed the fact that it is not a perfect
solid that helps it serve its purpose as a fracture-resistant support system for the body. In order to
sustain loading though a variety of modes and frequencies, bone must be a viscoelastic-viscoplastic
material that is capable of dissipating mechanical energy. Several modes of energy dissipation have
been proposed for bone, including homogeneous and inhomogeneous thermoelastic coupling, fluid
motion, inhomogeneous deformation (such as motion at cement lines between osteons), piezoelectric
coupling, and molecular-level effects such as unequal distribution of stress and strain between
hydroxyapatite nanocrystals and collagen fibrils.9,10 Though these each contribute to different levels
(piezoelectric effects being negligible, and molecular-level effects being difficult to measure), their total
effect can be summarized as viscous elements in spring-dashpot model of bone. A number of groups
have attempted to determine how many and what kind of elements best represent bone in a
mechanical model. Recently, one group drew upon the work of many others to determine that there are
at least two significant time constants to energy dissipation a
in bone, and thus at least two viscous elements in addition
to the spring(s) that we would expect for a solid.11
The viscoelastic system shown in Figure 8b is composed of a standard linear model with an additional,
parallel Maxwell model; this arrangement is known as a Wiechert model.12 Figure 8 shows the
viscoelastic model only, turned on its side and sectioned to reveal the internal forces. The spring
constants and damping factors given by Johnson et al. are Young’s moduli and viscosities; it is important
to note that these relate the stress, strain, and strain rate, rather than force, displacement, and
displacement rate. To use these constants, analysis must be done using σ and ε, but the methods and
relationships are exactly analogous to F and u. We can make several mathematical statements by
inspecting Figure 9:
This model cannot be easily solved in the time domain because there are not enough simultaneous
equations to express the stress components in terms of only σ without using at least one of the branch
stresses or its time derivative. The model can be easily solved, however, by converting to the Laplace
domain, then adding the impedances of each of the components according to the arrangement.
𝑑𝜀
In the Laplace domain, each element is expressed as its impedance times the rate of strain, 𝜀· =
𝑑𝑡
·
𝜀(𝑠)
·
𝜀(𝑠). Using 𝑠 = 𝜀(𝑠) , The expressions become:
𝐸0 ·
1. Branch (a): 𝜎𝑎 (𝑠) = 𝜀 (𝑠)
𝑠 𝑎
𝐸
2. Branch (b): 𝜎𝑏 (𝑠) = 𝑠1 𝜀·𝑏1 (𝑠); 𝜎𝑏 (𝑠) = 𝜂1 𝜀· 𝑏2 (𝑠)
𝐸 ·
3. Branch (c): 𝜎𝑐 (𝑠) = 2 𝜀· 𝑐1 (𝑠); 𝜎𝑐 (𝑠) = 𝜂2 𝜀𝑐2 (𝑠)
𝑠
The impedance of each component is the factor that relates the force in the element to the
displacement rate. We now sum the impedances of each branch. For elements in series, impedances
1 1
add as 𝑍 = ∑ 𝑍 , so:
𝑖
𝐸0
1. Branch (a): 𝑍𝑎 = 𝑠
1 1 1 𝑠 1 𝐸1 +𝜂1 𝑠 𝐸1 𝜂1 1
2. Branch (b): = + = + = ; 𝑍𝑏 = = 𝐸1 𝐸
𝑍𝑏 𝑍𝑏1 𝑍𝑏2 𝐸1 𝜂1 𝐸1 𝜂1 𝐸1 +𝜂1 𝑠 𝑠+ 1
𝜂1
1 1 1 𝑠 1 𝐸2 +𝜂2 𝑠 𝐸2 𝜂2 1
3. Branch (c): = +𝑍 = + = ; 𝑍𝑐 = = 𝐸2 𝐸
𝑍𝑐 𝑍𝑐1 𝑐2 𝐸2 𝜂2 𝐸2 𝜂2 𝐸2 +𝜂2 𝑠 𝑠+ 2
𝜂2
We can add the impedances of each of the parallel branches to get the total impedance for the system.
Parallel impedances add as 𝑍 = ∑ 𝑍𝑖 , so:
Maria Hudock BME 303 Fall 2017 Honors Project
𝐸0 1 1
𝑍 = 𝑍𝑎 + 𝑍𝑏 + 𝑍𝑐 = + 𝐸1 + 𝐸2
𝑠 𝐸 𝐸
𝑠 + 𝜂1 𝑠 + 𝜂2
1 2
·
Therefore, for the total viscoelastic system, σ(s) = Z ε(s):
𝐸0 1 1 ·
𝜎(𝑠) = ( + 𝐸1 + 𝐸2 ) 𝜀(𝑠)
𝑠 𝐸1 𝐸2
𝑠+𝜂 𝑠+𝜂
1 2
The time-domain response can be found by taking the inverse Laplace transform of the Laplace-domain
equation above:
𝐸1 𝐸2
· − 𝑡 · − 𝑡 ·
𝜎(𝑡) = 𝐸0 ∫ 𝜀(𝑡)𝑑𝑡 + 𝐸1 (𝑒 𝜂1 ) ⋇ 𝜀(𝑡) + 𝐸2 (𝑒 𝜂2 ) ⋇ 𝜀(𝑡)
where ⋇ indicates convolution, since multiplication of the impedance function and the strain rate
function in the Laplace domain indicates convolution in the time domain. By the definition of the
convolution integral, if the strain rate is constant, and we assume a stress response from time 0 to t,
𝐸 𝑡 𝐸1 𝑡 𝐸1
− 1𝑡′ · ′ ) ≡ ∫ 𝐸 (𝑒 −𝜂1 𝜏 ) 𝜀(𝑡 · ∫ (𝑒 −𝜂1 𝜏 ) 𝑑𝜏
· ′ − 𝜏)𝑑𝜏 = 𝜀𝐸
𝐸1 (𝑒 𝜂1 ) ⋇ 𝜀(𝑡 1 1
0 0
𝐸 𝐸 𝐸
· (− 𝜂1 ) 𝑒 −𝜂11 𝜏 |𝑡 = −𝜀𝜂
= 𝜀𝐸
1
· (1 − 𝑒 −𝜂11 𝑡 )
· (𝑒 −𝜂1 𝑡 − 1) = 𝜀𝜂
1 0 1 1
𝐸1
and similarly for the second convolution integral. Then the stress is related to the strain rate in the time
domain by:
𝐸 𝐸
− 1𝑡 − 2𝑡
𝜎(𝑡) = 𝐸0 𝜀 𝑉𝐸 𝑡 + 𝜀 𝑉𝐸 𝜂1 (1 − 𝑒 𝜂1 ) + 𝜀 𝑉𝐸 𝜂2 (1 − 𝑒 𝜂2 )
Where VE signifies the viscoelastic displacement rate. This precisely matches the solution put forth by
Johnson et al.11
11 𝑉𝑃
The solution for the viscoplastic element of the entire system · group as 𝜀 =
· is given by the same
𝜎 |𝜎| 𝑚
( ) , where S is a yield stress (from intercept of plot ·in Figure 10) and m is a power relationship
|𝜎| 𝑆0
0
determined from the inverse slope of the trend line in Figure 10.
* The units of strain rate are s-1. The expression on the righthand
side is unitless (S0 has units of stress), so it appears the expression
might require a constant with units of s-1 on the righthand side.
Figure 10: Yield stress vs. strain rate plot. This constant is omitted in analysis, however, because even if
log(S0) = y intercept; m = inverse of required, we do not have its value.
slope.
Maria Hudock BME 303 Fall 2017 Honors Project
Given a constant strain rate, a constant stress, and a sinusoidal strain rate with a frequency of 10/2π Hz
(order of magnitude of loading rate during walking), the respective responses are shown in Figure 11.
a b
c
Figure 11: Stress or strain rate responses to (a)
constant strain rate, which reaches half maximum
physiological strain at t = 10s, (b) constant stress,
approximated by half of the weight of a 150lb
person (68kg) on a femur of r = 1.5cm, and (c)
oscillating strain rate with a frequency of 10 rad/s.
All responses derived from model of bone as a
viscoelastic material only.
It is evident in Figure 11 that the viscoelastic effects of bone, responsible for energy dissipation, are
subtle compared to the solid response one more typically perceives. In 11a, bone appears to behave, for
all intents and purposes, as a perfect solid: constant strain rate (linearly increasing strain) leads to
linearly increasing stress, following σ = εe. In Figure 11c, we also see that the phase shift between strain
rate and stress response is π/2, implying that strain (integral of strain rate) and stress are perfectly in
phase (also characteristic of a solid). This would not be good news for energy dissipation in bone;
however, the energy dissipation effect is slightly more evident in Figure 11b. There, we see that for a
constant stress, the strain rate is not an impulse function (it is not infinite, then immediately zero, as for
the nearly instantaneous response of a solid). Instead, strain rate slowly decays to zero over a period of
about 10 seconds. Thus, strain itself increases rapidly at first before tapering to a maximum, instead of
occurring as a step function. Though the time constant for this process seems to be unrealistically large,
parameters or assumptions in the model that might lead to erroneous scale were not found. The fact
that deformation in bone does not occur instantaneously contributes to its durability. An applied stress
causes a change in momentum of the material. For this given change in momentum, taking place over a
longer time means that the average stress experience by the material is lower according to the force-
impulse relationship: 𝑖𝑚𝑝𝑢𝑙𝑠𝑒 = 𝑚Δ𝑣 = 𝐹Δ𝑡 = 𝜎𝐴Δ𝑡. The longer Δt, i.e., the longer the time constant,
the lower the average stress for a given change in momentum. One can see from Figure 11b that the
effect of the viscoelastic components of bone is to increase Δt. In combination with the viscoplastic
effect, the viscoelastic behavior of bone is what makes it so resistant to fracture and strong for its
relatively low mass.
Maria Hudock BME 303 Fall 2017 Honors Project
V. Conclusions
In this project, the concepts of BME 303 were extended to understand research describing the fluid and
viscoelastic mechanics in bone. Bone is an interesting biomaterial because it is so frequently perceived
as a solid, but it could not execute its biological role (growth and atrophy in response to mechanical
stimulation, dissipation of mechanical energy to prevent fracture) without fluid mechanics or its
viscoelastic-viscoplastic properties. The Navier-Stokes equation was used to derive an equation that
describes the flow profiles in canaliculi; with additional information about the viscoelasticity of bone
cells, these flow profiles can be used to calculate the magnitude of stress and strain in cell processes and
the cellular response to this mechanical signaling. Using knowledge of the effects of solid and viscous
components in relating stress, strain, and strain rate, a mathematical description of how bone responds
to various loadings was derived. From that mathematical result, the mechanism of durability in bone –
by reducing average stress for a given momentum change – can be demonstrated, and its response to
loading in pathological situations can be predicted.
VI. Acknowledgements
Dr. Dong, I would like to sincerely thank you for a wonderful semester. You are a talented lecturer, and I
truly enjoyed the dynamicity and first-principles explanations of the class. BME 303 has given me not
only a greater understanding of mechanics, but it has inspired me to look for ways to apply these
techniques and knowledge at the next level in my research. Thank you for sharing your lessons,
enthusiasm, and real-world perspectives. These will stick with me for a long time.
Maria Hudock BME 303 Fall 2017 Honors Project
References
1. Martin R.B., Burr D.B., Sharkey N.A., Fyhrie D.P. (2015) Skeletal Biology. In: Skeletal Tissue
Mechanics. Springer, New York, NY. DOI https://doi.org/10.1007/978-1-4939-3002-9_2
2. Giraud-Guille, M.M. Twisted plywood architecture of collagen fibrils in human compact bone
osteons. Calcified Tissue International 1988, 42:167-180. https://doi.org/10.1007/BF02556330
3. Sadat-Shojai, M.; Khorasani, M.-T.; Dinpanah-Khoshdargi, E.; Jamshidi, A. Synthesis methods for
nanosized hydroxyapatite with diverse structures. Acta Biomaterialia 2013, 9(8):7591-7621.
https://doi.org/10.1016/j.actbio.2013.04.012
4. Fritton, S.P.; Weinbaum, S. Fluid and Solute Transport in Bone: Flow-Induced
Mechanotransduction. Annual Review of Fluid Mechanics 2009, 41:347-374.
https://doi.org/10.1146/annurev.fluid.010908.165136
5. You, J.; Yellowley, C.E.; Donahue, H.J.; Zhang, Y.; Chen, Q.; Jacobs, C.R. Substrate deformation
levels associated with routine physical activity are less stimulatory to bone cells relative to
loading-induced oscillatory fluid flow. Journal of Biomechanical Engineering 2009, 122:387–93.
6. Weinbaum, S.; Cowin, S.C.; Zeng, Y. A model for the excitation of osteocytes by mechanical
loading-induced bone fluid shear stresses. Journal of Biomechanics 1994, 27(3):339-360.
https://doi.org/10.1016/0021-9290(94)90010-8
7. Wikipedia contributors. "Darcy's law." Wikipedia, The Free Encyclopedia.
https://en.wikipedia.org/w/index.php?title=Darcy%27s_law&oldid=809043775 (accessed Nov
20, 2017).
8. “General Differential Equation Solver.” Wolfram Alpha Widgets.
https://en.wikipedia.org/w/index.php?title=Darcy%27s_law&oldid=809043775 (accessed Nov
20, 2017).
9. Lakes, R.S.; Katz, J.L. Viscoelastic properties of wet cortical bone-II: relaxation mechanisms.
Journal of Biomechanics 1979, 12:679-687.
10. Nair, A.K.; Gautieri, A.; Chang, S.-W.; Buehler, M.J. Molecular mechanics of mineralized collagen
fibrils in bone. Nature Communications 2013, 4:1724. DOI: 10.1038/ncomms2720.
11. Johnson, T.P.M.; Socrate, S.; Boyce, M.C. A viscoelastic, viscoplastic model of cortical bone valid
at low and high strain rates. Acta Biomaterialia 2010, 6(10):4073-4080.
https://doi.org/10.1016/j.actbio.2010.04.017
12. Roylance, D. Engineering Viscoelasticity. MIT OpenCourseWare 2010.
https://ocw.mit.edu/courses/materials-science-and-engineering/3-11-mechanics-of-materials-
fall-1999/modules/MIT3_11F99_visco.pdf (accessed Nov 24, 2017).
BME 303 Honors Project
Supplementary Materials: Code for Figure Generation
Maria Hudock
Fall 2017
Navier-Stokes Description of Flow in a Cannaliculus: Velocity Profiles
% Now, scale by different values of cos(wt) assuming loading frequency of
% 1Hz => w = 2pi to examine velocity profile at different points in time
% With a load every 1 second, profile will repeat (backwards) after 0.5s.
% Plot only 0.5s of loading.
r = 0:0.01:3.5; % change r limits, since we know A = 0 occurs around r = 3
t = 0:0.1:0.5; % time vector for one half-second of loading
Afunct = zeros(length(t), length(r)); % preallocate amplitude function array
w = 2*pi; % assign loading frequency in radians
figure(2) % call a new figure
for ii = 1:length(t)
Afunct(ii, :) = (besselj(0,r) + bessely(0,r) -0.25)*cos(w*t(ii));
% fill a new row in amplitude array for each value of t
plot(r, Afunct(ii, :))
hold on % so that each resulting plot will be on the same graph
end
plot([0, 3.5], [0, 0.000001]) % plot a line to approximate y = 0
legend('t = 0', '0.1', '0.2', '0.3', '0.4', '0.5', 'line v = 0')
title('Velocity Profile for Selected Points in Time')
xlabel('radius (au)') % au = arbitrary units
ylabel('Velocity')
% annotations on graph were created manually using the plot editing tool
Sigma =
% note: the whole function result will not display in the PDF, but it is rewritten in
% the code below
E0 = 10*10^9; %Pa
E1 = 5*10^9; % Pa
E2 = 10*10^9; % Pa
eta1 = 120*10^6; % Pa s
eta2 = 100000; % Pa s
t = 0:0.01:2*pi; % s; time for simulation of sinusoidal strain rate
% Write a numeric double equation matching the symbolic result above,
% and dividing by 1000 so that the result is in kPa:
Sig = [((cos(10*t)*E1^2*E2^2*eta1 + cos(10*t)*E1^2*E2^2*eta2 + ...
0.1*E0*sin(10*t)*E1^2*E2^2 + 10*sin(10*t)*E1^2*E2*eta2^2 + ...
100*cos(10*t)*E1^2*eta1*eta2^2 + 10*E0*sin(10*t)*E1^2*eta2^2 ...
+ 10*sin(10*t)*E1*E2^2*eta1^2 + 1000*sin(10*t)*E1*eta1^2*eta2^2 ...
+ 100*cos(10*t)*E2^2*eta1^2*eta2 + 10*E0*sin(10*t)*E2^2*eta1^2 ...
+ 1000*sin(10*t)*E2*eta1^2*eta2^2 + ...
1000*E0*sin(10*t)*eta1^2*eta2^2)/((10000*E2^2 + 1000000*eta2^2)* ...
(E1^2 + 100*eta1^2)) - (E1^2*eta1*exp(-E1*t/eta1))/(10000*E1^2 + ...
1000000*eta1^2) - (E2^2*eta2*exp(-E2*t/eta2))/(10000*E2^2 + 1000000*eta2^2))/1000];
% Rewrite the strain rate in microstrain per second:
Epsdot = 0.1*cos(10*t); % microstrain per second
% Plot the strain rate and the stress response:
figure(5)
subplot(1,2,1)
plot(t, Epsdot, 'linewidth', 2)
xlabel('time (s)')
ylabel('Strain Rate (microstrain s^-1)')
title('Sinusoidal Strain Rate')
xlim([0, 1.01])
subplot(1,2,2)
plot(t, Sig, 'linewidth', 2)
xlabel('time (s)')
ylabel('Stress Response (kPa)')
title('Stress Response')
xlim([0, 1.01])