Pulsatile flow
≈ 9 min read
Names in other languages
No names in other languages are held for this entry yet.
Article bodies are English and are not machine-translated. A name in another language is not a translation of the article.
In fluid dynamics, a flow with periodic variations is known as pulsatile flow, or as Womersley flow. The flow profiles was first derived by John R. Womersley (1907–1958) in his work with blood flow in arteries. The cardiovascular system of chordate animals is a very good example where pulsatile flow is found, but pulsatile flow is also observed in engines and hydraulic systems, as a result of rotating mechanisms pumping the fluid.
In fluid dynamics, a flow with periodic variations is known as pulsatile flow, or as Womersley flow. The flow profiles was first derived by John R. Womersley (1907–1958) in his work with blood flow in arteries. The cardiovascular system of chordate animals is a very good example where pulsatile flow is found, but pulsatile flow is also observed in engines and hydraulic systems, as a result of rotating mechanisms pumping the fluid. u(r, t) = \mathrm{Re}\left\{ \sum^N_{n=0} \frac{i \, P'_n}{\rho \, n \, \omega} \left[ 1 - \frac{J_0(\alpha \, n^{1/2} \, i^{3/2} \, \frac{r}{R})}{J_0(\alpha \, n^{1/2} \, i^{3/2})} \right] e^{i n \omega t} \right\} \, , where: \alpha = R \left( \frac{\omega \rho}{\mu} \right)^{1/2}. For \alpha \lesssim 2 , viscous forces dominate the flow, and the pulse is considered quasi-static with a parabolic profile. For \alpha \gtrsim 2 , the inertial forces are dominant in the central core, whereas viscous forces dominate near the boundary layer. Thus, the velocity profile gets flattened, and phase between the pressure and velocity waves gets shifted towards the core. \lim_{z\to\infty} J_0(z) = 1 - \frac{z^2}{4} \,, which converges to the Hagen-Poiseuille flow profile for steady flow for \lim_{n \to 0} u(r, t) = - \frac{P'_0}{4 \mu} \left(R^2 - r^2 \right) , or to a quasi-static pulse with parabolic profile when \lim_{\alpha \to 0} u(r, t) = \mathrm{Re} \left\{ - \sum^N_{n=0} \frac{P'_n}{4 \mu} (R^2 - r^2 ) \, e^{i n \omega t} \right\} = - \sum^N_{n=0} \frac{P'_n}{4 \mu} (R^2 - r^2 ) \, \cos(n \omega t) \, . In this case, the function is real, because the pressure and velocity waves are in phase. \lim_{z\to\infty} J_0(z \, i) = \frac{e^z}{\sqrt{2\pi\, z}} \,, which converges to \lim_{z\to\infty} u(r, t) = \mathrm{Re} \left\{ \sum^N_{n=0} \frac{i \, P'_n}{\rho \, n \, \omega} \left[ 1 - e^{\alpha \, n^{1/2} \, i^{1/2} \left(\frac{r}{R} - 1 \right) } \right] e^{i n \omega t} \right\} = - \sum^N_{n=0} \frac{\, P'_n}{\rho \, n \, \omega} \left[ 1 - e^{\alpha \, n^{1/2} \left(\frac{r}{R} - 1 \right) } \right] \sin(n\,\omega\, t) \, . This is highly reminiscent of the Stokes layer on an oscillating flat plate, or the skin-depth penetration of an alternating magnetic field into an electrical conductor. On the surface, u(r=R,t) = 0 , but the exponential term becomes negligible once \alpha (1 - r/R) becomes large, and the velocity profile becomes almost constant and independent of the viscosity. Thus, the flow simply oscillates as a plug profile in time according to the pressure gradient \rho \frac{\partial u}{\partial t} = - \sum^N_{n=0} P'_n \,. However, close to the walls, in a layer of thickness \mathcal{O}(\alpha^{-1}) , the velocity adjusts rapidly to zero. Furthermore, the phase of the time oscillation varies quickly with position across the layer. The exponential decay of the higher frequencies is faster. \rho \frac{\partial u}{\partial t} = -\frac{\partial p}{\partial x} + \mu \left(\frac{\partial^2 u}{\partial r^2} + \frac{1}{r} \frac{\partial u}{\partial r}\right) \, and \frac{\partial u}{\partial x} = 0 , respectively. The pressure gradient driving the pulsatile flow is decomposed in Fourier series, \frac{\partial p}{\partial x} (t) = \sum^N_{n=0}P'_n e^{in\omega t} , where i is the imaginary number, \omega is the angular frequency of the first harmonic (i.e., n = 1 ), and P'_n are the amplitudes of each harmonic n . P'_0 (standing for n = 0 ) is the steady-state pressure gradient, whose sign is opposed to the steady-state velocity (i.e., a negative pressure gradient yields positive flow). Similarly, the velocity profile is also decomposed in Fourier series in phase with the pressure gradient, because the fluid is incompressible, u(r,t) = \sum^N_{n=0}U_n e^{in\omega t} , where U_n are the amplitudes of each harmonic of the periodic function, and the steady component ( n = 0 ) is simply Poiseuille flow U_0 = - \frac{P'_0}{4 \mu} \left(R^2 - r^2 \right) . Thus, the Navier-Stokes equation for each harmonic reads as i\rho n\omega U_n = -P'_n +\mu \left(\frac{\partial^2 U_n}{\partial r^2} + \frac{1}{r} \frac{\partial U_n}{\partial r}\right) . With the boundary conditions satisfied, the general solution of this ordinary differential equation for the oscillatory part ( n \geq 1 ) is U_n(r) = A_n \, J_0 \left( \alpha \, \frac{r}{R} n^{1/2}\,i^{3/2} \right) + B_n \, Y_0 \left( \alpha \, \frac{r}{R} n^{1/2}\,i^{3/2} \right) + \frac{i\,P'_n}{\rho\, n\, \omega}\, , where J_0(\cdot) is the Bessel function of first kind and order zero, Y_0(\cdot) is the Bessel function of second kind and order zero, A_n and B_n are arbitrary constants, and \alpha = R \surd( \omega \rho / \mu ) is the dimensionless Womersley number. The axisymmetric boundary condition ( \partial U_n/ \partial r|_{r=0} = 0 ) is applied to show that B_n = 0 for the derivative of above equation to be valid, as the derivatives J_0' and Y_0' approach infinity. Next, the wall non-slip boundary condition ( U_n(R) = 0 ) yields A_n = - \frac{i \, P'_n}{\rho \, n \, \omega} \frac{1}{ J_0 \left( \alpha \, n^{1/2} \, i^{3/2} \right)}. Hence, the amplitudes of the velocity profile of the harmonic n become U_n(r) = \frac{i\,P'_n}{\rho \, n \, \omega} \left[ 1 - \frac{J_0(\alpha \, n^{1/2} \, i^{3/2} \, \frac{r}{R})}{J_0(\alpha \, n^{1/2} \, i^{3/2})} \right] = \frac{i\,P'_n}{\rho \, n \, \omega} \left[ 1 - \frac{J_0(\Lambda_n \, \frac{r}{R})}{J_0(\Lambda_n)} \right] , where \Lambda_n = \alpha \, n^{1/2} \, i^{3/2} is used for simplification. The velocity profile itself is obtained by taking the real part of the complex function resulted from the summation of all harmonics of the pulse, u(r, t) = \frac{P'_0}{4 \mu} \left(R^2 - r^2 \right) + \mathrm{Re} \left\{ \sum^N_{n=1} \frac{i \, P'_n}{\rho \, n \, \omega} \left[ 1 - \frac{J_0(\Lambda_n \, \frac{r}{R})}{J_0(\Lambda_n)} \right] e^{i n \omega t} \right\} . \frac{d}{dx} \left[ x^p J_p(a\,x) \right] = a\,x^p J_{p-1} (a\, x) \quad \Rightarrow \quad \frac{d}{dx} \left[ x\, J_1(a\,x) \right] = a\,x J_{0} (a\, x) \, , then Q(t) = \iint u(r, t) \, dA = \mathrm{Re} \left\{ \pi\,R^2 \, \sum^N_{n=1} \frac{i \, P'_n}{\rho \, n \, \omega} \left[ 1 - \frac{2}{\Lambda_n}\frac{J_1(\Lambda_n)}{J_0(\Lambda_n)} \right] e^{i n \omega t} \right\} .
Velocity profile
To compare the shape of the velocity profile, it can be assumed that u(r,t) = f(r)\,\frac{Q(t)}{A} \, , where f(r) = \frac{u(r,t)}{\frac{Q(t)}{A}} = \mathrm{Re} \left\{ \sum^N_{n=1} \left[ \frac{\Lambda_n \, J_0(\Lambda_n) - \Lambda_n \, J_0(\Lambda_n \, \frac{r}{R})}{\Lambda_n \, J_0(\Lambda_n) - 2\,J_1(\Lambda_n)} \right] \right\} is the shape function. It is important to notice that this formulation ignores the inertial effects. The velocity profile approximates a parabolic or plug profile, for low or high Womersley numbers, respectively. \tau_w = \mu \left. \frac{\partial u}{\partial r} \right|_{r=R} \, . The derivative of a Bessel function is \frac{\partial}{\partial x}\left[ x^{-p} J_{-p} (a\,x) \right] = a\,x^{-p} J_{p+1}(a\,x) \quad \Rightarrow \quad \frac{\partial}{\partial x}\left[ J_0 (a\,x) \right] = -a\,J_1(a\,x) \, . Hence, \tau_w = \mathrm{Re} \left\{ \sum^N_{n=1} P'_n \frac{R}{\Lambda_n} \frac{J_1(\Lambda_n)}{J_0(\Lambda_n)} e^{i n \omega t} \right\} \, . \tilde{u}(t) = \mathrm{Re}(u(0, t)) \equiv \sum^N_{n=1} \tilde{U}_n \, \cos(n \, \omega \, t) \, . Noting that J_0(0) = 1 , the full physical expression becomes u(0, t) = \mathrm{Re} \left\{ \sum^N_{n=1} \frac{i \, P'_n}{\rho \, n \, \omega} \left[ \frac{J_0(\Lambda_n) - 1}{J_0(\Lambda_n)} \right] e^{i n \omega t} \right\} at the centre line. The measured velocity is compared with the full expression by applying some properties of complex numbers. For any product of complex numbers C = AB , the amplitude and phase have the relations |C| = |A||B| and \phi_C = \phi_A + \phi_B , respectively. Hence, \tilde{U}_n = \left| \frac{i \, P'_n}{\rho \, n \, \omega} \left[ \frac{J_0(\Lambda_n) - 1}{J_0(\Lambda_n)} \right] \right| \quad \Rightarrow \quad P'_n = \tilde{U}_n \left| i \, \rho \, n \, \omega \left[ \frac{J_0(\Lambda_n)}{1 - J_0(\Lambda_n)} \right] \right| and \tilde{\phi} = 0 = \phi_{P'_n} + \phi_{U_n} \quad \Rightarrow \quad \phi_{P'_n} = \operatorname{phase} \left( \frac{i}{\rho \, n \, \omega} \left[ \frac{1 - J_0(\Lambda_n)}{J_0(\Lambda_n)} \right] \right) \, , which finally yield \frac{1}{\rho} \frac{\partial p}{\partial x} = \sum^N_{n=1} \tilde{U}_n \left| i \, \rho \, n \, \omega \left[ \frac{J_0(\Lambda_n)}{1 - J_0(\Lambda_n)} \right] \right| \, \cos \left\{ n\,\omega\,t + \operatorname{phase} \left( \frac{i}{\rho \, n \, \omega} \left[ \frac{1 - J_0(\Lambda_n)}{J_0(\Lambda_n)} \right] \right) \right\} \, .
Text derived from https://en.wikipedia.org/w/index.php?oldid=1353886635 under CC BY-SA 4.0; modifications have been made.