Difference between revisions of "Wave Scattering by Submerged Thin Bodies"

From WikiWaves
Jump to navigationJump to search
Line 126: Line 126:
 
-\frac{1}{2}\sum_{m=0}^{M} b_m (m+1)U_m(x) + \int_{\partial \Omega} \partial_{n^{_{'}}n} G  \phi ds^{_{'}}- \int_{\partial \Omega}\partial_n G \partial_{n^{_{'}}} \phi ds^{_{'}} = 0
 
-\frac{1}{2}\sum_{m=0}^{M} b_m (m+1)U_m(x) + \int_{\partial \Omega} \partial_{n^{_{'}}n} G  \phi ds^{_{'}}- \int_{\partial \Omega}\partial_n G \partial_{n^{_{'}}} \phi ds^{_{'}} = 0
 
</math></center>
 
</math></center>
 +
  
 
Multiplying this entire expression by <math>(1-x^2)^{1/2} U_n(x) \,</math> and integrating allows us to obtain the matrix representation
 
Multiplying this entire expression by <math>(1-x^2)^{1/2} U_n(x) \,</math> and integrating allows us to obtain the matrix representation
Line 135: Line 136:
  
 
where
 
where
 +
 +
<center><math>
 +
\begin{align}
 +
\tilde{M}_{nn} &= -\frac{\pi}{4}n \\
 +
\mathbb{Y} &= \mathbb{U} \ \mathbb{W}^{(1)}\mathbb{W}^{(2)} \\
 +
\mathbb{U} &= [U_1(x),U_2(x), ... , U_M(x)]\\
 +
\mathbb{W}^{(1)}_{nn} &= (1-x_n^2)^{1/2}\\
 +
\mathbb{W}^{(2)} &= \mbox{weights matrix for numerical integration}
 +
\end{align}
 +
</math></center>
  
 
The two <math>\mathbb{G}</math> matrices are obtained using boundary element method, which is discussed below.
 
The two <math>\mathbb{G}</math> matrices are obtained using boundary element method, which is discussed below.

Revision as of 23:02, 28 September 2009


Introduction

We consider the problem of a thin submerged structure assumed to be rigid, submerged in a semi-infinite domain of constant depth. This submerged structure is then subject to some incident wave forcing, and the resulting scattering problem is investigated.


Equations

The governing equations for our problem are as follows:

[math]\displaystyle{ \begin{align} & \Delta \phi=0, &\qquad in \ \Omega\, \\ & \partial_z \phi=0, &\qquad at\ z=-h\, \\ & \partial_z \phi=\alpha \phi, &\qquad at\ z=0\, \\ \end{align} }[/math]

Plus radiation conditions:

[math]\displaystyle{ \partial_x \phi \pm i k \phi = 0, \qquad as \ x \rightarrow \pm \infty \, }[/math]


For the purposes of using Boundary Element Methods later in this article, we restrict our domain to a finite region by imposing artificial boundary conditions at [math]\displaystyle{ x= a_1 \, }[/math] and [math]\displaystyle{ x= a_2 \, }[/math]. We denote these conditions by

[math]\displaystyle{ \begin{align} & \phi=\tilde{\phi}_1(z), &\qquad at \ x=a_1\, \\ & \phi=\tilde{\phi}_2(z), &\qquad at \ x=a_2\, \\ \end{align} }[/math]

and will be considered in more detail later on.

Thin bodies

Additionally, we have a condition for the response of our structure (which is submerged at constant finite depth [math]\displaystyle{ -d }[/math]):


[math]\displaystyle{ \begin{align} & \partial_n \phi = f(x,t), &\qquad x \in (-L,L), \ z=-d \, \end{align} }[/math]

We denote this segment of the boundary by [math]\displaystyle{ \Gamma \, }[/math].


There are additional considerations to be made when dealing with thin obstacles - we need to split [math]\displaystyle{ \Gamma \, }[/math] into two regions ([math]\displaystyle{ \Gamma^{\pm} \, }[/math]) and distinguish between their respective normal derivatives ([math]\displaystyle{ \partial_n \rightarrow \partial_{n^{\pm}}\, }[/math]). This allows us to express the boundary condition in the form

[math]\displaystyle{ \begin{align} & \partial_{n^{\pm}} \phi = \pm f(x,t), &\qquad x \in (-L,L), \ z=-d \, \end{align} }[/math]


Observe that the normal derivatives are related in the following manner

[math]\displaystyle{ \begin{align} & \partial_{n^{-}} = - \partial_{n^{+}} & \, \end{align} }[/math]


From this, we can express our problem involving the submerged structure entirely in terms of [math]\displaystyle{ \Gamma^{+} \, }[/math], and omit the notation in the future (ie [math]\displaystyle{ \partial_{n^{+}} = \partial_n }[/math]).


For the case of a rigid structure, we define the associated boundary condition to be


[math]\displaystyle{ \begin{align} & \partial_n \phi = 0, &\qquad on \ \Gamma \, \end{align} }[/math]

Integral Equation formulation

Using Green's second identity, in tandem with the discussion of thin bodies (above), we can obtain an expression for [math]\displaystyle{ \phi \, }[/math] around the outer boundary ([math]\displaystyle{ \partial \Omega \, }[/math]) of our domain:


[math]\displaystyle{ \frac{1}{2}\phi(\textbf{x})=\int_{\Gamma} [ \ \phi \ ] \partial_{n^{_{'}}} G ds^{_{'}} +\int_{\partial \Omega} (\phi \partial_{n^{_{'}}} G - G \partial_{n^{_{'}}} \phi )ds^{_{'}} }[/math]


where [math]\displaystyle{ [ \ \phi \ ]=\phi(x^{+})-\phi(x^{-}) }[/math], and is typically referred to as the 'jump in [math]\displaystyle{ \phi \, }[/math]'.

Note carefully that [math]\displaystyle{ \phi \, }[/math] relates to points on the outer boundary, [math]\displaystyle{ \partial \Omega \, }[/math], whereas [math]\displaystyle{ [ \ \phi \ ] \, }[/math] relates to points on the inner boundary, [math]\displaystyle{ \Gamma \, }[/math].


In order to compute [math]\displaystyle{ [ \ \phi \ ] \, }[/math], we will restrict all x points in the expression above to be on [math]\displaystyle{ \Gamma \, }[/math], and then take the normal derivative of the expression to obtain


[math]\displaystyle{ \oint_{\Gamma} [ \ \phi \ ] \partial_{n^{_{'}} n} G ds^{_{'}} +\int_{\partial \Omega} (\phi \partial_{n^{_{'}}n} G - \partial_n G \partial_{n^{_{'}}} \phi )ds^{_{'}} =0 }[/math]

Note that in our case, [math]\displaystyle{ \oint \, }[/math] refers to a Hadamard finite-part integral, as opposed to a contour integral.

Hadamard Finite Part Integrals

A hadamard finite part integral is an example of a hypersingular integral equation (an integral equation whose kernel has high order singularities), and must be treated with caution.

In order to evaluate our integral above, it is necessary to express the jump in [math]\displaystyle{ \phi \, }[/math] as

[math]\displaystyle{ [ \ \phi(x^{_{'}}) \ ] = (1-x^{_{'}2})^{1/2} \sum_{m=0}^{M}b_m U_m(x^{_{'}}) }[/math]

where [math]\displaystyle{ U_m(x^{_{'}}) \, }[/math] denotes Chebyshev Polynomials of the Second Kind.

It is also necessary to employ the powerful result discussed in Martin and Rizzo 1989:

[math]\displaystyle{ \oint_{-1}^{1} \frac{(1-v^2)^{1/2}U_m(v)}{(u-v)^2}dv = -\pi (m+1) U_m(u) }[/math]

Note that the structure has to be non-dimensionalised accordingly.


From this we can obtain

[math]\displaystyle{ -\frac{1}{2}\sum_{m=0}^{M} b_m (m+1)U_m(x) + \int_{\partial \Omega} \partial_{n^{_{'}}n} G \phi ds^{_{'}}- \int_{\partial \Omega}\partial_n G \partial_{n^{_{'}}} \phi ds^{_{'}} = 0 }[/math]


Multiplying this entire expression by [math]\displaystyle{ (1-x^2)^{1/2} U_n(x) \, }[/math] and integrating allows us to obtain the matrix representation


[math]\displaystyle{ \tilde{M} \vec{b} + \mathbb{Y} (\mathbb{G}^{(1)}\vec{\phi} - \mathbb{G}^{(2)} \vec{\phi}_{n^{_{'}}} ) = 0 }[/math]

where

[math]\displaystyle{ \begin{align} \tilde{M}_{nn} &= -\frac{\pi}{4}n \\ \mathbb{Y} &= \mathbb{U} \ \mathbb{W}^{(1)}\mathbb{W}^{(2)} \\ \mathbb{U} &= [U_1(x),U_2(x), ... , U_M(x)]\\ \mathbb{W}^{(1)}_{nn} &= (1-x_n^2)^{1/2}\\ \mathbb{W}^{(2)} &= \mbox{weights matrix for numerical integration} \end{align} }[/math]

The two [math]\displaystyle{ \mathbb{G} }[/math] matrices are obtained using boundary element method, which is discussed below.

Boundary Element Methods

refer to the wikiwaves page etc

Eigenfunction Matching

refer to the wikiwaves page etc

Matlab Code

References

Article by Yang? Linton McIvor?

Bibtex

The bibtex reference for this webpage is below:

@MISC{referenceID,

 author = {wikiwaves},
 title = { Wave Scattering by Submerged Thin Bodies },
 month = November,
 year = { 2024 },
 url = { https://wikiwaves.org/Wave_Scattering_by_Submerged_Thin_Bodies }

}