Solution of a boundary value problem

168 Views Asked by At

I have a rectangular region where $-1<x<1$ and $-1<y<1$. The boundary value problem for this region is given by $$ \kappa^2\frac{\partial^2u_1}{\partial x^2}+\delta^2\frac{\partial^2u_1}{\partial y^2}+\delta (\kappa^2-1)\frac{\partial^2u_2}{\partial x \partial y}=A,\tag{1} $$ $$ \delta^2\kappa^2\frac{\partial^2u_2}{\partial y^2}+\frac{\partial^2u_2}{\partial x^2}+\delta (\kappa^2-1)\frac{\partial^2u_1}{\partial x \partial y}=0,\tag{2} $$

$$ \left.\kappa^2\frac{\partial u_1}{\partial x}+\delta(\kappa^2-2)\frac{\partial u_2}{\partial y}\right|_{x=\pm 1}=B^{\pm },\quad \left.\delta\frac{\partial u_1}{\partial y}+(\kappa^2-2)\frac{\partial u_2}{\partial x}\right|_{y=\pm 1}=0,\quad u_2(x,\pm 1)=0. \tag{3} $$ where $u_1$ and $u_2$ are displacements and $\kappa$, $\delta$, $A$ and $B^{\pm}$ are known constants.

Eqs (1) and (2) can be transformed to a fourth-order PDE in respect to one of sought for displacements. For example, in terms of $u_2$, we obtain $$ u_{2,1111}+ a\, u_{2,1122} + b\, u_{2,2222}=0\tag{4} $$ where $a=2\delta^2$ and $b=\delta^4$. Therefore, it can be presented in operator form as $$ (\partial_1^2+\lambda_1^2\partial_2^2)(\partial_1^2+\lambda_2^2\partial_2^2)=0\tag{5} $$ $\lambda_1^2+\lambda_2^2=a$ and $\lambda_1^2\lambda_2^2=b$. Hence, the solution of (5) can be expressed $$ u_2=\sum_{i=1}^{2}\phi_i(\lambda_i x,y) $$ where $\phi_i$ are arbitrary plane-harmonic function. Similarly, the other displacement $u_1$ may be found again the sum of two plane-harmonic functions which are harmonic conjugate functions of $\phi_i$. Unfortunately, after applying the boundary conditions (3) I couldn't find any proper functions for the solutions. Does anyone have an idea about the solution to the problem or suggest any book I can use?

1

There are 1 best solutions below

3
On BEST ANSWER

(I kept the previous clarification now below, for the sake of the comments)


So motivated by the third set of boundary conditions one can attempt the following Ansatz:

$$u_2=\sum_{k=1}^{\infty}\sin(\tfrac{k\pi}{2}(x+1))\cdot r_k(y)$$ Since everything is linear we solve for one fixed $k$ first and then take the sum in the end. To this end we plug $u_2=\sin(\tfrac{k\pi}{2}(x+1))r_k(y)$ into the forth order equation and get $$ 0=\frac{\kappa^2}{16}\sin(\tfrac{k\pi}{2}(x+1))\left(k^4\pi^4r_k(y)-8\delta^2k^2\pi^2r_k''(y)+16\delta^4r_k^{(4)}(y)\right)\\ \Rightarrow r_k(y) = (C_2y+C_1)e^{\frac{k\pi}{2\delta}y} + (C_4y+C_3)e^{-\frac{k\pi}{2\delta}y} $$

Next, we use equation $(2)$ to calculate $u_1$. This is achieved by simply integrating once over $x$ and once over $y$.

$$ u_1(x,y)=\cos(\tfrac{k\pi}{2}(x+1))\left((C_4y+C_3+C_4\tfrac{2\delta(\kappa^2+1)}{k\pi(\kappa^2-1)})e^{\frac{k\pi}{2\delta}y}-(C_2y+C_1-C_2\tfrac{2\delta(\kappa^2+1)}{k\pi(\kappa^2-1)})e^{-\frac{k\pi}{2\delta}y}\right)+F_1(y)+F_2(x) $$

In the next step, we use the first equation and arrive at

$$ F_2''(x)\kappa^2+F_1''(y)\delta^2=A\\ \Rightarrow F_2(x)=F_{22}x^2+F_{21}x+F_{20}\\ \Rightarrow F_1(y)=\frac{A-2\kappa^2F_{22}}{2\delta^2}y^2+F_{11}y+F_{10} $$ Which altogether yields:

$$ u_1(x,y)=\cos(\tfrac{k\pi}{2}(x+1))\left((C_4y+C_3+C_4\tfrac{2\delta(\kappa^2+1)}{k\pi(\kappa^2-1)})e^{\frac{k\pi}{2\delta}y}-(C_2y+C_1-C_2\tfrac{2\delta(\kappa^2+1)}{k\pi(\kappa^2-1)})e^{-\frac{k\pi}{2\delta}y}\right)+\frac{A-2\kappa^2F_{22}}{2\delta^2}y^2+F_{11}y+F_{10}+F_{22}x^2+F_{21}x+F_{20}\\ u_2(x,y)=\sin(\tfrac{k\pi}{2}(x+1))\left((C_2y+C_1)e^{\frac{k\pi}{2\delta}y} + (C_4y+C_3)e^{-\frac{k\pi}{2\delta}y}\right) $$

At this point we would normally take the sum over $k$, and then try to fit the parameters (all dependend on $k$ of course) to solve the remaining boundary conditions.

At this point however (and that is exactly why i wanted to know the exact problem!) we run into two problems:

  1. The system is underdetermined in the sense of boundary conditions. This means we will have some degrees of freedom left even after fitting.
  2. The existing boundary conditions do not allow a solution for all cases. Namely, we have the condition $2A=B^+-B^-$. This is NOT a byproduct of the Ansatz, this can be proven rigorously, let me know if you want details (it is a bit lengthy)

However, for $2A=B^+-B^-$ due to the first point we can easily deduct many solutions which do not even require the sum for example:

\begin{align} u_1(x,y)&=\frac{Ax^2+2(A+B^-)x}{2\kappa^2}\\ u_2(x,y)&=0 \end{align}


This is not an answer but its too long for a comment:

I don't see how you obatin the stated coefficients of the forth order equation:

$$ \kappa^2\frac{\partial^2}{\partial x^2} (2)+\delta^2\frac{\partial^2}{\partial y^2} (2) -\delta(\kappa^2-1)\frac{\partial^2}{\partial x\partial y} (1)\\ \Leftrightarrow \delta^2\left(\kappa^2u_{2,xxxx}+(\delta^2+2\kappa^2-1)u_{2,xxyy}+\delta^2\kappa^2u_{2,yyyy}\right)\\ \Leftrightarrow u_{2,xxxx}+\underbrace{\frac{\delta^2+2\kappa^2-1}{\kappa^2}}_{=a}u_{2,xxyy}+\underbrace{\delta^2}_{=b}u_{2,yyyy} $$