Numerical method to solve an 'almost' convolution integral

31 Views Asked by At

I'm trying to solve the following Volterra integral of the first kind for $w(x)$.

\begin{equation} P(x)=\int_{-\infty}^{\infty}[d_1(x-x_0)+d_2(x)]w(x_0)dx_0 \end{equation}

My attempt so far: I'm aware that I can break this integral up into a convolution integral and a simpler integral. If I define the constant $C = \int_{-\infty}^{\infty}w(x_0)dx_0$ then

\begin{equation} P(x)=\int_{-\infty}^{\infty}d_1(x-x_0)w(x_0)dx_0 + d_2(x)\int_{-\infty}^{\infty}w(x_0)dx_0 \\ P(x)=\int_{-\infty}^{\infty}d_1(x-x_0)w(x_0)dx_0 + Cd_2(x) \end{equation} If I Fourier transform both sides then: \begin{equation} F[P(x)-Cd_2(x)]=F[d_1(x)]\times F[w(x)] \\ w(x) = F^{-1}\frac{F[P(x)-Cd_2(x)]}{F[d_1(x)]} \end{equation}

But this is not an ideal format because $C$ is a depends on $w(x)$ so I guess I'd have to solve this iteratively somehow. Perhaps there are better ways? An approximate solution is fine.

1

There are 1 best solutions below

3
On BEST ANSWER

Just follow what you have, and assume the Fourier transformations are well-defined. Note your $C = \hat{w}(0)$.

$\hat{P}(\zeta) - \hat{w}(0) \hat{d_2}(\zeta) = \hat{d_1}(\zeta) \hat{w}(\zeta)$,

You can first solve $\hat{w}(0)$ by setting $\zeta = 0$, which means

$$\hat{w}(0) = \frac{\hat{P}(0)}{\hat{d_1}(0) + \hat{d_2}(0)}$$

Then you can solve the rest of $\hat{w}(\zeta) = \frac{1}{\hat{d_1}(\zeta)}(\hat{P}(\zeta) - \hat{w}(0) \hat{d_2}(\zeta))$.

Note: if $\hat{P}(0) = \hat{d_1}(0) + \hat{d_2}(0) = 0$, then you can set $\hat{w}(0)$ as an arbitrary number.