# Calculate displacement from acceleration signal data using Fourier transformation

I recorded acceleration data from an accelerometer attached to a vehicle, and I want to calculate displacement using the Fourier Transformation Integration method.

I used software called vibration Data ToolBox. This software has a built-in Fourier transformation which lets me save transformed acceleration data into a complex number (real + imaginary).

According to the rules of the Fourier transform, I can perform integration on the time-domain data by dividing the transformed acceleration data by the scale factor (-2πf) where f is the frequency in Hz.

However, the transformed data is a complex number. Should I divide the real part or the imaginary part or the magnitude by the scale factor?

Also, the software can perform Inverse Fourier transformation just for complex transformed data to obtain displacement in the time domain. how to calculate displacement?

I attached transformed data as Excel sheet.Excel sheet

• "However, the transformed data is a complex number. Should I divide the real part or the imaginary part or the magnitude by the scale factor?" Both. As a number you can't separate them. In computers they are stored in two different parts. $x+iy=re^{i\theta}$. Same number, two different representations. The DFT holds the left side form, but is usually interpreted with the right side form. Aug 18, 2020 at 20:12
• I divided −4π2f2 to the magnitude of (real and imaginary part) then I used this formula z=|z|(cos(ϕ)+isin(ϕ)) where z is the magnitude and ϕ is phase angle then I performed inverse fourier transformation then good result appeared Aug 19, 2020 at 11:35

I can perform integration on the time-domain data by dividing the transformed acceleration data by the scale factor (-2πf) where f is the frequency in Hz.

Well, not quite, but it looks like you have pieces of it. To integrate once, you divide by $$2i \pi f$$ -- note the fact that the divisor is purely imaginary. To integrate twice you just divide by that factor twice. You can do that explicitly, or you can note that $$\left (2i \pi f\right)^2 = -4 \pi^2 f^2$$

However, the transformed data is a complex number. Should I divide the real part or the imaginary part or the magnitude by the scale factor?

The key phrase here is "complex number". Almost always with complex numbers, you do arithmetic on the whole thing, real and complex parts equally. If you're just plucking out the real or the imaginary parts out to do math on, you're almost certainly taking some shortcut, and you should understand the "right" way to do that task.

In this case, you're just doing plain old arithmetic, so you should divide the entire number (meaning: the real and the imaginary parts) by $$-4 \pi^2 f^2$$.

Also, the software can perform Inverse Fourier transformation just for complex transformed data to obtain displacement in the time domain. how to calculate displacement?

Theoretically, divide throughout by $$-4 \pi^2 f^2$$ and do an inverse Fourier transform. In reality, you'll have some non-zero signal content at low frequencies, and dividing that by 0 will cause problems.

So you need to multiply it by some function $$A\left(f\right)$$ that is $$A\left(f\right) = 0$$ for $$f = 0$$, $$A\left(f\right) = -\frac{1}{4 \pi^2 f^2}$$ for $$f$$ above some threshold frequency (which will be determined by your accelerometer, BTW), and that transitions smoothly in between.

At worst, just make a function $$A(f) = \begin{cases} 0 && |f| < f_0 \\ -\frac{1}{4 \pi^2 f^2} && |f| \ge f_0 \end{cases}$$

This will hopefully work well enough to tell you that you're doing your work right, but will suffer from time-domain ringing around $$f_0$$. Then you can either try out ways to make a smooth transition, or ask back here and we'll suggest something.

• I divided −4π2f2 to the magnitude of (real and imaginary part) then I used this formula z=|z|(cos(ϕ)+isin(ϕ)) where z is the magnitude and ϕ is phase angle then I performed inverse fourier transformation then good result appeared ... is it the correct way? Aug 19, 2020 at 11:32
• I think so, it's hard to tell from the amount of detail you gave. You can always check by taking the signal from the accelerometer and double integrating it by hand, or start with a known displacement, calculate the acceleration, and run that through your tool. Aug 19, 2020 at 14:55