Measuring the deflection of a beam

1. Jan 27, 2017

saybrook1

Hi guys, I have a set of 2-D data in excel, an X and Y column. The X column ranges from ±200 and the Y column is a Gaussian distribution ranging from about 6 to 12. In total there are about 500 data points. I need to normalize the data so that the Y column values have endpoints at 0 for X=±200. I know this requires a rotation of the data or a linear transformation but I'm really pretty foreign to that process. If anyone could point me in the direction of being able to do this either on paper or within excel or other software, I would really appreciate it. Please let me know if you need any more information or pictures. Thanks!

2. Jan 27, 2017

Staff: Mentor

What does that mean?
How would that look graphically?

3. Jan 27, 2017

saybrook1

What I mean is that Y currently has values at approximately 6.7 and 6.8 at X=±200. Since I need the values of Y to be zero at X=±200, I believe that I need to transform/rotate the entire set of data in order to keep the shape/integrity. Did this clear it up at all? I can try to explain further or attach a graph of what I mean if that helps. Thanks for the quick reply.

4. Jan 27, 2017

Staff: Mentor

Well, set all Y values to zero, and you achieved your goal. I'm sure that is not what you want. Which brings me back to the question: What do you want to do? What is the purpose of the transformation you are looking for? What should it preserve, what can it change, and how exactly do the Y values have to be zero "at X=+-200" (given that you have a discrete dataset)?

5. Jan 27, 2017

saybrook1

Okay, right, so the Y values(slope/displacement) need to be zero at the endpoints of +- a given length(say L) because I am trying to reproduce the curve based on an equation that I derived from a differential equation of a bent beam (Bernouilli-Euler eqn). In deriving the equation I set the boundaries to be zero at +-L/2. So in order to match the curve from the data to the curve that I'm producing from my closed-form equations, I need the boundary conditions of the data to be the same.

The data currently looks like the attached plot.

Attached Files:

• Capture.PNG
File size:
5 KB
Views:
118
6. Jan 27, 2017

saybrook1

Do you think that this thread might be more appropriate elsewhere?

7. Jan 27, 2017

Staff: Mentor

Ah, that's not what I expected. I expected a Gaussian distribution of your y-values.

If those values are measured displacement values, then I don't see how they would be compatible with zero slope. You must have some weird source of measurement error that you have to identify first before you can correct it.
Zero offset is easy - subtract 6.8 from all values.

Let's try the mechanical engineering forum. This is first a question about measuring the deflection.

8. Jan 27, 2017

saybrook1

I did some research and it looks like a 2D conformal transformation might work? I understand the subtraction will work to bring us down to zero but only for one point, they are not both at 6.8. Also, they won't be compatible with zero slope - it's the case that my equation contains variables for the moments C_1 and C_2 and I intend to make a small application that will find the best ratio of C_1 by C_2 to give the smallest slope error when combined with the data. There also shouldn't be a measurement error - it's from a really well put together FEA model - the asymmetry comes from an asymmetric heat load deposited on the beam. Let me know if you need more details.

Thanks a ton for your help; I'm really excited to make progress on this.

9. Jan 27, 2017

Staff: Mentor

There is an infinite set of mathematical transformations that will work, but most of them won't accurately reflect your beam deflection.

If your models are really good, they shouldn't have a non-zero slope where you have exactly zero slope.

10. Jan 27, 2017

saybrook1

Hmmm... are you saying that I can't rotate this data so that the endpoints respect my boundary conditions while maintaining the same shape? I'll attach a short derivation of the equation for deformation(y) with respect to beam/mirror length(x). I may not have explained it properly.

Attached Files:

• Eliptical fitting derivation2.pdf
File size:
80.8 KB
Views:
56
11. Jan 27, 2017

Staff: Mentor

Currently the shape is parabola-like. The two ends point in different directions. You cannot make them point in the same direction (well, 180° apart) without changing the shape.

12. Jan 27, 2017

saybrook1

Okay, right; I don't want them to point in the same direction, it's just that one end is slightly lower than the other so after a linear translation, I should be able to rotate them a bit so that the ends are at y=0 for x=±L/2. I guess the x-coordinates will transform as well in this case. I understand that once I add the my own analytical(and opposite) curve, I will not be able to get it completely flat.

13. Jan 27, 2017

Staff: Mentor

I still don't see the physical motivation behind it, and how you expect anything useful from transforming the parabola to a shape that has to look completely different.

14. Jan 27, 2017

saybrook1

I'm just leaving the place where I have that data and forgot to put it on the web although I think I figured it out. I just did a rotation and a linear transformation on both sets of coordinates which I got from defining a new origin where the endpoints equal zero. The shape appeared the same although I didn't have time to properly see how much different it is. I know people have done this sort of thing before in this type of application. If you'd like, I can explain it to you in more detail on Tuesday. I really appreciate your help/input today.

15. Jan 27, 2017

Stephen Tashi

What specifically is the equation?

How do the variables in the equation relate to the measured values of X and Y?

(Presumably the measured values of X and Y are not exactly the same quantities as the variables in the equation since you are looking for some way to modify the measured values so they fit the variables in the equation.)

Last edited: Jan 31, 2017
16. Jan 31, 2017

Tolklein

I may be waaay of the mark, but this looks like you are taking the data from a LIDAR and are trying to normalize it's values into usable distance data?