Plotting of graphs and applying the data into an equation

**mickey_gsn** · 03-19-2018, 09:32 AM

Hi.

I am a civil engineering student. Currently I am doing data interpretation of some soil testing graphs.
The graphs are plotted as depth against the parameter (soil resistance pressure, in kPa):
i.e. X-axis is the parameter, Y-axis is depth (downward, negative).

In a cone penetration test (a soil testing method), two pressure parameters are obtained.
To explain the problem in a simpler manner, let's say the two parameters are A and B.
So in each test, there are a A-depth graph and B-depth graph.

As I do not do laboratory test, I get the data from existing graphs plotted by different researchers for different research purposes.
What I do is to digitise the graphs into excel-format data.
I use the Web Plot Digitiser.

The problem I am facing is:

I need to use the data interpreted of A and B in an equation:
C = A^n * B^m
(just a format)

As the plot digitisation is random, the data in A and B do not have the same depths (Y-coordinates).
So I cannot simply input the data of A and B against depth.

How I try to correct the problem is to manually interpolate the depth values for the corresponding A and B values, so as to obtain a common depth value for each A and B value.
But it takes so much time. I have dozens of these graphs and many equations to be input.

Is there any quicker method? Such as directly use the graph to get the common values of A and B?
Thank you.

Attachment 566329
Attachment 566330

**MrShorty** · 03-19-2018, 10:57 AM

Your attachments did not attach correctly.

I am not aware of any way to directly read A and B or m and n from the pictures of the charts.

I don't know if this is any quicker than the interpolation approach you are currently using, but the approach I might use would be:

1) Get A, B, and C data from the pictures in a nice list in Excel. ie

Please Login or Register  to view this content.

2) A little algebra C=A^n*B^m take ln() of both sides
ln(C)=n*ln(A)+m*ln(B) -- a relatively simple multivariate linear regression equation. (A little logarithm refresher if you need to remember rules for manipulating logarithms: http://www.purplemath.com/modules/logrules.htm ).
2a) I like to use helper columns to perform the transformation, so add ln(A), ln(B), ln(C) columns.
3) Use LINEST() to obtain the m and n parameters (https://support.office.com/en-us/art...a-fa7abf772b6d ).

At this point, I have a good, best fit equation for the data and can proceed with whatever I need to do.