Two Phase regime determination

[Shaheryar]

All,
I have been trying to develop a program that would predict the flow regime by programming one of the regime charts like dukler charts. However cannot think of an algorithm that would do the job. Has anyone tried this before ? Or any ideas ?

 

[UmeshMathur]

You have to break up the plot computationally into the different regimes:
(1) Use a cubic spline function for each curve that delineates each boundaries.
(2) Establish the coordinates of the limiting points on each curve.
(3) Develop IF-THEN-ELSE logic to decide which flow regime you're in.

I did this coding almost 27 years ago and am happy to report that it worked fine. The coding is not complex, just tedious.

Have fun.

 

[siretb]

Define a 3 column array ; x , y and code the regime in the third. (use log scale).
Within each area select several points (say 9). Add 3 points for each "corner" or the domains
Then from the actual x & y you wish to know what egime tou are in, select the closest point in you reference matrix.

 

[panduru]

I feel the original post was about what may be the governing function(s) - the algorithm - (for each of the sub-regimes, if need be).

 

[Shaheryar]

UmeshMathur and siretb,
I have been trying to read x and y coordinates off the regime plot and what i do is , just read the full envelope for a certain regime.

The problem is that for example if you look at the graph, stratified and wave flow have one common boundary (right side of stratified and left side of wave flow). So even if I have the x and y coordinates, one line is common so programmatically I do not know how to handle, it probably means like a thousand if, then, else statements.

Is there a better way ?

and by the way how does the cubic spline function come into the picture ?

Thanks.

 

[siretb]

My approach, I did it and I know it works, is certainly not as clean as the one proposed by UmeshMatur.
What I propose is to select in each domain, and not at the boundaries, several points inside the domain and define the x,y and code. It is necessary to have several point near the boundary. YOu should not locate points on the boundary.
then when the algorithm runs, for az given point, it will look for the closest referenced point and assign the corresponding domain. Again, all points are inside the domains, most of them close to a boundary
As an example (only), for the "Wave" domain, I would select
2 points on the left x=0.1 y=2 and 20
3 points near the bottom boundary x=0.5 y=2, x=2 y=1.9, x=9 y=1.
3 points near the upper boundary
1 point in the center.
I think you get the idea. Since the "tip" of the Wave domain is bent, you may need more points inside the tip.
I can e-mail a sample Excel spreadsheet.

Again, to program the splines is better, and not that difficult. Hope this helps.