EQUATIONS FOR API-650 CURVES LIKE P-2A/B/C/D etc

[UbaidUllah65]

Friends,

I am a piping engineer and am working on a software for evaluating Nozzles connected to API-650 based tanks.
Anybody worked on developing the equations for various curves?
It will be useful if some body could share.

BR,

 

[gr2vessels]

Yes, a number of companies have developed the equations you are asking (software like E-Tank, Tank by Coade/Integraph, Tank by Chempute). I'm sure if you look up the internet, someone else might have developed their own software and wiling to share.
Cheers,
gr2vessels

 

[Aodwan]

Thanks gr2 as i am also looking for the same, But E tank do not have this facility as we are the user. we have to perform manually.

 

[brubakerj]

UbaidUllah65,
Where are the source curve located? Is it a solution to an equation or are the curves available?

-Jeff

 

[UbaidUllah65]

Jeff,

The CURVES are in the Appendix-P of API-650.

UbaidUllah

 

[brubakerj]

I'm looking at these curves now. It doesn't look like there is enough data. There is just two L/2a curves?

I think the way to do this is the following

do a linear curve fit for every curve.

Calculate all the curves at the given R/t Then interpolate using L/2a and a/r to get an appropriate value.

 

[UbaidUllah65]

Jeff,

You are right.
I have already done it for my software.

 

[brubakerj]

UbaidUllah65,
I'm working through the procedures myself at the moment. I curvefit every given curve and evaluated at a given L/2a and r/t. I am interpolating on a_r and debating which interpolation method to use. Is it better to use linear or to use quadratic interpolation?

-Jeff

 

[UbaidUllah65]

I used Linear.

I spent more than two months in using different approaches later had to do the curve fitting for curves between the ones given in the figures.

Best of luck.

 

[brubakerj]

The linear doesn't seem to match the example problem. (2.36E-04
versus 3.0e-4). Looking at the graph I wonder if a quadratic scheme would work better. I wish they would give the derivation for these curves, that may give some insight into the best interpretation scheme. These codes so often seem to lack an evident background. Does anyone know the author is? Is anything published?

It doesn't look there is enough data points (just two L/2a) for the other interpolation. What do you do if it's above or below?

 

[UbaidUllah65]

I faced the same problem.

Some of the values (stiffness coefficients) given in the example are wrong. You can check yourself by careful manual plotting.

There are not just two L/2a but also only three a/R.

I have done curve fitting for intermediate a/Rs.

I used linear interpolations with both L/2a & a/R, gives me good results.

Send me the inputs and I will send you the stiffness coefficients calculated by my software. But please send them in the units followed in the example.

 

[brubakerj]

UbaidUllah65,
Are the numbers matching?

-Jeff

 

[jeffbru]

UbaidUllah65,
I got blocked, I'm not sure why. I'm not sure if they think I'm spamming. I'm just trying to get some of these calcs online as a portfolio and maybe attract some attention from the authors to get a live version of some of the appendixes online. It's a bit of a hobby anyway and I use it at interviews.

Anyway, I'm not going to post the link again!

R/t=1173
a/R=0.008
L/2a=1
Radial-Shell 0.00044 (unitless)
Radial-Neck 0.00023 (unitless) (3.1E-4 from API 650)
Longitudinal-Shell 0.000527 (unitless)
Longitudinal-Neck 0.000364 (unitless) (3.0e-4 from API 650)
Circumferential-Shell 0.0012383 (unitless)
Circumferential-Neck 0.001172 (unitless) (5.0e-4 5 from API 650)

 

[jeffbru]

Hey in Regards to this Appendix P,
See the following thread

http://65.57.255.42/ubbthreads/ubbthreads.php?ubb=showflat&Number=32030#Post32030

Richard Ay at Codade wrote
"I didn't find it easy, the examples contradicted the body of the Code and there were curves mis-labeled. I couldn't get through it (computation wise) which is why I never put it in our software. Will something like this come back - I don't know. This was a massaged WRC107 method and I'm not a big fan of that bulletin either. Whatever they put in the Code has to be complete, correct, and computationally possible."

It doesn't look it's very popular. At another point he says it's based upon two FEA studies.

-Jeff

 

[chicopee]

Do a polynomial regression analysis. about twenty years ago, I had enough time to develop a regression analysis program in basic using DOS. My program can do polynomial, reciprocal, powers, logs and exponentials which to this date I still use.I think MS Excel can do that for you.

 

[UbaidUllah65]

Jeffbru:
Sorry for a very late reply. I got busy in a very tight scheduled project which is not over yet.
I can understand your feelings as I have been through it...

An override of my system inputs for your indicated values:
R/t=1173
a/R=0.008
L/2a=1

gives me the following values.:
KR=0.00017619091441700257
KL=0.00033535179923315476
KC=0.00079268642564544762

Please check them as I could not get time to authenticate due to lack of time.

Sorry once again...