Comparing the results from AutoPIPE against other programs, why are they the same or different?
Answer:
Comparing results between two applications is always tricky and are bound to be different apart from very simple cases. Although the basic theory and assumptions behind the two applications may be the same, but as the products evolve the different components may start to diverge slightly based on selected options. These divergences may compound when the models depart from being very simplistic models.
AutoPIPE's calculations are subjected to the most rigorous testing standards in the industry - ensuring the most accurate output results are produced. AutoPIPE is currently certified against ASME NQA-1; US Code of Federal Regulations, Title 10, Part 50, Appendix B; 10CFR21/10CFR50; ISO 9001; and ANSI/ASME N45.2. AutoPIPE has been and continues to be, successfully tested and audited by most major companies in the piping arena, consistently outperforming the accuracy rate of our competitors.
Genereal Reasons
Reason #1: Analysis Method
One of the major contributors in answering this question is directly related to how analysis are performed in AutoPIPE compared to how they are performed in other programs. There are only 3 general methods used by most stress analysis computer programs to solve non-linear analysis (see the WIKI page here for details about these 3 general methods). Therefore it goes without saying if the programs are using different analysis methods than the results may be close but may never be exactly the same.
Apart from the basic analysis method employed by a program, there could be other assumptions made by an application which could be different to other applications. For example, for any component which is set as rigid, AutoPIPE would multiply the modulus values of that component by 1000.0. Other applications may have a slightly different approach for definition of rigid components. This may impact the analysis results slightly. Other examples could be how an application defines curves elements, reducers, or flexible joint.
Reason #2: Model input is identical
Are the models represent in both programs exactly the same (i.e. Pipe properties, support bearing directions, soil spring data, analysis settings, load cases i.e. Wind ,Seismic, dynamic loads, etc..). Any deviation between the input listings would significantly alter the results comparisons. Again, the entire input listings must match exactly.
Reason #3: Code Calculations are performed differently
Are the code calculations performed exactly the same between both application? When implementing a code, application engineers need to make decisions about the intent of a code calculation. There maybe footnotes that are overlooked, difference of interpretation for a variable / statement, etc... that could lead different results.
Detailed Reasons
Comparing AutoPIPE and caesar:
1. A major difference between these 2 programs that leads to different results is the analysis approach used by each application, see General Reason #1 above for details. AutoPIPE uses Operation Condition approach, while caesar uses Algebraic Subtraction approach (which is a questionable approach to accurate stress analysis).
2. Some bearing support directions are not exactly the same. Confirm bearing direction supports.
3. Analysis Settings / Configuration options in both applications must match.
4. Rigid support and anchor stiffness match.
5. Element weight and stiffness (i.e. valve, reducer, tee, bend, etc..) match exactly
7. Pressure thrust load vector formulation
8. Conversion of units from English to Metric / SI
9. Manual or Calculated SIF values on all piping components
10. Nozzle flexibility
11. The CAE development team has proven with many example models that through careful attention to adjusting settings within AutoPIPE, caesar result can match AutoPIPE's results within a very small percent error. Example:
if you run very simple cantilever models in both the applications with a tip force applied, you will get exactly the same results:
|
Length of pipe (ft) |
Free end force (lbf) |
AutoPIPE Displacement (in) |
caesar displacements (in) |
. |
|
0.5 |
10,000 |
0.0008 |
0.0008 | |
|
1.0 |
10,000 |
0.0021 |
0.0021 | |
|
2.0 |
10,000 |
0.0085 |
0.0085 | |
|
5.0 |
10,000 |
0.0946 |
0.0946 | |
|
10.0 |
10,000 |
0.7135 |
0.7135 |
This could tell us that both applications are using the same beam theory with shear deformation included. When systems depart from simplistic to more complex piping arrangements, then some assumptions / implementations like those mentioned above may come in to play:
Comparing AutoPIPE and Abaqus:
Some information is available in AutoPIPE reference information in the AutoPIPE help. SAP IV acted as the basis for AutoPIPE analysis engine. There had been modifications and improvements to the analysis engine over the years. For a list of major improvements, please refer to the Revision History page in AutoPIPE help. The general implementation for the stiffness matrix for the Timoshenko beam theory is available in text book with assumptions mentioned below.
i. Plane section remain plane but not necessarily perpendicular to the centerline of the beam
ii. Effects of shear deformation are included
iii. Only valid for small deformations (the cross section or properties of the beam are not affected by the deformations)
iv. Only valid for stresses under elastic range (no plastic deformation)
v. Modulus is considered the same in tension and compression
Summary:
As shown above, there could be any number of reasons why 2 applications performing a pipe stress analysis on the same piping arrangement may or may not have matching results. The largest contributor to these differences are typically model input, analysis settings, and method of analysis. On individual level, there may be slight differences in these areas. However, when the whole system is analyzed, the differences may get compounded due to interaction of these individual differences.
One can trust any application if one understands the assumptions involved in the solution and understands that the results are a fair approximation of the actual system. Some applications may have a better representation of a particular component / analysis, while the other may be good in another area.
Recommend not to blindly use a given application just because but to gain more confidence in any application results by starting with simple models, perform a few tests on the component / analysis, and confirm that the results simulate the actual system reasonably well. AutoPIPE QA program backs up the results generated by the program with a rigorous set of hand calculations and published results.