| Applies To | |||
| Product(s): | AutoPIPE, | ||
| Version(s): | ALL | ||
| Area: | Reports | ||
| Original Author: | Bentley Technical Support Group |
Problem:
One of the following has occurred:
a. Have 2 separate models with identical input, but the results are different using the same version of AutoPIPE
or
b. Have 1 model with identical load-cases / analysis sets and the results are different
or
c. Results applicably change in one part of the model after adding a single node point in a different area of the model
or
d. Same model opened in 2 different versions of the application but has different results in each version
Why has this happened and how to fix?
TR Number: none
Product: AutoPIPE
Area: General
Problem ID#: 71085
Solution:
Check the model output reports to verify input is truly identical.
A. Comparing 2 AutoPIPE models
1. Start AutoPIPE with a valid license.
2. Open model DAT
3. Create a complete model input listing report, check on all sub sections to be printed, and press OK button
4. Analyze the system
5. Next, on the combinations dialog, Combination Options are identical between models, delete all code and non-code combinations. Press the "Reset Defaults Only" button to automatically recreate the default combinations to be compared (see Note #1. below).
6. Now create a complete output report; Check all sub reports ON, except "Sorted Stresses" should be disabled.
6. Perform steps 2-5 for both models.
7. Use a program like "Beyond compare" or "Ultra Compare" to compare the files. This will quickly highlight and differences between the models.
Note: take extra care to be sure that all Analysis summary settings are exactly the same (Tools> Model Option> General, Edit, Results, Buried Pipe Options, etc..). In addition for comparison purpose only, set Tools> Model Options> Results> Sustain Margin (Y/N/E) = N. This setting will affect the allowable calculation.
8. Expect to find different time stamps and maybe some info that may be out of sync. However, everything else should be identical.
9. Fix any differences between model input until they are identical and then run an analysis to compare results again.
10. If the models still produce different results, follow the procedure for logging a Service Request with Bentley, add model APC & supporting files to run the analysis in a ZIP file, and send to Tech support for review.
Video of procedure above:
B. One AutoPIPE Model with similar analysis sets, but different results
1. Start AutoPIPE with a valid license.
2. Open model DAT
3. Analyze the system, Analyze> Analyze all.
4. Open the model's input grid, select the Press/Temp/PipeID tab, copy the entire grid to Excel, Use Excel to compare each Temp/Press loadcase to confirm identical settings (could have different Press, Temp, Expansion, Hot Mod,Hot Allow, or Yield values).
5. After the analysis is completed, on the Load Combinations dialog, delete all code and non-code combinations. Press the "Reset Defaults Only" button to automatically recreate the default combinations to be compared (see Note #1 below). Now create a complete output report; Tools> Model Input listing> check on all sub reports. Then under, Results> Output report> Check all sub reports ON, except Sorted Stresses.
6. Review model listing to confirm no additional Xtra Data is assigned to a load-case (i.e. Force, distributed load, concentrated force, cut short, etc..)
7. If the models still produce different results, follow the procedure for logging a Service Request with Bentley, add model APC & supporting files to run the analysis in a ZIP file, and send to Tech support for review.
C. Results appreciably change in one part of the model after adding a single node point in a different area of the model
Most likely the cause is related to the 1 of 2 things.
1. Non-linear analysis
A Non-linear analysis is an iterative process that requires the application to guess at a solution, test that solution against a tolerance value, if the value is not within the tolerance iterate the guess, repeat the calculations again and again until the value is within the tolerance, and then move onto the next load case in the analysis. These tolerance values are located the Nonlinear Analysis dialog (imaged below) that appears after creating a Non-linear analysis set from the Static Analysis dialog .
Keep in mind time required to run a Static Analysis is inversely proportional to the tolerance and calculation's accuracy. Another words, the lower the tolerances, the higher the calculation accuracy and the longer it takes AutoPIPE to complete a static analysis. Verses, the larger the tolerance, the lower calculation accuracy and shorter time it takes AutoPIPE to complete a static analysis. Thus the results can be different with just adding a node point(s) or with different tolerance settings.
2. Dynamic Analysis
When performing a dynamic analysis (response spectrum, force spectrum, harmonic, or time history analysis), first the application must perform a modal analysis. Modal analysis command enables the program to calculate the natural frequencies and mode shapes for all piping and framing modeled. Anything that causes the model analysis results to change will impact the dynamic analysis results. Please take a moment to read about "Modal Analysis Theory" and "Mass Discretization" in AutoPIPE help before continuing.
Case in point, adding a single node point in one part of the model, what will happen to results throughout the mode? For some node points results will be fairly similar, while for other node points, load case results may significantly change. From the reading above, recall that modal analysis is dependent on the mass matrix and discretization in the AutoPIPE model. Adding a node to the model will further discretize this mass matrix / mass equations, improving results.
Note, while "Mass points per span" setting equals 0, all mass is averaged across each pipe and framing span, and then lumped directly at the node points. A value greater than 0 will allow mass discretization for dynamic analysis or adds additional mass points per span to the model. Thus the mass from a. node point to mass point or b. mass point to mass point, are then averaged and lumped at each node point and mass point making a much more even distribution of mass across the model, again improving results.
The best solution would be to add an infinite number of mass points per span, however this would add considerably more time to an analysis. Hence the trade off, adding mass points increases accuracy but takes longer for the analysis to be completed. At some point, adding additional mass points per span will have very little effect on results comparatively.
Finally, it is up to the user which option to use:
a. Add zero mass points between node points
b. Let AutoPIPE calculate the best number of mass points between node points
c. Manually enter the number of mass points per a span between node points.
Now you know why adding a single node point can affect AutoPIPE's results. And, knowing is half the battle.
D. Same model opened in 2 different versions of the application but has different results
When it comes to comparing the results between 2 different versions of the AutoPIPE; Yes, the results may or may not be different.
The CAE development team is constantly adding new features and correcting existing bugs in the program. Some of these problems have a clear connection to the findings of the analysis or to modifications to piping code calculations. As a result, yes, the outcomes can vary from one version to the next version. As a result, the most recent version of the software should be regarded as the most reliable because it contains the most recent code updates, new features, and fixes for known issues.
Highly suggest that you perform the instructions as mentioned above: Comparing 2 AutoPIPE models. This should highlight all the differences between the models with respect to geometry and settings. Next, and more difficult, consider known issues fixed by the development team between the 2 versions, see known issue here. Additional information about enhancements and defects can be found in version release notes. Lastly, remember to thoroughly review code calculations to see if there were any changes between related code years.
AutoPIPE has a QA program, one of the perks about being apart of the QA&R program is access to the AutoPIPE Error reports and notification of all publicized issues.
Notes:
#1. Under certain circumstances what looks to be a default code combination may in fact not be a default code combination.
Steps to Repeat:
Model: Walkthru.dat (included in AutoPIPE example folder)
1. In tools->Model options->General: Set operating cases to 2
2. Modify the Press/Temp data in case 2 for all points: P2 = 400 psi, T2 = 550 deg F
3. In static analysis set, select GR/T1/P1 (via pressure extension)
4. Create a second analysis set with only GR case
5. Analyze the model, and to filter the results, select Max P{1} for print only
6. In code compliance report, the Stress/Allow is 6447/14500
7. Delete all combinations
8. Create a user combination hoop category and sum method, Max P{1} (keeping the name same as a default combination name)
9. Now press ‘Reset Defaults Only’, the user case gets converted to ‘Default Type’, also uncheck the Auto Update flag
10. Viewing the results for Max P{1} gives Stress/Allow of 8596/14500 which differs from the results calculated in step. 6
This issue has been logged under TFS-D172835: A user defined code combination with same name as a default code combination can produce erroneous results. (logged Feb 2015)
Avoidance:
When creating user combinations avoid using Default Combination names. If you are not sure if this has happened in a particular model, delete all Default Code and Non-Code combinations from the model and press the "Reset Defaults Only" button to have AutoPIPE recreate the official default combinations.
#2. Large world coordinates can cause inconsistent offsets, Confirm that coordinates size do not exceed the programs. Reduce the magnitude of the offsets of the first point from the global origin (example, first point in model is located at 0,0,0) (TFS-93164).
#3. Anchor movements stiffnesses and loads are not rotated when model is rotated (TFS-E101701).
See Also
Results Post Processing in AutoPIPE