Applies To
Product(s):	AutoPIPE
Version(s):	ALL;
Area:	Analysis
Date Logged & Current Version	Oct. 2018 11.04.00.10

Problem:

How many node points are needed in a model?

While animating the displacement load cases in AutoPIPE, the maximum displacement appears to be greater at a position located between the existing physical node points.

Is it possible for the pipe displacement values to exceed the values at the existing node points?

 

Solution:

From AutoPIPE help for displacement report:

In a linear static load case, each represents the linear force displacement response of a point due to the loads deformed for that case alone. However, a nonlinear static load case represents the force-displacement response of a point due to the loads deformed for that case alone, and considering that the points' initial global coordinate position (prior to application of load) is the deflected position of the preceding (initial state) load case.

Therefore Linear and Non-liner analysis reported displacements are only calculated at node points in the model. No results are given for intermediate locations between existing node points during a Static analysis. The deformed shape of the pipe would be graphically approximated between node points based on influences of applied loads at node points or piping displacement beyond adjacent node points. Many question how many node points are required to achieve high accuracy of displacement. Yes by adding more node point the closer to actual displaced shape will be displayed / calculated. However, adding to many unnecessary node points would significantly increase the analysis time and would not greatly increase the accuracy of the results. See example below.

Example, pipe between 2 anchors:

Seg A has no node points between the anchors

Seg B has 1 node point between the anchors

 

Seg C has 2 node points between the anchors

Seg D has 3 node points between the anchors

 Seg E has 19 node points between the anchors

Notes:

1. 1. 1. Seg A does not have any displacement in between the anchor points. Recall from above, reason there was no node point for which the program to apply the load-case nor influence of pipe displacement beyond adjacent node points.  '
      2. Seg E shape of the displaced curve is smoother as the number of node points increase, however the amount of accuracy is for maximum displacement is no better than when segment has 1 added node point. 

Conclusion

When performing a:

• • • • Static analysis, no results are generated for piping between node points.
      • "Non-Code" combination deflected shape plots would be graphically approximated between node points based on applied loads at existing nodes and influences of piping displacement beyond adjacent node points. 
      • Adding more node points between existing unsupported nodes does increase the displacement accuracy of movement at a specific pipe location. However adding too many nodes into the model significantly increases analysis time without much gain in precision. Suggest to use an iterative approach of adding a few additional nodes between supports / existing node points, run a Static analysis, create an Displacement output report (Run1), add more node points into the model, again run a Static analysis, create an Displacement output report (Run 2) and compare the results. If significant difference, repeat iteration of adding more node points as needed until the difference was within an acceptable tolerance. 

---

General Questions and Answers:

Q1. For static piping analysis, is manual node refinement/discretization generally required in AutoPIPE in the same sense as mesh-independence studies in ANSYS/APDL?

• • Answer: 
  • As illustrated in the five example segments above, if information is required at a particular point in a model, it is advisable to add a node point at that location. If not, one should utilize the iterative approach suggested earlier. After executing this process several times, an individual should develop a sense of the number of node points necessary between supports, bends, tee ect..

Q2.  For dynamic piping analysis such as modal, response spectrum, seismic, or time-history analysis, what is Bentley’s recommended approach for discretization refinement?

   * Mass Points per Span = Auto and increase the "cut-off frequency" until results converge?

   * Or add more physical pipe nodes along long spans/runners?

   * Or both approaches be used together?

• • Answer: 
  • In short, there is no quick / sure solution. Most models will require examination to determine the best option; increase the actual nodes in the model (ex. between supports, bends, tees, valves, etc..), increase the "Mass points per span", increase the "cut-off frequency", etc...  Ensuring that the number of modes and the cutoff frequency in the modal analysis are high enough to capture the relevant dynamic behavior.
  • Longer answer, performing a dynamic analysis  is a completely different calculation than Static Linear or Non-Linear analysis. Individuals are strongly urged to attend AutoPIPE's Advanced training class (contact your Account manager for Training options) where this topic is discussed at length. For people who have not taken the class, first refer to Bentley LEARN site or AutoPIPE's YouTube channel for content on dynamics analysis. Otherwise the most important points:
  • Make sure that Static Analysis results are sufficient for locations in the model (see above question).
    • The mass of the pipe, components, and contents is lumped at the associated node points, and there are three mass degrees of freedom per node (rotational mass is generally ignored except for eccentric weights)  
  • Suggest "Mass Points per Span" = Auto and "cut-off frequency" = (determine an appropriate value for your  type of analysis)
  • Run modal analysis and check the Frequency and Modal results looking for Total Mass participation.
    • The accuracy of modal / dynamic analyses depends on the number of modes extracted and the cutoff frequency specified. If the discretization (node spacing) is too coarse, higher modes may not be captured, potentially missing important dynamic behavior. Again, AutoPIPE’s default settings and automatic mass point insertion are generally sufficient for most piping models.
  • Employing an iterative method by modifying "Mass points per span" and "cut-off frequency" until the model reaches a Total Mass participation of 70% or higher, while ensuring that Total Mass participation remains relatively stable.
  • Refer to AutoPIPE's help and community pages for information on this topic.
  • Summary Table

    Step/Aspect	AutoPIPE Recommendation
    Node/Mass Point Placement	AutoPIPE can automatically insert mass points; manual refinement rarely needed
    Modal Analysis	Ensure enough modes and high enough cutoff frequency to capture dynamic behavior
    High-Frequency Loads	For loads near supports, increase cutoff frequency and number of modes, or supplement with static analysis

      

Q3. Does the automatic mass discretization have any practical cap or limitation for very large models?

  For example, if the model already has a large number of nodes, can AutoPIPE stop adding automatic mass points because of model size, solver limits, or performance limits?

• • Answer: 
  • Yes and No,
    • AutoPIPE’s automatic mass discretization does have practical limitations for very large models, primarily related to computational resources for a given model size and time to run the calculation will significantly increase. Another words, users should be aware that increasing the number of mass points (especially with high cutoff frequencies) can take a substantial amount of time. During such operations, best to not touch the keyboard while the system focus is on AutoPIPE. The application will interpret this as a new command and may interrupt the analysis adding a notification (Not Responding) in the window header. Otherwise, there is no limitation. 
    • Refer to AutoPIPE help for details about this command, "Mass points per span" = Auto. The program will perform the function of adding mass points per the setting defined. 
    • Typical values for this setting include Zero, Auto, and 9 (maximum). Individuals should determine the correct value to enter based on using the iterative method described earlier.

Q4. Our model includes some pipe runners/spans longer than approximately 8 ft and we perform both static and dynamic analyses. For this type of model, should the refinement be based primarily on:

   * maximum physical node spacing,

   * Auto mass point generation,

   * cut-off frequency,

   * modal mass participation,

   * or a result sensitivity/convergence check?

• • Answer: 
  • Once more, it is important to note that no single bullet point above provides the complete answer. After conducting various analyses, the application of sound engineering experience and judgment will assist in determining which option is most appropriate for reaching a suitable design and results.

Q5. What output quantities does Bentley recommend monitoring during a node/mass discretization sensitivity study?

   For example:

   * support reactions,

   * natural frequencies,

   * mode shapes,

   * modal mass participation,

   * response spectrum results,

   * time-history peak responses.

For context, in ANSYS/APDL, a piping model may require mesh/discretization independence checks by refining 1D pipe elements or shell/solid elements and monitoring convergence of results. I understand AutoPIPE is a specialized piping program based on 1D pipe/beam-type elements, so I would like to understand Bentley’s recommended equivalent practice: whether it should be considered a “node discretization sensitivity study” or “mass discretization sensitivity study,” and how it should be performed for static versus dynamic piping analysis.

• • Answer: 
  • This question is fairly similar to the question above. The response remains unchanged. Certain systems may exhibit sensitivity to Natural Frequencies, so it is advisable to monitor this aspect. It is suggested to ensure at least 70% mass participation, so it is advisable to monitor this aspect. Support reactions and Response Spectrum results are critically significant, potentially necessitating a redesign of the support system or pipe routing, so it is advisable to monitor this aspect. Ultimately, there is no single item of importance; all factors are crucial and should be considered prior to finalizing the design.
  • Summary Table

    Output Quantity	Why Monitor?
    Support reactions	Ensure loads at supports/anchors are stable with refinement
    Modal mass participation	Ensure sufficient modal mass is captured in each direction, suggest 70% or better
    Dynamic responses	Confirm peak dynamic responses are not sensitive to further discretization

 

See Also

"Displacement" sub report

Bentley AutoPIPE