BrDR 3D FEM Analysis - Hardware/Modeling/ Troubleshooting
- Mark Mlynarski, P.E.
Date | May, 2022 |
Product | AASHTOWare BrDR |
Version | 7.2, 7.1, 7.0, 6.8.4 |
Description | I am having problems with my 3D FEM models. |
What do I do if having issues with my BrDR 3D FEM models?
This article covers the following topics related to 3D FEM modeling:
Recommended hardware and system requirements
Modeling tips
Troubleshooting your model
Recommended hardware and system requirements for 3D FEM analysis
The following is excerpted from the AASHTOWare Catalog located at:
FY-2023-AASHTOWare-Catalog_FINAL.pdf
For the latest supported configurations, please visit: https://www.aashtowarebridge.com
The Core i7 Processor is a minimum requirement for 3D FEM analysis. The faster the better. Currently only one core of the processor is utilized for analysis.
Tests have been performed with multi-threading portions of the analysis, but because the current analysis algorithm heavily uses the hard drive, the performance upgrade was minimal and in some cases, worse. AASHTO is considering looking at reworking these algorithms to make them run faster with multithreading and to get them cloud ready.
While 32 GB is the minimum recommended, for 3D FEM, the more the better. Also, as noted, for 3D FEM, a Solid State Drive (SSD) is recommended as it will help improve run times.
Modeling tips for 3D FEM modeling in BrDR
The factors that influence the speed of 3D FEM analysis include:
Number of shell elements in the deck between girders or in the web between flanges,
Target aspect ratio for shell elements,
Vehicle increment for longitudinal loading.
Vehicle increment in lane and lane increment for transverse loading,
Number of load cases,
Number of vehicles used in the analysis,
Number of degrees of freedom
Some ways to speed up the 3D FEM analysis include visiting the Superstructure Definition: Analysis tab and revising the “number of shell elements” and “target aspect ratio for shell elements” to produce a coarser (larger) mesh. In addition, increase the vehicle increment for longitudinal loading, vehicle increment in lane and lane increment for transverse loading. Note that these changes will also affect the accuracy of the solution so engineering judgment should be used in making such changes.
Other recommendations include:
Operating System
The 64-bit operating system and 64-bit BrDR are recommended for 3D FEM analysis.
Analysis Output Folder
Choose a local folder as the Analysis Output Folder. Writing analysis output to a network folder will degrade the performance of the analysis. This can be set on the ‘Analysis’ tab of the ‘Preferences’ window. (see below). Note that the ‘My Documents’ folder may NOT necessarily be located on a local drive. Check with your System Administrator if not sure.
Selecting Output
Select only the necessary analysis output for the analysis. Disk operations are expensive from the analysis performance perspective.
Run the DL Only Analysis Option
Initially, run the ‘DL Only’ Analysis Option first (see below) to fine tune your model. This will allow for debugging your model before running the longer live load analysis.
This option runs considerably faster and allows the user to review the DL model with ‘Model Viewer’ (see below) before running the Live Load (which is considerably longer). Look for ‘irregularities’ in the model (e.g. missing diaphragms, irregular diaphragm connections, overhangs,etc). Make sure the model ‘looks right’ before continuing).
Check the moments to see if they look ‘Normal’. Run coarser meshes initially and only then fine tune the model. Things to look for: spikes in the moment curves, moments at exterior supports (see XXXX – reference other confluence article).
This will allow for debugging your model before running the longer live load analysis.
Tips for running the LL Analysis
An initial analysis should be made first to check the running of the model. When doing the initial run, you should:
Run a single vehicle only (one without a variable axle spacing)
Do the first specification checking of 1 girder only. Launch the 3D analysis from the member alternative from the Bridge Workspace (BWS). Only ‘Existing’ alternatives have influence surfaces loaded and are spec checked. (see below)
Avoid spec checking of diaphragms initially (uncheck these boxes)
Troubleshooting your model
Evaluating “Pinned” support conditions
See the following confluence article
Moments At Supports For 3D Models - BrDR Support Service Desk - Confluence (atlassian.net)
on evaluating your pinned/roller conditions
Different Number of Nodes
Review the ‘3DGirderNodes.txt’ file. This file was initially developed for the developers but is now shared with the users and it provides a list of the node locations and the reason why the nodes were generated (e.g. diaphragm location, section change, support, tenth point, etc.). If a node is generated as “bitmask 0” it means the node was added to either get the same number of nodes in a span or added based on the number of shell elements selected on the Structure Definition: Analysis tab.
The file is located at:
Different Number of Nodes – Fix
Step 1: Use Paste/Use Text Import Wizard to copy each girder’s data into its own column in Excel
Step 2: Examine total number of nodes per girder looking for differences
Step 3: For girders with difference, examine number of nodes per spans. Note in the example below that G10 has 30 nodes in this span, but other girders have 29 (e.g. G9).
Diaphragm at 70.536605’ (support 2) + 259.018115’ = 329.55472’
Pier 4 is at 329.650012’
Change the number of diaphragm spacing from 15 to 14
Add a new row at 0’ from Support 5(Pier 4) with 0’ spacing
Move Span 5 data down a row