ALGOR FEA Predicts Mechanism Failure and Confirms Design Modification
By Richard Helms
Industrial Construction Company
This article was published in Machine Design,
FEA Diagnoses Machine Failures, April 1, 2005.
As an engineering consultant and president of the Industrial Construction Company, I regularly use finite element analysis (FEA) to help
my customers design better and safer products. Recently, I performed a linear static stress analysis using ALGOR FEA software to verify the
failure mode of a flange-to-shaft joint, which was loaded in combined torsion and bending.
ALGOR was chosen due to its ability to easily model and analyze surface-to-surface contact between parts.
After validating the FEA process by accurately predicting the failure mode and stresses, the next step was to use FEA to
analyze potential design changes before they were implemented.
A new design was developed where loads were "buried" inside the flange rather than transferred to the flange-to-shaft interface.
Validation of this concept required creating a model in Alibre Design with very close tolerances suitable for FEA in ALGOR.
With the new design, only a portion of the shaft and all of the flange and weld were modeled.
A combination of meshing options, including automatic refinement points, was used to obtain a finite element model suitable for analysis on my computer.
After meshing, loads and constraints were applied to the new flange-to-shaft joint model.
The ALGOR linear static stress analysis results indicated that maximum stresses had been reduced from 83 ksi to 25 ksi and
the maximum stress locations were successfully "buried" inside the center fiber of the flange as intended.
First, I created a CAD model in Alibre Design, a parametric, feature-based solid modeling system. The solid model was then easily
opened within FEMPRO, ALGOR’s easy-to-use single user interface.
I initially meshed the model using a very coarse mesh utilizing ALGOR's automatic hex-dominant solid meshing option and its default settings.
This step was performed to establish a baseline for further iterations and to validate the model.
After meshing, materials were selected from ALGOR's extensive material library with just a right click of the mouse on the menu tree.
It only took a couple of hours to set up my entire FEA model and produce results.
The linear static stress analysis ran to completion on the first try and creating a report was also just a click away.
After my initial analysis, I meshed the model with a substantially finer mesh – again utilizing the automatic solid meshing option.
ALGOR's mesh dialog is very easy to use with a simple slider to adjust the mesh to a finer option.
Several iterations of the analysis were performed using increasingly finer meshes to establish convergence
(i.e., level of mesh that produces results that have "settled down," indicated by very little change in results between meshes).
Noting the convergence mesh size, I re-meshed the model using the default settings and then used ALGOR's automatic mesh refinement feature.
Mesh refinement is very easy to perform, needing only a slider to adjust the mesh refinement points to a finer option.
After a couple of tries, I was able to produce mesh refinement points corresponding to the level of mesh that I had determined by my convergence studies.
This feature is a great time saver when working with large models that have many components and/or models that need a very fine mesh in many locations.
Final results of the failure analysis were produced in less than three weeks (from start of modeling to completion of analysis).
At this stage of the project, my customer was particularly interested in the ability to accurately predict the failure mode that had been observed in the actual part.
Sections of the final analysis visualizations were output and overlaid upon actual photographs of the part that had broken.
The points of failure and the propagation of the failure throughout the shaft accurately corresponded with the actual photos.
In addition, the reported stresses at the points of failure were 83 ksi in 4130 HT steel – indicating potential failure.
Now that the failure had been accurately predicted, it was time to propose and validate modifications to the design.
Joining a shaft of any shape to a flange is known to cause stress risers, so an attempt was made to get the loads to "bury"
inside the flange rather than transfer at the flange-to-shaft interface.
This new design relies on running the gear from off the end of the shaft.
The shaft will protrude through the flange and will be welded on the back side.
In addition, the flange will be broached to meet the shape of the flange with a very tight (but not interference) fit.
Validation of this concept required creating a high-quality CAD model with very close tolerances suitable for use with FEA.
Again, I used Alibre Design to model a portion of the shaft and all of the flange and weld and then opened the model in ALGOR and applied materials.
With this model, I used the "bonded" contact option, which gives node-for-node connection over the entire area of the weld.
Using the "welded" contact option would have limited contact to the edges of the weld, which would not have been accurate for this model.
Before meshing this model, I opened the mesh dialog and selected the "Options…" box.
There, I specified a user-defined tolerance of 0.001 inches.
This was done to prevent bonding of the shaft gear to the broached gear hole in the flange.
Had the default settings been accepted for this operation, the meshes of the shaft gears would have been forced to match
the meshes of the broached hole in the flange and the parts would have been bonded.
The goal here was to create the situation where the weld on the back side was the only connection between the shaft and the flange.
Having access to advanced mesh settings provided added flexibility for specifying the mesh exactly how I wanted.
Since the minimum feature length of interest was 0.161 inches, a mesh size corresponding to 0.161/2.000 was utilized (0.0805 inches).
Using this mesh size, a very large model of approximately 500,000 solid brick elements was created.
To limit the size of the analysis, I again refined the mesh using the automatic refinement feature.
Utilizing this process resulted in a quality model with less than 0.5 million degrees of freedom.
On a problem of this size and complexity, manual mesh refinement would have been prohibitive.
ALGOR's mesh refinement tools easily made this into a problem that could be solved on my personal computer though.
After successful meshing, loads and constraints were applied and the model was verified.
A solid mesh was created in approximately 18 hours while the actual analysis required only 90 minutes and 700 MB of disk space to run.
The computer had an 850 MHz Athalon processor with 382 MB of RAM running on the Window 98 SE operating system.
After completion, the solution results indicated that maximum stresses had been reduced from 83 ksi in the failure analysis to 25 ksi in the new design.
In addition, the maximum stress locations were successfully "buried" inside the center fiber of the flange as intended.
I have had considerable experience working with other popular FEA programs.
However, this was my first experience with ALGOR.
I found it easy to get up and running with ALGOR – with many tutorials and generous help items available for my use.
In addition, the personnel at ALGOR extend friendly and competent assistance via on-line help or by telephone as necessary.
I would highly recommend ALGOR FEA to the novice and experienced user alike.
Industrial Construction is a consulting engineering house specializing in unique solutions.
In addition to strong analytical capabilities, we feature machine and tooling design as well as project management.
Customers include: Government (DoD, NASA, Bureau of Lands and Corps of Engineers),
Aerospace (Boeing, Lockheed-Martin and Alliant Space Systems) and Commercial (Caterpillar Tractor, Barnard Construction, Peak Oil and Weldco Beales) among others.