Axle Bracket Weld Fatigue Analysis Using Verity And Node Based Submodeling

Axle Bracket Weld Fatigue Analysis Using Verity and Node Based Submodeling

Marcus Peng Chen & Jerry Chung

Analytical Engineering, American Axle & Manufacturing

FeSafeUGM 2011

ChenmVerity & Submodeling

2Table of Contents

* Part I: Background & Challenge;

* Part II: Submodeling & Case Study;

* Part III: Summary & Discussion;

FeSafeUGM 2011


3Axle Failure Mode

* Seam weld is extensively used on Beam/Banjo axles to attach brackets;

* In addition to the global axle bending/distortion, brackets are subject to loading as well;

* Therefore, weld fatigue failure is one of the most dominant failure modes in axles.

Part I: Background & Challenge

FeSafeUGM 2011


4Limitation of Weld Stress Analysis

Issue 1: a slight change in element size/orientation results in significant difference in the weld end stress.

Different weld mesh of the same modelDifferent weld mesh of the same model

Issue 2: actual weld failure occurred on another weld with much lower stress.


FeSafeUGM 2011


5Weld Fatigue Process & Common Failure Modes

1. Crack Initiation

The initiation fatigue life is usually short in the welded structure because it can be assumed that sharp-edged discontinuities exist in a welded structure. [Almar 1985; Maddox 1991, Nykanen 1993]

2. Crack Propagation

The fatigue life estimations for welded structure are often based on the fatigue life propagation using fracture mechanics. [Almar 1985; Niemi 1995]

Toe Failure(1), (2), (3) Throat Failure

(4) Root Failure

The most common failure mode.The most common failure mode. Does not occurs very often, and is usually caused

by weld process imperfections.

Does not occurs very often, and is usually caused

by weld process imperfections.

[Verity® in fe-safe V2 User Manual] [Verity® in fe-safe V2 User Manual]


FeSafeUGM 2011


6Verity Showed a Good Correlation


1.47X cycles

0.21X cycles to crack penetration

@ the arrow marked spot

Test: X cycles to visible leakage

* As shown above, Verity was able to pinpoint the failure location.

* Prediction indicates crack penetration at one spot, which is not exactly equivalent to the visible leakage time seen from the test.

FeSafeUGM 2011


7Challenges & Proposed Solution


Solution: Node Base Submodeling (Global/Local Method)* Eliminate the necessity of fine-meshing the whole axle so that more efforts could be focused on the interested areas (such as weld).* Avoid rebuilding full axle model if a bracket or weld has been redesigned.

Solution: Node Base Submodeling (Global/Local Method)* Eliminate the necessity of fine-meshing the whole axle so that more efforts could be focused on the interested areas (such as weld).* Avoid rebuilding full axle model if a bracket or weld has been redesigned.

Interested area:

Bracket Weld Toe

Challenges:* Due to the special mesh requirements imposed by Verity, brick element mesh for the whole tube turned out to be labor-intensive and error-prone. In addition, the inclusion of numerous welds makes the meshing even more difficult.

* Full axle model takes long time to solve.

* With the inclusion of the Verity required Nodal Force output, the FEA output file can easily exceed 5 gb, which does not only consume more FeSafe File Loading/Solving time, but also causes FeSafe Reading error in some cases.

FeSafeUGM 2011


8Submodel Setup: Case Study I

Part II: Submodeling & Case Study

Step 2: Local model (conforms to Verity mesh requirement in the weld toe area), which only includes the interested welds and brackets. And the Nodal Displacement of the Global model is used as the boundary condition for this model. The solve time for this local model is very short (1 second for this case).

Step 2: Local model (conforms to Verity mesh requirement in the weld toe area), which only includes the interested welds and brackets. And the Nodal Displacement of the Global model is used as the boundary condition for this model. The solve time for this local model is very short (1 second for this case).

Step 1: Global model (1st/2nd Tet/Brick),which does not have to include welds and brackets. The Spring Seats are included in this model for Boundary Condition purpose. No special mesh treatment is needed.Step 1: Global model (1st/2nd Tet/Brick),which does not have to include welds and brackets. The Spring Seats are included in this model for Boundary Condition purpose. No special mesh treatment is needed.

FeSafeUGM 2011


9Submodel v.s. Full Model: Case Study I Stress Comparison


Full Model Submodel

FeSafeUGM 2011


10Submodel v.s. Full Model: Case Study I Weld Fatigue Comparison


Full Model Submodel





Test: X cycles to visible leakage

FeSafeUGM 2011


11Case Study II Weld Fatigue Correlation


X cycles to tube fracture.

[Wei Li, 2010]

* Both test and FEA prediction indicate that the tube will crack from the same location shown above.

* The prediction indicates crack penetration at one spot, which is not exactly equivalent to the visible tube fracture time seen from the test.

* As shown above, the actual weld geometry is not well controlled, which does causes large variations in weld fatigue life.

* Both test and FEA prediction indicate that the tube will crack from the same location shown above.

* The prediction indicates crack penetration at one spot, which is not exactly equivalent to the visible tube fracture time seen from the test.

* As shown above, the actual weld geometry is not well controlled, which does causes large variations in weld fatigue life.


@ the arrow marked spot.

FeSafeUGM 2011


12Advantages of Submodeling Combined with Verity

Part III: Summary & Discussion

Full Model Submodel

FEA Pre-processing Difficult to mesh; time-consuming Easy to mesh the global model; weld toe mesh is treated separately

FEA Solver Time A few hours Global model takes a few hours; submodel run for a few seconds

FEA Output File Size A few GB A few GB for global model, a few MB for the submodel

FeSafe Pre-processing 10-20 minutes to load the model; possible to exceed FeSafe limit Less than 1 minute to load the model

FeSafe Solver Time 0.5-1 hours A few minutes

Post-Processing Takes a few minutes to load the model

Takes a few seconds to load the model

Accuracy Better Less accurate

Bracket Design Change Have to rebuild the full model Can use the same global model, and rebuild the submodel only

FeSafeUGM 2011


13

Thank You

[email protected] & [email protected]

Axle Bracket Weld Fatigue Analysis Using Verity And Node Based Submodeling

Automotive