View source for Failure Analysis of an AZ91 Automotive Shock Tower

== Background and Initial Observations ==

Contributing Author: Wilburn Whittington

<table width="100%" cellspcing="7" cellpadding="7">
<tr>

<td>
AZ91 automotive shock towers, supplied by General Motors (GM), were received for mechanical testing and failure analysis for use in ISV/damage model validation at the Center for Advanced Vehicular Systems (CAVS).  The shock towers were designed to sustain loading from the suspension system and transmit the load to the vehicle chassis.  Figure 1 illustrates the shock tower’s with relation to the chassis frame components of the GM vehicle.

Also received with the shock towers was a computer aided design (CAD) drawing of the ideal shock tower after machining.  To compare the geometry of the as-received shock towers with the ideal shock tower, several locations were examined with regard to material thickness.  The ideal shock tower thickness was obtained from the CAD drawing at nine locations.  For the received shock tower, a Mitutoy micrometer ( ±0.001 mm error) was used to measure locations near the edge of part and then used to calibrate an Olympus 35DL ultrasonic thickness tester ( ±0.01 mm error) to measure locations near the center.  The results from four measured shock towers are shown in Table 1 with an illustration of the shock tower CAD drawing showing ideal dimensions and examined locations (A-I) on the real shock towers.

Notice in Table 1 that many of the as-received locations show a mean difference in thickness of about 0.20 mm from the ideal shock tower.  A difference of this thickness is not uncommon in casting, however, due to the shock tower being only 3 to 5 mm in thickness, some measurements showed up to 7-10 % change in thickness from the ideal condition.

Having locations with 7-10% difference in thickness can allow deviation from the ideal condition especially when, in the case of a shock tower, many of the loading conditions are in bending.  Being that the shock tower deviated significantly from the CAD drawing, consideration of the effects of geometry could play an important role in the mechanical performance.

Although the difference between the ideal and real shock tower thicknesses were significantly different, the real sample-to-sample variation was small.  For instance, in Location G, the difference between the ideal and real shock tower showed that the real shock tower could be as much as 0.3 mm thinner than the ideal shock tower, however, in measuring four different shock towers the maximum variation was only 0.02 mm showing that all of the shock towers showed the same trend.  The large sample to ideal  and small sample to sample difference shows that the shock tower casting process can produce consistent geometry and that an alternate CAD drawing would provide an accurate geometry of many shock towers.

<table style="border:1px solid black">
<tr>
<th>Location</th>
<th>Mean Sample<br/>Thickness<br/>(mm)</th>
<th>Sample-to-Sample<br/> Variation<br/>(± mm)</th>
<th>Ideal Sample<br/>Thickness<br/>(mm)</th>
<th>Ideal-to-Real<br/>Mean Difference<br/>(mm,%)</th>
<th>Ideal-to-Real<br/>Maximum Difference<br/>(mm,%)</th>
</tr>
<tr><td>A</td><td>4.97<td>0.08</td><td>5</td><td>-0.03, -0.6</td><td>-0.11, -2.2</td></tr>
<tr><td>B</td><td>4.79<td>0.04</td><td>5</td><td>-0.21, -4.2</td><td>-0.25, -5.0</td></tr>
<tr><td>C</td><td>2.95<td>0.05</td><td>3</td><td>-0.05, -1.7</td><td>-0.01, -3.3</td></tr>
<tr><td>D</td><td>2.84<td>0.10</td><td>3</td><td>-0.16, -5.3</td><td>-0.26, -8.7</td></tr>
<tr><td>E</td><td>2.78<td>0.02</td><td>3</td><td>-0.22, -7.3</td><td>-0.24, -8.0</td></tr>
<tr><td>F</td><td>3.04<td>0.06</td><td>3</td><td>+0.04, +1.3</td><td>+0.01, +3.3</td></tr>
<tr><td>G</td><td>2.72<td>0.02</td><td>3</td><td>-0.28, -9.3</td><td>-0.03, -10.</td></tr>
<tr><td>H</td><td>3.15<td>0.05</td><td>3</td><td>+0.15, +5.0</td><td>+0.02, +6.7</td></tr>
<tr><td>I</td><td>3.19<td>0.04</td><td>3</td><td>+0.19, +6.3</td><td>+0.23, +7.7</td></tr>
<tr>
<td colspan="6" align="center">Table 1.  Shock tower CAD drawings and locations A-I for thickness measurements (images), and the mean and maximum thicknesses measured from four shock towers at Locations A-I as compared to the CAD drawing thicknesses (table). </td>
</tr>
</table>
</td>

<td width="510">

<table width="100%">
<tr><td align="center">[[Image:ShockTower.jpg|500px]]</td></tr>
<tr><td align="center">Figure 1.  Shock tower and connecting parts.</td></tr>
</table>

<table width="100%">
<tr><td align="center" colspan="6">[[Image:ShockTowerCAD.jpg|500px]]</td></tr>
<tr><td align="center">CAD drawings of the shock tower</td></tr>
</table>

</td>
</tr>
</table>

== Microstructure Analysis ==

Being that cast components typically show varying microstructure due to a myriad of environmental effects, investigation into the porosity of the shock towers was undertaken.  Specimens taken from seven different locations were extracted, mounted, and polished using Struers recommended polishing procedures, and then examined for porosity distribution.  Table 2 shows an illustration of the measured locations and their respective porosity volume fraction.

<table width="100%">
<tr><td align="center">[[Image:ShockTowerMeasuredPorosity.jpg]]</td></tr>
<tr><td align="center">Table 2.  Shock tower measured porosity locations (image) and volume fraction results (table).</td></tr>
</table>

Notice in Table 2 that the porosity volume fraction was significantly higher in Locations 1 and 3, than in other locations.  In all locations, with the exception of Location 5, the porosity volume fraction was greater than 1% showing that the shock tower exhibits high porosity content.  Although many locations showed a high volume fraction of porosity, the high porosity was mainly due to concentrated cracks and defects on the millimeter scale.  Notice the optical mosaic image of Location 3 in Figure 2.  Much of the image shows large cracks or pores that are overshadowed with a constant low porosity zone.  The low porosity zone is precisely the cause of the low volume fraction in Location 5, as shown in Table 2.  The increased porosity is from a system of defects that are large.  Because the defects are large (major dimension> 1 mm) attaining an accurate volume fraction of porosity is difficult.  Therefore, as shown in Figure 2, specimen locations of at least 25 mm2 were used for imaging.  The optical mosaic, shown in Figure 2, used sixty four images taken at 500x magnification to construct the mosaic used for porosity volume fraction values.

<table width="100%">
<tr><td align="center">[[Image:ShockTowerOpticalMosaic1.jpg]]</td></tr>
<tr><td align="center">Figure 2. Optical mosaic image of Location 3 showing porosity.</td></tr>
</table>

The large mosaics were imported in a CAVS image analysis code which uses grey scale threshold values to separate the porosity zones from the fully dense material zones.  Shown in Figure 3 is a typical result from a single image analyzed with the image analysis software.  

<center>
<table>
<tr>
<td>[[Image:ShockTowerImage2a.jpg|400px]]</td>
<td>[[Image:ShockTowerImage2b.jpg|400px]]</td>
</tr>
<tr>
<td colspan="2" align="center">Figure 3.  Single optical mosaic image from Location 3 (left) and porosity detected image from software (right).</td>
</tr>
</table>
</center>

Because the images gathered from optical microscopy show the revealed porosity on the surface of the polished sample, being able to relate the values of porosity area fraction (2D distribution) to volume fraction (3D distribution) is difficult.  Therefore, two locations, Location 3 and 1, were examined with a Phoenix X-Ray CT scanner which takes over six hundred 2D x-ray images of the sample cross-section and then reconstructs at 3D part showing the true volume fraction of the samples.  The samples were then polished and imaged on a Zeiss Optical Microscope along with the other sample locations.  Figure 4 shows the two images gathered from the different techniques.  The calculation of porosity volume fraction from the 2D images compared well with the 3D scans and therefore showed that the 2D imagery provides an accurate means of attaining the porosity volume fraction.

<center>
<table>
<tr>
<td>[[Image:ShockTowerOpticalMosaic4a.jpg|400px]]</td>
<td>[[Image:ShockTowerOpticalMosaic4b.jpg]]</td>
</tr>
<tr>
<td colspan="2" align="center">Figure 4. Optical mosaic image (left) and x-ray scanned image (right) of Location 3 showing porosity.</td>
</tr>
</table>
</center>


Specimens taken from Locations 1 and 2, measured at the University of Michigan as part of another project, were extracted from several shock towers to show the sample-to-sample variation.  The conclusion of that project was that at a constant location, sample-to-sample variation was small.  This means that not only does the geometry of the shock tower remain consistent from sample-to-sample, but that the microstructure also exhibits this trend.

Because the porosity images in the shock tower microstructure show large pores and high volume fractions, porosity appears to be a main concern in the mechanical performance of the shock tower.  Notice in Figure 4 that Locations 3 not only has 1 mm large pores, but that there are numerous large cracks that extend beyond 5 mm.  Due to the shock tower having thickness of nearly the same values as the cracks exhibited throughout the thickness, defects due to high porosity could be a major concern.

=== Animations ===
<table width="400">
<tr>
<td> </td>
<td>Location 1</td>
<td>Location 5</td>
<td>Location 8</td>
</tr>
<tr>
<td>Axial</td>
<td>[https://icme.hpc.msstate.edu/mediawiki/images/c/cb/220-58-1_Axial.avi 220-58-1]</td>
<td>[https://icme.hpc.msstate.edu/mediawiki/images/c/c4/220-58-5_Axial.avi 220-58-5]</td>
<td>too large to load<td>
</tr>
<tr>
<td>Frontal</td>
<td>[https://icme.hpc.msstate.edu/mediawiki/images/e/e5/220-58-1_Frontal.avi 222-58-1]</td>
<td>[https://icme.hpc.msstate.edu/mediawiki/images/b/b6/220-58-5_Frontal.avi 222-58-5]</td>
<td>too large to load</td>
</tr>
<tr>
<td>Sagittal</td>
<td>missing</td>
<td>[https://icme.hpc.msstate.edu/mediawiki/images/8/8a/220-58-5_Sagittal.avi 220-58-5]</td>
<td>too large to load</td>
</tr>
</table>

== Material Mechanical Response ==

Mechanical tests were carried out on the shock tower in tension and compression.  The tension specimens were extracted from Locations H and I, shown in Table 1.  The tension specimens were flat dog-bones with a gage section of 10 mm wide, 3 mm thick, and 50 mm long, with a fillet radius of 10 mm.   The compression specimens were extracted from the rib sections near location 3 and were machined to right cylinders of 10 mm in diameter.  The specimens were examined and shown to be free of surface cracks.  Three specimens of each test method were tested and a representative curve from each is shown in Figure 5.

<table width="100%">
<tr><td align="center">[[Image:ShockTower StressStrain.jpg|400px]]</td></tr>
<tr><td align="center">Figure 5.  Quasi Static (0.001/s) compression and tension tests results from shock tower material.</td></tr>
</table>

Figure 5 also shows that the material mechanical behavior in tension and compression is extremely brittle.  The tension specimens all failed between 1-1.5% true strain and the compression specimens failed between 9-11% true strain.  The yield strength of 175 MPa is not significantly different than other cast AZ91.

To further investigate the brittle failure of the tension specimans, the fractured surface of a tension specimen was collected and examined using an EVO-SEM manufactured by Zeiss.  Figure 6 shows the fractured surface of the tension specimen at four different magnifications.  Notice that the fractured surface is flat and smooth.  This is indicative of brittle failure.  The large defects presented earlier could be the driving force for the catastrophic failure in the tension tests at such a low strain.  The fractured surface did not show a specific fracture initiation location, which could be due to the numerous large cracks that were present in the material’s initial state.

<table width="100%">
<tr><td align="center">[[Image:ShockTowerFracturedSurface.jpg]]</td></tr>
<tr><td align="center">Figure 6.  Fractured surface of a quasi static tension specimen, at four different magnifications, from an AZ91 shock tower.</td></tr>
</table>

The tension and compression failure strains are remarkably low even for magnesium alloys. The fractured surfaces of the tension specimens also show that the cracks provide the fracture characteristics at the low strain values.  Brittle material response, such as shown here, could lead to reduced energy absorption and subsequent crashworthiness performance during the shock tower component deformation.

== Component Tests ==

Two shock towers were utilized for component testing.  The test setup, shown in Figure 7, was chosen to ensure that the failure of the shock tower was not near the fixtures, but that the specific failure location could not be easily predicted.  As shown in Figure 7, both the legs and the frontal portion of the shock tower were allowed rotation though their planar axis with a vertical load applied at the frontal portion.  The vertical load’s line of action was positioned such that it crossed the leg’s axis of rotation.  This allowed the shock tower to deform without significant rotation of the members.

<center>
<table>
<tr>
<td>[[Image:ShockTowerSetupA.jpg]]</td>
<td>[[Image:ShockTowerSetupB.jpg]]</td>
</tr>
<tr>
<td colspan="2" align="center">Figure 7. Illustration of test setup (left) and actual test setup (right).</td>
</tr>
</table>
</center>

Two strain gages were placed on the shock tower at arbitrary locations to compare the sample-to-sample strain variations due to both the shock tower variability and the test setup variability.  The strain gages used were large strain (10% full scale) rectangular rosette gages provided by Micro-measurments.  The shock tower was loaded at a displacment rate of 1 mm/min and displacment was gathered from the Instron 5882 cross-head displacement.

<center>
<table>
<tr><td>[[Image:ShockTowerLoadDisplacement.jpg]]</td><tr>
<tr><td>Figure 8.  Load versus displacement (left) and calculated principal strain (right) from the two shock tower tests.</td></tr>
</table>
</center>


Figure 8 shows the shock tower load versus displacement as well as the principal strains calculated from the two rosette locations.  The results show little variability in mechanical response between the two tests which  further illustrates the low sample-to-sample variation.  Because the geometry, microstructure, and material mechanical response had consistent results, the component tests were expected to also exhibit this trend.  The gage data show that the sections where the gauges were placed exhibited less than 0.5% strain.  While this is not a lot of strain, the failure strain is only about 1.5% strain in uniaxial tension as shown previously.  The low displacement-to-failure of approx. 36 mm  is also a result of the low strain-to-failure of the material.

During the test, several video cameras were placed in various locations to capture the onset of fracture.  A large crack was observed in one of the legs and was accompanied with the load drop at 36 mm, as shown in Figure 8.  The test was stopped and images of the fracture was gathered and are shown in Figure 9.

<table width="100%">
<tr><td align="center">[[Image:ShockTowerFailureLocation.jpg]]</td></tr>
<tr><td align="center">Figure 9.  Fracture location from the shock tower component test.</td></tr>
</table>


The fracture initiated at the center of the crack and then grew in two directions to form the crack seen in the images in Figure 9.  Both shock towers experienced failure at the same location, however, the crack lengths were different.   This is because in Test 1 the Instron machine was not stopped as soon as the crack initiated.  For Test 2, however, knowledge of the crack location and the failure displacement was known and therefore the test was stopped sooner.  This also explains the deviation of the principal strains from the two tests at the end of the tests, as shown in Figure 8.

The location of failure was close to the measured porosity of Location 7 and the measured thickness of Location G.  Although Location 7 had a median value of porosity as compared to the other locations, Location G has one of the highest reductions in thickness as compared to the ideal CAD drawing received.  Because of thinness of the section, the cracks become large, as compared to the thickness, and the stress becomes much higher, especially in bending.  Therefore, increasing the thickness of the material in this location would increase the maximum load and strain-to-failure in tested loading condition.

== Conclusions ==
AZ91 shock towers, provided by GM, were examined and tested to provide insight into the mechanical performance and failure mechanism.  The initial examination of the geometry of the shock towers revealed large variation in thickness, with many locations having a reduced thickness as compared to the ideal shock tower geometry.  Sample-to-sample variation in thickness was shown to be low for a single location which showed constituent geometry between samples.

When investigating the microstructure of the shock towers, large defects in the form of pores and cracks were observed.  Many locations exhibited large cracks on the order of 1-5 mm in length and large pore volume fractions on the order of 1%.  Testimony from the UM project groups led to understanding that the sample-to-sample variation in porosity distribution was low and showed that the microstructure between shock towers were similar for a single location.

Experimental testing of the AZ91 material in tension and compression revealed low strain to failures and low sample-to-sample variation which substantiated the testimony from the UM group’s conclusion on low microstructure variation.  The brittle failure in the tension specimens were analyzed using an EVO SEM and showed a very flat and smooth fracture without necking.  This led to the conclusion that the overwhelmingly high porosity and large crack lengths led to the materials inability to accommodate deformation at large strains.

Component testing to the shock towers showed low displacement to failure exhibited by a large vertical crack in one of the shock tower legs.  The failure location and load versus displacement data showed consistency between two samples which showed that the low sample-to-sample variability led to the components performance being similar.  The failure location was found to exists in a location where the material was much thinner than was originally intended and could have played an important role in the low displacement to failure of the component.

In conclusion, the microstructure of the shock tower played a key role in the material mechanical response.  The material mechanical response, along with the large geometry variation from the ideal case, governed the low performance of the shock tower component.  Therefore, to increase the shock tower performance, the shock tower thickness must be kept more uniform  and the material microstructure defects must be reduced substantially.




[[Category:Magnesium]]