Performing failure analysis involves determining the root cause or causes of the failure of a product and involves a careful examination of potential contributing factors and their interdependence. To a large extent, the process of failure analysis depends on the experience of the analyst and needs to be tailored to each project.

As part of the accreditation procedure, Bodycote Polymer – Broutman Laboratory has developed a formalized test protocol for performing failure analysis investigations.1  This procedure documents the steps necessary to perform and report a failure analysis project on a part made from a polymeric material, such as plastic, rubber, composite, adhesive, coating. There also are several ASTM Standard Practices that discuss failure analysis in general terms, but, in practice, each failure investigation is different.2

Fig. 1A. Magnified view of the fracture at screw hole on a housing, with arrow pointing at plastic deformation and stress whitening.

Summarized below are the 10 most common root causes for failures. This list is not inclusive, but it is intended to serve as a general guideline for an analyst.

• Design error.


• Mechanical failure.


• Compound.


• Polymer type.


• Polymer grade.


• Additives.


• Compounding/mixing.


• Processing variables.


• Environmental damage.


• Combination of several factors.

In the past, the failure analysis process has been described as following a 12 part protocol,3  reproduced below:

1. Collection of background and selection of samples.


2. Review of safety considerations.


3. Establishment of record keeping.


4. Identification and cleaning of samples.


5. Macroscopic examination and analysis.


6. Microscopic examination and analysis.


7. Determination of failure mechanism.


8. Mechanical testing.


9. Chemical, thermal and rheological analysis.


10. Stress analysis (including fracture mechanics).


11. Testing under simulated service conditions.


12. Analysis of evidence and formulation of conclusions.

Clearly not all of the investigations require all of the above steps. However, the reason for skipping any of the 12 steps should be known by the analyst, and it should be recognized that such skipping may produce misleading answers or result in repeating an investigation.4

The following are a few examples where analysis of a failed polymer part helped determine the cause of the failure and suggested a means to prevent such failure in the future.

Housing redesign

Fig. 1B. Redesigned housing, after testing.

In this case study, a polycarbonate control housing was submitted for failure analysis. According to the information provided, the failure had been rarely observed, but was always localized at the screw holes, where the housing with controls was attached to the cover plate. There was no commonality between failed samples otherwise. Macroscopic and microscopic examination established that cracks on the failed samples showed significant plastic deformation. Ductile fracture, with stress whitening, but no evidence of chemical attack (Fig. 1A), suggested that failure was likely due to overloading.

Given the ductile nature of the failure that was observed, it was believed that testing mechanical properties of material from the housing was unnecessary. However, mechanical testing of one exemplar assembly was performed in a fixture, configured and manufactured specifically for the control box. The assembly fractured at the screw holes on samples at a load of approximately 60 lbf.  Failure was ductile, with stress whitening, again indicating overload. 

Fig. 2A. Bushing from original supplier, with flow lines at the meeting of two flow fronts, coming from two different gates.

Since the housing material was a standard, injection-molding-grade polycarbonate, with flexural yield strength of around 13 ksi, the material was not believed to be the cause of the problem. A noticeable deficiency in the screw hole design was obvious, along with an easy design remedy.   The customer decided that a redesign of the control box and corresponding tooling modifications were warranted, rather than a search for a new material.

Five of the redesigned assemblies were tested in a slightly modified compression testing setup.  Average load at failure was just over 200 lbf. Failed samples after testing are shown in Fig. 1B.  

Bushing evaluation

Fig. 2B. Bushing from original supplier, with incomplete filling of the mold.

A number of injection-molded, vibration-isolation HNBR rubber bushings were experiencing high permanent set and fatigue failures. The bushings from the supplier with failures (Fig. 2A) and from a candidate replacement supplier were submitted for a direct side-by-side comparison.

Shore A durometer hardness testing per ASTM D2240 (with deviations) was performed on samples taken from three bushings supplied by each manufacturer. The bushings were cut open, and at least three rubber sections were removed from the inner metal surface of each bushing.  Rubber samples were stacked to guarantee that the required minimal height and Shore A durometer measurements were taken. The values for the supplier with failures averaged 77 Shore hardness units. For the replacement supplier, values averaged 81. Both were within specification calling for Shore A durometer of 80.

Axial and radial stiffness of both the current and the replacement bushings were measured on an Instron tester. Both measurements were found to be approximately 10 percent higher for the replacement bushing.

Compression-set testing was performed per ASTM D395 Method B, with modifications. Spacers for the compression set jigs were machined based on the thicknesses of the samples. The aging period was 22 hours at 100 DegC. The compression set for the original supplier averaged 30 percent, while for the candidate replacement supplier it was 7 percent.

Fig. 2C. Bushing from candidate replacement supplier, with correlation of flow patterns on the core surface with the position of the gates.

The bushings from both the original supplier and the replacement candidate had manufacturing problems. The original supplier had incomplete fills (Fig. 2B) and core tearing. The replacement supplier had poor packing and incomplete fusion of the flow fronts (Fig. 2C).

Since the customer had an in-house capability of running functional tests on replacement bushings, he declined dynamic mechanical testing on a DMA, or fatigue testing on a servo-hydraulic system. The customer believed that once the molding problems of the candidate replacement supplier were fixed, they would be able to validate the conversion to replacement bushings by the functional test.

Housing stress crack

Fig. 3A. Hairline crack on ABS/PC knob housing. High magnification view shows area of stress whitening and probable brittle fracture.

An ABS/Polycarbonate housing with a hairline crack in it was submitted for failure analysis investigation. The problem appeared to be an isolated occurrence, with similar ABS/PC housings from this and other runs performing without problems. Fig. 3A gives a magnified view of the crack. Stress whitening on the edges suggests that the material is not degraded. This, and the isolated nature of the failure, steered investigators away from looking for sub-spec or degraded material in this case.

The view in Fig. 3A shows that, even though there is an area of stress whitening, the hairline appearance of the crack suggests brittle fracture. Typically ABS material deforms in a ductile manner, exhibiting stress-whitening behavior. The brittle fracture of ABS is often due to exposure of ABS to an incompatible chemical. Upon the removal of the housing, the crack on its inside surface showed a small amount of oily residue, as seen in Fig. 3B.

Micro FTIR was used to identify the oily substance. The spectrum gave a very good match for silicone oil, which is a known Environmental Stress Crack (ESC) agent for polycarbonate, and would be expected to be such for ABS/PC. 

Fig. 3B. The arrow marks an oily residue at the crack on the inside surface of the housing.

To confirm the findings, a fractographic examination of the crack was called for. The housing was sectioned, preserving the crack, which was then cryogenically fractured. The appearance of the fracture surface is given in Fig. 3C.

The elliptical shape of the crack’s frontal markings and orientation of the hackle marks confirmed that the crack initiated from the inside surface. The crack initiated along the inside surface and propagated through the thickness, reaching the outside surface at one location.

Confirmational ESC resistance testing of the PC/ABS material in silicone was a possibility, however, the customer felt that the findings were sufficiently conclusive in this isolated instance, so declined to proceed.

Conclusion

Fig. 3C. The fracture surface, showing crack initiation on the inside surface of the housing (bottom of the picture). Arrow marks the microcrack visible on the outside surface of the part.

Failure analysis is a process that frequently takes time and thought to reach the right conclusions.  In the preceding three examples, the investigators generally followed the 12-step protocol. The first case study led to a redesign of the part. The second case led to a change in the vendors. And the third was determined to be an isolated failure.

For more information, email: info@bodycoteusa.com



Note: Original expanded version of this article first presented as paper at 2006 International Appliance Technical Conference.