Failure Analysis

When good engineers first start to design a product, one of their first concerns is to determine the harshest conditions that the product will experience in service. The product is then designed to minimize the hazards associated with this "worst case scenario." Discerning the worst case scenario requires a complete understanding of the product, its loading and its service environment. Prior to the product entering service, a prototype will often undergo laboratory testing which proves the product withstands the worst case scenario as expected.

Unfortunately, a "complete" understanding of a product's service environment can be extraordinarily difficult to obtain. Unforseen vibrations develop; unanticipated temperatures are encountered; manufacturing tolerances can not be held and components do not quite mate as intended; the product is used or serviced in unintended ways. Failure is often the end result of these unexpected factors.

Of course, bad engineering, poor manufacturing, inadequate quality control, fraud, even typographical errors in specifications have caused failures.

In engineering, a failure occurs when a device or structure is no longer able to function as intended. The failures can be mundane, as in the case of an electroplated table pedestal which became discolored during service. This incident cost the table manufacturer thousands of dollars and jeopardized its contract with a major hotel chain. The failures can also be tragic on a monumental scale, as in the case of a wide body aircraft which suffered an engine departure on take-off ( "engine departure" is a euphemism for an engine falling off). The end result of that particular incident was a death toll in the hundreds and tens of millions of dollars in damage.

Failures can be costly and tragic. They can also demonstrate the difference between a product's expected and actual service environment. Once the actual service environment is determined, the weakness which caused the failure can be corrected and future failures avoided.

In many ways, a failure is also a full scale test of a product in actual usage. No responsible engineer would intentionally place 250 human beings, their luggage, and enough fuel to carry them 6000 miles into a $100,000,000 aircraft and then crash it into a crowded neighborhood during a thunderstorm. Analyzing the aftermath of such a disaster, however, can provide insight into the stresses experienced by the aircraft's components; aerodynamic characteristics of the aircraft in extreme conditions; the capabilities of weather detection equipment; flame retardance of cabin components; human factors from both the cockpit crew before the crash and the passengers during the impact; the fire resistance of buildings on the ground; etc. The knowledge gained by analyzing such a failure can prevent additional failures and minimize the hazards associated with "the worst case scenario."

Because engineering failures can be so costly in terms of human suffering and financial losses, parties harmed by a failure often seek recompense for their damages. Thus, failure analyses are also performed to assign blame.

The most common forms of material failures are fracture, corrosion, wear, and deformation. Each of these failure modes will be explored in detail with examples in subsequent discussions. Before the actual failure mode can be determined, however, a failure analysis must be performed. The next series of discussions will concern the procedures for a proper failure analysis and the techniques utilized during such an evaluation.

The outline below details future discussions in this treatise. Please visit this site in the future as we continue our discourse on engineering failure analysis.

Brian G. Brady, P.E.
Roger Harvey, P.E., P.C.
25 Kinkel Street
Westbury, NY   11590
Phone: 516 333 2520
Fax:	516 333 2530
© 1999 by Roger Harvey, P.E., P.C. All rights reserved. The Materials Science and Engineering Department of the State University of New York at Stony Brook is allowed by Roger Harvey, P.E., P.C. to post this document on its website. Individuals may download one copy of this document for scholarly purposes. This document is not to be used for instructional, training or similar purposes without the written consent of Roger Harvey, P.E., P.C., its successors or Brian G. Brady.

Outline for Treatise on Failure Analysis

I. Introduction

Describe Failure Analysis:

     A study performed to determine why a component, subassembly, assembly, system, etc. failed.

Define Failure:

     Cessation of function or usefulness of a part, component, etc...
     The most common modes of failure are fracture, corrosion, wear and deformation.

State Reasons for Performing Failure Analysis:

     Most important: To prevent future failures.

     To assign blame.

II. Discussion

1. Procedure for Failure Analysis

The following is an ideal procedure. Often, various constraints prevent each step from being performed.

  • Perform preliminary examination of part. Determine preliminary failure mode. Note part numbers, serial numbers, supplier or manufacturing markings. Photograph part with special attention paid to anomalies (fractures, scratches, unusual marks, etc.). Preliminary examination may include selecting samples, noting location of pieces separated from the failed item, questioning witnesses, etc.

  • Collect as much information as possible concerning the part, including engineering drawings, specifications, product literature, life history of part including servicing, etc....

  • Perform non-destructive testing.
    • Check for cracks and internal flaws
      • x-ray
      • magnetic particle
      • fluorescent penetrant
      • ultrasound
      • eddy current

    • Dimensional inspection
      • micrometer, vernier, plug gauges, thread gauges, radius gauges,...
      • surface finish
      • optical comparator
      • coordinate measuring machine

    • Check coating thicknesses
      • magnetic induction
      • eddy current

    • Residual stress measurement
      • photoelastic studies
      • x-ray diffraction studies

  • Collect samples of residue for chemical analysis

  • Clean and preserve fractures (use acetate replication to preserve residue on fracture if necessary) or other surface of interest. Perform examination (visual and binocular microscopic) and document same. If required, prepare samples for SEM, microprobe analysis, etc.

  • Prepare samples (with proper identification) for determination of mechanical properties, chemistry (alloy identification), macrostructure, hardness, conductivity, etc. Insure proper testing of these samples, as well as samples collected under steps iii. and iv. Compare actual values with required values (if applicable) or typical values.

  • Prepare samples for metallographic examination and review microstructure, finishes, etc. Document same.

  • Perform testing on identical or similar parts. Such testing can include corrosion testing, fatigue testing, vibration testing, dimensional studies, etc.

  • Determine actual failure mechanism (e.g., second stage turbine blade fracture initiated by fatigue at a machining score, fracture propagated by fatigue until final failure in overload. Separated portion of blade then contacted third stage turbine section, resulting in destruction of third and fourth stage turbine sections).

  • Perform stress analysis, fracture mechanic analysis, etc. if required.

  • Analysis of all evidence, formulation of conclusions, preparation of report with recommendations.

  • Examine Techniques of Particular Interest
    • Fractography
    • fatigue
    • corrosion cracking
    • embrittlement
    • creep

  • Metallography

  • Evaluate processing of part
    • fabrication processing (cast, rolled, forged, etc.)
    • thermal processing (heat treatment)
    • nature and condition of coatings (plating, paint systems, etc.)

  • Evaluate damage to microstructure
    • Grinding/machining damage
    • Overall overheating of microstructure

  • Compare actual microstructure with required microstructure

  • Evaluate nature of cracking
    • transgranular
    • intergranular

  • Wear Evaluation
    • adhesive wear
    • abrasive wear
    • fretting wear

  • Contact Stress Fatigue

  • Visual, binocular, SEM examination

  • Non-Metallic Materials
    • Concrete
      • Differential settlement of foundations
      • Sulfate attack
      • Improper mixture
    • Wood
      • Microbial attack
      • Insect damage
    • Glass-failure related to internal flaws.

III. Examples of Failures

1. Overload failure of rivet squeezer yoke due to improper heat treatment

1. Rivet squeezer yoke used to squeeze rivets

2. Failed in overload during first use

3. Hardness varied with thickness. Thickest section had hardness of Rockwell C 51, thinnest section had hardness of Rockwell C 31, Rockwell C 42 to 45 was required.

4. Microstructure was non-uniform. Lightly banded as well as blocky ferrite was present in martensite matrix. This indicates a slack quench.

5. Part was the propr material but poorly machined. Improper machining and heat treatment indicate poor quality at shop.

6. Failure was attributed to poor mechanical properties due to improper heat treatment.

2. Fatigue Failure of Hip Prosthesis

1. Ti-6A1-4V hip prosthesis installed for one year

2. Topographically flat fracture with clamshell markings surrounding a single point. SEM noted individual fatigue striations. These indicate fatigue cracking.

3. SEM found numerous cracks along surface both near and well away from through-crack. Through-crack probably initiated as one of these small surface cracks.

4. Surface was ion-nitrided. It was claimed that ion-nitriding improved fatigue life. This claim was based on tests involving a total of four samples all at the same stress ration (a fifth sample was tested but the results were not reported). Mil-HDBK-5, Section recommends 8 to 16 specimens at each of three stress rations for adequate fatigue testing. Fatigue testing of ion-nitrided surfaces was inadequate.

5. Some researchers found that ion-nitriding caused a thin layer of alpha phase (alpha case) to develop on the part's surface. Alpha case is notoriously brittle, leading to premature fatigue failure. It was theorized that such an alpha case was present on the prosthesis, resulting in the numerous cracks along the length of the prosthesis. One of these cracks eventually became a through-crack, resulting in failure of the prosthesis.

3. Failure of Fitting Due to Exfoliation Corrosion

1. Three inch long crack noted in 7075-T651 aluminum beam support fitting during overhaul. Aircraft had been in service for 20 years (5,807 flight hours, 1,841 catapult launches and arrested landings). Some corrosion was noted around the crack.

2. White and gray corrosion product present on the surface of the crack. Small segments of material appeared to be flaking away, giving fracture a leafy or laminar appearance. This is typical of intergranular corrosion.

3. Metallographic studies through the fracture found both the main crack and secondary cracks were intergranular. Cracking was generally in the transverse and long transverse directions, indicating it was exfoliation cracking.

4. Failure could probably have been prevented with a better coating system and sealing of bolt holes. Another and better solution was to change the alloy from 7075-T651 (which is susceptible to exfoliation and stress corrosion cracking) to 7050-T76511 (which is highly resistant to such cracking).

4. Evaluation of Plastic Bead Blasted Ti-6A1-4V Fatigue Specimens.

1. Plastic bead blasting was being evaluated as a possible paint removal technique. There was some concern that the blasting would induce damage to the surface which could cause premature fatigue damage. Possible failure scenarios included: roughening of the surface with rough spots acting as fatigue initiation sites; heat damage due to friction between the workpiece and the beads; solid metal embrittlement due to lead and cadmium in recycled blasting media. Clad aluminum, cadmium plated steel, and Ti-6A1-4V material was tested. Only the Ti-6A1-4V test pieces exhibited a decrease in fatigue life compared to control samples.

2. Binocular microscopy and SEM found machining induced damage (flowed metal) on the surface of the bead blasted samples. The control samples were free of such damage.

3. Metallographic examination found a white layer on the machined surfaces of the bead blasted samples. The white layer indicated it was an alpha case induced either by overheating or excessive strain. No such layer was noted on the control samples.

4. Profilometer (surface finish) studies found that the control specimens had a surface finish of 7.33 microinches while the bead blasted samples had a surface finish of 22.7 microinches. 8 microinches was required.

5. It was concluded that the premature failure of the bead blasted specimens was related to improper machining, not the bead blasting. New bead blasting samples were prepared and tested with no decrease in fatigue life due to the bead blasting.

5. Fretting Wear (False Brinelling) of Aircraft Rudder Bearing

1. Vibration noted as rudder was moved, ball bearing found to be cause.

2. Bearing was disassembled and evenly spaced indentations were noted around the entire bearing race. Such indentations are termed false brinelling, after the Brinell hardness test.

3. Indentations were caused by fretting wear. Such wear is generally related to vibration.

4. Spalling was noted on rolling elements. Spalling is the result of contact stresses. It is believed that the false brinelling in the raceway caused excessive contact stresses in the rolling elements.

6. Tack-Welding of Silver Cadmium Contacts in a Relay

1. Contacts of relays were found to be occasionally tack welding shut, causing malfunctioning of various systems. One such relay was submitted for analysis. None of the contacts were tack-welded when the relay was submitted for analysis.

2. The relay incorporated three sets of contacts. Severe tack-weld marks were noted on two of the three sets of contacts.

3. The relays utilized silver cadmium (85% Ag-15% Cd) contacts, which require an internal oxidation procedure. The internal oxidation procedure causes cadmium oxides to precipitate in the microstructure. These precipitates greatly improve the contacts' resistance to tack welding. No such precipitates were noted in at least one contact of each contact set with severe tack-weld marks. It was concluded that the contacts which tack-welded had not been properly oxidized.

7. Failures of Non Metallic Components

1. Collapse of masonry building during differential settlement and sulfate damage.

2. Failure of wood beam due to rot. Decay is related to moisture from a plumbing leak.

IV. Conclusions

- Failure analyst is to engineering what the pathologist is to medicine.

- Failure analysis requires familiarity with many branches of engineering.

Check out the Failure FAQ

    A Web Page brought to you by the friendly folks at:
Department of Materials Science and Engineering
State University of New York at Stony Brook

04/30/99 Brian Brady and JQ