17.3.2 Destructive Testing
Differentiating destructive testing from non-destructive testing appears simple at first glance, but is actually not always so clear. Of course, tests in which the part is destroyed, altered in a way that makes its later use impossible, or has samples taken from it, should be classified as destructive. However, what about tests that only preclude reuse in case of insufficient part quality? One example of this is glass shot peening as a bond strength test for silver coatings (Fig. "Methods of qualitative bond strength testing"). Even a so-called proof test that relies on overstressing parts is not clearly classifiable. An example of this is the spinning of ceramic rotor parts. Parts with insufficient quality are eliminated through fracturing, while those that pass the test are installed.
There are also testing processes that can be classified as destructive on non-destructive, depending on the application and approach. Typical examples are microscopic inspections using metallography (Ills. 17.3.2-5 and 17.3.2-6) and scanning electron microscopes (SEM, Fig. "Scanning electron microscopy (SEM)").
If samples must be removed from the part, the process is generally destructive. If impressions (Fig. "Non destructive microscopic inspection") suffice or if the surface can be directly inspected without unallowably influencing the part, it is undoubtedly a non-destructive procedure.
All of these processes that can potentially be classified as either destructive or non-destructive are dealt with as destructive processes in this chapter.
Residual stress measurements (Table 17.3.2-1) can be non-destructive (on the surface using X-ray bending) as well as destructive (depth measurement using a bore hole), depending on the approach (Fig. "Measuring residual stresses and its interpretation").
Bond strength tests on coatings (Table 17.3.2-2) can usually only be conducted destructively; i.e. they require recoating following the test. Therefore, the normal approach is to use separate specimens that were coated at the same time as the part (Ills. 220.127.116.11-5 and 18.104.22.168-6).
Material characteristic values for the design of parts are determined using destructive tests on specimens. This includes the LCF strength of rotor components, as well as the thermal fatigue life and the creep resistance of hot parts. The HCF strength (dynamic fatigue strength in the range of fatigue resistance), on the other hand, is usually determined/verified on actual parts. This usually involves dynamic fatigue tests on blades that are made to resonate (Fig. "Axf testing dynamic strength").
Destructive procedures are generally used in the framework of quality assurance for scrap pattern testing, random testing of incoming goods (Fig. "Axf testing dynamic strength"), life span verification (Fig. "Integral technological testing"), rating tests (e.g. cyclical overspeed tests, Fig. "Cyclical spin tests of rotating parts"), as well as the testing of integral specimens (Ill. 17.3.2-1).
Destructive metallographic inspections of cross-sections (Ills. 17.3.2-5 and 17.3.2-6) are used in the context of scrap pattern testing and testing of incoming goods in order to ensure that the internal structure
of parts (e.g. forged blanks for rotor disks) conforms to specifications and does not have potential flaws, such as shrinkage cavities in cast parts. Inspections of very fine structural components, such as the g' phase of hardenable nickel alloys, can be done by examining specimens from the part with an electron microscope (REM, Fig. "Scanning electron microscopy (SEM)").
Figure "Testing on integral specimens": The incoming goods inspection of highly stresses series parts such as rotor disks involves taking specimens from excess material on the blanks that is specially provided for this purpose. The destructive testing of these specimens should be completed before the finishing process begins. Otherwise, it is possible that high finishing costs will have already been incurred and/or reordering of the raw parts is no longer possible by the time the incoming goods inspection reveals unallowable deviations. Because these tests can be very time-consuming, it is essential to budget sufficient time for them. Naturally, the testing specimen must be positioned on a part in a way that the specimen test results (usually LCF and/or creep; specimen diagram) are representative for the life-determining part zone. The left diagram shows an integral specimen ring that is separated from the raw part. The specified (integral) specimens are taken from this ring.
The specimen location and orientation are selected in such a way that the characteristic values that form the basis of the design are properly represented.
Figure "Integral technological testing": In many cases, it is not possible to avoid quality verification with the aid of technological testing. This is destructive and must therefore act as a representative for a larger production batch. A technological test is a test that simulates typical life-determining operating loads and damage mechanisms on a completed part. In this case, the test specimen may have technologies such as coatings that act in combination with the base material to influence the part behavior. In these situations, the operating loads and/or the interaction of the various technologies are so complex that they cannot be represented by simple specimens. The results are therefore qualitative, and suitable merely for comparisons with the desired behavior.
Typical examples of these tests include:
Cyclical overspeed tests (left diagram) with temperature gradients on coated spacers or rubbing tests on abradable coatings in housings.
Thermal fatigue tests on turbine stator blades that are coated with ceramic thermal barriers (right diagram). In this case, static tests of the finishing process are conducted by taking single parts from the series and subjecting them to defined thermal cycles. If the part reaches a minimum number of cycles without unallowable damage occurring, the relevant production batch is cleared for shipment.
Example 17.3.2-1 (Fig. "Axf testing dynamic strength"): During a finishing inspection, a large number of compressor rotor blades were subjected to an axf test. It was noticed that the blades had low axf values (dynamic strength), but there were no apparent anomalies that could explain this. There were no dimensional deviations outside of the drawing specifications. There were no explanatory differences in the surface quality (roughness, hardening, residual stresses) or material. A statistical analysis revealed that the blades in question behaved in apparently systematically subordinated ways relative to normal blades during the axf test.
- The resonant frequency (testing frequency) was somewhat lower (within one percent).
- The cracks that led to dynamic fatigue fractures were slightly closer to the root platform.
A very exact measurement of the blade geometry (this procedure was not series-typical) revealed a trend that clearly explained the unique test results:
The blades with the higher axf values had a slightly thinner profile at the tip and a somewhat thicker cross-section in the transition to the root platform (Eiffel tower effect). These blades then had a lower mass at the the tip and greater stiffness at the root. Both characteristics raised the resonant frequency of the fundamental flexural mode as it is used in the axf test. The different locations of the dynamic fatigue cracks could be explained by the lower loads and a more even transition to the root platform.
These realizations led to the blade geometry with greater dynamic strength being specified in the drawing requirments.
Figure "Axf testing dynamic strength": There are some parts, such as blades, from which it is not possible to remove satisfactory representative specimens for destructive testing, especially dynamic fatigue strength testing in the HCF range (Volume 3, Ill. 22.214.171.124-6). This is because, for example, the most highly-stressed area of the blade surface (i.e. the transition of the blade to the root platform) cannot be sufficiently simulated by a specimen. In this case, finishing parameters and material characteristics (e.g. forged structure with grain orientation and grain size) play an important role that cannot be sufficiently realistically reproduced on a “standard” HCF specimen. This is clear when one considers the relatively small bending radius of a round specimen or the different stiffnesses under the influence of chipping forces. These affect decisive hardening and residual stress effects.
For this reason, a more common procedure is to statistically select parts in the context of a incoming goods inspection of compressor blades and to subject them to an axf test (Volume 3, Ill. 126.96.36.199-9). This test at resonance not only permits a realistic verification of sufficient HCF strength. Unusual crack positions in the profile and/or in radial directions can be indications of production deviations or material anomalies. The resonant frequency also permits conclusions regarding deviations in the complex blade geometry (bottom frame).
Figure "Cyclical spin tests of rotating parts": When a engine part is certified, it undergoes procedures that verify that all following series parts will have the determined properties. This is only possible if the finishing process steps that were optimized during the development phase are sufficiently documented and safely reproduced. The prerequisite for this is so-called stable finishing processes. In other words, the deviation of process parameters must be sufficiently safely prevented. If changes are necessary for reasons of workflow rationalization or in the context of technical improvement, extensive certification work may be required for safety-relevant parts.
A typical example is rotor disks (right diagram). Even changing the raw part supplier or changes in the raw part production process (e.g. melting) require renewed verifications such as cyclical overspeed tests. These determine the safe operating load cycles (left diagram) that serve as the basis for certification. The same is true for seemingly harmless changes in finishing. These include changes to process parameters. Even the use of a new type of machine tool, new cutting media (e.g. from hard metal to ceramic), or cutting lubricants can necessitate the “verification process”.
Table 17.3.2-1: Residual stresses (Chapter 188.8.131.52) are induced (e.g. chipping, hardening through particle blasting, coating, welding) and/or altered (e.g. material-removing processes such as ECM or etching, stress reduction annealing) by many different finishing processes. Because residual stresses influence the dynamic fatigue strength of the parts and therefore also their operating safety, they can be seen as important quality indicators. Residual stresses must therefore not deviate unallowably from the specified and certified part conditions. Unfortunately, even today there is no satisfactory non-destructive measuring method that can encompass the entire volume of complex parts.
One of the most common measuring methods (Fig. "Measuring residual stresses and its interpretation"), due to its non-destructive nature, is X-ray inspection (“1”). Its greatest handicap is that it only measures the micro residual stresses in a very narrow surface layer. These measurements are useful in the context of process optimization (e.g. chipping, hardening).
Destructive measuring processes include those that require material removal (cross-section) to change the residual stress conditions (breakdown). This is used to determine the macro residual stress conditions. A small part zone outfitted with strain gauges can be removed for this process (“2b”). A more elegant and detailed method, which can determine the residual stress patterns in the part, is the borehole method (“2a”).
Neutron diffraction (“3”) has provided indications that it may be possible to measure the macro residual stresses in the entire volume of a part (e.g. a turbine disk). However, this procedure is still very cost- and work-intensive. For this reason, the procedure is only used in exceptions with very high potential risk.
Because residual stress changes (increases or decreases) lead to warping of the part, this can be used to draw conclusions regarding the residual stresses. This principle is used in “Almen intensity measurement” (“4”, Fig. "Blasting (Almen) intensity") in the framework of peening for hardening. The object of measurement is the bending of separate flat metal specimens that were sufficienly free of residual stresses before treatment. These undergo the hardening process along with the part. Electroplated coatings can have very high dangerous tensile residual stresses. In order to control these, process optimization and control is done with the aid of measuring procedures (“5”) that use separate specimens to determine the residual stresses during the coating process. In this case, as well, the deformation of a thin metal cross-section (Fig. "Measuring in situ residual stresses in galvanic coatings") under the influence of the residual stresses is used as the basis for measurement.
The measurement of the quantitative bond strength of coatings and adhesive connections is only possible using destructive methods in most cases. If measurement is done on separate specimens (“1”) it is non-destructive for the part. However, the information this provides is only indirect to the part or is at best suited for process monitoring purposes. In exceptional cases such as elastomer coatings ( “2”) it is possible to replace the area that was tested and found to be satisfactory. This could be called a “quasi-non-destructive” test.
In the case of thin, tough metallic coatings, direct testing on the part is possible with the aid of strength-increasing peening (“3”).
Direct, qualitative, comparative testing of bond strength on parts can be done through thermal cycles (“4”, Fig. "Integral technological testing"). These usually lead to destruction of the part and can aid statistical process monitoring.
In lacquers, cross-cut testing (“5”) is widespread. In this process, the lacquer coating is damaged by an oriented “scratch grid” and any delamination is analyzed.
Bond strength testing with ultrasound can be done in various ways. Echo testing can only be used to verify already separated areas (bond strength = 0) in suitable coatings (dense, sufficiently thick; “6”). In this case the entire coating surface can be tested insofar as it is accessible. This testing method can be considered non-destructive.
Table 17.3.2-2: The table contains bond strength tests with corresponding diagrams that are dealt with in various chapter.
A qualitative, locally limited indication of the bond strength of brittle ceramic coatings and/or porous coatings such as thermally sprayed thermal barriers or abradables can be obtained with the aid of ultrasonic vibrations. The sufficiently thin-walled part or specimen is intensively excited from the back side by an ultrasonic probe (“7”) until the coating separates.
A process that can relatively simply and non-destructively test the entire coated surface for local coating separations is thermography (“8”). With sufficient optimization, this procedure is simple and not overly expensive. The results can be photographically documented.
Figure "Damage analysis using metallography": Even in the age of electron microscope inspections (scanning electron microscope = SEM, Ill. Bild 17.3.2-7) metallography in the finishing area is essential. In this case, especially, technical expertise is required in order to obtain accurate results. Without this, there is a risk of mistaken and/or unusable results and/or interpretations (Fig. "Metallography and SEM").
The required so-called metallographic cut generally requires the removal of samples from the part (“1”, top frame). In this case, it is destructive process. If, however, an impression (Fig. "Non destructive microscopic inspection") is sufficient for indirect inspection (e.g. structure), then the process can be seen as non-destructive. This is also true for “ambulant” inspections on parts with accessible surface areas (polishing, etching, visual evaluation) using a portable microscope.
Typically, metallographic inspections proceed as follows. First, suitable testing samples must be removed from the part with sufficient care and on the basis of technical expertise (Fig. "Metallography and SEM"). In this case, it is important that the position of the samples is representative of the problem. Damages such as overheating or plastic deformations due to the separation process, which can influence the inspection results as artifacts, must be avoided. The samples are then embedded into plastic (“2.1”) in preparation for the following required steps (middle frame). The samples are then polished (“2.2”) and undergo material-specific etching to “develop” their structure (“2.3”). The analysis of the cross-section (bottom frame) is done under a microscope with incremental magnification up to about 1000x. If flaws such as cracks or pores are the focus, then the polished state is first examined microscopically. After this, the cross-section is etched in accordance with the questions asked (structure, coating). In this process, the different structural components are removed in different degrees depending on their type and location. The result is an appearance that can be evaluated by specialists, for example revealing the grain sizes and grain orientations.
Therefore, metallographic inspections can be used to relatively simply and accurately identify, verify, and provide long-term documentation for changes andcharacteristics in the parts.
The many possible uses of metallography include especially:
- Typical structural characteristics: Used, for example, to monitor the quality characteristics of raw parts and blanks.
- Testing for flaws such as casting porosity.
- Process-specific characteristics such as the thickness and structure of coatings. Changes (such as following diffusion processes) and damages (e.g. cracks: size, pattern, location). The findings permit conclusions regarding the creation mechanisms and causes.Hardness patterns in the micro-range; e.g. in surface hardening (case hardening, nitriding).
Figure "Metallography and SEM": Metallographic inspections form a core component of quality assurance and damage analysis. They aid process monitoring (e.g. coating evaluation) and contribute important information to problem analyses (Fig. "Systematic analysis of problems i"). For this, especially, it is vital that misinterpretations are avoided. Obtaining the most accurate inspection results demands a suitable environment.
Microscopic inspections such as metallography and SEM inspections require teamwork, in this case primarily with the technically knowledgeable client. The client has the task of providing sufficient information regarding backgrounds that could aid a targeted investigation. This can ensure that the cross-section is taken from the proper location of the part and is at a suitable angle. These are prerequisites for practical results. Important information includes the meaning, goals, effects, previous history, characteristics for interpreting the findings, and the reasons for deadlines. This contribution is motivating in that it provides the inspecting personnel with motivation by giving clear meaning to the investigation/test and allowing them to understand the importance of their work. In addition, seemingly secondary findings such as structural anomalies or signs of plastic deformation can be better detected and evalueated. The client should ideally be present during the analysis of the cross-sections, and certainly during interpretable results and inspections outside of the routine (e.g. in the cases of deviations or damages). This does not mean that the client must necessarily have a technical understanding of microscopic analysis. On the contrary, an outside perspective may be helpful. This provides the opportunity to recognize causal influences from the surrounding environment. This ensures that there will be optimal information regarding aspects that may only be understood in their relationships and influences once the findings have been obtained. In this way, a necessary critical discussion is ensured. Naturally, the finaly analysis must be left up to the inspection personnel with their technical expertise. An important requirement for longer-term trusted cooperation is the ability to meet agreed inspection appointments. It is equally important that the client recognizes these. If tightly scheduled appointments are not observed or even forgotten, it sends a message that the inspection results are of a rather low priority, which can also affect later cooperation.
In a team, the investigator also has an important role above and beyond his technical work. He or she should encourage the client to take an interest in the evaluation of the cross-section or SEM inspection. One should avoid the use of “killer phrases” such as: “what (specific test result) do you expect?” In many cases, the answer will only be clear after the test. Experience has shown that in many situations, information gained during the evaluation is the catalyst for further explanatory and evaluatory developments.
The following addresses some typical potential problems related to microscopic inspections:
Realistic evaluations of findings (top left diagram): It is not easy to estimate the effects of findings on part behavior based on highly-magnified images. For example, flaws that are only a few mm long and not capable of growth under part-specific operating conditions may be classified as weak points. However, their appearance under a microscope may make them intuitively seem very dangerous and lead to their interpretation as unallowable flaws.
Findings and cross-section location (top right diagram): A prerequisite for problem-specific relevant inspection results is that the sample removal takes place in the problem zone of the part. This appears trivial, but experience has shown that it is not. If, for example, the grain size is to be determined in connection with LCF strength, it is essential that the life-determining part area is clearly defined and understood during sample selection. If necessary, the client may have to provide the applicable specifications. One example of this is rotor disks, which can have very different types of structures between the hub and annulus due to the forging conditions (Fig. "Finishing semi-finished parts").
Even if the location of the cross-section is correct, it is not always a given that the cross-section plane will be correct. This can determine the inspection results. If a crack is cut in its plane (from the side), the findings may be very similar to those of shrinkage porosity (cavities) and the crack
may be confused for the latter (Fig. "Metallography and SEM"). If the material is a cast material with a likelihood of this type of flaw, misinterpretations are almost preprogrammed.
An example is the friction weld seam of a cast Ni-based turbine disk with a steel shaft (Fig. "Flaws of friction welding"). Penetrant testing revealed flaws in the area of the weld seam on the casting side. A metallographic cross-section was made across the weld seam in an axial plane. In this process, the existing axial welding cracks were cut horizontally, leading to them being mistakenly classified as typical shrinkage cavities. The result was ineffective corrective measures in the casting process with high costs and considerable lost time. Only a repeated inspection using cross-sections in the proper plane was able to identify the indications as welding cracks.
Edge sharpness (bottom left diagram): The surface area is the focus of many microscopic inspections. Typical examples include process monitoring for:
- Coatings: thickness, structure, strength patterns
- Hardened zones: depth, strength patterns (Ills. 184.108.40.206-3 and 220.127.116.11-2)Also, flaws that can influence dynamic fatigue strength, such as:
- Damages from a chipping process: Comma cracks (Ill. 18.104.22.168-2), overheating from grinding, excessive blasting (Fig. "Problems by angular shot peening").If it is not possible to use a special technology (cutting, preparation) or adapted polishing processes to ensure sufficient sharpness of the edges, then unsatisfactory findings can be expected. A possible remedy is suitable metallic coating (Ni coating) before cutting the cross-section. Unsuitable edge preparation typically results in underestimation of coating thicknesses or undetected flaws.
Artifacts resulting from unsuitable technology when creating a cross-section (bottom right diagram): Sample removal from a part for a metallographic cross-section uses various separating processes. This is often done using cutting machines with rotating cutting disks. There is a danger that these could overheat the sample (e.g. structural changes, cracking) and/or confusing plastic deformations. Brittle materials (e.g. ceramics) and coatings (e.g. thermal barriers, Al diffusion coatings) tend to cracking and spalling. If these damages on the cross-section are not removed by the subsequent polishing steps, which is a possibility, then misinterpretations are likely. Another source of misinterpretations can be excessively intensive sample removal with the aid of electric discharge machining, which can distort findings. If the effects of the process-specific overheating, such as the recast layer and brittle cracking (Fig. "Electric discharge machining features") are not sufficiently removed by the polishing process, they can simulate flaws.
Even a metallographer needs motivation. This means that the client is responsible for:
- providing background information,
- explaining the importance of the case,
- providing information regarding anomalies
A prerequisite for optimal results from metallography and SEM testing is trustworthy cooperative evaluation of the specimens. In this case, the client may contribute background information.
Illustrations 17.3.2-7: Scanning electron microscopes (SEM, Refs. 17.3-11 and 17.3-12) are an essential tool in the inspection of part surfaces and fractures. For the client, they have the characteristics of a binocular with the potential of very high magnification (up to several 104) and an extreme depth of field. In addition, they make it possible to analyze compositions in the micro range. This can, for example, permit the identification of damaging fouling (Fig. "Damaging by foreign materials on parts"). SEM inspection does not compete with metallography, although the relationship between the two is often viewed this way. Instead, the two process can complement each other very effectively. For example, structural components such as the g'-phase, which cannot be recognized under light optics, can be analyzed by a SEM using metallographic cross-section with suitable etching. On the other hand, SEM findings can often only be safely confirmed and explained by an additional metallographic cross-section.
Even more than in the case of metallographic inspections (Fig. "Metallography and SEM"), SEM inspections require close cooperation between the investigator (operator) and a client with relevant technical expertise. Experience has shown that in the context of problems and damages, maximum understanding can only be expected if the client witnesses the investigation. The outstanding spatial image qualities of SEM microscopes can convey a realistic impression and evaluation of the situation even to non-specialist observers. This presents an opportunity to incorporate important background information, the importance of which may only be recognized in connection with the findings.
SEM inspections can also be done on impressions (Fig. "Non destructive microscopic inspection") insofar as the surface topography is significant. This makes the procedure non-destructive.
Illustration 17.2.3-8: Metallographic inspections and SEM inspections can be non-destrucive if it is possible to make and analyze impressions of the parts. For metallographic inspections in which the primary task is evaluating the structure, cross-section (polishing, etching) is done on the part. This can usually be done non-destructively.
For SEM inspections, the impression is usually taken from the unprepared, only sufficiently cleaned surface. The non-conductive impressions must then be metallized for the inspection through vapor deposition. This is a routine process in SEM inspections.
Three methods for making impressions have been shown to be especially effective:
Impressions with synthetic resins that are poured onto a previously delimited (plasticine) part zone and allowed to harden. These synthetic resin impressions are relatively hard and brittle and not suitable in cases with positive fitting.
Impressions can be made of surfaces with complex shapes. The prerequisite for this, however, is that there is no positive fitting, such as special surface roughness or burrs, which prevents the intact removal of the impression. During this process, it is important to avoid any wearing movements that could damage the surface being inspected in the micro range. If the imprint must be made in a positive fitting part area (e.g. fir tree root), it is possible to use a hardenable, rubber-elastic elastomer (usually silicon rubber). After hardening, the impression can be carefully removed by taking advantages of its elastic properties.
Foil impressions can be made anywhere where a sufficiently flat surface is available. The purchasable synthetic foils are “softened” in a supplied solution and pressed onto the part surface, where they quickly harden. After this, they are pulled off. This method can be used to take impressions of very fine structures, such as for analyzing material structure.
Impression techniques have proven themselves for the evaluation of deviations in expensive parts. They can be used to test or verify uncertain results from non-destructive testing methods such as eddy current testing or penetrant testing.
When reworking flawed areas periodic inspections with the aid of impressions can be used to determine whether the damage was completely eliminated. Examples include small melted areas from arc burning (Fig. "Dynamic fatigue lowered by electric arcs") or damages related to tool fractures (Ill. 22.214.171.124-2).
17.3-1 P.Adam, “Fertigungsverfahren von Turboflugtriebwerken”,Birkhäuser Verlag, 1998, ISBN 3-7643-5971-4, pages 193-221.
17.3-2 D.M.Walker, R.Y. Hom, “Residual Stress Analysis of Aircraft Aluminium Forgings”, periodical “Advanced Materials & Processes” , June 2002, pages 57-60.
17.3-3 J.Horowitz, “Das `Shot-peening'-Verfahren”, periodical “Metalloberfläche” 32 (1978) 7, pages 285-292.
17.3-4 R.E.Green Jr, “Non-Destructive Methods for the Early Detection of Fatigue Damage in Aircraft Components”, AGARD Lecture Series No. 103 “Non-Destructive Inspection Methods for Propulsion, System and Components”, 23-24 April 1979, London, UK, and 26-27 April Milan, Italy, pages 6.1-6.31.
17.3-5 K.Nitzsche, “Schichtmesstechnik”, VEB Deutscher Verlag für Grundstoffindustrie, Leipzig.
17.3-6 W.Dettner, I.Elze, “Handbuch der Galvanotechnik”, pages 258-295.
17.3-7 A.K.Koul, R.V.Dainly, “Fatigue Fracture of Aircraft Engine Compressor Disks”, “Handbook of Case Histories in Failure Analysis”, pages 241-250.
17.3-8 W.D.Rummel, P.H.Todd Jr.,R.A.Rathke, W.L.Castner , “The Detection of Fatigue Cracks by Nondestructive Test Methods”, periodical “Materials Evaluation”., 1974, No. 32, pages 205-212.
17.3-9 D.S.Forsyth, A.Fahr, “The Sensitivity and Reliability of NDI Techniques for Gas Turbine Component Inspection and Life Prediction”, NRC/IAR Report, LTR-ST-2055, July 1996.
17.3-10 N.G.H. Meyendorf, P.B.Nagy, S.I.Rohklin, “Nondestructive Materials Characterization with Applications to Aerospace Materials”, Springer-Verlag, 2004, Springer Series “Materials Sciences” 67, ISBN 3-540-40517-8.
17.3-11 ASM, “Metals Handbook Ninth Edition, Volume 11, Failure analysis and Prevention”, 1986, ISBN 0-87170-007-7, pages 11-71.
17.3-12 L.Engel, H.Klingele, “An Atlas of Metal Damage”, Wolfe Science Books/Carl Hanser Verlag Munich, 1981, ISBN 0-7234-0750-9.