0 valutazioniIl 0% ha trovato utile questo documento (0 voti)
11 visualizzazioni8 pagine
A new method has been developed for detection and quantification of fretting fatigue damage in gas turbine engine alloys. Profilometry data taken by white light interference microscopy was used to evaluate the likelihood of crack initiation in fretted surfaces.
A new method has been developed for detection and quantification of fretting fatigue damage in gas turbine engine alloys. Profilometry data taken by white light interference microscopy was used to evaluate the likelihood of crack initiation in fretted surfaces.
A new method has been developed for detection and quantification of fretting fatigue damage in gas turbine engine alloys. Profilometry data taken by white light interference microscopy was used to evaluate the likelihood of crack initiation in fretted surfaces.
E.B. Shell * , D. Eylon ** * Center for Materials Diagnostics, University of Dayton, Dayton, OH 45469-0121, USA ** Graduate Materials Program, University of Dayton, Dayton, OH 45469-0240, USA ABSTRACT A new method has been developed for detection and quantification of fretting fatigue damage in gas turbine engine alloys. Profilometry data taken by white light interference microscopy was used to evaluate the likelihood of crack initiation in fretted surfaces. Crack initiation in fretting samples was preceded by two precursors. First, an increase in the surface roughness in the slip region followed by an increase in the population and power of closely spaced asperities. The combination of these two factors has led to a detection and quantification method capable of differentiating between the slip and stick regions in a fretted component. A Fretting Fatigue Damage Parameter (FFDP) has been established that can be measured and used to assess the localized degree of damage of a component. When used in cooperation with a life prediction methodology, the FFDP may be used to calculate residual life in laboratory samples. In addition, the FFDP may be used in field inspections during maintenance and as a criterion for removal of parts from service. This could lead to the development of more reliable depot nondestructive evaluation (NDE) techniques for fretting fatigue prone components. 1. INTRODUCTION Fretting fatigue has long been recognized as an important mode of failure in the service of mechanical components that are not intended to have a relative motion, but because of vibrational loads or deformations experience minute cyclic relative motions [1]. Such damage can result in a considerably lower component life. Fatigue strength under fretting fatigue conditions may be as low as 30% of the plain fatigue strength, depending upon the material and loading conditions [2]. This increased rate of damage accumulation, along with the unpredictability of the actual conditions experienced in use, necessitate a means for non- destructive examination of the surface and prediction of the remaining service life. Fretting fatigue is a surface and near-surface phenomena. Thus, surface NDE techniques may be particularly useful in the detection of fretting fatigue damage. However, this may be complicated by the initial surface conditions of the material. Because fretting damage is a very localized phenomenon, techniques developed to detect fretting damage must have a reasonably high spatial resolution. Although much research has been devoted to the understanding of fretting mechanisms, no characterization methodologies have been established for quantifying the level of fretting damage. E.B. Shell, D. Eylon, A New Method for the Detection of Fretting Fatigue Damage, LiMAT-2001, 2nd International Conference on Light Materials for Transportation Systems, Pusan, Korea, May 2001 1.1 Need for an Improved Fretting Damage Detection Method Currently, the procedure for the inspection of fretting fatigue critical components (e.g., as turbine engine fan blades and disks) is simply a visual inspection for the presence of any large surface irregularities during depot inspections. This led to the removal of many blades and disks either: (1) before there was a substantial amount of fretting damage, or (2) due to less dangerous surface deformation such as sliding wear. In addition, this criterion has been proven ineffective when compared to laboratory-tested samples. In fact, laboratory-testing conditions frequently lead to failure in samples where the roughness of the surface is less than 3 m, well below what could be called a large surface irregularity during inspection. A new method that would allow for a better understanding of the fretting process is, therefore, needed. This may be achieved through observation of the inherent changes that may be universally observed, leading to the ability to quantify fretting damage and calculate residual component life. 1.2 Concept of Precursor Damage Detection Because fretting fatigue is a surface and near-surface phenomena, surface NDE techniques, particularly optical techniques, lend themselves to the detection and characterization of fretting damage. Most surface techniques are specifically designed for the detection of surface cracks. However, because high strength materials, such as titanium and nickel-based alloys, have relatively low fracture toughness, methods that are used to detect cracks are inherently unsafe and require frequent inspection. In fact, some high strength materials may not form a detectable crack before 95-99% of life has expired. Instead of crack detection, it may be more beneficial to assess the material on the basis of the likelihood of or potential for crack initiation. This may be accomplished by evaluating the condition of the surface through such parameters as roughness and asperity spacing. These parameters relate, on a microscopic level, the material surface finish to the very localized stress concentration at the bases of the asperities geometry and spacing. It is postulated that through fretting action, the morphology of these asperities changes. This change allows, through an analysis of the roughness and the spacing between adjacent asperities, an inference to be made as to the condition of the material and the changes that have occurred due to the vibrational loading of the sample. The fretting damage level may then be evaluated through profilometric observations and analysis based on the non- destructive characterization of the asperities on the surface. 1.3 Fracture Mechanics Basis for Profilometric Damage Criteria The theoretical basis for a profilometric description of fretting damage accumulation lies in the fracture mechanics that may be used to describe the stress field, on a microscopic level, near the surface of the material. An equation that may be used to describe the stress intensity at the front of a semi-elliptical flaw on the surface of a sample is: K (a/), where is the applied stress, a is the length of flaw, and is the radius of curvature of the elliptical flaw. In a two-dimensional analysis of asperities on the surface of a sample this equation may be applicable. In this case, we cannot accurately measure the radius of curvature of the asperity bases. However, we can make an inference into the curvature at the bases of the asperities based upon the spacing between asperities. The average radius of curvature between adjacent asperities is then directly proportional to the average spacing between asperities. For every sample there will be a continuous range of asperity spacing present on the surface of the material. However, the range of most interest is that of a low asperity spacing that would lead to a smaller radius of curvature and, thus, higher localized stress intensity near the surface. As for the investigation of the asperity height, it is impractical to analyze each individual asperity. But, we can use the calculated roughness of given area of analysis to provide a statistical inference of the average asperity height, which is directly proportional to the surface roughness of the material. Combining the asperity spacing and surface roughness over a given area leads to a better understanding of crack initiation from fretting fatigue-related surface damage. 2. MATERIALS AND PROCEDURES 2.1 Materials The forging stock for the material used on this work was a double VAR melted Ti-6Al-4V 63.5mm diameter bar stock from Teledyne Titanium produced in accordance with AMS4928, and supplied in mill-annealed condition: 705C/2 hr/AC [3]. Chemistry was acceptable in accordance with AMS4928 as shown in Table 1, as well as the beta-transus temperature as determined by DTA. Table 1: Chemistry of Teledyne Titanium Heat No. TE01 Element Ti Al V Fe O N Transus C Top bal. 6.27 4.19 0.20 0.18 0.012 990 Bottom bal. 6.32 4.15 0.18 0.19 0.014 1003 AMS4928 bal. 5.50-6.75 3.50-4.50 0.30 max 0.20 max 0.050 max ------- The 63.5mm diameter bar stock was cut into 200 400mm long forging performs forged in one campaign in a 400mm long closed-end channel-die, with the intended plate size of 400x150x20mm on an 8,000 ton mechanical press. Dies were initially heated to 150C. Glass-lubricant coated bars were preheated to 940C +/- 10C for 30 minutes in a continuous furnace and rapidly transferred to the press. After a one stroke forging, pieces were air- cooled. The average strain rate at impact (for plain strain case) was calculated to be 9/sec. Forging was followed by a vacuum anneal at 700C/2hr to stabilize microstructure and normalize hydrogen content that might have been picked up during de-scaling, and followed by 930C/1hr ST and another 700C/2hr vacuum anneal. This practice is similar to the one used for forging gas turbine engine fan-blades, and resulted in a duplex microstructure with 60vol% equiaxed primary alpha (Fig.1) with good fatigue strength. Tensile results of one plate at 5x10 -4 (s -1 ) are shown in Table 2. The higher modulus in the transverse direction indicates processing texture. Table 2: Room temperature tensile results of plate #7 Location Orientation UTS (MPa) YS (MPa) El (%) Modulus (GPa) Center-long L 976 929 21.2 116 Center-long L 981 931 19.0 124 center-trans T 997 947 21.0 121 center-trans T 993 937 19.2 134 2.2 Mechanical Testing Details of the fretting fatigue procedures are provided elsewhere [4]. In brief, the test system is an axial fatigue test machine in which the gripping system allows the development of a slip region on the samples surface. The samples measure 100 mm in length, 10 mm in width, and 2 mm in thickness. The sample is clamped at each end by two flat pads (25.4 mm in length), also machined from the same Ti-6Al-4V plate. When a cyclic stress is applied, a slip and stick zone is developed on the sample. The samples were tested using a variety of surface finish conditions. These included samples that were finished with a RMS#8 finish and samples that were carefully polished, both tested with Ti-6Al-4V fretting pads. 2.3 Profilometry White light interference microscopy was used to measure the surface topography before and after the samples were subjected to load. This fast technique is capable of a lateral surface resolution of 0.2 m and a vertical/height resolution of 3 nm. The details of the instrument are provided elsewhere [5]. From the surface height maps, two evaluations of the data are performed. (a) A Fourier transform of the surface from which the Power Spectral Density (PSD) can be plotted against Spatial Frequency is calculated. It has been postulated that fretting fatigue cracks initiate when the surface features reach a critical spatial frequency. As the concentration of high spatial frequencies in the surface increases, cracks are able to initiate easier. (b) The roughness of the sample is evaluated; either quantitatively by a roughness calculation or qualitatively by viewing a line-scan of the sample surface. From the profilometry data, the inherent spatial frequencies of the surface can be investigated. Looking at the asperity spacing gives some insight into the likelihood of crack initiation. So, instead of looking at the data in the spatial domain, it would be more useful to investigate the spacing of the surface asperities by transforming the data into the frequency domain. This enables the observation of local decreases in the asperity spacing that are present in the slip region of fretting samples. The PSD-spatial frequency plot is generated by a Fourier Transform of the 3-dimensional height data. This effectively transforms the data from the space domain to the frequency domain, which enables an easier evaluation of the inherent frequencies of the data set. Because the sample surface is nominally flat, the maximum PSD occurs as the spatial frequency approaches zero, which represents a perfectly flat surface. The higher range of spatial frequencies corresponds to small, closely spaced asperities. This PSD plot represents a spatial frequency spectrum over the entire field of view Fig.1 Ti-6Al-4V microstructure used in this study. of the sample. So, when using this data to determine the presence of slip regions, care must be taken to ensure that the sampling area and the reference area are uniform. If this is not the case, the PSD levels for the data set will only be an average of the entire field. Likewise, when determining fretting damage levels in a given area, the percentage of slip versus stick regions must be taken into account to determine the true degree of damage. 3. RESULTS AND DISCUSSION The profilometry data resulted in two types of information; (a) the power spectral density, which is the Fourier decomposition of the measured surface into its component spatial frequencies, and (b) the surface roughness. While an increase in surface roughness due to asperity yielding and microwelding is often associated with fretting fatigue damage [6], the samples used in this study already had a machining roughness on the order of that expected due to fretting wear. So, while a slight increase in roughness was seen, more promising results are shown based upon the PSD data. 3.1 PSD-Spatial Frequency Analysis Profilometric data has been proven effective in the qualitative detection of fretting fatigue and fretting wear. By taking profilometry data from non-contacted, slip, and stick regions of the same sample, differences in the Power Spectral Density (PSD) of the surface data may be seen. PSD-Spatial Frequency plots, such as that shown in Fig. 2, have shown that regions that are damaged by fretting action contain a higher PSD value at higher spatial frequencies than areas subjected solely to fatigue (non-contact region) or areas that were in the stick regime. As can be seen, at very low spatial frequencies, all regions have high PSD values, indicating that all regions are nominally flat. However, as higher spatial frequencies are considered, an order of magnitude difference is seen between the slip and non-contact regions. It is postulated that high spatial frequency asperities may be associated with fretting fatigue crack initiation. So two related conclusions may be formed: (a) an area that has been fretted will have a higher density of high spatial frequency asperities, and (b) an area that has a Fig.2 PSD data for the partial slip, stick, and non-contact regions of a fretting fatigue laboratory sample. higher density of high spatial frequency asperities is likely to initiate cracks sooner. This is justified by imagining each area between asperities as a stress concentration area. As the frequency of the asperities gets higher, the average radius of curvature of the valleys between the asperities decreases, thereby increasing the localized stress concentration factor at the surface. These characteristics enable the detection of possible crack initiation sites at the surface of the material through spatial frequency analysis before crack initiation occurs. The detection of localized fretting damage can be seen in the following example. A scan of the surface of a laboratory fretting fatigue sample is taken. The fretting scar is clearly visible in the data shown in the background of Fig. 3. The data set is then separated into subsets as shown by the gridlines. A PSD plot is generated from each of these regions. To better compare the surface from stick, slip, and non-contacted regions, the PSD is compared for discrete spatial frequencies (0.5, 1.0, 1.5, and 2.0 m -1 ). This enables a relative comparison between adjacent regions in the sample. The data clearly shows that the slip region may easily be detected through profilometry measurements. The PSD of the slip region is 2.5 to 7 times higher than the non-contact region, decreasing as the measurement is obtained back into the stick region. 3.2 Fretting Fatigue Damage Parameter Beyond using PSD data for the detection of fretting fatigue damage, PSD analysis is promising in the quantification of fretting damage. This quantification is accomplished by observing the relative differences between the fretted region and the region subjected solely to plain fatigue. As seen in Fig. 4a, scans taken from samples of similar damage levels are not identical, possibly the result of slightly different initial surface conditions. However, by normalizing the fretting data with respect to data from the non-contact region (Fig. 4b), a Fretting Fatigue Damage Parameter (FFDP), is derived: FFDP = PSD slip / PSD NC . The deviation of this parameter above unity is representative of the damage level of the area examined and may be used to quantify the damage. It is thought that this parameter relates directly to the likelihood of crack initiation relative to the starting condition of the material. Fig.3 PSD data plotted at discrete spatial frequencies. The slip zone can be seen in the background of the chart. No Contact Slip Stick 3.3 Quantification of Damage and Residual Fatigue Life Predictions The development of a refined quantification tool based upon PSD analysis presents a number of challenges. The primary difficulty in developing a calibrated technique for the prediction of residual life in components is the inability to actually determine the true fractional life of a specific area. Because most fretting fatigue experiments do not yield uniform damage across the slip line, there is a question as to what damaged area should be considered for determination of the overall sample residual life. In fact, due to non-uniform damage, only the area that is the most damaged will ultimately initiate the final failure and should be considered at the nominal fractional life. This leads to another obstacle to making quantitative measurements for life prediction purposes: the difficulty involved in using an area analysis to search for a relatively localized phenomenon. In any sample or component subjected to fretting fatigue, generally only a small region is subjected to the most severe conditions. As a result, manually searching for this most damaged area is extremely time consuming. However, a methodology based upon these techniques could implement computer analysis of a large image to: first, break up the image into smaller, discrete regions; second, perform a Fourier transform on each of these smaller regions; finally, through comparison of the PSD curves determine the region representative of the highest degree of damage. This method has been (roughly) performed manually for one sample. In this case, the sample was scanned along the leading edge of the slip line as seen in Fig 5. The resulting data, shown for a few spatial frequencies, shows that the FFDP is highest within 2 mm of both edges of the sample. This increased level of damage near the edge corresponds to the location of crack initiation as seen in failed test samples and should be used for residual life calculation [7]. 4. CONCLUSIONS The objective of this work was to develop a new tool for the detection of fretting fatigue damage in Ti-6Al-4V. In addition, characteristics of fretting fatigue damage were investigated in order to develop damage quantification techniques. From this work, the following conclusions were made: Fig.4 (a) Data taken from the slip and non-contact regions from two sides of a specimen. Both sides were subjected to identical loading conditions but experienced slightly different machining conditions. (b) Data from two slip regions normalized with respect to the non-contacted surface data. Both sides exhibit identical degrees of damage. (a) (b) 1. PSD-spatial frequency analysis has proven effective in the detection of fretting fatigue damage. Differentiation between the slip, stick, and non-contact regions is possible with this NDE technique. 2. The surface roughness increases in the slip zone compared to the non-contact and stick zones. However, the machining lines and pre-existing roughness of the samples prevent this measurement from being a reliable method of slip region detection. 3. Initial attempts at quantification have proven difficult due to the extremely localized nature of failure initiation due to fretting fatigue. The technology is available that would permit this method to be used as a NDE tool. However, further refinement and computerized data analysis of the damaged region must be utilized for this method to be a mature, reliable damage quantification technique. 4. A Fretting Fatigue Damage Parameter (FFDP) has been developed that may be used to quantify the level of damage in a component. ACKNOWLEDGEMENTS Effort sponsored by the Defense Advanced Projects Agency (DARPA) Multidisciplinary University Research Initiative (MURI), under Air Force Office of Scientific Research grant number F49620-96-1- 0442. The cooperation of A. Hutson, Dr. T. Nicholas, Dr. S. Mall, Prof. M Niinomi, and Prof. S. Gustafson in providing samples for examination is also greatly appreciated. REFERENCES [1] J.A. Collins, Failure of Materials in Mechanical Design. John Wiley & Sons, NY, 1981. [2] R.A. Antoniou, T.C. Radtke, Mat. Sci. Eng. A237 (1997) 229-240. [3] D. Eylon, "Summary of the Available Information on the Processing of the Ti-6Al-4V HCF/LCF Program Plates"; UDRI report, October 1998, UDRI, University of Dayton, Dayton, Ohio. [4] A.L. Hutson and T. Nicholas, Int. J. of Fatigue 21 (1999) 663-669. [5] J.L. Schroeder, D. Eylon, E.B. Shell and T.E. Matikas, in Nondestructive Methods for Materials Characterization, MRS, Fall Meeting, Symposium S, (1999) Paper No. S2.3. [6] O. Vingsbo and S. Soderberg, Wear 126 (1988) 131. [7] A. Hutson, unpublished research conducted at Wright-Patterson Air Force Base, 1999. Fig.5 Variation of the FFDP across the slip region of a fretting sample. The schematic shows the location of the scans in the slip region.