Nondestructive Inspection Method for Jet Engine Turbine Blades

Transcription

1 73-GT-92 $3.00 PER COPY $1.00 TO ASME MEMBERS The Society shall not be responsible for statements or opinions advanced in papers or in discussion at meetings of the Society or of its Divisions or Sections, or printed in its publications. Discussion is printed only if the paper is published in an ASME journal or Proceedings. Released for general publication upon presentation. Full credit should be given to ASME, the Professional Division, and the author (s). Copyright 1973 by ASME Nondestructive Inspection Method for Jet Engine Turbine Blades I. R. KRASKA General American Research Division, General American Transportation Corp., Niles, Ill. W. L. BERNDT Oklahoma Air Material Area, Tinker Air Force Base, Okla. Nondestructive inspection of jet engine turbine blades during field and depot maintenance is accomplished normally with fluorescent penetrant and/or visual techniques. In spite of the widespread use of the penetrant process, it has many disadvantages as a maintenance inspection method. The reliability of penetrant inspection depends upon the preparation of the blades prior to actual penetrant processing and the inspector's skill in detecting and evaluating defect indications. The process can miss cracks if they are filled with material that blocks the penetrant from entering the defect. An eddy current technique which can detect leading and trailing edge cracks in turbine blades and an instrument based upon this technique were developed under Air Force sponsorship. The instrument has been tested in a rework facility and field maintenance. Results of the evaluation and photomicrographs of typical cracks detected in turbine blades are presented. Results are compared, in some instances, to results obtained with the fluorescent penetrant and/or visual examinations. A field penetrant inspection of leading trailing edges detect cracks at an inspection rate of one stage in 1 hr and 50 min. Contributed by the Gas Turbine Division of The American Society of Mechanical Engineers for presentation at the Gas Turbine Conference and Products Show, Washington, D. C., April 8-12, Manuscript received at ASME Headquarters January 31, Copies will be available until February THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS,- UNITED ENGINEERING CENTER, 345 EAST 47th STREET, NEW YORK, N.Y

2 r Nondestructive Inspection Method For Jet Engine Turbine Blades I. R. KRASKA. W. L. BERNDT In recent years, a need has developed for a good non-destructive inspection method to find cracks in the leading and trailing edges of turbine blades. Some in-flight failures of blades have occurred with disastrous results. Non-destructive inspection (NDI) of turbine blades in the field is normally limited to a visual examination. Such an evaluation of blade integrity is subjective. It is affected by variation from operator to operator. An objective technique for blade evaluation in the field is necessary. During overhaul periods, the cracks are commonly detected using fluorescent dye-penetrant inspection methods. In spite of the widespread use of the penetrant process, it has many disadvantages as a maintenance inspection method. The reliability of penetrant inspection depends upon the preparation of the blades prior to actual penetrant processing and the inspector's skill in detecting and evaluating defect indications. The process can miss cracks if they are filled with material that blocks the penetrant from entering the defect. In an effort to provide a solution to this inspection problem, the Non-destructive Testing and Mechanics Branch and Aeronautical Systems Support Branch of the Air Force Materials Laboratory co-sponsored a program to select and develop an improved non-destructive inspection method for the examination of turbine blades. 1 An eddy current technique was developed to inspect turbine blades without removing them from the rotors. In this paper, an analysis, the results of laboratory evaluation and field tests, and resulting technique modifications are discussed. EDDY CURRENT ANALYSIS current is brought near a conductive metal, eddy currents are induced in the metal by electromagnetic induction. The eddy currents generate a secondary magnetic field that, in turn, opposes the first field produced by the coil. It is the effects produced by the secondary field which provides information about the metal. Significant changes to the nature of the metal will modify eddy current flow detected by the coil. One of the most obvious metal properties affecting the z AIR FERRITE \\\.\\N:\ 0 GRAPHITE \\1/4\ N N N \\ N CtRAPHITE Al -4 Background When a coil of wire carrying alternating 1Kraska, I. R., and Kamm, H. W., "Eddy Current Inspection of Turbine Blades," Technical Report AFML-TR , June COIL RESISTANCE Fig. 1 Complex impedance plane of a typical eddy current probe coil adjacent to thin metal plates 2

3 ,... SIMIN MIER.. Resistance MI Air SS 304 Bronze Aluminum Air Turbine Blade Material Fig 2(a) Oscilloscope display of an eddy current probe output complex impedance of several common metals Fig. 2(b) Oscilloscope display of eddy current probe output of a new turbine blade 0 a M a k Blade Good Turbine Blade Edge Fig. 2(c) Expanded curve and change in impedance when the probe intercepts a crack Fig. 2 Oscilloscope display of an eddy current probe output eddy currents flowing in the material is the electrical conductivity. Other parameters affecting the flow of eddy currents are magnitude and frequency of the induced field, material chemistry, magnetic permeability, shape of the specimen, discontinuities and inhomogeneities present. When inspecting a sample for cracks or other properties with eddy currents, the measurements are made with a search probe. No other connections are made to the sample. All information must be obtained from measurements at the coil itself. The easiest coil measurement is coil voltage. Such voltage changes are influenced by the changes in impedance of the coil. Impedance can be broken down into two terms known as reactance and resistance. If we graph the changes of reactance and resistance for a search coil when we vary conductivity, we obtain a curve shown in Fig. 1. This diagram shows these components of the impedance which different metals offer the eddy current flow. The heavy solid curve represents the locus of specimen conductivity values. The lighter spiraled curve represents the locus of specimen thickness values, and the dashed line represents the locus of lift-off values. The diagram answers the question: What happens to the reactance and resistance when a specific type of change occurs in the test specimen? This diagram shows several important facts: (a) increasing lift-off (probe to sample spacing) or decreasing conductivity causes the coil impedance to approach that of air; (b) the coil impedance locus of changing lift-off and changing conductivity are different, separating at the angle, 43 ; and (c) the impedance of magnetic materials (ferrite, etc.) are very different from the impedance of non-magnetic materials. Since the cracks in a sample have an effect 3

4 sigmo Ii... Air Crac k / Crack Blade 0 j Angle lade 0 --B Resiatance Resista - Fig. 3(a) Impedance plane curve including crack indication Fig. 3(b) Impedance plane curve showing lift-off locus in the horizontal direction, crack indication Resistance Fig. 3(o) Oscilloscope display of crack with lift-off locus in the horizontal direction, crack indication Fig. 3 Oscilloscope recordings of impedance plane curve showing lift-off locus, separation angle, and crack indication on coil impedance comparable to a combination of lift-off and conductivity variation, measurement of coil impedance will provide an indication of crack presence. Turbine Blade Considerations Turbine blades have low electrical conductivities (typically 1 to 2 percent IACS), complex geometries, rough surfaces, variable section thicknesses, and, in some cases, changes in magnetic properties. These variables, which will cause the eddy currents in the test specimen to vary, make crack detection difficult. There are several different states of blades which have to be considered in a test evaluation. The following is a definition of them as used in this report. New blades are those which have never been used in an engine. Uncleaned blades are blades as removed from the engine. Cleaned blades are used blades which have been cleaned. Turbine Blade Inspection An eddy current probe output can be displayed on the face of a storage oscilloscope. The oscilloscope can display the coil signal on the equivalent of the impedance plane discussed in the foregoing. Such a presentation was used to obtain the photographs shown in Fig. 2. Fig. 2(a) shows the change of probe impedance as it moves from air to the various metals 4

5 Crack Indication Air Magnetic Oxide indication Fig. 4(a) Oscilloscope display of a clean blade Fig. 4(b) Oscilloscope display of a blade having which does not contain magnetic oxide surface magnetic oxide Fig. 4 Oscilloscope recordings of a clean and unclean J-57 blade indicated in the figure. This provides a qualitative feeling for the range of impedance changes possible from a given coil. Fig. 2(b), taken at the same gain settings as Fig. 2(a), shows the change in coil impedance as the probe moves from air to a new turbine blade. Because the turbine blade material has a low conductivity (1 to 2 percent IACS), the impedance change (and the distance between the dots) is smaller than that of the materials in Fig. 2(a) For close analysis of small blade effect, it is obviously necessary to increase gain in the system. Fig. 2(c) shows such an expanded curve and the change in impedance when the probe intercepts a crack on the new blade. In Fig. 1, it was pointed out that the angle, 0, separates the paths of lift-off and conductivity. 0 is the important to consider when impedance analysis is used to minimize the effects of lift-off. An experimentally generated 0 is shown in Fig. 3(a). The impedance diagram in this case has been rotated to approximately 45 deg by use of phase rotation. It shows the crack effect on the eddy current signal in the vertical direction. Thickness changes and lift-off cause the 45-deg signal locus. To help in minimizing lift-off and thickness variation effects, the diagram in Fig. 3(a) can be rotated to a new separation angle of approximately 90 deg. The result is shown in Fig. 3(b). In summary, it can be seen that, with the foregoing processing, the following conditions have been achieved: (a) the impedance plane coordinates have been reoriented so the air to turbine blade lift-off locus lies in the horizontal direction, (b) 0 has been enlarged, and (c) the locus of the apparent conductivity change caused by the crack is in the vertical plane. Minimizing thickness effects is desirable because small changes in blade thickness would otherwise appear to the eddy current test as cracks. Minimizing lift-off is necessary to provide an inspection system which will be relatively insensitive to probe to blade geometry changes which can easily occur in a hand-held inspection. The Fig. 3(b) signal, a slow oscilloscope sweep, and an a-c presentation will, as shown in Fig. 3(c), allow a simple crack presentation while suppressing the undesirable effects of lift-off and gradual thickness changes. EVALUATION OF EDDY CURRENT METHOD FOR FIELD USE Background The major difference between the application of the eddy current inspection as described in the preceding section and that required in field turbine blade inspection is that the uncleaned blades, while of themselves nonmagnetic, have surface magnetic oxide coatings. These coatings are formed in the central part of the blades which are exposed to the hottest gases passing through the engine and are normally known as "heat affected zones." It is in this zone that essentially all blade cracking occurs. Such magnetic coatings seriously influence any eddy current field inspection attempts. A typical oscilloscope pattern of a crack in a special ultrasonically cleaned blade, which does not contain the surface magnetic oxide, is shown in Fig. 4(a). Fig. 4(b) is a recording of

6 Fig. 5(a) Photograph of the leading edge of a J-79 blade having a crack Fig. 5(b) Cross section of the crack found in the leading edge of the above cleaned J-79 blade Fig. 5(c) Microcracking in the trailing edge of a J-79 blade Fig. 5 J-79 cleaned turbine blade a similar size crack in an uncleaned blade having surface magnetic oxide. Note, by comparison, that the inspection technique is strongly influenced by the magnetic oxide on the uncleaned blade. The signal response is much greater to the oxide-included and surrounded crack than to a clean crack and blade. It can be assumed that the thickness of the oxide in a crack in an uncleaned blade will be greater than the surface thickness. This causes a larger concentration of oxide in the crack as compared to the surface background. It permits a greater amount of stored electrical energy and produces a good signal from a crack. Setting up eddy current equipment to sense for localized oxide buildup will, therefore, produce greater crack detection capability on an uncleaned blade than a new blade. In the case of normally cleaned blades, the surface magnetic oxide is removed and only the 6

7 III Mil Fig. 6(a) Oscilloscope recording of leading and trailing edge noise in an uncleaned blade Fig. 6(b) Oscilloscope recording of leading and trailing edge cracks in an uncleaned blade Fig. 6 Oscilloscope recordings showing magnetic oxide and crack indications oxide in the crack remains. A better signal-tonoise ratio is, therefore, expected than that produced by a crack in a new blade. A number of blades were inspected which verified this. For eddy current inspection of ultrasonically cleaned blades, cracks down to in. (0.750 mm) long, in. (0.025 mm) wide and in. (0.500 mm) deep have been detected. This limiting value is imposed by the low electrical conductivity of the blade. For normally cleaned blades, cracks of about in. (0.500 mm) in length, x in. (0.012 mm) in width x in. (0.125 mm) in depth have been detected. This increase in sensitivity is due to the oxide in the crack. Based on laboratory results, of which the foregoing are typical, the eddy current inspection technique was deemed ready for field evaluation on both cleaned and uncleaned blades. FIELD RESULTS Initial Tests The eddy current technique was tested at Oklahoma City Air Material area (OCAMA), Tinker Air Force Baae, Nellis Air Force Base, Lambert Field, Pratt and Whitney and the Air Force Materials Laboratory (AFML). With the help of Tinker, Nellis, AFML, and Missouri Air National Guard personnel, it was possible to inspect blades and verify the results obtained through optical analysis, penetrant inspection, and metallographic verification of the inspected blades. A variety of built up rotors were inspected. Blades which gave defect indications were removed, inspected optically, and then sectioned for further evaluation. The following examples do not contain details on every type of turbine blade inspected, but have been selected mainly as a representative sample of the blades tested. J-79 Rebuilt and Inspected Rotor Assemblies At Nellis AFB, four overhauled J-79 builtup rotor assemblies (each consisting of first, second, and third stage) were inspected. This inspection showed one of the first-stage wheels of the rotor assemblies contained six cracked blades. No other defects were detected. The first stage was then completely inspected with fluorescent penetrant. The penetrant inspection found only one cracked blade. An on-rotor microscopic inspection confirmed that all six blades contained cracks. These blades were removed for metallographic inspection to determine crack depth. Examples of the cracks found are shown in Fig. 5. Fig. 5(a) is a photograph of the leading edge of a J-79 blade which contained a crack found by the eddy current system. Fig. 5(b) is the metallographic verification of the crack. The depth of this crack, 7

8 Fig. 7(a) Oscilloscope recording of leading and trailing edges in an uncleaned blade Fig. 7(b) Oscilloscope recording of a number of leading and trailing edge cracks in an uncleaned blade Fig. 7 Filtered oscilloscope recordings of the leading and trailing edges in J-57 blades (differential probe) here shown at its largest, was in. (0.300 mm). The width of this crack was in. (0.025 mm). Cracks such as these are typical of the tight cracks found in the leading edges of the J-79 blades. The predominant defect found in the trailing edges of the J-79 blades have been microcracks. An example of this defect is shown in Fig. 5(c). As can be seen, the network of cracks is small, on the order of in. (0.008 mm) in width. The ability to detect them by eddy currents might be surprising. However, the presence of microcracks is detectable, due to the large volumetric density of magnetic oxide which can be present at the blade surface due to the many oxide susceptible surfaces present in a typical microcrack structure. J-57 Rebuilt Rotor Assemblies At the Missouri Air National Guard stationed in Lambert Field, in St. Louis, Missouri, several overhauled J-57 built-up rotor assemblies (first stages) were inspected. Results of this inspection showed on overhauled first-stage wheel contained two cracked blades. The two blades were then in situ inspected with a standard fluorescent penetrant technique. This did not confirm these cracks. An overnight dwell time, however, confirmed the presence of the indicated cracks in both blades. To demonstrate the utility of the system in the hands of maintenance personnel, two NDI technicians were trained to operate the inspection equipment. The technicians then inspected two J-57 first-stage sections (as removed from the aircraft). The results of this inspection showed a wheel contained eight blades which had defect indications. These anomalies appeared to be cracks. The blades were removed from the disks, cleaned, and visually inspected. Cracks were verified in five blades. It is believed that the other three were either blended over, microcracking or surface magnetic oxide. United Aircraft At Pratt and Whitney, Division of United Aircraft Corporation, in East Hartford, Connecticut, we first became aware of the magnitude of the eddy current impedance change caused by surface oxides. Actual probe crack and noise responses are shown in Fig. 6. The top photograph is a recording of the signal in an uncleaned blade. The bottom photograph is a recording of a number of cracks in an uncleaned blade having surface magnetic oxides. The signal response is about equal 8

9 for the surface magnetic oxides and the cracks in the blade. This fact and a noted excessive wear of the eddy current probe face indicated modifications to the inspection system were required. SYSTEM MODIFICATION Based on the Missouri Air National Guard and Pratt Whitney results, it was decided to modify the eddy current system to eliminate the interference caused by surface oxides. GARD -11"4 Electronic Filtering Filtering to enhance an inspection system signal-to-noise ratio in eddy current testing is a commonly applied technique. For the filtering to be an effective technique of noise suppression, there must be a frequency range in the spectrum of the system 1 s output signal where a crack produces a higher amplitude than does noise. A spectrum analysis was made of the signal produced by the magnetic surface oxide changes and of the signal produced by a crack. The wave form produced by the magnetic changes is composed primarily of low-frequency components, with highfrequency components which are comparatively small. The crack wave form is composed of primarily highfrequency components. Therefore, it was possible to select a window in the frequency spectrum which would pass most of the wave form produced by the crack, yet block all but a small high-frequency portion of the magnetic change wave form. With proper filtering, the interfering effects of magnetic oxide was lowered and crack response enhanced. The crack detection signal-tonoise ratio was increased by 3. Probe Configuration Another method of improving the detection of small defects is the use of a differential pair of small coils. In this technique, the reference coil is placed directly on the test specimen along with the test coil, and the difference of the two signals is read and may be highly amplified. In addition, the sensitivity to gradual metallurgical variations is lowered because the probes are too close to each other to observe the change. Thus, the differential method has definite advantages in the exclusive measurement of cracks. Such an approach was used in conjunction with the filtering, and the results are shown in Fig. 7. Fig. 7(a) is a typical filtered differential probe produced oscilloscope pattern of the unclean Pratt and Whitney blade, which contains surface magnetic oxides. Fig. 7(b) is a recording of a similar unclean blade having a number Fig. 8 Photograph of the GARD-100 turbine blade inspection system of cracks. The vertical amplifier of the oscilloscope was set at 1 v per division in Fig. 7(a) and 2 v per division in Fig. 7(b). A comparison :f Figs. 7(a) and 7(b) illustrates the effective suppression of the interfering variable of the magnetic oxide on the uncleaned blades. The signal response is much greater to the cracks than to the magnetic oxide variations. In summary, it can be seen that, with the signal processing and the differential probe, the following condition is achieved. The noise generated by surface magnetic oxides has been reduced: The crack detection signal-to-noise ratio has been enhanced to about 10:1. The advantages or the filtering and use of a small field differential coil are obvious. Another advantage of this technique is that no mechanical probe-to-part coupling is now required during inspection, because the increased sensitivity to cracks allows the use of a non-contacting probe. This eliminated the wear problem. SYSTEM DESCRIPTION Based upon the favorable results obtained during the initial tests conducted under this AFML contract, GARD designed and fabricated a turbine blade inspection system known as the GARD-100 which incorporates the filtering and the new probe discussed in the foregoing. Further field tests, together with favorable Air Force personnel acceptance, demonstrated the GARD-100's applicability to turbine blade inspection. Sever- 9

10 al systems, with appropriate operational instructions, were provided to the U.S. Air Force Logistics Command for an extensive field evaluation at Oklahoma City (OCAMA) and San Antonio (SAAMA). Briefly, it incorporates both phase analysis/modulation analysis eddy current techniques, which are used to suppress the interference caused by the magnetic surface oxides. The equipment contains a bridge circuit and has the capability to establish various bridge balancing conditions through phase control. It has a signal-processing unit which contains filtering which reduces background noise, a variable rejection network which discriminates between signal and noise, and an adjustable alarm network with audio readout (and override) which can be used to provide "gono go" part quality indications. In addition, the unit contains a storage cathode ray tube for visual defect readout A photograph of the system is shown in Fig. 8. EVALUATION PROGRAM The Air Force evaluation program has been divided into various phases. The first is to provide turbine blade crack standards containing "natural" defects for calibration of the eddy current inspection equipment. Standards In this phase, a laboratory evaluation using natural occurring defects for reference standards with the eddy technique was performed. "Natural" defects were chosen as reference defects instead of artificial defects because they would provide some insight to the relationship between eddy current instrument response and defects which would be encountered during inspection. Since the equipment response, for the most part, will determine a blade reject level, some correlation needed to be establishes between crack size and instrument output. To achieve the aforementioned goal, five types of engines were chosen for inspection. Blades were first grouped by engine type and visually (about 50 to 100 blades per engine) inspected for similar surface crack dimensions (length and width). Blades having similar crack dimensions were then eddy current inspected for similar electrical response. For each blade type, five blades with cracks, which generated similar eddy current and visual indications, were acquired. Two blades of each type were metallurgically examined to determine if crack size can be reliably predicted, and the remainder were used to make up three reference packages. Each of the two blades metallographically examined correlated well in their indications of crack size. The correlation between crack size and eddy current response permits reference crack size assignment with an accuracy on the order of ± 10 percent. 2 As discussed previously, eddy currents respond to the effects of microcracking, surface magnetic oxides, and foreign object damage. The level of "background noise" encountered in any one inspection affects the inspection sensitivity that can be employed. Background noise can be defined as any signal which affects the detectability of defects and is caused by test specimen variables which are not considered defects by the inspection specifications. These background variables impose the lower limit of crack detectability. Thus, background reference blades are required for eddy current equipment setup. Such blades were chosen for inclusion into the standards set developed. Based on the favorable metallurgical results, GARD prepared three sets of eddy current reference blade packages. One set was given to SAAMA, one to OCAMA, and the third was kept at GARD. The package contains ten reference blades -- one crack standard and one noise standard from five different engines. Personnel Familiarization Field Optimization and Production Testing The second phase was a meeting with the evaluation personnel to give them appropriate operational instructions and provide them time to become familiar with the equipment. In phase III, an on-line inspection of about 20 engines is currently being performed by the Air Force evaluation team. It is in this phase that the evaluation of the technique will be accomplished and field optimization achieved. The results are expected to confirm the system's capability and give sufficient information to define production procedures. This phase has just begun, and no data is available at this time. However, results of this part of the evaluation program are expected to be given in the oral presentation at the 18th Annual International Gas Turbine Conference. In phase IV, production testing of blades will be accomplished. The purpose of this phase is to demonstrate the utility of the equipment in the hands of production personnel. The results of this thorough testing program should 2 Kraska, I. R., "Development of Eddy Current Turbine Blade Standards," General American Research Division, Analysis Report GARD-AR , June

11 give us sufficient information to confirm the (MXA) of the Air Force Materials Laboratory, capability of the technique to successfully per- Wright-Patterson Air Force Base, Ohio; MSgt. E. form the desired inspection. Hammond, Hdq. TAC (DMMS6), Langley Air Force Base; and MSgt. L. Manville, 57th TFW (FWEP/NDI) Nellis ACKNOWLEDGMENTS Air Force Base; and MSgt. B. L. Brannum, 131st TAC FTR. Group Lambert Field; and F. Vickey, The authors gratefully acknowledge technical Pratt and Whitney Division of United Aircraft assistance provided by the Air Force Project Corporation for their assistance in the field Engineers R. R. Rowand (LLN) and E. W. McKelvey testing. 11