Table of Contents ECNDT '98
Session: Electromagnetic Methods
SQUID Eddy Current Technique Applying Conformable Eddy Current ProbesLudwig Bär, Gabriel M. Daalmans, Lars Ferchland
Siemens AG, ZT MF 1, D-91052 Erlangen
Erich Becker, Hans-Peter Lohmann, Annette Sperling
|Table of Contents|
An Eddy current method applying a High Temperature Superconductor ( HTS ) DC SQUID sensor operating at liquid nitrogen temperature (77K) is presented. The method is developed for the detection of surface or surface near defects. We compare the performance of the SQUID system with the performance gained from a commercial Eddy current system, while using identical probes. The experimental data are obtained on defects in gas turbine blades. The advantage of planar conformable probes for the use with the SQUID is discussed.
The SQUID (Superconducting Quantum Interference Device) is a magnetic flux sensor with an ultimate flux resolution, based on the superconducting Josephson effect. Combined with a flux antenna a field resolution better than 1 pT is possible. The SQUID sensor has to be cooled and special care has to be taken to reduce the influence of background fields on the field resolution. It has been shown that a field resolution of about 20 fT can be obtained in a magnetically well shielded environment (laboratory) with a HTS-SQUID. For non destructive testing the system has to be mobile and has to be operated at locations with typical background fields in the order of 100 pT. The influence of these mainly homogeneous fields can be reduced by applying flux antennas with a differential nature. In this manner, a field gradient resolution of about 1 pT/cm Hz could be achieved without magnetic shielding.
The outstanding figure of merit of the SQUID is its high signal to noise ratio over a wide frequency band, ranging from DC up to tenth of MHz. When thinking about magnetic probing with a SQUID three different applications, associated with different frequency regimes, are possible.
|Fig 1: principles of measurement|
In fig.2 a schematic representation of the performance of the SQUID methods and the conventional method is given. We compare the signal voltage at the first preamplifier stage for similar probes at a given excitation. For a completely superconducting flux antenna the SQUID response is nearly frequency independent down to about 1 Hz. The frequency response of a SQUID sensor with a normalconducting flux antenna is frequency independent above a cut off frequency given by the ratio of the resistance and the inductance of the flux antenna. Below this frequency the behaviour is similar as for a usual induction coil. However if the flux antenna and the SQUID are proper designed, the SQUID sensor acts as a very low noise amplifier. This is a quite interesting property, taking into account the low noise of a low ohmic induction coil compared with the noise of a low noise room temperature preamplifier. The best broad band preamplifiers offer a noise level of about 1 nV/Hz. Low ohmic induction coils show a noise level of typically 0,1 nV/Hz. So the direct connection of a low ohmic induction coil with a low noise preamplifier would result in a reduction of the signal to noise ratio by about one order of magnitude. This is prevented by the application of a SQUID stage between the induction coil and the preamplifier. In contrast to conventional transformer solutions, the SQUID amplifier solution is very broad banded. The band limit is determined by the probe itself. The application of either a normalconducting flux antenna or a superconducting flux antenna depends upon the kind of defect to be examined. Deep defects, at least 10 mm from the surface, are usually detected with a superconducting flux antenna using all the advantages of a SQUID sensor. The thermal insulation between the cold sensor and the room temperature test object is also in the order of one cm. So the decay of the signal due to a larger distance between sensor and test object is not so dramatic in this case.
|Fig 2: Performance of Eddy current read out schemes|
In the case of surface defects however the signal decays rapidly at a distance in the order of one mm from the surface. So the application of a cold flux antenna would reduce the signal by some order of magnitude. In this case the application of a normalconducting room temperature flux antenna is much more appropriate. Surface defects in gas turbine blades are usually examined with an Eddy current technique at frequencies up to one MHz. This can be realized with the SQUID amplifier by applying low inductance ( 1 µH ) Eddy current probes, taking into account the small resistance value needed for low noise purposes. These small inductance coils are preferably designed as planar coils in order to couple the relatively few turns as good as possible to the test object. Moreover three dimensional groove structures in the fixing parts of the blades demand for Eddy current probes with a three dimensional conformable character.
|Fig 3: Eddy current SQUID system|
We use several types of evacuated stainless steel cryostats of about 1 litre capacity to provide the 77 K cooling with liquid nitrogen. The simplest, best valued and easiest to handle one is a ordinary thermos flask available from the supermarket (see fig.3). It is sufficient for a measurement period of more than 12 hours. A specially designed stainless steel vessel gains cooling periods of 30 hours.
The SQUID converter and the coupling coil are thin film devices prepared of the high temperature superconductor Y1Ba2Cu3O7 on SrTiO3 substrates. Typical substrate dimensions are 10 mm by 10 mm. In one version , fig.4, the SQUID converter and the coupling coil are integrated on one substrate to improve the coupling; in another version two substrates have been applied and the magnetic coupling has been made by a flip chip technique. This results in a looser coupling on the one hand but a much simpler preparation process on the other hand. The SQUID converter is a one layer device applying an artificial crystalline grain boundary on a bicrystalline substrate for the preparation of the Josephson junctions. A flux to voltage conversion of about 100µV/###o could be obtained. A first stage of the flux antenna (see fig.4) is integrated in the same layer in order to couple properly to a coupling coil. This is of practical value for the flip chip version as well as for the integrated version. The design was optimized for Eddy current probes of a few µH by designing a coupling coil of the same inductance.
|Fig.4 : Layout of the SQUID sensor|
|Fig 5: electronic scheme of the SQUID Eddy current system|
Fig 6: 2 mm deep hairline crack in Ni-based alloy:
a) commercial Eddy current system
b) SQUID system
Fig 7: groove (0.2x1x2mm3) under Sermetel in X4CrNiMo16.5 / ferritic core probe:
a) commercial Eddy current system
b) SQUID system
Fig 8: groove (0.2x0.5x2mm3) under Sermetel in X4CrNiMo16.5 / ferritic core probe:
a) SQUID system at 10 kHz
b) SQUID system at 3 kHz
Fig 9: groove (0.2x1x2mm3) under Sermetel
in X4CrNiMo16.5 / planar film probe:
a) commercial Eddy current system
b) SQUID system
Fig 10: Shaped probe
Several types of Eddy current probes were used with the SQUID system and the commercial system as well. High inductance wire wound probes with a ferritic core and low inductance planar thick film coils were applied. The wire wound probe is the commonly used probe for high resolution conventional testing. The low inductance planar coil is more suited to be applied in combination with the SQUID system. It is well adapted for surface defects and shallow defects.
Generally the signal to noise ratio can be enhanced by increasing the excitation field. But especially at higher frequencies the signal amplitude, the SQUID can amplify, is restricted by the speed - so called slew rate - of the flux locked loop. This rate is determined by the flux to voltage conversion of the SQUID. With a conversion factor of about 100µV/###o we could manage a gradient field change of about 10-3 T/cm·s at 1 MHz. Further SQUID development will allow to improve these data by one order of magnitude. In order to maximize the excitation, precautions have to be taken to avoid cross-talk between excitation and signal. Therefore differential probes are commonly used with a SQUID system. Nevertheless, for the discussed defects the SQUID system has a lower excitation field by a factor of about 100 compared with the commercial system. This we must keep in mind, when we compare measured signal to noise ratios. There is a potential to improve for small defects, when cross-talk is managed very well.
Surface defects, high frequency:
We investigated defects on the surface of Ni-based alloys. This material is used for high performance gas turbine blades. The prefered excitation frequency for surface detection is about 1 MHz for this material. Although the electronics for demodulation and filtering differ for the SQUID read out and the conventional read out, we took care to have equal filter characteristics for the demodulated signal. The example we show in fig.6 is the signal of a 2 mm deep hairline crack.
The probe for both measurements is a ferritic core coil. The signal to noise ratio in both cases is nearly identical at a level of about 75 dB. In the scheme of fig.2 we are well above the cut off frequency R/L, where no benefit from the SQUID is to be expected. When applying a low inductance planar coil, we can preserve the resolution of the SQUID system. We will refer to the advantage of planar pick up coils later.
Under surface flaws, moderate frequencies:
Gas turbine compressor blades made of ferritic steel (X4CrNiMo16.5) and coated with 300 µm Sermetel were used as samples. Fatigue cracks and artificial grooves (0.2 mm wide, 0.5 - 1 mm deep and 2 mm long) in the base material lay underneath the defect free coating. Excitation frequencies for these tests were 50 and 100 kHz. Fig.7 shows the result for the groove, achieved with a commercial ferritic core probe at 50 kHz. For this probe the frequency is about the cut off frequency R/L (R/L 35kHz) (see fig.2). The SQUID read out is by a factor of about 6 better in the signal to noise ratio. As shown in fig.8 the signal to noise ratio nearly can be preserved with the SQUID system down to a frequency of 3 kHz.
Since the compressor blades are curved with a large radius of curvature, planar, surface adapted probes are the appropriate means for testing. In fig.9 we show the signal of a fatigue crack taken with a planar differential probe at 50 kHz. While with the commercial equipment the signal hardly exceeds the noise the SQUID system gains a signal to noise ratio better than a factor of 30. For this probe R/L is about 1 MHz. So we are working far below the cut off frequency (see fig.2).
As stated above the SQUID amplifier demands a low inductance Eddy current probe in order to be able to amplify signals up to 1 MHz. Low inductance Eddy current probes can be obtained by reducing the number of turns and by loosing the magnetic coupling between the turns. So magnetic cores should be avoided as well as tight wounded turns. For this purpose planar coils are the best solution.
This type of coil was prepared from copper cladded printed circuit board material by applying photolithographic techniques. The p.c. board material is available with different copper thicknesses and with either a stiff or a flexible carrier. The flexible material offers the opportunity to adapt the planar coil to a curved three dimensional test object. In our turbine blade application this is a major advantage. The thickness of the copper layer was chosen to be 17 µm. The period of the coil was 100 µm. The coils were patterned by wet etching. A major advantage of this approach is the parallel processing with narrow tolerances, resulting in many identical Eddy current probes. An example of such a probe is shown in fig.10.
Two types of defects in gas turbine blades were examined. One type of defect concerns cracks under a thin Sermetel coating located on places with a large radius of curvature (see fig.9). For this application the flexible Eddy current probe is mounted on a conformable carrier allowing a good fit to the curved test object by pressing gently by hand. The other type of defect concerns fatigue cracks in turbine blade mounts ( see fig.10)with small radius of curvature. In this case the Eddy current probe has to be fitted to the three dimensional groove structure. This is accomplished by in situ mounting the flexible coil to a polymeric carrier shaped as the groove.
One of the advantages of planar, conformable probes is the reliable handling of the Eddy current head. Tilting effects can be reduced considerably. Another one is the large amount of area, that can be tested with one scan. Together with the good access to curved surfaces, this makes the application of conformable probes to a very fast method of Eddy current testing. Though the SQUID read out is the appropriate means for high resolution testing, a favourable use with conventional systems is possible, if defect size and probing frequency is suitable.