2nd International Conference on NDE in Relation to Structural Integrity for Nuclear and Pressurized Components, New Orleans May 2000.

TABLE OF CONTENTS

Abstract Introduction Materials and test conditions Structure investications by means of diffraction experiment X-ray diffraction (XRD) Testing for LCF degradation by means of magnetic NDT method Measurements of magnetic stray field strengt Measurements by means of eddy current technique Conclusion Acknowledgement References

Abstract

The microstructural changes in the pre-crack stage of low-cycle fatigue damage (LCF) in austenitic piping steels are characterised by neutron and X-ray diffraction. The LCF damage evolution in the metastable austenitic steel causes a deformation-induced phase transformation from austenite to martensite. Thresholds exist for the formation of martensite as a function of both, the load amplitude and the number of LCF cycles. Magnetic stray field and eddy current measurements were chosen to transfer the results of material characterisation to NDT methods. The density and distribution of martensite obtained from neutron diffraction experiments were used to adjust the NDT signals. Both NDT techniques were able to detect the very low amount of martensite in the different aged specimens (0.5 - 3.1 vol.% martensite, usage factors from 0 up to 1.0).

Introduction

The development of advanced diagnostic systems that are able to identify and detect material degradation in the microstructures of steels is a new challenge for non-destructive testing (NDT). Especially in nuclear power plants, early detection of material degradation can contribute to improve safety and reliability of the primary circuit boundary as well as plant life time management of certain components. Lifetime extension of nuclear power plants (NPP) has become an important topic and resulted in the requirement for advanced safety management tools. Therefore systems for early detection of material degradation of pressurised primary loop components are demanded.

With regard to leaks in pressurised water pipes of nuclear power plants [e.g., 1-3], which were caused by thermal fatigue damage, deformation-induced changes in the microstructure of metastable austenitic steels were investigated. It is well known that varying temperatures and thermal stratification in piping of power plants might cause fatigue damage by thermal strains. Such strains promote the phase transition from a cubic face centred (fcc) austenite to cubic body centred (bcc) martensite [4].

Our structure investigations using metallography as well as neutron and X-ray diffraction are primarily intended to identify and quantify martensite formation in the pre-crack stage of low-cycle fatigue damage (LCF) in austenitic piping steels and its influencing parameters. Martensite is a ferromagnetic phase distributed in the austenitic matrix which itself is paramagnetic. This opens the possibility to detect the martensitic content by means of magnetometers. To monitor the pre-crack damage state, the application of magnetic NDT methods based on magnetic stray field and eddy current measurements is investigated.

Materials and test conditions

The investigations were performed on hour-glass specimens (according to ASTM E606, Fig.1). The material under investigation was the titanium stabilised austenitic metastable steel X6CrNiTi18-10, which is widely used for vessels and piping. The chemical composition (in wt.-%) is given in Tab. 1. In order to homogenise the microstructure, the specimens were annealed at 1040 °C (1 h) and afterwards quenched in oil. The fatigue tests were performed with alternating loading with a test frequency of 2 s^-1 at room temperature and were controlled by the total strain amplitude. A first test with a strain amplitude of ±2.55 mm/m did not result in crack initiation. Therefore a strain amplitude of ±2.95 mm/m was applied referring to the position of the smallest cross section. Different usage factors were realised by well defined cycles of strain loads in series of hour-glass specimens. Tab.2 gives the cycle number for some investigated specimens and the corresponding usage factor D. The usage factor indicates how many of the life time is passed. D is defined to be equal to 1 for specimens, where crack initiation has already occurred. The LCF tests were stopped when a 5% force drop was reached. Therefore the corresponding cycle numbers vary for specimens with D=1 due to differences in the damage progression. For specimens without a crack, D is defined as the ratio of the applied cycle number to an averaged cycle number representing crack initiation. This averaged value for crack initiation is obtained in pre-testing on 5 specimens.

C	Mn	P	S	Cr	Ni	Ti
0.05	1.08	0.039	0.019	17.55	9.86	0.39
Table 1: Chemical composition in wt.-% of the X6CrNiTi18-10 specimens

Fig 1: Hour-glass specimens (ASTM E606).

For the metallographic investigations, the specimens were cut lengthwise. After mechanical grinding and polishing, the specimens were electrochemically etched using Beraha-II solution [5]. Fig.2 shows the microstructure of specimen 1.6 at three axial positions. At middle positions dark streaks were found. With increasing distances from the middle of the specimen the width of the streaks decreases down to zero. At positions far from the middle, the microstructure is homogeneous.

Fig 2a: -30mm from the middle of the sample	Fig 2b: middle of the sample	Fig 2c: +30mm from the middle of the sample
Fig 2: Microstructure of specimen 1.6 at different axial positions

The microstructure of the dark streak regions consists of martensite needles in an austenitic matrix (Fig.3a). Outside these streaks only austenite grains can be found (Fig.3b).

Fig 3a: inside the streaks	Fig 3b: outside the streaks
Fig 3: Microstructure of specimen 1.6 in- and outside the dark streaks

Structure investigations by means of diffraction experiments

The martensite content formed by phase changes was measured by means of neutron and X-ray diffraction experiments. Austenite and martensite differ in their microstructure. Austenite has a face centered cubic lattice cell whereas the lattice cell of martensite is body centered cubic. This makes it possible to determine the phase composition in the steel by neutron and X-ray diffraction experiments.

Neglecting textures, the volume fraction of a phase v₁ in a two phase system (austenite and martensite) is given by [6]:

(1)

I_{h,k,l} is the intensity scattered at the lattice plane {h,k,l}. K_1,2 is a factor which considers the different structure type of martensite and austenite and the different angular positions of the reflexions. K_1,2 factors are tabulated for X-rays in [6] and can be calculated for neutrons [7].

Neutron diffraction (ND)
The neutron diffraction experiments were performed at the powder diffractometer DMC at SINQ (Swiss Spallation Neutron Source, PSI Villigen) [8]. The instrument is optimised for high intensity. It allows the determination of the phase content down to values below 1 %. With a neutron beam wave length of l  = 0.38 nm, a range of the scattering angle 2Q of 68° £ 2Q £  147° was analysed. The measurements on the LCF specimens were performed with a beam cross section of 40 mm (width) x 10 mm (height). The measuring position was in the middle of the LCF specimens (±5 mm in height). Results of martensite content obtained by these measurements are summarised in Tab.2. At four specimens also the axial distribution of the martensite was measured. For this purpose a reduced beam height of 2 mm was used and the specimens were scanned in axial direction.

Fig.4 shows the specimen cross section variation, the corresponding strain amplitude and the measured martensite content in dependence on the axial coordinate of the specimen. It can be concluded that the martensite formation is limited to the area of the smallest cross section. In Fig.5, the axial distribution of the martensite content measured with the reduced beam height of 2 mm is given for four specimens. Three of them, 1.6, 2.3 and 2.5 have D = 1. The maxima of the martensite content are detected at the crack position. It seems that the crack formation results in an additional martensitic transformation. In the specimen 1.7 (without crack) the maximum of the martensite content was found at the smallest cross section. The martensite distribution in axial direction is unsymmetric for this specimen. This is also proved by the eddy current measurements (see 4.2).

Sample No.	Cycle number	Usage factor D	Crack position	Martensite (vol.-%)
1.2	0	0	-	< 0.50
2.4	0	0	-	< 0.50
1.8	11200	0.4	-	< 0.50
1.4	15850	0.6	-	0.61
2.2	15850	0.6	-	0.54
1.3	22400	0.8	-	0.95
1.7	22400	0.8	-	0.93
2.5	26000	1.0	5 mm away from the middle	1.49
2.3	32000	1.0	in the middle	1.49
1.6	63200	1.0	in the middle	3.13
Table 2: Martensite content obtained by neutron diffraction

Fig 4: Axial distribution of the martensite content for specimen 1.6 together with the axial dependence of the specimen cross section and the local strain amplitude .

Fig 5: Axial distribution of the martensite content for specimens 1.6, 1.7, 2.3 and 2.5

In Fig.6, the measured martensite content is plotted as a function of the strain amplitude. Below a threshold strain amplitude (about 2.2 mm/m for specimens 1.7, 2.3 and 2.5 and 2.5 mm/m for specimen 1.6), no martensite was detectable by neutron diffraction (detection limit of these experiments is about 0.5 vol.-%). For strains higher than the threshold strain amplitude, a nearly linear dependence of the martensite content on the strain amplitude was found if the data measured at the crack locations were not included. The slope of the curves depends on the cycle number. The dependence of the martensite content on the cycle number N is given in Fig.7. Also for the cycle number a threshold value (N_c = 13'000) exists, below which no martensitic transformation can be detected. At higher cycle numbers the martensite content depends nearly linearly on the cycle number.

Fig 6: Martensite content in dependence on the strain amplitude for the specimens 1.6, 2.3, 2.5

Fig 7: Martensite content in dependence on the cycle number

X-ray diffraction (XRD)
X-ray diffraction experiments using synchrotron light sources enable to analyse martensite distribution in two dimensions. The method guarantees a higher local resolution than neutron diffraction in shorter measurement periods, but is limited to the surface area. X-ray diffraction results give additional information on the radial distribution of martensite.

The X-ray diffraction experiments were performed at the ROBL beamline at ESRF Grenoble (France) [9]. In order to use the K_1,2 values from [6], the wavelength of the Mo-Ka radiation (l = 0.07107 nm) was applied. The investigated specimen 1.6 was cut into two halves parallel to the sample axis. A mapping of the axial and radial martensite distribution was gained by scanning the specimen through the fixed beam.

With the applied wave length and the beam cross section of 0.2 mm x 0.5 mm, the analysed volume was about 1 mm (axial) x 0.5 mm (radial) x 0.005 mm (in depth). For the determination of the martensite content, the intensity relations between the {111}-austenite and the {110}-martensite peak measured in XRD experiments were used.

Due to the small beam size and the very small beam divergence in the XRD experiments and the relative large grain size of the austenite (diameter about 20 - 30 m m), the statistic condition for a polycrystal (infinite number of grains N_G seen by the beam, in practice N_G > 1000) is not completely fulfilled. A wide scatter of the intensity values of the {111}-austenite reflexion between the several specimen locations is the consequence. The grain size of the martensite is small enough to meet this condition. As the content of martensite is very small compared with the content of austenite, the mean value of the austenite {111}- intensity can be used for the phase analysis. Using this mean value, the relative error of the martensite content (x_M) is smaller than 6% (Dx_M = 0.06 x X_M).

Fig 8: Mapping of the radial and axial distribution of the martensite content and typical XRD patterns

Fig.8 shows the axial and radial mapping of the martensite content of specimen 1.6 determined from the X-ray diffraction data. Some typical scattering patterns are included. In agreement with the ND measurements, in the axial direction, the martensite is found at the smallest cross section of the specimen. In the radial direction, the martensite content is concentrated near the surface (y= -6 and 7mm) and, as a broad area, in the center of the specimen (x,y = 0mm).

Testing for LCF degradation by means of magnetic NDT methods

By means of neutron diffraction experiments it was demonstrated that the martensite content depends on the usage factor for the realised loading conditions at room temperature. Based on these results, it appears challenging to develop methods for the use in screening tests to identify areas with an increased amount of martensite. In this way, an early detection of LCF damaged zones could be managed.

In case of LCF damage in metastable austenitic steels, the application of magnetic methods is possible as the deformation-induced martensite is of ferromagnetic and the austenitic matrix of paramagnetic nature. The problem is to measure very low amounts of martensite between 0.5 and 3 vol.-%. Therefore sensitive magnetic sensors must be used to identify local differences in magnetic properties. For an advanced NDT technique it is also expected that the local distribution of the martensite can be visualised. In principle it is possible to apply active and passive magnetic techniques.

Concerning the martensite content, results of the neutron diffraction experiments are used to adjust the NDT signals. Note, that for the usage factor D = 1, the LCF damage is different with regard to the cycle numbers.

Measurements of magnetic stray field strength
Due to the ferromagnetism of martensite it is possible to apply a passive magnetic technique. After magnetising of the specimens, the remanent field strength can be measured. The magnetisation of the specimens was carried out in a field with a strength of about 11 kA/m. Afterwards all specimens were examined using high sensitive SQUID and fluxgate sensors. Measurements of the axial and normal components of the magnetic stray field were performed in a distance of 20 mm from the bar axis. The fluxgate sensor was sensitive enough to detect martensitic areas and the results did not show any differences compared with the SQUID measurements.

The fluxgate output (axial component) is plotted as a function of the axial scanning position in Fig.9. This output shows a pronounced positive peak when the sensor is directly over the position of the smallest cross section of the LCF samples. In the case of measurements on samples with cracks (D = 1), the amplitude of field strength depends on the cycle number. For the pre-crack stages the amplitude of the fluxgate output increased with the usage factor or with the corresponding cycle numbers. The LCF damage in the specimens was detected down to the usage factor of D=0.6. The fluxgate output for the usage factors 0.6, 0.4 and 0.0 do not differ significantly.

Fig 9: Stray field measurements (fluxgate output) at LCF specimens of different usage factors, cycle numbers and martensite content.

Measurements by means of eddy current techniques
Eddy current techniques seem to be very promising for detecting LCF degradation. If the eddy current field is locally disturbed by changes of permeability, the eddy current signal is changed. The induced voltage amplitude referred to a defined phase angle of the receiving coil was recorded. It is favourable to apply miniaturised sensors. The distance between the measuring sensor and the surface should be small. The applied manipulator system was able to follow the hour-glass shape of the specimens, thus the distance between sensor and surface was always 1 mm. An essential aspect for eddy current measurements is the choice of the exciting frequency. For martensite visualisation a frequency of 50 kHz was used which provides a reasonable balance between high local resolution and depth of the eddy current field in the steel. The lower and upper bound of 0.5 and 3.15 vol.-% martensite content define the adjustment of the eddy current sensitivity.

For eddy current measurements, the end pieces of the LCF specimens were cut off and only the middle sections 75 mm in length were used. Additionally, the specimens were cut lengthwise into two halves. Thus it is possible to determine the martensite distribution also in the radial specimen direction and obtain information about the progress of the martensite formation. Some examples of eddy current output obtained on the circumferencial surface and on the axial cross section are plotted in the Figs.10 and 11. The plots visualise the martensite distribution on the circumferential surface (Fig.10) and on the axial cross section (Fig.11). The axial specimen direction corresponds to the horizontal coordinate axis, the circumferential and radial specimen directions correspond to the vertical coordinate axis. Plots are shown for different usage factors (specimen numbers 1.4, 1.7 and 2.3 with the corresponding usage factors 0.6, 0.8 and 1.0).

In Fig.10 the eddy current signal amplitudes at the circumferential surface are plotted for an angle range of 180°. The martensite content of 0.61 vol.-% for specimen 1.4 was clearly detected. The martensite is homogeneously distributed in the circumferential direction. The plot shows the beginning of the martensite formation during LCF loading. The changes of the martensite density in the axial direction show that the formation of martensite was really deformation-induced. In the following damage stages D = 0.8 and D = 1.0, the martensitic pattern is the darker as the usage factor increases. For specimen 1.7, the martensite is not as homogeneously distributed as for specimen 1.4. It is possible that the axial loading was accompained by a small amount of bending. For specimen 2.3, loaded up to crack initiation, a single large crack was detected at the position of the smallest cross section. Furthermore, some local areas with a very high martensite concentration were found. These are probably new sources for macroscopic cracks. The martensite is homogeneously distributed. At the crack position an additional formation of martensite was observed.

Fig 10a:	Fig 10b:	Fig 10c:
Fig 10: Eddy current measurements on the circumferencial surface of LCF specimens1.4 (D=0.6, xM=0.61 vol.-%)1.7 (D=0.8, xM=0.93 vol.-%)2.3 (D=1.0, xM=1.49 vol.-%)

In Fig.11, the eddy current signal amplitudes at the axial cross section surface are plotted. The sensor output for specimen 1.4 shows, that the formation of martensite is limited at the surface in the position of the smallest cross section. No martensite was detected in depth.

Fig 11: Eddy current measurements on the axial cross section of LCF specimens1.4 (D=0.6, xM=0.61 vol.-%)1.7 (D=0.8, xM=0.93 vol.-%)2.3 (D=1.0, xM=1.49 vol.-%)

For further damage progression (specimen 1.7), the martensite tends to spread in depth. The distribution of martensite in the axial direction is wider at the surface than in the bulk. A macroscopic crack initiation will be effected by stress relaxation around this location. This results in a decrease of the martensite spreading in this area which is shown for specimen 2.3. The shape of the martensitic pattern is broader at the opposite side of the crack.

Conclusion

All applied methods (neutron and X-ray diffraction, magnetic stray field measurements with SQUID and fluxgate sensors, eddy current measurement) allowed the detection of martensite. While the amount of martensite could be quantified with the diffraction methods, the magnetic techniques yield only relative results; that means, they have to be calibrated somehow, i.e. with results from neutron diffraction experiments.

Neutron and X-ray diffraction are powerful tools to determine the deformation-induced martensite content and its distribution in hour-glass specimens. Neutron diffraction is the only method to obtain the absolute value of martensite content in a specimen section covering the whole thickness.

It was shown that the content of martensite depends on the usage factor. The existence of thresholds for the cycle number and the total strain amplitude to form deformation-induced martensite could be concluded (total strain amplitude: 2.2 .. 2.5 mm/m, cycle number: 13'000). These thresholds are useful when examining industrial components under service conditions. It should be clarified whether thresholds also exist for other conditions of loading and temperature.

Both applied magnetic techniques, stray field and eddy current measurements, were able to detect material degradation due to low-cycle fatigue. For our testing conditions the eddy current technique was able to detect LCF damage down to D=0.6, the stray field measurements down to D=0.8. No significant differences of measuring signals for both techniques were found for D < 0.6, which corresponds to the results obtained by neutron diffraction experiments. For practical application, eddy current measurements in combination with neutron diffraction investigations are recommended.

The correlation between the usage factor and the NDT signal is clearly demonstrated which is rather promising for further investigations. However, further investigations namely for higher temperature ranges are needed before the application of the method for real components can be tested.

Ackowledgements

This work was supported by the Swiss Federal Nuclear Safety Inspectorate, HSK.

References

1. Experience with thermal fatigue in LWR piping caused by thermal mixing and stratification. Paris: OECD Nuclear Energy Agency, 1998. OECD NEA CSNI R(98)8
2. Leak on the Residual Heat Removal System at Civaux 1. Vienna: International Atomic Energy Agency, 1999. IAEA TC 485.27
3. R. Rantala, Thermal Fatigue Experiences and Countermeasures in Finland. Paris: Meeting of the OECD Nuclear Energy Agency CSNI Principal Working Group 3, 1999
4. R.W.K. Honeycombe, Steels, Microstructure and Properties. Edward Arnold Publishers Ltd., 1981
5. Metallographische Anleitung zum Farbätzen nach dem Tauchverfahren, Band2: Farbätzung nach Beraha und ihre Anwendungen. Düsseldorf: Deutscher Verlag für Schweisstechnik, Herausgeber, E.Weck, E.Leistner, 1983
6. G.Fanninger, U.Hartmann, Physikalische Grundlagen der Quantitativen Röntgenographischen Phasenanalyse (RPA). Härterei Technische Mitteilungen HTM 27, 1972, p. 233
7. G.E.Bacon, Neutron Diffraction, Oxford: Clarendon Press, 1975.
8. http://www1.psi.ch/www_sinq_hn/SINQ/instr/ DMC.html
9. W.Matz et al., ROBL- a CRG Beamline for Radiochemistry and Materials Research at the ESRF. Journal Synchrotron Radiation 6, 1999, p. 1076

© NDT.net - info@ndt.net

|Top|