MULTI-FREQUENCY EDDY CURRENT TESTING OF FERROMAGNETIC WELDS

Transcription

1 U-FQUCY DDY CU SG OF FOGC WDS ODUCO C. W. Gilstad,. F. Dersh and. Deale David aylor esearh Center etals and Welding Division nnapolis D, Single frequeny phase analysis eddy urrent tehniques have limited potential for inspetions for surfae raking in inhomogeneous ferromagneti welded surfaes [1]. Signals from these raks are masked in noise produed by lift-off, probe wobble and loal hanges of the permeability of the weld, the most troublesome noise soure. Surfae raks and hanges in magneti permeability reate signals with phase angles that are too lose to be disriminated from eah other. nother method, in addition to phase disrimination, must be used to distinguish rak signals from signals reated by hanges in magneti permeability. his paper explores the use of the omparison of phase and amplitude data from two probe frequenies to distinguish raks from hanges in loal magneti permeability. OD ddy urrent surfae inspetion an be modeled by the transformer iruit shown in Fig. 1. he primary iruit represents the eddy urrent instrument whih onsists of a urrent soure and a probe oil whih has a resistane, 1, and indutane, 1. he seondary iruit represents the weld surfae. he indutor, 2 has a ferromagneti ore with a variable permeability, ~. that imitates the permeability of the weld. he resistane, 2, represents the resistane of the weld surfae to urrent flow. Changes in the input impedane of the primary iruit, Z, are measured. he input impedane of the iruit is mathematially desribed as follows: Z = 1 + j w 1 + w 2 k (1) 2 + jw 2 where < k < 1 is the oupling oeffiient of the two oils and w is the frequeny of the applied urrent in radians. Sine Z has both real and imaginary omponents, j donotes an imaginary omponent. detailed derivation of the input impedane desribed by q. (1) an be found in ibby [2]. eview of Progress in Quantitative ondestrutive valuation, Vol. 9 dited by D.O. hompson and D.. Chimenti Plenum Press, ew York,

2 FOGC CO WH VB J.- < k < 1 Fig. 1. ransformer iruit modeling the eddy urrent weld surfae inspetion. he plot of the imaginary versus the real omponents of the input impedane, Z, is shown in Fig. 2. ll values are normalized to w 1 a, where 1 a is the value of the probe indutane in air and 1 has been subtrated sine it only auses a horizontal shift of the impedane plane. he arrows point in the diretion of an inrease in the labeled variable. ore detailed information of the impedane diagram an be found in ibby [2]. qualitative understanding of the effet of the inspetion parameters on the impedane diagram may be obtained by independently varying the values of 2, 1, 2, w, or k. he semi-irular loi shown are obtained by varying w, 2, or 2 Changes in w represent hanges in the exitation frequeny of the inspetion probe. Changes in 2 orrespond to hanges in the resistivity of the metal in the weld speimen. he hanges in the weld permeability are modeled by varying the indutane of both 1 and 2 he value of 2 hanges proportionally with ~. while the extent to whih 1 is affeted is dependent on its proximity to the ore. he effet of lift-off an be observed by varying k. derease ink is analgous to inreasing the probe lift-off. n terms of the atual inspetion, the extent to whih 1 hanges depends on the type of probe being modeled and how well the probe is oupled to the weld surfae; however, 1 will always be affeted less than 2 beause the oupling effiieny is always less than one. he predited response to a derease in permeability of two types of probes oils (panake - oil axis perpendiular to the weld surfae, and vertial oils - the oil axis is parallel to the weld surfae) is shown in Fig. 3. Sine panake type probes have better oupling to the weld surfae, 1 is made to vary more than it would in a model of a probe with vertial oils. ote that the probe with better oupling, that is, the panake probe, has a response that is more in the diretion of a derease in 1 (movement away from the semiirle lous.) 1364

3 OZD SSC Fig. 2. ransformer iruit impedane plane. z D VC POB CO PCK POB CO C - + ~ OZD SSC Fig. 3. he effet of a derease i n ~ probe oupling effiienies. for two different 1365

4 SUS FO H OD n this paper it is hypothesized that raks an be modeled as a simultaneous derease in the ondutivity and permeability of the weld. o test this hypothesis, probes were onneted to a vetor impedane analyzer whih was set to step through several frequenies. he lous of impedane values obtained at different frequenies is the ondutivity lous. Permeability loi were obtained by plaing the probe on a defet free, flush ground weld sample at various loations and plotting the impedane values at a number of frequenies. hese loi are shown in figure 4a. he ondutivity lous is in the shape of a portion of a semiirle represented by the dashed line. he permeability loi are radially direted and are shown for the ases of 125, 25, 5, and 1kHz. o test the hypothesis that raks an be modeled as a simultaneous derease in the ondutivity and permeability the probe was plaed adjaent to and then on a rak in a flush ground weld sample. hese data are plotted in figure 4b for the same frequenies previously identified. o support the hypothesis, an arrow pointing from the impedane value obtained adjaent to the rak to the value obtained on the rak would be expeted to point in a diretion between dereasing ondutivity and dereasing permeability. Figure 4, a ombination of figures 4a and 4b, demonstrates that rak signals lie in a diretion that supports the hypothesis. he data from the highest frequeny tested (1 khz) does not onform to the predition as well as the data from the lower frequenies. his is due to resonane of the probe near this frequeny. he probes have small apaitanes whih in ombination with the indutane of the system ause the probe to resonate. Future efforts should inlude the apaitane in the model sine nothing prevents a probe from operating as an inspetion tool while in resonane. he predited phase and amplitude of a rak signal and its relation to a signal from a permeability derease were alulated for the frequenies w-.1 and w- 1. he initial values of 2, 1, and 2 were 2 was dereased to.5 and 1 to.99. simultaneous inrease in 2 and derease in 2 and 1 was used to the set to 1. o model a hange i n ~. model a rak; 2 was dereased to.5 and 1 1 and 2 were dereased as they were to model a derease i n ~ he value fork was set arbitrarily to.5 as it would not effet the relationship of the results to eah other. he model resulted in the preditions shown in table 1 for the amplitude and phase of signals from a rak and a derease in permeability. he results presented in table 1 indiate that for a given frequeny the model predits only a small hange in phase regardless of whether the ondition modeled is a permeability derease or a rak. Out of the total 36 degrees on the phase display sreen that must be evaluated, a 1 or 12 able 1 - Predited signal amplitudes and phases Frequeny Condition mplitude Phase w modeled derease ~.1 rak derease ~ 1. rak

5 PCK CO DCS PBY z ~ < Q \-... _..,_ +++ '. + + \... *... *... 8:J \ ~ SSC OZD 125KHZ + 26KHZ * 5KHZ 1KHZ X 1KHZ 2KHZ Fig. 4a. Condutivity lous and permeability loi. PCK CO CCK SG DCO 1.15, ,. z ~ SSC OZD 125 KHZ * 5KHZ + 25KHZ 1KHZ Fig. 4b. Crak signal diretions. 1367

6 PCK CO ~ z D X <>......: ~ SSC OZD D Fig KHZ 1KHZ + 211KHZ X 1KHZ * 11KHZ 2 KHZ Comparison of rak, ondutivity, and permeability signal diretions. degrees differene is very small - only a 3% hange. On the other hand, there is an 18% differene in the high to low frequeny amplitude ratios of the rak signal and the hange in permeability signal. he amplitude of the rak signal at w = 1 is 5.6 times its value at w =.1, while the amplitude of the higher frequeny permeability signal exeeds its lower frequeny value by a ratio of 4.5. he ombination of phase similarities and a large high to low frequeny amplitude ratio an be useful in diserning raks from permeability hanges. DVOP OF WO FQUCY SG COPSO GOH two frequeny signal omparison algorithm was developed to determine if signal amplitudes in addition to phase ould provide a means to distinguish raks from hanges in permeability. n this algorithm, two time varying signals with X and Y omponents are sreened for phase diretion. f both signals simultaneously have the harateristi phase of a rak (and as a result also permeability) the amplitudes of the signals are measured. f the ratio of the high frequeny to low frequeny exeeds a ertain threshold, the signal an be determined to be the result of a rak in the weld. ests were onduted on ferromagneti flush ground weld samples, known to ontain short raks, in order to determine a threshold value for three different types of probes. One probe simply onsisted of a singular panake oil. seond probe onsisted of two oaxial panake oils operated in a differential mode. he third probe was also operated differentially and onsisted of two vertial oils plaed with their axes orthogonal to eah other. Sine the samples were flush ground the only 1368

7 other major signal soure was hanges in permeability. Signals from five randomly hosen raks and five hanges in permeability were obtained from eah probe type. he frequenies used depended on the probe. ah probe was driven by two frequenies, differing by an order of magnitude, that were both sensitive to raks. he mean and standard deviations of the high to low frequeny amplitude ratios are shown for eah signal type and probe type in Fig. 5. t is readily apparent that the algorithm does not perform well with the two types of differential probes. he distribution of amplitude ratios for raks and hanges in permeability overlap eah other in a manner that a threshold value that will disposition with an adequate ertainty annot be determined. he amplitude ratios from the orthogonal probe behave oppositely to what was predited, that is, the mean ratio of permeability signals are higher than that of rak signals. Better results are obtained from the single panake probe. ssuming that the amplitude ratios are Gaussian-distributed about their mean value, an optimum threshold value an be determined for the panake oil that allows raks to be dispositioned with a good degree of ertainty. threshold value of 1.39 applied with a single panake probe dispositions raks and permeability hanges orretly 89.5% of the time assuming that raks and permeability hanges are equally likely. f raks are assumed to be less likely, for example, to our only 1% of the time, the threshold value an be set higher at 1.58 and the overall perentage of orret dispositions will inrease to a maximum of 94%. he determination of the optimum threshold value and the resulting perentage of orret dispositions are obtained by using the method explained by Stremler [3] to determine the probability of error in transmission of a binary pulse-ode modulation system. S r p u D 7 sr Fig. 5. o _ ~ ~ _ - - -_- _ ~ CCK PCK u CCK u DFF PCK CCK u OHOGO Signal amplitude ratios obtained from various probe types. 1369

8 COCUSOS two frequeny amplitude ratio omparison algorithm used in onjuntion with phase angle analysis will improve eddy urrent inspetion for raks in ferromagneti welds by disriminating against the similar signals aused by permeability hanges. Probes with panake oils perform better with the algorithm than probes with vertial oils beause they have better oupling to the weld surfae. single panake probe design has disriminated raks from permeability hanges in flush ground welds; however, this probe is too sensitive to lift-off to be used on as-welded surfaes. norporation of this probe's design into a differential arrangement is neessary for the algorithm to be applied to as-welded surfaes. FCS 1. P. Holler,. Beker, and. S. Sharpe, "he ppliation of ddy Currents in Weld esting," W e l d i n ~ in the World, 22 (1984). ondestrutive est 2. H.. ibby, ntrodution to l e t r o m a ~ n e t i ethods (obert. Kreiger, Huntington, Y, 1971). 3. F.G. Stremler, ntrodution to Communiation Systems (ddison-wesley, eading,, 1982). 137