A Company Committed to Problem Solving
and Research in the Engineering and Physical Sciences


Harold A. Sabbagh, Elias H. Sabbagh, R. Kim Murphy
Victor Technologies, LLC
P.O. Box 7706, Bloomington, IN 47407-7706 USA
John Nyenhuis
Purdue University
West Lafayette, IN 47907-1285

(Presented at ASNT Fall Conference and Quality Testing Show--2000
Indianapolis, Indiana November 14-17, 2000)


Advanced turbines, such as the GE Frame 7FA/9FA, are used in applications ranging from aerospace to land-based power generators. These turbines are fired at higher temperatures (1850° F-1950° F), and utilize optimum cooling of hot section components. Because of the higher operating temperature, the performance and durability of the first stage blades has become one of the prime life-limiting factors. Individual blades are nickel-based GTD 111 alloys, that are protected by sacrificial metallic coatings to extend service life. The first-stage blades are especially important, and it is desirable to develop an in-situ NDE system to monitor, evaluate, and predict remaining coating life. The coatings used on the turbine blades include CoCrAlY and NiCoCrAlY, with a top aluminide coating (GT29+, GT33+, respectively), and a NiCoCrAlY coating, called PWA 286 (EPRI(1), 2000).

Victor Technologies has been performing research into the nondestructive characterization of in-service high-temperature metallic (such as MCrAlY) thermal barrier coatings (TBC) applied by vacuum plasma spray on Ni-based superalloy turbine blades. Figure 1 illustrates an as-coated PWA286 coating on a GTD111 substrate and the same coating after aging.

PWA286 Coating
Figure 1: An as-coated PWA286 coating on a GTD111 substrate (left) and the same coating after 2400 hours (right).

All coatings form a thin protective adherent layer of Al2O3. As the protective oxide spalls off during service, aluminum in the coating diffuses out to re-form the protective oxide layer, and also diffuses into the substrate and causes the interdiffusion zone to increase in thickness. Thus, the coating degradation is represented by the transformation of beta-phase NiAl into a gamma-matrix of solid Ni solution. The gamma-matrix is represented in the figure as beta-phase depleted zones. Therefore, the diffusion zone thickness increases from the addition of the beta-phase depletion zone (Zone 1), that is located just below the coating (EPRI(1), 2000).

The primary objective is to estimate the equivalent thickness of the aluminum beta-phase content of the PWA286 coating. This information is essential to maintaining the integrity of blades, because it allows the timely repair or refurbishment of coatings to extend the service-life of operating blades. Further, it is desirable to obtain the interdiffusion layer thickness, since this information indicates the level of blade exposure to service temperature. The overall remaining coating thickness indicates the reduction of the coating thickness caused by the oxidation-induced degradation of the top beta-phase depleted layer (EPRI(1), 2000).

We have performed a number of model calculations of the TBC problem, using our proprietary eddy-current NDE code, VIC-3D®, and have determined important system features, such as operating frequency, coil characteristics, electronic test equipment considerations, and means of accelerating the computations. As a result, we have concluded that the optimum frequency range for performing the inversions required of the TBC problem is 10MHz to 200MHz. Successful inversions in this range will allow us to achieve the desired resolution for these extremely thin coatings.

Because there are no commercially available eddy-current instruments that operate in this frequency range, we used the Hewlett-Packard HP 3577A Network Analyzer, that is designed to operate over the frequency range of 5Hz to 200MHz. This instrument measures the reflection coefficient of a one-port network (namely, the loaded coil), from which the impedance data, that are the input to the inversion algorithm, are determined.

There are several steps to be carried out during the testing. First will be to characterize the probe in free-space, then to collect the impedance data, and finally to invert these data using our `8layer-algorithm.'

Determining Coil Parameters

The coil used is a Zetec Z0000595-1 ultra high frequency pancake coil, that is designed to operate in the range of 10 to 100 MHz. Its dimensions are 0.090 inch outer diameter, 0.025 inch inner diameter, 0.008 inch height, and is wound with 19 turns. For the purpose of this demonstration, the coil was housed in a Plexiglas block, and the test pieces were laid on the block, over the recessed coil.

VIC-3D® assumes that the probe coil is an ideal inductor that carries a uniformly distributed current within a rectangular window. This is the basic assumption in most theoretical models for eddy-current NDE. In practice, any real coil exhibits self-capacitance and resistance, as well as additional capacitance associated with coaxial cables or other connections to the coil. Furthermore, the assumption of a uniform current-distribution cannot be supported in a real coil, because such coils use windings of wire with circular cross-sections, and these windings are never uniformly distributed, as shown in Figure 2. Furthermore, skin and proximity effects give rise to nonuniform current distributions within the wires themselves. These deviations from ideal behaviour must be taken into account if good agreement between theory and experiment is to be obtained over a significant frequency range.

Coil Turns
Figure 2: Showing a nonuniform distribution of turns within a typical real coil.
A real eddy-current probe can be modeled by the equivalent circuit of Figure 3, in which L0 and R0 are the (low-frequency) inductance and resistance of the probe, ZW is the impedance of the workpiece (with or without the flaw), coupled back into the probe circuit, and YP is the parallel admittance that accounts for the remainder of the probe and its connecting cable. For example, YP could be the self-capacitance of the coil, as in Figure 4, which would account for the resonant behavior of the probe. In any case, we assume that Yp goes to zero as the frequency goes to zero.
Equivalent Circuit
Figure 3: The equivalent circuit of a real eddy-current probe. We assume that YP goes to zero as the frequency goes to zero.
Equivalent Circuit of Real Coil
Figure 4: The equivalent circuit of a real eddy-current probe, showing the presence of the self-capacitance of the probe, which accounts for the resonant behavior of the probe.

VIC-3D® has a filter that removes the effects of YP, thereby correcting the data obtained when using a real coil, and transforming them into data that would be obtained by using the equivalent ideal coil.

Data Collection

Using the HP network analyzer and the Zetec coil, we measured the reflection coefficient, S11, from one to fifty megahertz, in one-megahertz steps, for each of the fourteen test samples (top and bottom of each sample). S11 is recorded as a magnitude and phase (in degrees), as shown in the following figures. Figure 5 shows S11 for the coil in air, i.e., away from the test samples.