The nondestructive evaluation (NDE) of high-temperature coatings is
one of the important factors in achieving a high-level of structural
integrity in advanced gas turbines. This problem is of great interest
to users of gas turbines, including power utilities. Not only are the
"established" public utilities relying more and more heavily on
combustion turbines for power generation, but the independent producers
that have emerged since deregulation are relying exclusively on
land-based engines, and they can ill afford a blade failure.
Victor Technologies, LLC has been performing research into the
nondestructive characterization of in-service high-temperature metallic
(such as MCrAlY) coatings applied by vacuum plasma spray on Ni-based
superalloy turbine blades. When new, such coatings are not diffusive --
they present quite sharp interfaces with the base metal; the
interdiffusion layer is thin (~10µm) as compared to the coating
thickness(>100µm) (see G. Antonelli, M. Ruzzier, and F. Necci,
"Thickness Measurement of MCrAlY High-Temperature Coatings by Frequency
Scanning Eddy Current Technique," Presented at the ASME 97-GT-1
International Gas Turbine & Aeroengine Congress & Exhibition,
Orlando, Florida, June 2 to June 5, 1997.)
Coatings, such as MCrAlY, possess physical and chemical properties
quite similar to those of the Ni-superalloys. This means that the
electrical conductivity may vary by only 3% to 10% between the two
metals. Thus, part of the challenge is to develop an eddy-current-based
algorithm that is able to resolve such small changes. A further
challenge arises because it is found that the coating may become
ferromagnetic with use, which means that the algorithm must also
determine magnetic permeabilities, as well as electrical
conductivities. Finally, the blade curvature must ultimately be taken
into account.
In the problem
displayed here, we reconstruct a two-layer system, in which there are
four variables to be reconstructed, the conductivity and permeability of
each layer. The top layer is 0.2 mm thick, and the bottom layer is an
infinite substrate. The data comprise ten impedance measurements at
frequencies, 1E5, 2E5, 4E5, 8E5, 1.6E6, 5E7, 6.25E7, 7.5E7, 8.75E7, and
1.E8 Hz. The data were corrupted by the addition of zero-mean random
noise, with a variance of 0.01. This is probably a rather large
variance, but it allows us to demonstrate the algorithm, anyway. The
left-hand portion of the figure shows the original structure, and the
right-hand shows the reconstruction, using the data just described.
Clearly, eddy-current measurements can allow the simultaneous
determination of conductivities and permeabilities with excellent
precision.