| Abstract: | Nondestructive testing (NDT) of highly inhomogeneous materials, such as carbon-filled plastics, composites, especially fiber-reinforced
and using coarse-grain fillers, etc., is a quite difficult problem. Thus, when these materials are scanned by sensitive elements
of conventional systems for NDT, the instability of initial signals caused by the structure of the material may exceed the
signal emitted by the defect which, therefore, remains undetected. We investigate a principal possibility of detecting defects
by representing them in the form of finite regions whose spatial signals are characterized by distribution functions whose
shapes differ from the shape of the same function for the intact material. We analyze the most difficult case of complete
overlapping of these functions. By applying both theoretical methods and numerical simulation, we demonstrate the possibility
of detection of defect zones in highly inhomogeneous materials which are not visible in the process of human inspection even
under the most favorable conditions for observation. We present a method for the calculation of the optimal statistical filter
and a system of scanning of the investigated material in local zones. The experimental data of automatic detection are compared
with the data of human control according to the Neumann-Pearson risk criterion. The regions of the optimal application of
automatic and manual inspection are indicated.
Karpenko Physicomechanical Institute, Ukrainian Academy of Sciences, L'viv. Translated from Fizyko-Khimichna Mekhanika Materialiv,
Vol. 32, No. 4, pp. 97–108, July–August, 1996. |
