Title :
Eddy current crack extension direction evaluation based on neural network
Author_Institution :
Jiangsu Key Lab. of New Energy Generation & Power Conversion, Nanjing Univ. of Aeronaut. & Astronaut., Nanjing, China
Abstract :
In this paper we study the nondestructive evaluation of crack extension direction by using the differential eddy current testing sensor which is composed of two planar circumferential gradient winding spiral coils. The experiment test is set up and a series of cracks with different widths are detected. We apply a multi-layer feed-forward error-back propagation neural network for the inverse quantitative evaluation of crack extension direction. The results present that the estimation error by using BP neural network is less than 2° which meets the test requirement.
Keywords :
backpropagation; coils; computerised instrumentation; crack detection; eddy current testing; feedforward neural nets; inverse problems; mechanical engineering computing; sensors; windings; backpropagation neural network; crack extension direction evaluation; differential eddy current testing sensor; estimation error; inverse quantitative evaluation; multilayer feedforward error BP neural network; nondestructive evaluation; planar circumferential gradient winding spiral coil; Coils; Eddy currents; Impedance; Neural networks; Surface cracks; Surface impedance; Windings; circumferential gradient winding; crack extension direction; eddy current sensor; neural network;
Conference_Titel :
Sensors, 2012 IEEE
Conference_Location :
Taipei
Print_ISBN :
978-1-4577-1766-6
Electronic_ISBN :
1930-0395
DOI :
10.1109/ICSENS.2012.6411149
