Nondestructive and quantitative evaluation of wire rope based on radial basis function neural network using eddy current inspection

Wire ropes have been widely used in elevators, lifting machinery, passenger aerial ropeways, and other related fields. Such ropes often deteriorate during their lifetime due to external or internal corrosion and abrasion, and dynamic mechanical stresses. Nondestructive evaluation methods are being increasingly applied to monitor wire ropes. In this paper, an adjustable, annular testing device, consisting of probes arranged in radial symmetry, is designed using low frequency transmission eddy current testing. The testing device is designed to overcome the usual limitations of eddy current techniques, namely the lift-off effect, edge effect, and skin effect. Peak-to-peak difference and phase difference of the response signal to the excitation signal are used as signal features, and are extracted using a numerical algorithm. A radial basis function neural network (NN) is proposed for the identification of broken wires within a wire rope. The NN models are established by offline training, with three different rope types and signal features being NN inputs, and number of wire-breaks being the output. The experimental eddy current sensor and computer measuring system has been developed to obtain characteristic data for rope samples made in our laboratory. The characteristic data are indentified by the RBF network, and the identification results show the proposed evaluation method to test if wire ropes is feasible and practical.