A method of damage detection in conveyor belts with steel cords by NDT signal processing

The paper deals with a method of NDT (non-destructive testing) signals based damage detection in steel cords used in conveyor belts. The formulation of the problem and the goal of research is followed by the description of experiment. Main body of the paper is devoted to procedures of signal processing and condition recognition. Proposed method consists of two parts: the training part and the testing part. For both parts we use signal adaptive standardization and frame forming: for a sliding window along the signal, two statistical features: kurtosis and standard deviation of signal are extracted. In the training part two sets of frames have been constituted: one with impulses (damaged steel cords) and another one without impulses (good condition). For both sets of frames mentioned features have been extracted and then probability density functions for both features and both conditions have been estimated. Final results of training part are threshold values for both features that minimize false acceptance and false rejection errors. The proposed detection algorithm relies on comparison of unknown values of features with thresholds estimated in the training part.