Feature Extraction and Quantitative Inspection for Broken Wires in Wire Ropes Based B-Wavelet and WNN

摘要

A data compression (at a ration of 1:4) method based on linear B-wavelet decomposition is proposed for wire rope signals without professional coding. This method saves the natural information of signals and reduces the high frequency noise of signals. WNN has been selected for the quantitative inspection of broken wires. WNN models and weights study algorithms suitable for quantitative inspection of broken wires are discussed as well. The input layer of WNN is the inner product of feature vector and wavelet using wavelet nonlinearity. The experiment results show that the WNN model displays a much higher level of generalization, shorter compution time and accurate inspection in comparison with BP-Network.