Wire Rope Broken Wires Damage Quantitative Identification Of Key Technology Research

In general, the working environment of wire ropes is severe, wire ropes is easilyinfluenced by a variety of uncertain loads and harsh environmental factors, resulting in broken,wear, corrosion and other defects. The event of a rupture of wire ropes can cause seriouspersonal and economic loss; at the same time, due to the complex rope damages, the differentconditions of the working environment of wire ropes and a variety of types of wire ropes, therope testing instruments has many problems, such as low accuracy and intelligence, poorreal-time dynamic and the difference effect from different personal experience. The limitationis strong. Therefore, the study of intelligent and efficient rope detection technology andequipment is necessary.This paper focuses on the detection of the damage of wire ropes. Concerning magneticflux leakage principle, wire ropes damage detection system is designed. The main part is thewavelet delousing of the signal and the selection and extraction of the characteristic value;quantitative recognition methods based on neural networks and support vector machines isconcerned, and wire ropes injury quantitative recognition system is designed.First of all, the form of wire rope damages and principle is analyzed, the characteristicsof the leakage magnetic field of wire ropes is discussed. A damage detection system isestablished, including the design of exciter, magnetizing steel wire rope to the magneticsaturation for making the broken wire to produce a leakage magnetic field; the design ofdetector, detecting the leakage magnetic field and converting it to electrical signal forfacilitating the subsequent acquisition; the design of signal acquisition system, converting thecontinuous signal to digital signal to enter into the computer for subsequent processing.Secondly, the theory of wavelet denoising is studied. With Matlab a wavelet delousingprogram is created. Feature extraction of the signal is selected, including the signal peak, thewidth, the area of wave and wavelet energy. Considering the characteristic distinctionbetween the degree of its characteristics, the signal peak, the area of wave and wavelet energyis retained. The three characteristics of the signal is got for subsequent quantities pattern recognition processing.Again, the quantitative recognition system based on neural network is designed. Neuralnetwork theory is studied. The paper discusses eigenvalue analysis and optimization of neuralnetwork, selection method of network training set and test set, the selection of networktraining objective, and from the three aspects discussed the method to optimize networkparameters and performance. Then BP neural network model is built. With real signal of wireropes quantitative recognition system for broken wires based on the neural network is trainedand tested, which can prove the system has a good performance.Finally, the quantitative recognition of the broken wire rope based on support vectormachines is designed. The theory of support vector machine pattern recognition is analysised.The choice of the kernel function and search of SVM parameter is discussed. A support vectormachine model is built. With real signal of wire ropes the broken wire quantitativerecognition system based on support vector machine is trained and tested, and found that thequantitative recognition of the broken wire ropes based on support vector machines had betterperformance. Support vector machines and neural networks were analyzed and compared inquantitative recognition, and discusse the reasons for this result.