The Research Of Derrick Damage Identification Based On The Wavelet Analysis And The Neural Network

During the drilling process, the derrick may appear the damage because of many reasons. In order to insure the security,integrity and the durability of the derrick, it's necessary to use the some effective methods to monitor and to appraise the existing derrick's healthy condition. In other words the derrick must carry on the health monitoring, and the most coral content is the derrick damage identification.This paper introduces the research methods of the structural damage identification, and the wavelet analysis theory is introduced, takes the advantage of its superiority on the processing unstable signal, the wavelet analysis is introduced to the derrick damage identification, a damage evaluation index based on normalized energy characteristic vector was proposed. A test by simulate the single and many place and the different extent of damage on the derrick model is being accomplished, obtained the data of the vibration signal of the derrick model during the different load. After the signal was carried on the pretreatment, three-layer wavelet-packet energy analysis to the test data followed, the normalized energy vector was obtained. The damage evaluation index which based on the normalized energy vector was calculated, it is proved that the normalized energy characteristic vector index can localize the derrick damage accurately. After contrast the index of the different damage extent, it is proved that the index could reflect the trend of the derrick damage. Besides, through analysis the de-noise signal, the analysis outcome revealed that the index have strong ability to avoid the infection of the noise. The wavelet function category has many series, maybe the analysis outcome have a great contrast with different wavelet function during the wavelet-packet energy analysis. By compared the haar wavelet function with others, the effect of the damage identification which with the haar wavelet function is excelled the others. Finally, the normalized energy characteristic vector which obtained from the wavelet-packet analysis is taken as the training samples of the neural network, two BP neural networks which identify the derrick damage location and the damage severity are established, and they processes the derrick damages information accurately, visually and fast.