Research On Acoustic Emission Detection And Signal Analysis Methods Of BOP

BOP is the core of well control equipment during the process of oil drilling,which is used for controlling wellhead pressure,and is an important technical support to ensure safe and efficient operation of oil drilling production.In using,the shell is easy to produce fatigue crack,even some new production BOP,there are cracks in the shell for casting,machining,heat treatment and welding.The BOP with active defects prone to leakage or cracking in working condition,once the BOP failure,which will lead to a blowout fatal accidents.So timely inspection and found the defects existing in the equipment,and evaluation generation and propagation of cracks in the entire pressure process,which is very important.Internal defects and the crack in BOP shells are mostly in the position of the geometry changes.due to the complexity of the shape and structure of BOP shell.The magnetic and permeation methods can not detect the internal defects of the shell.Although conventional ultrasonic flaw detection or ray non-destructive testing methods can identify the defects,some static defects in suing is not necessarily extended,which are security defects.If all repairing,will bring huge economic losses.According to the theory of AE test,the active defects in the BOP shell can be found by AE detection,only potentially hazardous defects are repaired,to avoid unnecessary losses,and increase the security.The paper studies firstly the acoustic emission characteristics of the BOP shell materials,ZG25CrNiMo,produce four specimens which can embody the stress state in BOP shell,study their AE characteristics during the stretching process,and the results show that the AE parameters characteristics is coincide with material tensile damage process,according to which can evaluate the state of the BOP shell,and the signal spectral characteristics of the different damage stages is different.For the the difficulties of the rectangular shell positioning of ram BOP,research three-dimensional positioning method based on SVM,selected support vector machine classifier,determine the kernel function of SVM,and research the regression principle based on the of multi-output support vector.Since the performance of SVM is related with the kernel function parameters and the error penalty parameter,this chapter researches the selection method.Improved for the the premature convergence shortcomings of the PSO algorithm,and established the mathematical model based on fitness-based calibration and NPSO.Carried out three-dimensional positioning method experimental study of ram BOP rectangular shell based on SVM,supported vector model,and the positioning results show that ram BOP shell housing defects can be very well positioned by the support vector machine,the positioning accuracy is greatly improved.Against the distinction between the small specimens made of BOP shell material and BOP shell used in the field,AE characteristics test BOP of BOP shell during pressure and damage is carried out,AE characteristics of crack propagation in the shell with cracks during pressure is researched.The result shows that the AE parameters and spectral characteristics of BOP shell is same with the small specimens,which means that AE characteristics of the shell material can reflect the AE characteristics of actual shell structure.Finally,the methods of AE signals analysis of the BOP is researched,including feature extraction and pattern recognition.Put forward that take wavelet spectrum coefficients,energy spectrum coefficients,and margin factor representing Signal strength characteristics as BOP AE signals characteristic parameters.Establish Pattern recognition model of K-means clustering algorithm based on SOFM,which overcome the disadvantage of determining of the Cluster centers unreasonable in Traditional K-means clustering algorithm.Carry out pattern recognition to the test data of the BOP shell material tensile process and BOP shell pressure damage process,and the result show that the accuracy of Improved K-means clustering algorithm is high.