Drill String Defect Automatic Indentification Method Based On Magnetic Memory Detection

Drill string is an important drilling equipment.Drill string damage will cause huge losses,and may even lead to abandonment.Therefore,the detection of drill string damage is an important part of drilling operations.Now,the massive data generated by drill string damage detection rely on manual processing,which is inefficient.Therefore,the effective and accurate identification and evaluation of drill string defects is the goal of drilling inspection.The realization of automatic defect recognition is of great significance to eliminate defective drilling tools,reduce drilling accidents and improve drilling efficiency.In this thesis,automatic recognition of drill string defects is studied based on magnetic memory testing data.(1)The stress analysis and failure analysis of the drill string are carried out.The characteristics,causes and types of typical damage,such as crack,corrosion and puncture defects are analyzed.(2)The feature extraction method of magnetic memory signal was studied,and a multi cycle scanning test was performed on four typical defects of prefabricated longitudinal grooves,pass defects,lateral engraving and circumferential injuries by using three dimensional magnetic memory detection platform.The test analyzes and processes the data to obtain the features of magnetic memory and gradient signals of various defects.The magnetic memory was able to well characterize the defect.The characteristics of magnetic memory and its gradient signal are obtained by analyzing and processing the data.It is proved that magnetic memory can well characterize defects.(3)The difference between the magnetic memory detection results of different defects is analyzed,and a 9-dimensional feature system which can well characterize the type of defect is proposed,and the characteristics of features are analyzed by the method of random forest.Two defects identification models based on SVM and random forest are established.The model is trained through experimental data,and the accuracy rate of defect recognition is 92.3% and 84.6% respectively.(4)A software that can automatically extract and recognize the characteristics of the defects is developed by MATLAB,which provides an effective tool for identifying drill string defects.It provides a reference for solving the problem of identification of real-time drilling string defect.