Research On PDC Bit Working Condition Intelligent Identification Method

In oil drilling engineering,PDC bit is the main tool for rock breaking.Because it is difficult to accurately judge the working condition of the drill bit in real time through the empirical method,there may be cases where the drill bit is replaced in advance before it is fully functioning,or the drill bit is replaced when the drill bit is worn or damaged.At the same time,the formation is complex and changeable will also cause unpredictable and complex working conditions of the drill bit during the drilling process,which will seriously affect the construction efficiency,even necessary to stop and trip out for processing,delaying the working hours and causing a huge impact on the overall economic benefits of the project.Therefore,it is of great significance to accurately identify the PDC bit working conditions in real time for determining the tripping time and adjusting drilling measures in time,thereby increasing the drilling speed and reducing the drilling cycle.This thesis applies the machine learning method,and focuses on the research on the intelligent identification method of PDC bit working conditions.Through literature research and combined with field logging and well history data.First,the working conditions of PDC bit are divided into bit wear,bit mud bag and other conditions,the wear condition is divided into three wear levels: initial wear,normal wear,and severe wear by using the IADC bit wear evaluation method.Then according to the multiple influencing factors that lead to the working condition of the drill bit,the principal component analysis method is applied to reduce the dimension,and the six principal components that affect the working condition of the PDC bit are selected.The six principal components are used as the input parameters of the model,and the PDC bit conditions are used as the output parameters,three machine learning methods,BP neural network,GA-BP neural network and decision tree,are used to establish the intelligent recognition model of PDC bit working condition respectively.Use 3000 sets of valid samples to train the algorithm and optimize the model parameters,through the analysis of training accuracy,prediction accuracy and result error,the performance of three intelligent identification models for drill bit working conditions is evaluated,and a new method for intelligent identification of PDC bit working conditions is obtained.Among them,the recognition model of PDC bit working condition based on GA-BP neural network has the best performance,the training accuracy is 96.12% and the prediction accuracy is 94.88%.Based on the above model,the software for the intelligent identification model of the PDC bit working condition was designed by using the App Designer tool in the Matlab R2020 a software,and an example was verified.Based on field logging parameters,the accurate identification of common working conditions of PDC bits under the efficient coupling of multiple data is realized.The research results of this thesis can provide an important reference for intelligent identification of PDC bit conditions and timely adjustment of drilling measures.