Abnormal Vibration Warning Method Based on Near-Bit Measurement Data

During the drilling process,abnormal vibration of bottomhole drilling tool is often caused by factors such as improper drilling parameters and mismatch between bottomhole assembly and formation,leading to damage of drilling tool,reduction of drilling efficiency and unacceptable wellbore quality.First,an abnormal downhole vibration warning model based on Informer time series was built.Second,based on the time-frequency domain characteristics of near-bit vibration data,the normal and abnormal vibration data sets were labeled,and the mean and root mean square values of downhole vibration after wavelet conversion were used as input values to conduct warning model training.Finally,the test set data were used to test the effectiveness of the warning model.The research results show that compared with LSTM model,this model reduces E_MS by 70% and has higher prediction accuracy in terms of long series prediction results; meanwhile,aimed at long series prediction of downhole vibration mean,the occurrence of stick-slip vibration can be judged 90 s in advance.This warning model can effectively identify and warn abnormal downhole vibration,reduce drilling risk,and provide a certain technical basis for further establishing advanced intelligent drilling system.