Incident early warning based on sparse autoencoder and decision fusion for drilling process

Complicated geological environments lead to a high risk of drilling incidents. Incident early warning for drilling process is in demand for industry field. An incident early warning method for loss and kick based on sparse autoencoder and decision fusion is proposed in this paper. Sparse autoencoder is employed to detect the abnormality of the drilling parameter time series. Mann-Kendall trend test approach is performed to extract the trend of the time series that is detected as abnormal. The abnormality detection and trend extraction results of each drilling parameter are fused to get the final incident early warning result. Experiments are executed with the actual data collected from a practical drilling process. The experiment results indicate the effectiveness of the proposed method.