Data line change detection with application to Mud Logging data processing

Motivated by Mud Logging data processing in petroleum exploitation, the detection of changes in data line direction is studied in this paper. Because of noise corruption to all measured variables, the classical regression model is not suitable. After an appropriate formulation of noise corrupted data line, the problem of noise covariance matrix estimation is first considered, then a numerically efficient generalized likelihood ratio test is derived for direction change detection. This detection method, applied to Mud Logging data processing, is now integrated in INFACT, an industrial software for petroleum exploitation.