Improved First-Motion Algorithm to Compute High-Resolution Sonic Log

摘要

Estimation of formation slowness is of significant value forrnpetrophysical and geomechanical applications. To do this, arnsource excites a signal in the borehole, which couples with,rnand propagates within, the formation prior to being recordedrnby receivers. Generally, sonic tools are composed of one orrnmore transmitters and an array of receivers. A next generationrnsonic tool developed by Schlumberger is composed of 3rnmonopole and 2 dipole transmitters. The receiver array isrncomposed of 13 receivers separated by 6 inches. Waveformsrnare digitized at the sensor with a dynamic range of 16 bits.rnDigitizing the waveform at the sensor will guarantee optimalrndata quality as the waveforms will not be corrupted byrnelectronic noise or other undesirable effects. The generalrntechnique used to measure formation slowness is based onrnsemblance processing which was first proposed 20 years ago.rnSemblance processing is robust and provides reliable resultsrnin most cases but its vertical resolution is linked to the lengthrnof the tool array. Practically, it means it is not possible torndetect beds that are smaller than a tool’s array length. Hence,rnit can be difficult to compare slowness logs with other datasetsrnof much higher vertical resolutions.rnIn order to alleviate this problem, we propose an improvedrnfirst motion algorithm to compute a high-resolution sonic log.rnThe technique considers that the part of the waveform beforernand after the first break can be modeled as an AR process. Thernkey parameters of this approach are the “model order”, whichrndescribes the data and the part of the waveform to bernmodeled. In this paper we will show that the use of thernBayesian Information Criterion combined with therncomputation of the envelope of the signal will allow for anrnautomatic computation that is suitable for wellsite operations.rnWe will demonstrate that this first motion technique does notrnrequire any human intervention and in most cases provides arnrobust and reliable result. This algorithm will then be applied on real data and the high-resolution log obtained will berncompared with other high vertical resolution measurements.