Abstract
The use of logging-while-drilling (LWD) imaging tools in real-time decision making and post-drilling analysis has become commonplace. However, image noise and processing errors due to the inherent measurement physics can propagate errors and thus complicate interpretation of open-hole log data and images. For example, the standard density images tend to amplify image noise from small borehole irregularities. Among different borehole irregularities, spiraling is known to occur more frequently with conventional rotary assemblies, steerable motor assemblies, and rotary steerable assemblies. When the cyclic noise amplitude from spiraling becomes large relative to the measurement of the primary interest, it grossly affects measurement accuracy of bulk density, photoelectric, and neutron porosity.
This paper shows novel methods to remove cyclic noise from formation evaluation (FE) images by applying frequency-domain filtering. Although the initial attempt of fast Fourier transforms1 illustrates the straightforward concept, it is seldom used due to implementation issues requiring interactive filter design and intensive operator intervention. Recently, a new method2 has been designed to improve the filtering process. Additionally, the new adoptive filter designs automate the cyclic noise removal process from the FE images. Stand-alone software has been developed to process the entire image logs from one well, without human supervision. The software adoptively modifies the filter behavior as borehole oscillation noise characteristics change with formation, drilling assemblies, hole inclination, and depth.
To quantitatively examine its validity, the algorithm is then applied to various field data, not only on low-resolution density images but also on higher-resolution density images. The new algorithm has been proved by bringing considerable improvement to the image quality without any artificial interruptions. The rugosity effect is significantly reduced, and the apparent resolution of bed boundary features increased.
Furthermore, comparative field examples illustrate the improvement in image feature extraction. This new method is critical, for example, for small fracture detection and for accurate dip analysis in irregular boreholes. This new method is not only effectively used for density images but also applicable to any other borehole images, such as resistivity and ultrasonic images.
