A Data-Driven Method for Formation Slowness Estimation Behind Casing

As the number of cased boreholes is overwhelmingly large in each oil field, estimating slowness behind the steel casing is of practical significance in understanding and monitoring formation changes during production. In many cases (like coalbed methane), sonic logging can only be implemented after casing and cementing due to unstable borehole conditions. Quite different from situations in open boreholes, the acoustic waveform acquired in cased boreholes is often contaminated by the casing waves, especially when the casing is poorly bonded with the formation. Because of its larger amplitude and longer time duration, the co-existing casing wave blends the borehole wavefield dramatically, making the extraction of formation slowness a challenging task. Many methods have been tested to solve the problem with limited success. Simulation-based methods can reconstruct missing slowness features of the formation, but they require much prior information about the borehole, mud, and cement as input. Meanwhile, present data-driven methods work to resolve the waveform interference in the transformed domain, in which the deliberately tuned time window and frequency seem critical. This paper presents a new data-driven method to get formation slowness estimates behind casing, free of additional intervention. The proposed method first separates the casing waves from the original wavefield in the time domain by utilizing the constant slowness of the casing wave (i.e., 57 us/ft). Such an initial separation is essentially a slowness-filtering process, but it is prone to remove many useful features of formation waves, still causing considerable uncertainties for subsequent slowness analysis. Therefore, we developed a masking strategy to constrain the separation process with the aim of preserving more formation waves. Moreover, the proposed workflow can be combined with data-enhancing methods to further elevate post-separation S/N by exploring the data redundancy of acoustic waveforms. Processing the ultimately separated waveform using slowness-time-coherent (STC) or dispersion analysis reveals formation slowness characteristics behind the casing. The proposed method has already been tested by synthetic examples and many field examples (in both vertical and horizontal wells), which verify its effectiveness and time efficiency in differentiating, separating, and enhancing weak formation signals behind the casing. It provides a new way of overcoming poor bonding effects and getting reliable estimates of the formation slowness. In particular, a challenging example in a horizontal well is presented, in which the compressional slowness of the formation is quite close to that of the steel casing. Because of the dominant casing waves, it is hard to discern formation compressional slowness from the original STC spectrum. As a result, the application of this new technique shows that it is capable of recovering and estimating formation slowness in spite of the severely blended borehole waves.