Traitement des diagraphies acoustiques. Deuxième partie : séparation des ondes en diagraphie acoustique
Full-Waveform Acoustic Data Processing. Second Part: Wave Separation in Acoustic Well Logging
Institut Français du Pétrole
L'utilisation d'outils acoustiques à émetteurs-récepteurs multiples et enregistrement numérique permet de faire une microsismique de puits en utilisant des techniques de traitement dérivées du traitement sismique. Comme les enregistrements acoustiques sont composés de différents types d'ondes (ondes de volume réfractées ou réfléchies et ondes d'interface), une étape importante du traitement acoustique est la séparation des ondes. Cet article montre que la séparation des ondes peut être optimisée en fonction du choix du type de collection des enregistrements acoustiques et de la performance des algorithmes utilisés, dépendante du nombre de traces par collection. Les différents types de collection sont la collection émetteur ou récepteur commun et la collection à déport constant. Trois exemples de traitement de diagraphie acoustique sont présentés. Le premier exemple montre que les interférences des ondes qui conduisent à des anomalies sur l'estimation des logs acoustiques tels que le log de lenteur (Delta t) sont réduites après une bonne séparation des ondes. Le deuxième exemple est un exemple d'imagerie en puits vertical. Le traitement par filtrage de Wiener sur une section à déport constant permet de différencier les modes réfractés et d'interface des diffractions profondes (environ 4 m) créées par la présence d'intercalations dolomitiques en milieu argileux. Le troisième exemple est un exemple d'imagerie en puits horizontal. Le traitement est réalisé sur une collection point de tir commun. La combinaison de filtrage en vitesse apparente pour extraire les différents types d'ondes et de filtrage matriciel pour améliorer le rapport signal sur bruit a permis d'extraire un jeu de réflexions. La connaissance a priori de la zone réservoir a permis d'identifier les événements réfléchis en-dessous et en-dessus du drain. Cet exemple montre la nécessité d'utiliser des techniques spécifiques pour lever l'ambiguïté sur l'origine des réflexions.
The fule waveforms recorded by an array of receivers in a borehole sonic tool contain a set of waves that can be fruitfully used to obtain detailed information about the nearborehole lithology and structure. Each wave contains information that can be disturbed by the presence of the other waves. The great amount of full-waveform sonic data leads geophysicists and log analysts to implement algorithms for separating waves. The log analyst must pay attention to the choice of parameters for acquisition. The acoustic data must be recorded with well suited spatial and temporal sampling-rates to avoid spatial and temporal aliasings. The efficiency of the wave-separation algorithms depends on the choice of acoustic data gathers. Acoustic data are sorted either in a common-shot gather or in a constant-offset gather. The wave separation filters commonly used are apparent velocity filters. Less usual filters are a Wiener filter and a spectral matrix filter. The efficiency of an apparent velocity filter is enhanced by application of spectrum equalization. The signal to noise ratio is enhanced by spectral matrix filtering. We describe different processing sequences applied to a set of field examples. First field example: full-waveform sonic data in a vertical wellFull-waveform sonic data were acquired in a vertical well drilled in an anticline structure used for underground gas storage by Gaz de France. The sonic tool used was a Schlumberger Dipole Imaging tool. This tool is an eight-receiver/three-transmitter device. The receiver section contains eight dipole-monopole stations spanning 3. 5 ft, each station spaced 6 inches (15 cm) from its neighbor. The distance between the monopole transmitter and the first receiver station is 9 ft. The source was fired at equal spacings of 6 inches throughout the open hole section. For this experiment, the monopole transmitter was activated with a low frequency pulse for the purpose of generating low frequency Stoneley waves. The full waveforms taken at each receiver were recorded on magnetic tape. The data sampling rate was 40 µs, and 20 ms of the data were acquired for each trace. In the part of the well studied, the reservoir layers are clean sandstones or shaly sandstones overlain by impervious shale cap rock. The well is completed by a casing down to the top of the reservoir layers. The sonic data were recorded to study the reflected Stoneley waves. The borehole Stoneley waves are trapped modes that can be reflected when the direct borehole Stoneley wave encounters permeable fractures (Hornby, Johnson, Winkler and Plumb, 1989) or a major change in lithology (Hornby, 1989). Figure 6 shows the common shot gather processing applied to the waveforms recorded when the source was at depth 913. 3 m. The figure shows, from right to left, the raw data (A), the flattened raw data (B), the first eigensection obtained by matrix spectral filtering (C) and its associated residual eigensection (D). These sonic sections were time shifted to be directly comparable with the raw data, using the picked time of the direct wave arrival. Sonic section E shows the direct Stoneley wave and its downgoing reflected waves (mainly between 13 and 15 ms). Sonic section F shows the upgoing reflected Stoneley waves. Figure 7 shows a constant offset gather. On this section we can observe good correlation between the frequency log and lithology derived from an independent analysis. In clean sandstone, the Stoneley wave have an apparent frequency of 1150 Hz; in shaly sandstone the frequency ranges from 800 Hz to 1100 Hz. The value of the Stoneley frequency is directly related to shaliness. Maximum shalk ness is observed in the 898-903 depth interval. Figure 8 shows from top to bottom the slowness and the standard deviation of the slowness computed from raw sonic data and filtered sonic data (first eigensection) as well as the frequency log and its standard deviation. We can note that the standard deviation is very low (average value = 20 µs/ft) by comparison with the time sampling rate (40 µs) and the depth sampling rate (0. 5 ft). After filtering, the standard deviation is reduced whatever the depth, but mainly in the 897-900 depth interval. This depth interval is associated with maximum shaliness. The maximum standard deviation observed at 903 m is due to the interference of downgoing reflected waves and a direct Stoneley wave. This reflected Stoneley wave is not filtered in the commonshot gather space. Figure 9 shows the energy log derived from the set of first eigenvalues associated with the first eigensections. This log correlates in depth with the frequency log and exhibits the interval depth where the reflected downgoing Stoneley waves are very strong. Figs 10 to 13 show the sonic fullwaveform sections after constant offset gather processing. Fig. 10 shows a constant offset gather obtained from the set of the first eigensections (Fig. 6(E)). Fig. 11 shows a constant offset gather obtained from the set of the residual eigensections (Fig. 6 (F)). This sonic section mainly contains upgoing reflected Stoneley waves. The sonic section shown in Figure 10 has been filtered in order to separate the direct Stoneley wave and downgoing reflected Stoneley waves. The results are shown in Figs. 12 and 13. Second field example: full-waveform sonic data and VSP (vertical seismic profiling)The sonic tool used was a SEMM monopole sonic imaging tool. This tool is a four receiver/one transmitter flexible device. In the configuration used, the distance between two receivers was 50 cm. The distance between the monopole transmitter and the first receiver station was 3. 25 m. The source was fired at equal spacings of 5 cm throughout the open hole section of the well. The data sampling rate was 5 µs and 5 ms of the data were acquired for each trace. The sonic data were used to compute an accurate compressional-wave velocity-log obtained after picking of the refracted arrivals (Fig. 14) and then processed to obtain a near-borehole image. For that purpose, the sonic data were collected in constant offset gathers. For each gather (Fig. 15) versus depth, the wavefield separation was performed using a set of trace-pair Wiener filters. Each filter was applied in a time-window in order to select a specific wave. After filtering, the residual sonic section contains the noise and the other waves. In this field case, wave separation was performed in two steps. In the first step, a Wiener filter was applied in the 2-to-5 ms time-interval. It was used to enhance the interface modes (Fig. 16). The interface modes are very different in, sandstones and in shales. The transition between these two geological formations are clearly marked at 1 018 m. In shales (1 002 - 1 018 m depth interval), the interface modes have a high frequency content and propagate with a constant velocity, which is the mud velocity. They are mud waves. In sandstones, the interface modes have strong amplitudes and a low frequency content. They are Stoneley waves. After filtering, the residual sonic section shows the refracted modes and the leaky modes (Fig. 17). In sandstones, we can notice a residue of Stoneley modes at about 2. 5 ms. In shales, dipping sonic events appear and the residual waves and noise have much stronger amplitudes than they can have in sandstones (1 018-1 060 m depth interval). In the second step, a Wiener filter was applied to the constant-offset sonic-data gather, after elimination of interface modes. The filter was applied in the 0 to 2. 5 ms time interval to enhance the refracted and leaky modes (Fig. 18). The residual constant-offset section was corrected for normal moveout using the velocity function derived from the acoustic log. At about 1 010 m, diffractions can be observed on the filtered sonic section (Fig. 19). The interpretation of these diffracted patterns shows that they were created by heterogeneities in the shale medium due to dolomitic intercalations situated about 4 m from the well. Note that the diffractions, at the acoustic log scale, appear as variations in the shape of the reflected event associated with the dolomitic zone, at the VSP-CDP stack scale (Fig. 19). Third exemple: full waveform sonic data in horizontal well on test siteA highly deviated well and a vertical well were drilled in a limestone quarry situated in Burgundy, France. The vertical well was drilled to establish a vertical geologic cross-section of the quarry and to identify the acoustic impedance variation boundaries. For this purpose a VSP and a set of logs were recorded in the vertical well using a SEMM well seismic-logging unit. The set of logs includes : caliper, density, resistivity, neutron and full waveform acoustic logs. Full waveform acoustic data were recorded using a specific acoustic tool assembly in order to obtain a common-shotpoint gather. The source-receiver distance varied from 3. 60 to 7. 88 m. The sampling rate was 10 µs. The recording length was 10 ms. The receiver spacing was 1 cm. In the common-shot gather, the refracted arrivals and borehole modes have linear moveout across the offset range, while the reflected events have hyperbolic moveout. The acoustic data must be recorded with well suited spatial and temporal sampling-rates to avoid spatial and temporal aliasings. Under these conditions, all the refracted arrivals and borehole modes can easily be filtered by a set of apparent velocity filters. Each filter is designed to cancel a selected wave, characterized by its velocity. The processing included spectrum equalization in the 15-35 kHz frequency bandwidth (Fig. 22) and filtering of the linear moveout arrivals. After processing, the acoustic data (Fig. 26) show reflected events up to 6 ms. These reflections come from reflectors situated in the oolithe blanche reservoir layer up to a distance of approximately 10 m. The use of information coming from vertical well (VSP and log) and synthetic modeling leads us to differentiate the reflections coming from discontinuities situated above (A) and below (B) the horizontal well. Without a priori information, the problem of above/belowambiguity must be solved before stacking. For this purpose, the above/below reflection-separation can be performed by comparison of a set of acoustic runs obtained with an off centered logging tool (Mari, 1991). Another way is to use an acoustic logging tool with a modified radiation-pattern in order to illuminate either aboveor belowdiscontinuities.
© IFP, 1992