Sonic Slowness Research Articles

Highest resolution high signal-to-noise ratio sonic slowness and application to Brothers Volcano, Kermadec Arc, New Zealand

High resolution and accurate acquisition of underground sonic properties is essential for energy detection and earth exploration. An effective method is developed to accurately calculate sonic slowness at the highest resolution (receiver spacing of 0.5 ft) with a high signal-to-noise ratio. The method uses supervirtual interferometry to reconstruct a large number of waveforms for slowness extraction using redundant information from overlapping interreceiver spacing. Synthetic calculations are used to verify the validity of the method, especially in the limited case of strong noise interference and the presence of damaged channels in the array receivers, to accurately calculate the formation of sonic slowness. The theory is applied to process and interpret low data-quality field measurements acquired at Brothers Volcano from the International Ocean Discovery Program. The processing analysis indicates that in the case of very weak formation signals, the residual tool waves from the wireline sonic logging can still interfere with the high-resolution processing, whereas this method can effectively suppress this interference and accurately obtain the formation sonic velocity.

Uncertainty and explainable analysis of machine learning model for reconstruction of sonic slowness logs

Logs are valuable information for oil and gas fields as they help to determine the lithology of the formations surrounding the borehole and the location and reserves of subsurface oil and gas reservoirs. However, important logs are often missing in horizontal or old wells, which poses a challenge in field applications. To address this issue, conventional methods involve supplementing the missing logs by either combining geological experience and referring data from nearby boreholes or reconstructing them directly using the remaining logs in the same borehole. Nevertheless, there is currently no quantitative evaluation for the quality and rationality of the constructed log. In this paper, we utilize data from the 2020 machine learning competition of the Society of Petrophysicists and Logging Analysts (SPWLA), which aims to predict the missing compressional wave slowness (DTC) and shear wave slowness (DTS) logs using other logs in the same borehole. We employ the natural gradient boosting (NGBoost) algorithm to construct an Ensemble Learning model that can predicate the results as well as their uncertainty. Furthermore, we combine the SHAP (SHapley Additive exPlanations) method to investigate the interpretability of the machine learning model. We compare the performance of the NGBosst model with four other commonly used Ensemble Learning methods, including Random Forest, GBDT, XGBoost, LightGBM. The results show that the NGBoost model performs well in the testing set and can provide a probability distribution for the prediction results. This distribution allows petrophysicists to quantitatively analyze the confidence interval of the constructed log. In addition, the variance of the probability distribution of the predicted log can be used to justify the quality of the constructed log. Using the SHAP explainable machine learning model, we calculate the importance of each input log to the predicted results as well as the coupling relationship among input logs. Our findings reveal that the NGBoost model tends to provide greater slowness prediction results when the neutron porosity (CNC) and gamma ray (GR) are large, which is consistent with the cognition of petrophysical models. Furthermore, the machine learning model can capture the influence of the changing borehole caliper on slowness, where the influence of borehole caliper on slowness is complex and not easy to establish a direct relationship. These findings are in line with the physical principle of borehole acoustics. Finally, by using the explainable machine learning model, we observe that although we did not correct the effect of borehole caliper on the neutron porosity log through preprocessing, the machine learning model assigned a greater importance to the influence of the caliper, achieving the same effect as caliper correction.

Estimation of Petrophysical Parameters of Carbonates Based on Well Logs and Laboratory Measurements, a Review

The purpose of this review paper is to show the possibilities of carbonate reservoir characterization using well logging and laboratory measurements. Attention was focused on standard and new methods of well logging acquisition and interpretation including laboratory experiments to show a part of the history of carbonate rock investigations as hydrocarbon or water reservoirs. Brief information on the geology, mineralogy and petrography of carbonate rocks was delivered. Reservoir properties, i.e., porosity (including fracturing), permeability, and saturation, were defined to emphasize the specific features of carbonates, such as fractures, and vugs. Examples of methodologies were selected from the commonly used laboratory techniques (thin sections examination, mercury and helium porosimetry, X-ray diffraction—XRD) combined with the standard well logs (bulk density—RHOB, neutron porosity—NPHI, sonic slowness—DT, and deep resistivity—Rd) to show the methods that have been used since the very beginning of the scientific and engineering studies of carbonates. Novelty in well logging, i.e., resistivity and acoustic imaging, nuclear magnetic resonance–NMR, dipole shear sonic imaging–DSI, and a spectral neutron-gamma log-geochemical device–GLT combined with modern laboratory investigations (NMR laboratory experiments, scanning electron microscopy SEM), showed how continuous information on mineral composition, porosity and saturation could be obtained and juxtaposed with very detailed laboratory data. Computed X-ray tomography (CT) enabling the 2D and 3D analyses of pores and fractures was presented as a quantitative methodology, effective in pore space characterization, revealing rock filtration abilities. Deep learning and artificial intelligence were used for joining various types of data. It was shown that thanks to new computational technologies original data from very small samples (micro scale), extensively describing the flow ability of the reservoir, could be extended to mezzo scale (core samples) and macro scale (well log images). Selected examples from the published papers illustrated the review. References cited in the text, together with the issues included in them, were the rich source of the practical knowledge processed These were checked by the authors and could be used in other projects.

Unsupervised time series clustering, class-based ensemble machine learning, and petrophysical modeling for predicting shear sonic wave slowness in heterogeneous rocks

Shear sonic logs are critical for formation evaluation, rock physics, quantitative reservoir characterization, and geomechanical studies. Although empirical and conventional machine learning (ML) have been used for shear sonic slowness estimates, both approaches suffer from multiple fundamental problems. New approaches to ML, namely, unsupervised multivariate time series clustering (Toeplitz inverse covariance-based clustering) and class-based ensemble ML, are integrated with geologic information and petrophysical inversion-based multimineral models to predict rock properties of the Wolfcamp Formation in the Midland Basin, United States. These new approaches consider the interdependence of wireline log attributes, temporal or depth dependence of log responses, multimodal geologic factors, and variability of individual ML model-based results to generate the best possible ML model. The results from the proposed approach indicate highly accurate and consistent shear sonic log predictions with a minimal error for thousands of feet (hundreds of meters) in several wells in the Wolfcamp Formation, covering a large study area in the basin. The results are used in estimating elastic properties, such as Poisson’s ratio, lambda-rho (incompressibility), and mu-rho (rigidity) to delineate brittle and ductile units in the formation.

Rock-mechanical properties and in situ stress state of the Upper Cenomanian Mishrif limestone reservoir, Zubair oil field, southern Iraq

Rock-mechanical properties and tectonic stress state have crucial implications for reservoir stability, drilling, and production optimization. We have developed the first ever in situ stress analysis focused on the Mishrif reservoir from the supergiant Zubair oil field, southern Iraq. We interpreted an overall hydrostatic pore pressure gradient of 9.8 MPa/km and a vertical stress ([Formula: see text]) gradient of 22.17 MPa/km in the Upper Cenomanian Mishrif limestone formation. We inferred a Poisson’s ratio of 0.27–0.31, a Young’s modulus of 27–38 MPa, a coefficient of internal friction range of 0.66–0.76, and estimated a uniaxial compressive strength (UCS) range of 35–70 MPa with a low cohesive strength of 7.8–9.8 MPa from compressional sonic slowness data. We established fitting relationships between the dynamic and static elastic properties from the core-based triaxial test measurements and used the same for estimating the minimum ([Formula: see text]) and maximum ([Formula: see text]) horizontal stress magnitudes by poroelastic horizontal strain model. We validated the reservoir [Formula: see text] with the mini-frac-based fracture closure pressure and interpreted its gradient as 14.5–15.9 MPa/km, whereas we calculated the [Formula: see text] gradient as 19.5–23 MPa/km. We interpreted C-quality breakouts from four-arm caliper data and deciphered a northeast–southwest [Formula: see text] orientation (N45°E–N70°E). Using the compressive failure criteria, we constrained the reservoir [Formula: see text] by stress polygon analysis against the estimated UCS ranges and concluded a normal fault to strike-slip transitional tectonic stress state ([Formula: see text] ≥ [Formula: see text] > [Formula: see text]) in the Mishrif reservoir.

Comprehensive workflow to quantify diagenetic cement from wireline logs of deep water east coast of India wells and rock physics modeling

Presence of diagenetic cement is common in deep-water siliciclastic reservoirs and largely affects reservoir quality. Hence, the quantification of cement volume is an integral part of the reservoir characterization workflow. Lab analysis of core samples provides the cement volume information used as input for rock physics modeling. In the absence of core sample analysis, use of analogous offset well information provides an approximate estimate of cement. In reality, most deep-water wells do not have core samples for lab analysis. Consequently, the availability of offset well information is sparse and sometimes unavailable. Hence, getting an accurate cement volume estimate for reservoir characterization is challenging. For a detailed analysis, it is desirable to have a continuous measurement of cement volume across the reservoir sand. Presence of a calibrated cement volume log for the entire reservoir improves rock physics modeling. In this work, petrophysical M-N lithology cross plot, which is widely used to identify lithologies, has been used to quantify the diagenetic cement present within the reservoir, where M is the density normalized sonic values and N is the density normalized neutron values. Here, the M and N has been used as M-Lithology (MLITH) and N-Lithology (NLITH). In estimating MLITH and NLITH values, compressional sonic, bulk density and neutron logs are used. Porosity estimated from neutron/density is useful to identify the type of cement. Whereas, sonic slowness values are more sensitive towards the amount of grain contact cements. Hence, the estimated values of MLITH and NLITH varies differently in the presence of diagenetic materials. These variations of MLITH and NLITH values are captured and converted to cement volume fractions. In the presence of core data, the lab analysis measurements are used to calibrate the cement volume log. Further, rock physics modeling results helps to validate the output. The results of the study add value by quantifying diagenetic cementation in the reservoir characterization process by mitigating the associated limitations due to absence of sufficient core data.

S-wave anisotropy from two dipole sonic data processing methods, confronted with fracture permeability, logs and cores

The present paper consists in two parts, determined by the historical emerging production of Dipole Sonic Imager (DSI) measurements and results in the early 1990’s. The DSI data were processed following two methods simultaneously developed in France and in USA by Schlumberger. In the first part the early dipole sonic S-wave velocity results obtained in late 1993 are confronted with the other borehole data obtained in the scientific borehole MM-1, entirely cored and extensively logged, as part of the comprehensive scientific project named Géologie Profonde de la France (GPF), conducted by the Bureau de Recherches Géologiques et Minières (BRGM, i.e. the French Geological Survey), in Ardèche, southern France. In 1994, José Perrin summarized and integrated all the borehole information including the preliminary results from an azimuthal “rotation scan” of S-wave sonic slowness determination method quickly developed in Schlumberger-France and aiming at detecting only the presence of S-wave velocity anisotropy in a first step. The initial results were presented to the French industrial logging community in April 1994, prior to the commercialization of any S-wave splitting computer detection routine applied to dipole sonic data. The second part focuses on the comparison of the dipole sonic S-wave anisotropy detection results from two methods produced at a later time by Schlumberger, namely: a) results from the commercial S-wave anisotropy detection routine based on cross energy minimization, obtained in October 1994, and b) principal S-wave azimuth results sorted from the “rotation scan” azimuthal method, produced in 1995 and further improved in July 1997. After discussing the discrepancies of the principal fast S-wave azimuth derived from the two methods with diverse specialists in Schlumberger, over several years, and on a spare time basis, the authors expose constructive explanations in the present paper. A limited overview of the latest dipole sonic data processing developments has also been attempted to better understand the differing S-wave birefringence results obtained in MM-1, suggesting that the rock formation in the immediate borehole vicinity, up to three times the borehole radius, may not be homogeneous along the borehole depth depending on the local geological context. Besides, the Fast Azimuth split S-wave (FAZ) fits with the strike of major regional faults and parallel to the maximal horizontal palaeo-stress, which happens to be nearly orthogonal to the local present stress direction accepted by the geologists! The present case study suggests that the S-wave anisotropy results ought to become more reliable, mainly on the accuracy and precision of the FAZ. Additionally, the efficiency of the semblance parameter for S-wave attenuation anisotropy detection is pondered, where no S-wave velocity anisotropy is detected over the dipole sonic receiver array.

Generation of Synthetic Sonic Slowness Logs From Real-Time Drilling Sensors Using Artificial Neural Network

Abstract The acoustic data in terms of compressional and shear wave velocity provide important petrophysical information about the rock. The sonic data are a significant input that is commonly used for deriving geomechanical parameters. Understanding the geomechanical properties of reservoir rock is essential during the drilling, development, production, and stimulation of an oil or gas reservoir. Among them, Young's modulus and Poisson's ratio are the most important elastic parameters. These properties are usually estimated from bulk density, compressional, and shear wave velocity log data. Sonic data acquisition is usually achieved through dipole sonic imager log or laboratory testing on core samples which is costly and time-consuming. Acquiring sonic data from wireline logs is not a feasible approach all the time, as the wireline log, specially shear-wave log, may not be recorded for every well. However, drilling data are available in real-time for every well using real-time drilling sensors. The main objective of this paper is to predict sonic slowness logs in real-time based on the drilling data using artificial neural network (ANN). The data used in this study were recorded during the drilling of 12¼″ hole sections from two wells. Many formations of different lithology were penetrated while drilling these sections of over 3000 ft vertical interval. The drilling and sonic data sets were recorded and preprocessed before using them for the ANN model. About 2900 data points from the first well were used for building and testing the model. The input parameters included weight on bit (WOB), torque (T), standpipe pressure (SPP), pipe speed (PS), rate of penetration (ROP), and mud flowrate (Q). Another data set of 2000 data points from the second well that was drilled in the same field was used to validate the model. The predictions were compared with sonic logs that were obtained after the drilling operation, and the results appear to be highly promising for future applications. The sonic slowness ANN models showed high accuracy for the model building (training and testing). Validation of these models was carried out using an unseen data set. The results using the validation data set for the compressional slowness model yielded a coefficient of determination (R2) of 0.983 and average absolute percentage error (AAPE) of less than 1.25%. For the shear slowness model, R2 was higher than 0.994 and AAPE less than 1.175%. The study offers empirical correlations that can be utilized to estimate the sonic slowness logs by engineers without the need to employ ANN software. The new shear slowness correlation was compared with other widely used correlations and the results showed high accuracy.

Comprehensive Rock Physics, Petrophysics workflow to correct Sonic Logs for improved seismic QI: Study using wells of Krishna Godavari basin

Rock physics analysis uses well log data as direct measurement of physical properties of the subsurface formations. Hence, in seismic reservoir characterisation they do not suffer the same level of scrutiny as seismic. Mud-filtrate invasion during drilling contaminates porous formation affecting log responses in non-intuitive ways. Ignoring invasion effect on well logs prior to quantitative petrophysical interpretations may led to erroneous reserve estimates as well as well-to-seismic misties. In addition, sonic logs are prone to ambiguity incorporated by weak waveform generation and non-unique coherence picking. This gives rise to non-unique compressional and shear slowness data. To correct the invasion affected logs, we have designed a workflow that effectively improves the recorded data. This workflow uses modified Gassmann (1951) fluid replacement modelling (FRM) method for sonic data correction. In addition, following an improved coherence picking we get a further updated sonic slowness. Using the corrected data in synthetic seismic generation enables us to achieve substantial improvements in synthetic-seismic correlation values. Furthermore, use of updated data results in a better three way matching with vertical seismic profiling (VSP) data.

Machine Learning Approach to Model Rock Strength: Prediction and Variable Selection with Aid of Log Data

Comprehensive knowledge and analysis of in situ rock strength and geo-mechanical characteristics of rocks are crucial in hydrocarbon and mineral exploration stage to maximize wellbore performance, maintain wellbore stability, and optimize hydraulic fracturing process. Due to the high cost of laboratory-based measurements of rock mechanics properties, the log-based prediction is a viable option. Nowadays, the machine learning tools are being used for estimation of the in situ rock properties using wireline log data. This paper proposes a machine learning approach for rock strength (uniaxial compressive strength) prediction. The main objectives are to investigate the performance of data-driven predictive model in determining this vital parameter and to select features of predictor log variables in the model. The backpropagation multilayer perception (MLP) artificial neural network (ANN) with Levenberg–Marquardt training algorithm as well as the least squares support vector machine (LS-SVM) with coupled simulated annealing (CSA) optimization technique is employed to develop the dynamic data-driven models. Capturing nonlinear, high dimensional, and complex nature of real field log data, the rock strength models’ performances are evaluated using statistical criteria to ensure concerning the model reliability and accuracy. The model predictions are compared and validated against the measured values as well as the results obtained from existing log-based correlations. Both the MLP-ANN and the CSA-based LS-SVM connectionist strategies are able to predict the rock strength so that there is a very good match between the model results and corresponding measured values. The input log parameters are ranked based on their contributions in prediction performance. The acoustic travel time and gamma ray are found to have the highest relative significance in estimating rock strength. New correlations are also developed to obtain the in situ rock strength of the siliciclastic sedimentary rocks using the most important log parameters such as dynamic sonic slowness, formation electron density, and shalyness effect. The developed correlations can be used to obtain quick estimation of dynamic uniaxial compressive strength profile using wireline logging data, instead of static data from the surface measurements or laboratory data. It is expected that the proposed models and tools will enable oil and gas engineers to better predict rock strength and thus enhance wellbore performance.

Application of Backus average to simulate acoustic logs in thinly bedded formations

Slowness logs acquired in layered formations are not only affected by spatial averaging associated with the borehole acoustic tool. Layers with thicknesses smaller than the acoustic wavelength can cause measurable effects on the associated wave propagation phenomena. While spatial averaging functions can be used to model tool averaging effects, computer-intensive numerical methods such as finite differences must be used to simulate slowness logs across formations with thin layers. We adopted Backus averaging as a faster alternative to model borehole slownesses when layer thicknesses are smaller than the acoustic wavelength (i.e., in the long-wavelength limit). Using synthetic models and numerical simulations via finite-element and finite-difference methods, we have determined that borehole slownesses of a stack of horizontal layers first approach the average slowness of the individual layers. However, as the layer thickness decreases, sonic slownesses approach the slowness of a homogeneous medium with elastic properties obtained from the Backus average. Therefore, to model acoustic logs acquired in layered formations, we first approximate thin layers as a single homogeneous layer with stiffness coefficients calculated using the Backus average. Next, we apply a spatial averaging function to reproduce the spatial averaging effect inherent to borehole acoustic tools. Results indicate that the latter method is accurate and efficient for fast modeling borehole slownesses of formations with thin layers that are isotropic and intrinsically vertical transversely isotropic. The fast simulation method decreases computation time by at least a factor of 10 and yields slowness logs with a relative error below 2% compared with finite-difference numerical simulations. We also determine that the moving Backus average that is typically applied to upscale acoustic logs for seismic applications is not accurate to model borehole acoustic logs acquired across thinly layered formations.

Present day In-situ stress magnitude and orientation of horizontal stress components in the eastern Illizi basin, Algeria: A geomechanical modeling

We analyzed drilling induced tensile fractures from resistivity image log data to ascertain the orientation of maximum horizontal stress (SH) from the eastern Illizi basin, Algeria. An average SH azimuth of 150°N (± 10°) has been interpreted from B-quality induced fractures, as per world stress map guidelines. The overall NW-SE orientation of SH translates to the relative plate motion of the African and Eurasian plates. Vertical stress (Sv) gradient of 1.07 PSI/ft has been derived from density log. Pore pressure estimated from sonic slowness reveals overpressure in Silurian shale, deposited in a transgressive depositional environment, whereas Devonian and Ordovician hydrocarbon reservoirs have been seen to be normally pressured. Poroelastic strain model has been employed to quantify maximum and minimum horizontal stress (Sh) magnitudes. An effective stress ratio of 0.6, interpreted from leak-off test has also been used to model Sh. Using frictional faulting theory, upper limit of SH has been quantified. SH/Sv ratio of 1.04 (1.01–1.26) has been seen in the study area. Based on the relative stress magnitudes (SH > Sv > Sh), a present day strike-slip faulting regime has been inferred in the eastern Illizi basin, Algeria. Fault reactivation potential at reservoir level has been inferred from stress polygon analysis.

Causes of disparity between predicted and measured pore pressure

The term pore pressure can be confusing. Pore pressure is predicted in relatively impermeable beds (shale and clay) and measured in reservoir quality rocks (sand). However, in many occasions, there is not a direct link between the pressure in the sand and in the sandwiching shale. Assuming such an immediacy can lead to serious drilling problems. The relationship between predicted (PPP) and measured (MPP) pore pressure is often complicated and establishing the relationship involves analysis of the geologic setting and expected hydrocarbons. This article will discuss the various causes for differing PPP and MPP and construction of an accurate pore pressure profile. In relatively clastic young sediments (Pleistocene-Tertiary), porosity indices (e.g., sonic slowness) are widely used to calculate PPP. Prior to drilling, seismic velocities are very helpful in estimating PPP. In reservoir type rock (sand, sandstone, oolites, etc.), pore pressure is measured using wireline tools and drilling stem gauges. MPP in wet sand usually follows the main hydrostatic gradient in the region as long as the formation water density stays the same. MPP in pay zones follows the gradient of the hydrocarbon which, in turn, depends on the density of the oil or gas column. The envelope shift in the pressure gradient (PG) between two compartments usually takes place in the seals (shale and clay). Most borehole problems (e.g., sloughing shale, enlarged and backed off-hole) take place when the PG slope changes from linear in the reservoir to exponential in the shale, especially near the seal base. Blow-outs, kicks, flow-kill-breakdown, and loss of circulation problems usually occur along the interface zone between the seal and the compartment below. The age of the deposits, rate of sedimentation versus rate of accommodation, structural setting of a prospect, and the fault plane lithology juxtaposition play a substantial role in pressure differential distribution in …

Acoustic wave velocity modelling in sedimentary sequences

Layered-clay shale density increases systematically with increasing depth of burial by progressive expulsion of pore water: ρ(z)= ρ0​+ (ρ∞ − ρ0)(1 − e−kz)Using elastic wave propagation theory, this shale density function can be transformed into a shale velocity prediction. Providing shale and water chemistry remain relatively stable, the progressive reduction in shale porosity and contained water also results in systematic changes to shale resistivity.More general formation (mixed lithology) sonic slowness can be modelled using established log interpretation transforms (the Archie porosity - resistivity and Raymer-Hunt-Gardner porosity - slowness relationships). The difference between measured formation resistivity and predicted shale resistivity is then interpreted in terms of departure of the expected slowness from that of pure shale. When added to the expected shale slowness, the combined slowness shows good agreement with observed sonic log data.This provides a new technique for prediction of formation velocities. The technique can be used to improve the quality of sonic log data, and can be applied where no sonic data were acquired. Numerous technical and commercial applications result.

Error bars for sonic slowness measurements

Semblance provides a quality indicator for beam‐former array processing methods such as slowness/time coherence (STC) and dispersive STC (DSTC). Semblance does not indicate the slowness variance because the array design, the processing time‐bandwidth parameters, and the signal spectral content are not taken into account. For the classical case of a single propagating wave with added Gaussian noise, semblance is distributed according to the noncentral beta distribution. This distribution is parameterized by the number of receivers, M, the processing time‐bandwidth product, BT, and signal‐to‐noise ratio. Given M and BT, the measured semblance can be inverted to give a high‐quality estimate of the signal‐to‐noise ratio. The Cramer‐Rao lower bounds on slowness variance for the classical case of signals in added Gaussian noise depend on the array dimensions, the signal power spectrum, and the signal‐to‐noise ratio. Estimates of the Cramer‐Rao bounds can be calculated from the estimated signal‐to‐noise ratio and the measured signal power spectrum. Monte‐Carlo simulations show that these bounds approximate the actual slowness variance for high, but practical, signal‐to‐noise ratios. Preliminary results on field data show that error bars based on Cramer‐Rao bounds are credible, and that error bars yield slowness standard deviations comparable to those from repeat runs.