First Results from the Maljamar Carbon Dioxide Pilot

The pressure and production behavior of a homogeneous cylindrical reservoir producing a single fluid through a centrally located vertical fracture of limited lateral extent was determined by using mathematical methods to solve the appropriate differential equation. It is assumed that there is no pressure drop within the fracture - that is, that the fracture capacity is infinite. It was found that the production-rate decline of such a reservoir is constant (except for very early times) when the flowing bottom-hole pressure remains constant. The production-rate decline increases as the fracture length increases. Thus, the lateral extent of fractures can be determined from the production-rate declines before and after fracturing or from the decline rate after fracturing when the properties of the formation and fluids are known. The production behavior over most of the productive life of such a fractured reservoir can be represented by an equivalent radial-flow reservoir of equal volume. The effective well radius of this equivalent reservoir is equal to one-fourth the total fracture length (within 7 per cent); the outer radius of this equivalent reservoir is very nearly equal (within 3.5 per cent) to that of the drainage radius of the fractured well. The effective well radius of a reservoir producing at semisteady state is also very nearly equal to one-fourth the total fracture length. It thus appears that the behavior of vertically fractured reservoirs can be interpreted in terms of simple radial-flow reservoirs of large wellbore. Introduction An earlier report has considered the effect of a vertical fracture on a reservoir producing an incompressible fluid. That investigation of the fractured reservoir producing an incompressible fluid was started because of its simplicity. Thus, pertinent behavior of fractured reservoirs was obtained at an early date, while experience was being gained of value in the solution of more complicated fracture problems. One of these more complicated problems, and the one discussed in this report, considers the effect of a compressible fluid (instead of incompressible fluids) on the production behavior of a fractured reservoir. In the incompressible-fluid work mentioned, it was shown that the production rate after fracturing could be described exactly by an effective well radius equal to one-fourth the fracture length whenever the pressure drop in the fracture was negligible. Because of the simplification in interpretation, it is a matter of much interest to determine whether the production behavior of reservoirs producing a compressible liquid could be described in terms of an effective well radius which remains essentially constant over the producing life of the field. The details of the mathematical investigation are given in the Appendixes. IDEALIZATION AND DESCRIPTION OF THE FRACTURED SYSTEM It is assumed that a horizontal oil-producing layer of constant thickness and of uniform porosity and permeability is bounded above and below by impermeable strata. The reservoir has an impermeable circular cylindrical outer boundary of radius r e. The fracture system is represented by a single, plane, vertical fracture of limited radial extent, bounded by the impermeable matrix above and below the producing layer (reservoir). It is assumed that there is no pressure drop in the fracture due to fluid flow. Fig. 1 indicates the general three-dimensional geometry of the fractured reservoir just described. When gravity effects are neglected, the flow behavior in the reservoir is independent of the vertical position in the oil sand. Thus, the flow behavior in the fractured reservoir is described by the two-dimensional flow behavior in a horizontal cross-section of the reservoir, such as the one shown in Fig. 2. SPEJ P. 87^

Lithology, porosity and water saturation were evaluated as part of a field-wide reinterpretation of log and core data for the Hemlock Reservoir, McArthur River Field, Cook Inlet, Alaska. To identify the conglomerates and sandstones which comprise this reservoir, a new method was developed for estimating the fraction of rock framework which is conglomerate. The method is based on the observation that, in this reservoir, the replacement of sand by nonporous pebbles results in poorer sorting and an overall decrease in porosity. Comparisons between core- and log-derived conglomerate estimates were used to calibrate and test the method. The reservoir properties of conglomerates and sandstones in the Hemlock Formation differ significantly, and while depositional factors have most likely resulted in a heterogeneous distribution of conglomerates and sandstones, the core data are too sparse to resolve this distribution. A map of the log-derived conglomerate fraction is sufficiently controlled to provide detailed information on reservoir heterogeneities; trends on this map are consistent with depositional models for the Hemlock Formation. The conglomerate map was integrated into the waterflood management of the field; for wells from one platform this resulted in a greater than two fold increase in incremental oil production per dollar spent on reperforations.

Summary. This paper presents the results of a CO2, flood pilot performed in the Permian Age carbonate rock formations of the Maljamar Cooperative Agreement (MCA) Unit. Field background and pilot development are reviewed. Injection and production history are presented, along with an evaluation of the pilot. A summary of the observation well logging results is also given. Introduction The 5-acre (20X10-3 -m2) Maljamar tertiary recovery pilot formally ended Jan. 1, 1986. Special features of the inverted five-spot pilot included dual (separate) completions in the Grayburg and San Andres zones and two fiberglass-cased logging observation wells for in-situ monitoring of oil, brine, and CO2 movement. Solutions to operational problems and a determination of the process performance were intermediate objectives of the pilot. The major objective was to provide a basis for commercial scale CO2 floodline economics for, the unit. CO2 retention by the reservoir has prevented major CO2 production. The observation wells indicated that CO2 contacted the entire vertical section. Operating problems and pilot performance have been different for the two zones. Problems, injectivity, oil response, and CO2 production will be discussed by zone. The pilot successfully met its objectives by providing data on field operations during CO2 flooding, as well as the process performance data needed to estimate largescale CO2 flooding results in the field. Operating problems have been largely resolved, incremental oil production has peaked, and performance has been encouraging. An expansion project is being designed. General Information The 8,040-acre [ 32 × 10 6 -M 2 ] MCA Unit occupies about 20% of the Maljamar field in Western Lea County, NM (Fig. 1). Wells produce from a Grayburg dolomitic sand (Sixth zone) and four San Andres dolomite pay zones (Upper Seventh, Lower Seventh. Upper Ninth, and Ninth Massive) at depths ranging from 3,600 to about 4,100 ft 1100 to about 1250 m], Fig. 2 presents a typical log. Oil gravity is 35 to 37deg. API [0.85 to 0.84 g/CM3]. Reservoir and fluid characteristics are shown in Table 1. The MCA Unit has performed well under waterflood, considering that it is a multizone flood in a heterogeneous reservoir. The better-quality zones. Sixth and Ninth Massive, have performed very well. At the economic limit for the existing waterflood, about 60% of the original oil in place (OOIP) would remain unrecovered. thereby making this a significant target for EOR operations.' The technical feasibility of CO2 flooding was demonstrated experimentally by determination of the minimum miscibility pressure (MMP) in slim-tube floods. At 1,515 psi [10.4 MPa]. the MMP is considerably lower than the average reservoir pressure of about 2.1500 psi [ 17.2 MPa]. Feasibility studies were performed for several different CO2 flood options in the field. Because of the economic risks associated with a commercial-scale project, the decision was made in 1978 to develop a pilot project for the MCA Unit. Information from the pilot would reduce the operational and economic uncertainties of a full-scale project. The pilot was expected to provide information on whether CO2 can mobilize oil in the reservoir, on how much CO2 would be required for a barrel of oil recovered. on how the CO2 injection rates would change with time, and on what values of process parameters should be used in simulations to predict field-scale CO2 flooding operations. Separate floods were conducted in the Sixth (the Grayburg sand interval) and the Ninth Massive zones (representative of the San Andres zones). The two floods were performed simultaneously by use of dual (separate) completions in the pilot wells. Two logging wells, completed with 680 ft [207 m] of fiberglass casing across the pay intervals and with no perforations, were also included. Logging data from these wells would allow determination of vertical variations in horizontal permeability and values for process parameters. An inverted five-spot pattern was selected to decrease the total volume of CO2 that would have to be purchased. At the time, there were indications that decisions about field-scale expansion would have to be made within 2 to 3 years to be confident of securing a CO2 supply. Five acres [20 × 103 M2) was chosen as the largest pilot that could be completed in a reasonably short period of time. Fig. 3 shows the final pilot pattern. Operating Plan The pilot development and prepilot planning and testing were reported previously. 1 Pilot development and operation consisted of the six phases shown in Fig. 4. JPT P. 1256^

Lithofacies description and petrophysical rock typing (RRTs) are critical in reservoir characterization as they provide an understanding of the spatial distribution of rock types and their petrophysical properties within the reservoir. Reservoir rock typing is a fundamental characterization element for integrated reservoir evaluation. RRTs are typically described at core, and then tied to logs and in some cases may be constrained with conceptual geologic models. The integration of core and log data is a robust approach in achieving comprehensive reservoir characterization. This study delves into this integration for a non-marine fluvial reservoir, aiming to furnish a detailed lithofacies description and establish petrophysical rock types. The study initiates with core lithofacies description, wherein 428 Laser Particle Size Analysis (LPSA), Thin Sections, and 421 X-Ray Diffraction (XRD) analyses contribute to defining eleven lithofacies within the core domain. Subsequently, the core data are integrated into Petrophysical Rock types (PRTs) through the examination of various electrical log combinations. This integration process involves grouping core-based resolution lithofacies to log scale studying log response and lithofacies similarity to generate PRTs. The outcome typically involves a reduction from core lithofacies to a more limited number of PRTs resolvable at the log scale. A proprietary supervised and unsupervised machine learning guided cluster analysis process is used to generate and calibrate the PRTs to core data. The output is a continuous log scale facies model solely using electrical log inputs for prediction (Gamma Ray Thorium /THOR, Radioactive Bulk Density/RHOB, Neutron Porosity/NPHIL, Photoelectric Effect Factor/XPEF) for successful consistent prediction and propagation away from core calibration where prolific electrical log measurements exist, and no core/rock data is present. A confident generation and prediction of PRTs at log scale allows components of petrophysical characterizations to be driven by lithofacies, ultimately providing better understanding of the reservoir. The reservoir under study corresponds to the proximal southeast part and the medial-distal northwest part of an Early-Middle Permian Gharif Formation, representing a non-marine fluvial deposit. It exhibits heterogeneity and layering, primarily comprising stacked fluvial channel deposits interspersed with flood-plain shales featuring paleosols. The variability in channel amalgamation rates is directly influenced by cyclic base-level variations, thereby dictating reservoir geometry. In conclusion, this paper presents a comprehensive methodology for describing lithofacies and establishing petrophysical rock types in non-marine fluvial reservoirs through core-log data integration. By providing valuable insights into reservoir behavior and enhancing the accuracy of reservoir characterization, this approach contributes significantly to more effective reservoir management. Through the detailed integration of core and log data, it offers a robust framework applicable to similar geological settings, aiding in better understanding and optimizing the exploitation of such reservoirs.

Exploration of the heterogeneous sandstone reservoirs presents a significant opportunity within the Nile Delta Basin. This study uses the well log, core, and petrographical data to describe the different rock types and characterizes the heterogenous sandstone of the Late Miocene Messinian Abu Madi reservoir as one of the main prolific reservoirs in the South Abu El Naga Gas Field in the Nile Delta. However, accurate assessment of the potential of these complex and heterogeneous sandstone reservoirs requires a meticulous approach. The available data was imported from four wells: SAEN-2, SAEN-4, SAEN-6, and SAEN-9. A total of 35 core plugs, which were derived from two cored intervals in the SAEN-2 well, were used in a well-integrated workflow for reservoir characterization, facies analysis, and rock typing. Core analysis (grain density ‘ρg’, helium porosity ‘∅He’, horizontal and vertical permeabilities ‘kH & kV’, and water saturation ‘Sw’) and well log data (caliper, gamma-ray, spontaneous potential, PEF, density, neutron, and resistivity logs) provided crucial insights into the lithology, pore systems, and textures. This information allowed us to define the dominant microfacies types as quartz arenite, feldspathic arenite, quartz wacke/wacke, feldspathic wacke, and subfeldspathic wacke. With the core data, it was also possible to estimate the reservoir quality index (RQI), flow zone indicator (FZI), and the effective pore radius (R35) from core data, while the net pay thickness, the effective porosity, the shale volume (Vsh), and the water saturation (Sw) from the well log data. It also enabled the identification of the potential zones of the gas-bearing reservoirs. Hydraulic flow units (HFUs) were established using well logs and core data. These units represent zones with similar fluid flow properties, facilitating the prediction of gas deliverability. Additionally, the flow zone indicator (FZI) that derived from the well logs further characterized the flow regime within the reservoir. Sedimentological studies, including thin section petrography, XRD, and SEM, complemented with the well log interpretation. This integrated workflow provided a comprehensive perspective of the reservoir, including pore structures, mineral composition, and textures. The Abu Madi Formation in the SAEN-9 well, to the northeast of the field, has the lowest net pay (7.3 m), while the SAEN-2 well, in the center of the field, has the highest net pay thickness (16.6 m). The core studies indicate that SAEN samples could be divided into four reservoir rock types (RRTs). The RRT1 has the lowest reservoir quality (0.12 ≤ ∅He ≤ 0.26, 2.4 ≤ kH ≤ 429 mD, 54.9 ≤ Sw ≤ 70.5%, 0.14 ≤ RQI ≤ 1.22 μm, 0.82 ≤ FZI ≤ 3.863 μm, and 1.055 ≤ R35 ≤ 11.41 μm), while the RRT4 has the best reservoir quality (0.25 ≤ ∅He ≤ 0.28, 2680 ≤ kH ≤ 4893 mD, 45.4 ≤ Sw ≤ 55.3%, 3.24 ≤ RQI ≤ 4.13 μm, 9.72 ≤ FZI ≤ 10.59 μm, and 34.668 ≤ R35 ≤ 44.78 μm). This study demonstrates the effectiveness of an integrated approach in comprehensively assessing the potential gas-bearing reservoirs and defining their quality in the Abu Madi Formation in the Nile Delta, which is characterized by very good reservoir quality (net pay thickness = 7.3–16.6 m, av. porosity = 23.3–30.35%, and av. water saturation = 31.7–64.0% for the various wells). The findings contribute significantly to optimizing exploration and development strategies for gas-bearing hydrocarbon resources in the Nile Delta Basin, especially for the Abu Madi reservoir.

This article describes the workflow used in continuous fracture modeling (CFM) and its successful application to several projects. Our CFM workflow consists of four basic steps: (1) interpreting key seismic horizons and generating prestack and poststack seismic attributes; (2) using these attributes along with log and core data to build seismically constrained geocellular models of lithology, porosity, water saturation, etc.; (3) combining the derived geocellular models with prestack and poststack seismic attributes and additional geomechanical models to derive high-resolution three-dimensional (3-D) fracture models; and (4) validating the 3-D fracture models in a dynamic reservoir simulator by testing their ability to match well performance. Our CFM workflow uses a neural network approach to integrate all of the available static and dynamic data. This results in a model that is better able to identify fractured areas and quantify their impact on well and reservoir flow behavior. This technique has been successfully applied in numerous sandstone and carbonate reservoirs to both understand reservoir behavior and determine where to drill additional wells. Three field case studies are used to illustrate the capabilities of the CFM approach.

Description of Paper. Selection among alternative log analysis models is based on the accuracy with which each can predict specific properties. The assessment of predictive accuracy involves core data not used for log calibration. Model selection is made using set of accuracy and vertical resolution criteria, based on both data quality and the requirements of the planned application (e.g., mapping, simulation). The log and core data used in this study are from deep marine reservoirs containing finely laminated sand/shale sequences in which the sand fraction exceeds 70%. Combined effects of thin laminations, variable sand properties, and high shale conductivity, complicate log analysis efforts. Several conventional log analysis models we evaluated provided satisfactory estimates of hydrocarbon pore volume, but none achieved the accuracy required for reservoir flow simulation. A new model was developed based on lithofacies relationships derived from integration of core, log, and geologic data. This model was subjected to the same calibration and testing as the conventional models, and achieved the desired level of predictive accuracy. The test results allowed the model to be applied with confidence to over 2000 feet of uncored pay. Technical Contributions:A method for selecting among alternative log analyses by evaluating predictive accuracyA case study of an integrated evaluation of log, core, and geological informationA simple and accurate log analysis model for finely laminated deep marine turbidites Introduction Deep marine turbidite deposits have become the focus of considerable exploration and development activity worldwide, and particularly within the "deepwater" area of the Gulf of Mexico. This study involves the Green Canyon 205 Unit (GC 205), located 150 miles south of New Orleans in an average water depth of approximately 2600 feet (Fig. 1). A development project for GC 205 is currently being pursued, with Chevron as Operator, and participation by partners Exxon and Fina. The geologic setting, facies distribution, and reservoir architecture for the major sands in this field have been previously described by Rafalowski, et. al. Most of the reserves in this field are contained within very finely laminated sand/shale sequences, with individual lamina often less than one inch thick. Even micro-imaging logs typically have insufficient resolution for quantitative determination of the net reservoir fraction in these formations. Since the shales in these reservoirs have high neutron porosities, moderate density porosities, and high conductivity, accurate assessment of the shale volume (VSh) is critical to all log analysis calculations. Early log analysis efforts were adequate for assessment of original hydrocarbon volumes in-place. However, they did not provide sufficiently accurate characterization of other reservoir parameters required for fluid flow simulation. The need for such simulation to assist with development planning prompted a review of the log analysis methods being used and available options. Data Availability and Quality Well Log Data. Of the ten penetrations drilled within the GC 205 unit, eight encountered oil sands in one or more of the six recognized development horizons. Figure 2 shows the locations of these penetrations with respect to the combined productive limits of the six horizons. Four of these reservoirs are grouped into the Lower Pleistocene "Nebraskan" series (Neb 1, Neb 2, Neb 3U, and Neb 3), while two sands are recognized within the Upper Pliocene series (14200' and 14800'). The Neb series sands contain the bulk of the developable resource, and were the primary focus of our study. P. 119

Drill String Vibration: A Proxy for Identifying Lithologic Boundaries while Drilling

Reservoir porosity and permeability are considered as very important parameters in characterizing oil and gas reservoirs. Traditional methods for porosity and permeability prediction are well log and core data analysis to get some regression empirical formulas. However, because of strong non-linear relationship between well log data and core data such as porosity and permeability, usual statistical regression methods are not completely able to provide meaningful estimate results. It is very difficult to measure fine scale porosity and permeability parameters of the reservoir. In this paper, the least square support vector machine (LS-SVM) method is applied to the parameters estimation with well log and core data of Qiongdongnan basin reservoirs. With the log and core exploration data of Qiongdongnan basin, the approach and prediction models of porosity and permeability are constructed and applied. There are several type of log data for the determination of porosity and permeability. These parameters are related with the selected log data. However, a precise analysis and determine of parameters require a combinatorial selection method for different type data. Some curves such as RHOB,CALI,POTA,THOR,GR are selected from all obtained logging curves of a Qiongdongnan basin well to predict porosity. At last we give some permeability prediction results based on LS-SVM method. High precision practice results illustrate the efficiency of LS-SVM method for practical reservoir parameter estimation problems.

The producing M. australis Sandstone of the Stag Oil Field is a bioturbated glauconitic sandstone that is difficult to evaluate using conventional methods. Well log and core data are available for the Stag Field and for the nearby Centaur–1 well. Eight wells have log data; six also have core data.In the past few years artificial intelligence has been applied to formation evaluation. In particular, artificial neural networks (ANN) used to match log and core data have been studied. The ANN approach has been used to analyse the producing Stag Field sands. In this paper, new ways of applying the ANN are reported. Results from simple ANN approach are unsatisfactory. An integrated ANN approach comprising the unsupervised Self-Organising Map (SOM) and the Supervised Back Propagation Neural Network (BPNN) appears to give a more reasonable analysis.In this case study the mineralogical and petrophysical characteristics of a cored well are predicted from the 'training' data set of the other cored wells in the field. The prediction from the ANN model is then used for comparison with the known core data. In this manner, the accuracy of the prediction is determined and a prediction qualifier computed.This new approach to formation evaluation should provide a match between log and core data that may be used to predict the characteristics of a similar uncored interval. Although the results for the Stag Field are satisfactory, further study applying the method to other fields is required.

The Maastrichtian (Upper Cretaceous) reservoir is one of five prolific oil reservoirs in the giant Wafra oil field. The Maastrichtian oil production is largely from subtidal dolomites at an average depth of 2500 feet. Carbonate deposition occurred on a very gently dipping, shallow, arid, and restricted ramp setting that transitioned between normal marine conditions to restricted lagoonal environments. The average porosity of the reservoir interval is about 15%, although productive zones have porosity values up to 30-40%. The average permeability of the reservoir interval is about 30 md. Individual core plugs have measured permeability up to 1200 mD. The currently available core data for the Maastrichtian is limited to six wells, five of which are essentially along the axis of the reservoir. As a result, the use of this core data set for stratigraphic and depositional interpretation is somewhat limited. This study was undertaken to determine: (1) if facies could be successfully predicted (classified) from well log data and (2) if the predicted facies could be used to extend the stratigraphic and depositional interpretation to the flanks of the reservoir. Efforts to predict facies from well logs in carbonate reservoirs is difficult due to the complex carbonate facies structures, strong diagenetic overprint, and challenging log analysis due in part to the presence of vugs and fractures. In the study, a workflow including (1) core description preprocessing (2) log and core data clean-up, and (3) probabilistic neural network (PNN) facies analysis was used to accurately predict facies from log data. First, the core descriptions were digitized and the number of facies reduced from twenty to six (mudstone, packstone, mud-rich packstone, grain-rich packstone, grainstone, and floatstone). Next, the log data for the cored wells were normalized and "cleaned" by removing intervals with incomplete data, evidence of significant vugs (or fractures in very limited cases), poor well bore conditions, or acoustic log or resistivity log spikes. For validation purpose, training sets were randomly selected from two cored wells. Hold-off data were used to validate prediction ability. The log data used for facies prediction include seven curves – GR, SP, DT, NPHI, RHOB, PE, and RT. GR spectrum logs, which are generally better lithological logs than GR were available for only a few wells; therefore, GR spectrum logs were not used in this study. After evaluation of a variety of statistical approaches, a probabilistic neural network PNN-based approach was used to predict facies from well log data. PNN was selected as a tool because it has the capability to delineate complex nonlinear relationships between facies and log data. PNN was shown to outperform multivariate statistic algorithms and in this study gave good prediction accuracy (above 70%). The prediction uncertainty was quantified by two probabilistic logs-discriminant ability and overall confidence. These probabilistic logs can be used to evaluate the prediction uncertainty during interpretation. Lithofacies were predicted for 15 key wells in the Wafra Maastrichtian reservoir and effectively used to extend the understanding of the Maastrichtian stratigraphy, depositional setting, and facies distribution.

In this paper we present an integrated approach to derive reservoir parameters from core and well-log data in clay–sand mixtures. This method is based on matching core and log data, and the linear and non-linear regressions are then used to build respective relationships between core and log data to determine formation parameters such as porosity, shale volume, clay content, permeability and fluid saturation. This information is then fed into a velocity prediction model to estimate seismic parameters such as elastic moduli, shear wave velocity and anisotropy coefficients. Finally, we test the method on real data from the North Sea and show that reservoir parameters can be accurately predicted.

Reservoir description for simulation studies requires good knowledge of the permeabilities. Unfortunately, reliable permeability is only available from laboratory tests on cores, which are usually taken from a small percentage of the wells. Frequently, this information is extrapolated to calculate permeabilities all over the field, but the lack of enough data points usually causes unreliable predictions. We propose a method to estimate formation permeabilities from standard well logs and core data. The analysis includes a first step consisting of the interpretation of the petrophysics and a characterization in lithofacies, electrofacies and hydraulic flow units. This step involves the use of modern mathematical tools to rationally classify each reservoir region into a given (discrete) hydraulic flow unit. As a second step the core permeability data is mapped with the well log data using neural networks and the restrictions found on the first step of the analysis. This approach allows the use of continuous hydraulic flow unit values and reduces the error arising from the discrete zonation technique. Also, it overcomes the error coming from the mapping between log data and hydraulic flow units. The method should be applicable to any kind of reservoir as long as sufficient core and log data are available. The method assumes that the Carman-Kozeny equation holds for the reservoir rocks, which is a fairly reasonable assumption, and that the well logs available contain intrinsic information on tortuosity, sand size distribution, cementing characteristics, etc., which ultimately determine the flow performance of the rock. This hypothesis is usually strong because the available logs are not able to fully read the physical phenomena that cover the complex dynamics of the flow on the reservoir rocks. The method was tested using available core and log data in a sandstone formation in Chihuido de la Salina, Neuquen Basin, Argentina. Some core data points were not used to train the neural network and therefore useful for validation and comparison. In spite of the cited drawbacks, the method has shown to outperform both the standard regression techniques and the hydraulic flow units approach.

Classifying rocks based on core data is the most common method used by geologists. However, due to factors such as drilling costs, it is impossible to obtain core samples from all wells, which poses challenges for the accurate identification of rocks. In this study, the authors demonstrated the application of an explainable machine-learning workflow using core and log data to identify rock types. The rock type is determined utilizing the flow zone index (FZI) method using core data first, and then based on the collection, collation, and cleaning of well log data, four supervised learning techniques were used to correlate well log data with rock types, and learning and prediction models were constructed. The optimal machine learning algorithm for the classification of rocks is selected based on a 10-fold cross-test and a comparison of AUC (area under curve) values. The accuracy rate of the results indicates that the proposed method can greatly improve the accuracy of the classification of rocks. SHapley Additive exPlanations (SHAP) was used to rank the importance of the various well logs used as input variables for the prediction of rock types and provides both local and global sensitivities, enabling the interpretation of prediction models and solving the “black box” problem with associated machine learning algorithms. The results of this study demonstrated that the proposed method can reliably predict rock types based on well log data and can solve hard problems in geological research. Furthermore, the method can provide consistent well log interpretation arising from the lack of core data while providing a powerful tool for well trajectory optimization. Finally, the system can aid with the selection of intervals to be completed and/or perforated.

The primary objective for drilling the Cote d’Ivoire-Ghana Transform Margin during ODP Leg 159 was to assess the sedimentary and deformation processes resulting from the different stages of continental break-up and related transform tectonism. In view of the structural importance of the leg, integration of logging and core data is important to help understand the main tectonic and deformation events that occurred. The effect of the transform deformation can be seen in physical properties data, for instance the porosity data derived from index properties measurements. Major breaks in porosity are associated with the tectonized lower Cretaceous and Cenozoic boundary, a trend also reflected in the P-wave velocity measurements. At each site, core and well log data show the presence of a major unconformity between the Cretaceous and Cenozoic, marked by an offset in porosity, density and P-wave data. The physical properties of log data are also heterogeneous, reflecting variations in consolidation, age and lithology. Another interesting aspect covered by core-log integration was the structural relationship within the sediments. As well as the direct measurements made on cores, in situ structural measurements have been obtained using the Formation MicroScanner (FMS; Mark of Schlumberger) logging tool in two of the holes. The measurements cover the Eocene to Turonian-upper Santonian limestones. Bedding planes dip predominantly towards NWNNW and show an increase with depth which can be interpreted to be the result of steady subsidence of the Deep Ivorian Basin. Break-outs and fracturing were also observed. Breakout occurrences depend on sediment type and their axes are perpendicular to the maximum compressive horizontal stress east-northeast west-southwest. Fracturing occurs as normal and reverse microfaults, with dispersion of dips and azimuth directions in these zones. The presence of fault zones are also correlated with changes in the physical properties of the sediments.