Nuclear Magnetic Resonance Logging Research Articles

Classification Method and Application of Carbonate Reservoir Based on Nuclear Magnetic Resonance Logging Data: Taking the Asmari Formation of the M Oilfield as an Example

The strong heterogeneity of carbonate reservoirs poses significant technical challenges in reservoir classification and permeability evaluation. This study proposes a new method for reservoir classification based on nuclear magnetic resonance (NMR) logging data for the Asmari formation of the Middle East M Oilfield, a carbonate reservoir. By integrating NMR T2 spectrum characteristic parameters (such as T2 geometric mean, T2R35/R50/R65, and pore volume fraction) with principal component analysis (PCA) for dimensionality reduction and an improved slope method, this study achieves fine reservoir type classification. The results are compared with core pressure curves and petrographic pore types. This study reveals that the Asmari reservoir can be divided into four categories (RT1 to RT4). RT1 reservoirs are characterized by large pore throats (maximum pore throat radius >3.8 μm), low displacement pressure (<0.2 MPa), and high permeability (average 22.16 mD), corresponding to a pore structure dominated by intergranular dissolution pores. RT4 reservoirs, on the other hand, exhibit small pore throats (<1 μm), high displacement pressure (>0.7 MPa), and low permeability (0.66 mD) and are primarily composed of dense dolostone or limestone. The classification results show good consistency with capillary pressure curves and petrographic pore types, and the pore–permeability relationships of each reservoir type have significantly higher fitting goodness (R2 = 0.48~0.68) compared with the unclassified model (R2 = 0.24). In the new well application, the root mean square error (RMSE) of permeability prediction decreased from 0.34 mD using traditional methods to 0.21 mD, demonstrating the method’s effectiveness. This approach does not rely on a large number of mercury injection experiments and can achieve reservoir classification solely through NMR logging. It provides a scalable technological paradigm for permeability prediction and development scheme optimization of highly heterogeneous carbonate reservoirs, offering valuable references for similar reservoirs worldwide.

All-optical filtering of nuclear magnetic resonance logging data based on a diffractive neural network

The signal-to-noise ratio (SNR) of nuclear magnetic resonance (NMR) logging data is very low; filtering methods based on U-Net and MsEDNet are always employed to extract information for logging stratigraphic evaluation. Since it is difficult to adjust the parameters of U-Net and MsEDNet for logging data, the filtered results suffer from low SNR and distortion. To address the problem, this paper proposes an optical diffractive neural network (DNN)-based filtering system for NMR logging data, which can protect the signal’s integrity and avoid degradation of the neural network. In this system, the Sinkhorn–Knopp algorithm upgrades one-dimensional echo data into two-dimensional data for optical diffractive computing. The proposed residue DNN separates the noise in NMR logging effectively. Therefore, the resulting SNR of our method is higher than that of U-Net and MsEDNet. Simulation and experimental results demonstrate the effectiveness of the proposed method.

Identification of the bound and free fluid pore types in an Iranian carbonate reservoir through the integration of well logs, rock physics modeling, and 3D seismic data

Characterizing pore types in carbonate rocks is essential for understanding their reservoir properties, such as water saturation, porosity, and permeability, as well as their reservoir quality and dynamic behavior. The aim and objective of this study are primarily to identify the pore types and then, as an innovative approach, to specify whether the fluid within the pores in the reservoir is producible or remains immobile. To achieve this goal, bound and free fluid pore types are defined using nuclear magnetic resonance logs and post-stack 3D seismic data from an Iranian oil reservoir for the Main Ilam carbonate Formation. Initially, the fullset of logs for two wells (A and B) were used to determine porosity, lithology, and fluid content, followed by the nuclear magnetic resonance log to identify macro, meso, micro, and clay pore types. Subsequently, in well A, the results were validated through pore-size distribution analysis of the available core samples. Afterward, a rock physics model for carbonated rocks was used to estimate compressional velocity and density based on petrophysical evaluation and pore-type calculation. The correlation coefficients for compressional velocity in wells A and B of the Main Ilam Formation were 92% and 80%, respectively. The strong correlation among measured and modeled values verified that nuclear magnetic resonance data effectively estimates pore-type variations. Furthermore, bounded fluid was calculated based on the clay and micro pores, while free fluid was calculated based on the meso and macro pores in both wells. Finally, this information was mapped into the 3D seismic cube using inversion and multi-attribute regression analysis.Graphical abstract

Petrophysical Evaluation of the Zubair Formation’s Main Pay Zone Using Conventional and NMR Log Data: Insights from the North Rumaila Field, Southern Iraq

This study investigates the calculation of petrophysical properties using both conventional logging tools and Nuclear Magnetic Resonance logs for the main pay zone of the Zubair Formation, Recognized as one of the most significant hydrocarbon-producing formations in the North Rumaila Field, located in southern Iraq, the main pay zone is categorized into three permeable units (AB, DJ, and LN) and two impermeable isolated zones (C and K) based on log responses. The petrophysical analysis of these zones indicates that the shale volume increases toward the northern part of the field due to the depositional environment. Effective porosity values across these zones range from fair to good; however, some wells in the AB layer exhibit excellent porosity, notably wells X-7, X-62, X-83, and X-110. Additionally, hydrocarbon saturation levels remain consistently high in the AB and DJ layers, whereas certain wells in the LN layer display lower saturation values, particularly wells X-85, X-78, X-63, X-110, X-84, X-19, and X-115. Furthermore, permeability values were calculated using the KSDR equation based on Nuclear Magnetic Resonance logs that span from good to very good, except well X-115, demonstrating excellent permeability in the AB layer.

A Method for Calculating the Relative Permeability of Tight Sandstone Oil Reservoirs Based on Nuclear Magnetic Resonance Logging

Summary Relative permeability is a key parameter in predicting production capacity. However, conventional core analysis methods, which rely on sampling at discrete depths, fail to capture the vertical heterogeneity of the reservoir along the wellbore and within the radial detection range of logging tools. To overcome this limitation, we propose a method for predicting oil-water relative permeability based on nuclear magnetic resonance (NMR) logging data. The method enables computation of relative permeability throughout the vertical extent of the reservoir within the limited radial detection range of NMR logging tools. In this study, based on the established empirical relationship between transverse relaxation time and capillary pressure, combined with fractal theory and resistivity experiments, and by improving the NMR porosity model, we consider the effects of bound water and residual oil on the permeability of water phase and oil phase to achieve a more comprehensive characterization of the vertical heterogeneity of the reservoir. We compare the proposed method with the traditional oil-water relative permeability calculation method based on core experiments. The results show that the new method can provide a relative permeability curve that matches the experimental results. Furthermore, we investigate the potential application of the proposed method in real oilfield settings. By analyzing the production performance of perforated intervals in two wells, the results show that the pseudocapillary pressure curves and oil-water relative permeability calculated from NMR logging align well with actual production outcomes. Specifically, perforated intervals with higher oil saturation and better productivity display pseudocapillary pressure curves indicative of favorable reservoir properties, as well as higher isopermeability crossover points on the relative permeability curves. In contrast, intervals characterized by higher water saturation and lower productivity exhibit poorer reservoir properties and lower crossover points. These observations suggest that the method offers considerable value in accurately forecasting production capacity, providing a new scientific basis for the evaluation and development of tight sandstone oil reservoirs.

Reservoir type classification and water yield prediction based on petrophysical conversion models

In the Chaixi region of the Qaidam Basin’s Qigequan tectonic zone, the compact sandstones are characterized by their low porosity and permeability, featuring intricate pore-throat formations, varied lithologies, assorted clay minerals, and pronounced unevenness among the reservoirs. There’s a weak link between reservoir metrics and logging reactions, making it challenging to assess these reservoir parameters. The microscopic pore structure of the reservoir can be illustrated through both the nuclear magnetic resonance relaxation time distribution and the capillary pressure curve. By using fractal dimensions to classify the reservoir, a conversion model between the transverse relaxation time in nuclear magnetic resonance logging and the capillary pressure in the mercury injection curve is established, enabling the conversion of pseudo-capillary pressure curves. Key elements of the pseudo-capillary pressure curve, specifically discharge and drive pressure, median pressure, and sorting coefficient, were analyzed and integrated with the generalized regression neural network for accurate reservoir type classification. An efficient categorization of reservoir types was accomplished by isolating three key elements from the pseudo capillary pressure curve—displacement pressure, median pressure, and sorting coefficient—and integrating them with the generalized regression neural network. Utilizing a rock physics framework, a correlation between transverse relaxation time of nuclear magnetic resonance and relative permeability conversion was formulated to accurately forecast the rate of water generation in the reservoirs of the western Qaidam Basin. The anticipated outcomes demonstrated a strong link with the real rate of water production. This technique presents an innovative method to forecast the comparative permeability of oil-water stages and the rates of water generation in compact sandstone reservoirs.

Study on Quantitative Characterization Method of Micro-Pore Structure of Carbonate Reservoir Based on Nuclear Magnetic Resonance Logging

Carbonate reservoirs have various types of reservoir spaces and complex pore structures, so the evaluation of microscopic pore structures is of great significance to favorable reservoir identification. In order to accurately characterize the micro-pore structure of carbonate reservoir, this paper uses the NMR experiment, high-pressure mercury injection, and NMR logging data to establish a conversion model between the NMR T2 spectrum and the capillary pressure curve by piecewise power function method. The nuclear magnetic T2 spectrum of the reservoir is mainly bimodal, with small pore T2 ranging from 0.1 to 6 ms, the peak value being about 2 ms, and the large pore throat T2 ranging from 100 to 6000 ms. The throat radius of small pores is 0.04–0.1 μm, the peak value is 0.08 μm, and the throat of large pores is 0.1–10 μm. The Bash layer has the smallest pore throat radius and median radius, higher median pressure, and poorer pore structure. By converting the T2 spectrum of nuclear magnetic logging into a pseudo-capillary pressure curve, the continuous and quantitative characterization of reservoir pore structure parameters was achieved vertically. The secondary method has important reference significance for the quantitative characterization of pore structure in reservoirs of the same type.

A region-partitioned nuclear magnetic resonance T2 spectra correction method in volcaniclastic reservoirs

Nuclear magnetic resonance (NMR) logging, one of geophysical exploration methods, is very sensitive to pore fluid and is promising in unconventional reservoir exploration. However, the volcaniclastic reservoir is characterized by tightness, low porosity, and the micro- and nano-pore development. A number of short relaxation signals acquired by NMR logging instruments may be lost, since the echo spacing is not sufficiently short for downhole NMR measurements. Additionally, paramagnetic minerals in volcaniclastic rocks, such as iron and manganese, increase the internal magnetic field gradient, enhancing diffusion relaxation of pore fluids, and thus leading to the complex relationship between T2 spectra and pore size. Based on the above factors, a region-partitioned T2 spectra correction method in volcaniclastic reservoirs was proposed in this paper. The T2 spectra were divided into short (T2 < 10 ms) and long (T2 ≥ 10 ms) relaxation parts. For short relaxation correction, the relationship model between the normalized interval porosity at TE = 0.6 ms and the normalized T2 spectra signal intensity at TE = 0.1 ms, as well as the relationship model between the signal intensity of the shortest relaxation peak at TE = 0.1 ms and TE = 0.6 ms, was constructed and employed. For long relaxation correction, the internal gradient Gint determined from T2–G spectra was utilized to calculate and remove diffusion relaxation from T2 spectra at TE = 0.6 ms. The proposed method's effectiveness is validated through simulations and field NMR data processing.

A novel Q-switch technique for ringing-down dissipation of borehole NMR measurement

Abstract Nuclear magnetic resonance (NMR) is a crucial logging technique for the unconventional and complex reservoir evaluation. However, the echo spacing is always an issue for borehole NMR measurement, which limits the performance of NMR tools when acquiring short relaxation components. In this paper, we proposed a novel Q-switch technique aimed at breaking through the limitation of the dead time of a borehole NMR logging tool, and achieving much shorter echo spacings. Instead of using resistors in parallel with the radio-frequency (RF) coil to reduce the dead time, an inductive-coupling Q-switch circuit was introduced to decrease the ringing-down time significantly after transmitting the RF pulses with a high voltage. The Q-switch circuit consists of an inductive-coupling coil, capacitors, resistors, and active high-voltage MOSFETs. The ringing-down time of the RF system was decreased by at least 10 times compared to a system that did not use the proposed Q-switch scheme, leading to echo spacings lower than 0.3 ms under the condition with a resonant frequency of ∼500 kHz. Both simulations and experiments were in great agreement, validating the feasibility and efficiency of the proposed Q-switch scheme, which proved to be promising for borehole NMR applications.

Measurement of vadose zone water content with direct-push nuclear magnetic resonance logging

Abstract Predicting the efficiency of managed aquifer recharge (MAR) operation requires understanding the distribution of sediment type and water content in the vadose zone. In this study we assess the use of a direct-push nuclear magnetic resonance logging technique which can rapidly image water content profiles without the need for permanent boreholes. By combining nuclear magnetic resonance logging with direct-push sediment texture logging using cone penetrometer testing, it is possible to estimate the water content in the vadose zone and relate differences in water content to sediment type. We demonstrate this approach using four co-located nuclear magnetic resonance and cone penetrometer testing logs acquired to depths of approximately fifteen meters at an active almond orchard in California’s Central Valley. The results show water contents typically greater than 25% in clay- and silt-rich sediments, and up to 10% in sand-rich sediments. The ability to rapidly assess sediment type and water content using two direct-push tools opens up many exciting opportunities to study complex groundwater systems, even active orchards and other agricultural environments.

Ground‐Truth Validation of T2 ${T}_{2}$ Estimates From Steady‐State Surface NMR

Abstract Steady‐state sequences are a new protocol for surface nuclear magnetic resonance (NMR), that can yield high quality data in a short time. In addition to faster acquisition, steady‐state surface NMR purports to measure the transverse relaxation time , which is correlated with hydrogeological parameters like pore‐size. This is in contrast to the effective transverse relaxation time associated with standard single pulse measurements, which may or may not be correlated with pore‐size. While previous studies suggest that steady‐state measurements have enhanced sensitivity to , a ground‐truth validation has not yet been done. We compare steady‐state surface NMR results against borehole NMR logs at four locations. For relaxation regimes of > 10 ms, profiles from steady‐state data are in excellent agreement with borehole measurements, provided the assumption of homogeneous horizontal layers is valid. Even in more complex geological environments, steady‐state surface NMR delivers representative estimates of .

Pore-scale T2-based numerical investigation on dynamics and wettability in mixed-wet shale oil reservoirs

Oil recovery in shale reservoirs is low due to the dynamics and wettability characteristics in mixed-wet shale oil reservoirs. Nuclear magnetic resonance (NMR) logging, a nondestructive and noninvasive technique, effectively evaluates the continuous dynamics and wettability in these reservoirs. The NMR numerical investigation can characterize the effects of dynamics and wettability, including varying wet regions and wet angles, on NMR responses, providing new insights into the frequency-dependent of T2-based petrophysical parameters. The NMR relaxation theory for mixed-wet shale oil reservoirs was proposed, and the relevant parameters were determined. The dynamics and wettability were characterized using the Shan Chen Lattice Boltzmann method, with constraints based on digital core technology. For the first time, the random walk method was employed to simulate the effects of water-wet regions with varying proportions, echo spacings, and wet angles on NMR responses in mixed-wet shale oil reservoirs at different frequencies. The proportions of water-wet regions, magnetic field frequencies, and echo spacings significantly influence porosity and T2LM, indicating that pore structure governs the dynamics and wettability and that petrophysical parameters can be characterized by their frequency dependence in mixed-wet shale oil reservoirs.

Determining Petrophysical Properties Using PLTs and Conventional Logs of the Middle Shale Member (4th Pay) of the Zubair Formation in the South Rumaila Oil Field, Southern Iraq

The petrophysical evaluation of the Zubair Formation's Middle Shale member (4th Pay) in the South Rumaila oil field has been conducted to characterize the reservoir and quantify its hydrocarbon prospects. Although the 4th Pay has a substantial oil reserve, it has never been studied. A gamma ray, resistivity, density, and neutron log were used to estimate the shale volume, porosity, water saturation, and hydrocarbon saturation for four oil-bearing subunits in the middle shale member (4th Pay), producing net pays of Middle Shale Member 7 (13 m), 10 (31 m), 20 (11 m) and 25 (16 m), respectively. Shale volume calculated using the Gamma-Ray Log method, averages between 0.05 % and 0.53 %. Effective porosity ranges from (0.08) % to (0.21) %, indicating good reservoir quality; permeability ranges from 0.9 to 193.6 mD, indicating well-sorted sands. Hydrocarbon saturation varies from 0.29 to 0.89, while water saturation ranges from (0.11) to (0.72). Production Log Tool and Nuclear Magnetic Resonance logs were used to estimate permeability. Production Logging Tool log permeability values ranged between 5.2 (Middle Shale Member 10, 3425 mD) and 49.7 (Middle Shale Member 25, 3448.5 mD), while Nuclear Magnetic Resonance log permeability values ranged between 5.81 (Middle Shale Member 10, 3425 mD and 49 (Middle Shale Member 25, 3448.5 mD). According to the correspondence between the two logs, the correspondence value equals 85%. Thus, the reservoir permeability can be calculated using the radial flow equation calculated from the Production Logging Tool log and generalized to areas where there is no core. Results show that permeability (K) is highest in subunits Middle Shale Member 20 and 25, which also have high porosity, low shale content, and fine grain sorting, making them key production subunits. Subunit Middle Shale Member, although lower in permeability, is the main reservoir with the most oil reserves. Subunits Middle Shale Member 5 and 7 have lower permeability due to high shale content and heterogeneity, with low porosity. Subunit Middle Shale Member 0 has very low permeability. The high oil shale content, due to its poor porosity and heterogeneity, renders it unproductive.

High-Precision Permeability Evaluation of Complex Carbonate Reservoirs in Marine Environments: Integration of Gaussian Distribution and Thomeer Model Using NMR Logging Data

Accurate evaluation of permeability parameters is critical for the exploration and development of oil and gas fields. Among the available techniques, permeability assessment based on nuclear magnetic resonance (NMR) logging data is one of the most widely used and precise methods. However, the rapid biochemical variations in marine environments give rise to complex pore structures and strong reservoir heterogeneity, which diminish the effectiveness of traditional SDR and Timur–Coates models. To address these challenges in complex carbonate reservoirs, this study proposes a high-precision permeability evaluation method that integrates the Gaussian distribution model with the Thomeer model for more accurate permeability calculations using NMR logging data. Multimodal Gaussian distributions more accurately capture the size and distribution of multiscale pores. In this study, we innovatively employ the Gaussian distribution function to construct NMR-derived pseudo-pore size distribution curves. Subsequently, Thomeer model parameters are derived from Gaussian distribution parameters, enabling precise permeability calculation. The application of this method to the marine dolomite intervals of the Asmari Formation, Section A, within Oilfield A in southeastern Iraq, demonstrates its superior performance under both bimodal and unimodal pore size distributions. Compared to traditional models, this approach significantly reduces errors, providing crucial support for the accurate evaluation of complex reservoirs and the development of hydrocarbon resources.

The Saturation Calculation of NMR Logging Based on Constructing Water Spectrum Function

Tight sandstone oil reservoirs are characterized by complex structures, poor pore connectivity, and strong heterogeneity, with features such as low porosity and ultra-low permeability. Conventional methods for calculating saturation cannot accurately evaluate the hydrocarbon saturation of these reservoirs. To address this, a study was conducted from the perspective of non-electrical logging methods, focusing on the inherent nuclear magnetic resonance (NMR) characteristics of different fluids to develop a saturation calculation method that avoids the influence of the rock matrix, thus enabling precise saturation measurement in tight sandstone oil reservoirs. The traditional NMR porosity model was modified by segmenting it using the clay-bound water cutoff value, aiming to identify the distribution pattern of fluids in pores outside the clay-bound water zone. Through theoretical derivation and water spectrum function simulation, a water spectrum function and its parameter range suitable for the NMR T2 distribution in tight sandstone reservoirs were determined. Using core-sealed core saturation as a reference, the particle swarm optimization (PSO) algorithm was applied to optimize the parameter range and construct the final water spectrum function tailored to tight sandstone oil reservoirs. The accuracy and practicality of this method were validated by applying the derived water spectrum function to NMR logging in the exploration block, allowing for precise saturation calculations and the accurate evaluation of tight reservoir saturation.

Application of nuclear magnetic resonance core logging technology in oil and gas exploration of deep/ultra deep low porosity and low permeability reservoirs

Abstract In order to improve the efficiency of deep/ultra deep oil and gas exploration decision-making and release production capacity, and to address the difficulties in implementing the physical and fluid properties of complex deep/ultra deep reservoirs, nuclear magnetic resonance logging technology is used to establish a permeability calculation method and a glutenite reservoir pore structure evaluation method so that the liquid production capacity of the reservoir is classified. Combing nuclear magnetic resonance logging with gas logging technology, an comprehensive hydrocarbon evaluation index SI is established to evaluate complex fluid properties. The results show that: (1) The permeability calculated by T2gm using nuclear magnetic resonance logging spectra is highly consistent with laboratory measurement results. (2) By combining the nuclear magnetic porosity and permeability of the test layer with the pore structure parameters calculated by nuclear magnetic, a classification standard for low porosity and low permeability glutenite reservoirs can be established, which can provide a preliminary evaluation of the reservoir’s liquid production capacity. (3) The comprehensive evaluation chart of nuclear magnetic resonance logging combining with gas logging can effectively distinguish oil layers, gas layers, and water layers, thereby distinguishing complex fluid properties. This research achievement is of great value for implementing the physical and fluid properties of deep/ultra deep complex reservoirs in Bohai Oilfield, and guiding efficient exploration and development operations.

Artificial Neural Network Application to Permeability Prediction from Nuclear Magnetic Resonance Log

Reservoir permeability plays a crucial role in characterizing reservoirs and predicting the present and future production of hydrocarbon reservoirs. Data logging is a good tool for assessing the entire oil well section's continuous permeability curve. Nuclear magnetic resonance logging measurements are minimally influenced by lithology and offer significant benefits in interpreting permeability. The Schlumberger-Doll-Research model utilizes nuclear magnetic resonance logging, which accurately estimates permeability values. The approach of this investigation is to apply artificial neural networks and core data to predict permeability in wells without a nuclear magnetic resonance log. The Schlumberger-Doll-Research permeability is used to train the model, where the model prediction result is validated with core permeability. Seven oil well logs were used as input parameters, and the model was constructed with Techlog software. The predicted permeability with the model compared with Schlumberger-Doll-Research permeability as a cross plot, which results in the correlation coefficient of 94%, while the predicted permeability validated with the core permeability of the well, which obtains good agreement where R2 equals 80%. The model was utilized to forecast permeability in a well that did not have a nuclear magnetic resonance log, and the predicted permeability was cross-plotted against core permeability as a validation step, with a correlation coefficient of 77%. As a result, the low percentage of matching was due to data limitations, which demonstrated that as the amount of data used to train the model increased, so did the precision.

A hybrid denoising method for low-field nuclear magnetic resonance data

A hybrid denoising method for low-field nuclear magnetic resonance data

Application of Two-Dimensional NMR for Quantitative Analysis of Viscosity in Medium–High-Porosity-and-Permeability Sandstones in North China Oilfields

The viscosity of crude oil plays a pivotal role in the exploration and development of oil fields. The predominant reliance on laboratory measurements, which are constrained by manual expertise, represents a significant limitation in terms of efficiency. Two-dimensional nuclear magnetic resonance (NMR) logging offers a number of advantages over traditional methods. It is capable of providing faster measurement rates, as well as insights into fluid properties, which can facilitate timely adjustments in oil and gas development strategies. This study focuses on the loose sandstone reservoirs with high porosity and permeability containing heavy oil in the Huabei oilfield. Two-dimensional nuclear magnetic resonance (NMR) measurements and analyses were conducted on saturated rocks with different-viscosity crude oils and varying oil saturation levels, in both natural and artificial rock samples. This study elucidates the distribution patterns of different-viscosity crude oils within the two-dimensional NMR spectra. Furthermore, the T1 and T2 peak values of the extracted oil signals were employed to establish a model correlating oil viscosity with NMR parameters. Consequently, a criterion for determining oil viscosity based on two-dimensional NMR was formulated, providing a novel approach for estimating oil viscosity. The application of this technique in the BQ well group of the Huabei oilfield region yielded an average relative error of 15% between the actual oil viscosity and the computed results. Furthermore, the consistency between the oil types and the oil discrimination chart confirms the reliability of the method. The final outcomes meet the precision requirements for practical log interpretation and demonstrate the excellent performance of two-dimensional nuclear magnetic resonance (NMR) logging in calculating oil viscosity. The findings of this study have significant implications for subsequent exploration and development endeavors in the research area’s oilfields.

Pore size classification and prediction based on distribution of reservoir fluid volumes utilizing well logs and deep learning algorithm in a complex lithology

Pore size classification and prediction based on distribution of reservoir fluid volumes utilizing well logs and deep learning algorithm in a complex lithology