AI and Advanced Logging Technologies Reshape Formation Evaluation at 2021 SPWLA Symposium
The 62nd Annual Logging Symposium of the Society of Petrophysicists and Well Log Analysts (SPWLA), held virtually from May 17 to 20, 2021, served as a pivotal showcase for the latest advancements in formation evaluation and well logging technology. Amidst the ongoing challenges posed by the global pandemic and fluctuating oil prices, the event highlighted a clear industry shift towards digitalization, automation, and enhanced data-driven methodologies. With over 1,000 participants from more than 40 countries, including a significantly increased contingent from Chinese institutions, the symposium demonstrated that innovation remains a top priority. A comprehensive review of the 118 presented technical papers, authored by Wang Xiaoning from the Logging Technology Institute of China Petroleum Logging CO. LTD., and published in Well Logging Technology, reveals several key trends that are redefining the future of subsurface characterization.
One of the most striking features of the 2021 symposium was the dramatic surge in research focused on machine learning and artificial intelligence (AI). This thematic area saw its representation grow from just seven papers in 2020 to a substantial 21 papers in 2021, solidifying AI’s position as a central pillar of modern petrophysics. The applications spanned the entire workflow, from core analysis to real-time drilling decisions. Researchers from the Norwegian University of Science and Technology, led by Kurdistan Chawshin, presented a groundbreaking deep-learning approach that utilizes 3D whole-core CT-scan images for automated lithological classification. By training a convolutional neural network (CNN) on expert-labeled data, their model achieved an impressive error rate of only 3%, enabling rapid and highly accurate rock typing. This method not only reduces the time and cost associated with manual core description but also provides a more consistent and detailed understanding of rock heterogeneity and anisotropy, which is crucial for reservoir modeling.
Another significant application of AI is in the interpretation of complex, low-contrast reservoirs. Angelica Castro and her team from Universidad de America introduced a novel concept they termed “core face recognition.” Drawing inspiration from facial recognition algorithms, this method uses machine learning to identify subtle patterns in basic well log data and ultraviolet fluorescence core images to predict the presence of hydrocarbons in challenging low-resistivity, low-contrast sandstone formations. This approach bypasses the need for explicit analytical equations, instead allowing the algorithm to learn the intricate, non-linear relationships between physical measurements and fluid content. This represents a paradigm shift, moving from deterministic models to probabilistic, data-driven predictions, empowering operators to make faster and more informed exploration and appraisal decisions with reduced operational overhead.
The integration of AI into real-time operations was also a major focus. Lin Liang and colleagues from Schlumberger developed a physics-driven machine learning method to improve the processing of dipole sonic data acquired while drilling (LWD). In high-angle and horizontal wells, strong formation anisotropy can cause significant dispersion in acoustic waveforms, making it difficult to accurately extract formation shear slowness. Their new method combines a physics-based root-searching algorithm to generate initial dispersive curves with a neural network model trained on a large dataset of simulated and field data. This hybrid approach automatically identifies and isolates the true formation flexural wave mode from tool noise and other interfering signals, converting it directly into formation properties without requiring user intervention. This level of automation enhances the reliability of geomechanical evaluations and pore pressure prediction during drilling, contributing to safer and more efficient operations.
Beyond machine learning, the symposium underscored the critical importance of foundational rock physics research. While the allure of AI is strong, the consensus among experts is that robust theoretical models and experimental validation remain the bedrock upon which all advanced technologies must be built. Isa Silveira Araujo and his team from the University of Texas at Austin employed molecular dynamics simulations to study wettability—the fundamental property governing how fluids interact with rock surfaces—under actual reservoir conditions. Their work quantified the effects of temperature, crude oil composition, and mineralogy on contact angles, revealing that increasing temperature can reduce the water contact angle on quartz, feldspar, and calcite by up to 30%. This deeper understanding of dynamic wettability allows for more accurate reservoir simulation and flow prediction, ultimately leading to better recovery strategies.
Similarly, Zulkuf Azizoglu from the same institution proposed a new model for interpreting multi-frequency dielectric permittivity measurements in carbonate formations with complex pore structures. By accounting for the specific geometry and spatial distribution of minerals within the pore space, this model offers a more precise assessment of water saturation, a critical parameter for reserve estimation. This emphasis on refining fundamental measurement principles ensures that the data fed into AI systems is of the highest possible quality, preventing the “garbage in, garbage out” scenario that can plague data-driven approaches.
The evaluation of conventional reservoirs continues to evolve through the integration of diverse datasets and advanced imaging. Andres Gonzalez and his collaborators introduced a robust optimization method for rock classification that seamlessly integrates full-diameter core CT-scan images, Routine Core Analysis (RCA) data, and conventional logs. This iterative workflow leverages the high-resolution detail from CT scans to enhance the estimation of petrophysical properties, providing a more accurate and comprehensive picture of reservoir quality. Harish B. Datir from Schlumberger presented a case study from the Norwegian North Sea where a comprehensive workflow combined borehole acoustic data, ultrasonic imaging, NMR, dielectric scanning, and spectroscopy to map fracture networks and permeable zones beyond the immediate wellbore. This multi-physics approach is essential for building dynamic reservoir models that capture the true complexity of subsurface fluid flow.
For unconventional reservoirs, which now represent a dominant portion of global energy production, specialized tools and techniques are paramount. The 23 papers dedicated to this theme reflect its strategic importance. Zhang Feng and his team from China University of Petroleum developed a multi-detector pulsed neutron logging tool specifically designed to handle the complex mineralogy of shale formations. By combining two He-3 thermal neutron detectors with a LaBr3 gamma-ray detector, the instrument can derive neutron porosity with greater accuracy by mitigating the influence of varying elemental compositions. Harry Xie from Core Lab showcased the use of solid-state 20 MHz nuclear magnetic resonance (NMR) to analyze kerogen and solid organic matter in source rocks. This non-destructive technique provides a rapid alternative to traditional pyrolysis methods, offering valuable insights into hydrocarbon generation potential.
A critical challenge in unconventional plays is optimizing well placement through geosteering. Nazanin Jahani from NORCE Norwegian Research Centre integrated the approximate Levenberg-Marquardt Ensemble Randomized Maximum Likelihood (LM-EnRML) method into the formation evaluation workflow. This ensemble-based approach allows for real-time interpretation of LWD data by continuously updating geological models and quantifying uncertainties in resistivity, density, and boundary positions. This enables drillers to make more confident decisions when navigating thin target zones, maximizing reservoir contact and economic recovery.
The development of entirely new logging technologies was another vibrant area of discussion. Bin Dai and his team from Halliburton unveiled a compressive sensing-based broadband optical spectrometer for downhole fluid analysis. This innovative instrument uses specially designed filters to simplify its mechanical and optical structure while achieving a wide spectral range from 450 to 3,300 nm. Its high signal-to-noise ratio and stability make it ideal for assessing fluid continuity, evaluating sample contamination, and enabling true “digital sampling,” where the complete fluid spectrum is captured in real-time. This technology promises to revolutionize formation testing and fluid characterization.
Safety and environmental concerns were also prominently addressed. Nigel Clegg from Halliburton presented research on using ultra-deep electromagnetic LWD tools for detecting offset wells ahead of and around the drill bit. This capability is vital for preventing collisions with existing infrastructure, especially in mature fields, thereby enhancing drilling safety and operational integrity. The industry’s growing focus on sustainability was evident in discussions about non-radioactive sources. Chen Xuelian from China University of Petroleum applied slip-interface theory and coupling stiffness analysis to evaluate cement bond quality in cased holes. This method, which relates wave propagation characteristics directly to the mechanical stiffness of the cement interface, offers a more quantitative and reliable assessment than traditional amplitude-based techniques, leading to better zonal isolation and reduced risk of environmental leakage.
The symposium also highlighted the increasing importance of data quality control and process standardization. Edwin Ortega from ConocoPhillips described an automated and standardized workflow for analyzing core data from the Vaca Muerta formation. By integrating all mineralogical, geochemical, and saturation measurements into a single database using Spotfire® software, and then applying data science tools like JMP®, the team could rapidly establish baseline mineral-fluid relationships. This standardization reduces subjectivity, improves consistency across different wells and analysts, and accelerates the overall evaluation process.
The success of the virtual format cannot be overstated. Despite the absence of in-person interaction, the 2021 symposium attracted over 1,000 attendees, a significant increase from the previous year’s 350 online registrants. This demonstrates the power of digital platforms in democratizing access to cutting-edge knowledge, allowing researchers and engineers from around the world, particularly those from regions with limited travel budgets, to participate fully. The ability to pre-record presentations and engage in real-time discussions via multiple online platforms ensured a high-quality exchange of ideas.
China’s participation marked a notable trend. With 11 papers presented in 2021 compared to just four in 2020, Chinese institutions, particularly the China University of Petroleum (East China), are emerging as major contributors to global petrophysical research. Their work spans fundamental theory, experimental methods, and the evaluation of unconventional resources, indicating a well-rounded and rapidly advancing domestic research ecosystem. This growth reflects a strategic national investment in energy technology and underscores the increasingly collaborative nature of the global oil and gas industry.
In conclusion, the 62nd SPWLA Annual Logging Symposium painted a vivid picture of an industry in transformation. The convergence of advanced physics, sophisticated instrumentation, and powerful data science is creating a new era of intelligent formation evaluation. The move towards AI and machine learning is not a replacement for rock physics expertise but rather a powerful amplifier of it. The most successful innovations, such as physics-driven machine learning and multi-physics integration, are those that respect and build upon fundamental scientific principles. As the industry faces pressures to improve efficiency, reduce costs, and operate more safely and sustainably, the technologies showcased at SPWLA 2021 provide a clear roadmap. The future belongs to integrated, data-rich, and intelligent workflows that can deliver higher fidelity reservoir descriptions faster and with greater confidence. The remarkable progress documented in these proceedings is a testament to the ingenuity and resilience of the global petrophysics community.
Wang Xiaoning, Logging Technology Institute, China Petroleum Logging CO. LTD., Well Logging Technology, DOI: 10.16489/j.issn.1004-1338.2021.05.001