巴西国家石油公司

Uncontrolled drill-string vibrations can lead to loss of productivity, with fatigue failure being a major factor that can lead to the rupture of these structures. The main contribution of this paper is to propose a probabilistic model to take into account uncertainties associated with the bit-rock interaction parameters, system parameters, and fatigue properties and evaluate its impact in multiaxial fatigue factors. To test the proposed methodology, stress signals are obtained from a coupled axial-torsional drill-string model, which is calibrated with experimental data. Two multiaxial fatigue models are considered, and, for the configurations analyzed, the model proposed by Findley is more conservative than the one proposed by Papadopoulos. Regarding uncertainties, system parameters have more influence in regions with a greater tendency for occurrence of stick–slip oscillations, and fatigue properties uncertainties are more important in regions with lower oscillations.

This work presents an atmospheric-pressure protocol for generating and characterizing cyclopentane hydrate slurries. Water-in-oil emulsions containing 30 and 40

Steam reformer furnaces for large-scale hydrogen production in the petrochemical industry consist of a set of catalyst tubes made from state-of-the-art centrifugally cast heat-resistant modified HP-Nb or HP-NbTi stainless steel. These tubes are designed for a typical lifespan exceeding 100,000 hours, usually limited by creep damage. Currently, continuous service times can exceed 200,000 hours due to improvements in manufacturing methods, alloying additions, and confidence in nondestructive testing. Long-term in-service exposure to temperatures up to 1000 °C and internal pressures of 1-5 MPa presents severe operating conditions that lead to microstructural evolution. The relationship between creep and aging is indirect, but periodic assessment of microstructural evolution during operation provides important information for residual creep life assessment. This study proposes a general microstructural classification criterion for Nb-modified and Nb-Ti micro-alloyed HP alloys based on six characteristic aging states derived from systematic microstructural observations of tubes from different furnaces, manufacturers, and service histories, with aging times ranging from 70,000 to 135,000 h. Aging states are closely related to service temperatures, with limited dependence on operating time. After approximately 5000 hours of operation, the aging states are defined and remain in that configuration until the end of the tube’s service life. During scheduled shutdowns, the proposed criterion can be used as a basis for a comprehensive industrial procedure for nondestructive examination of extraction replica by optical microscopy. Complementary characterization of interdendritic phases and secondary precipitates was carried out using scanning and transmission electron microscopy, respectively. Long-term creep data obtained under different stress levels at 900 °C, 950 °C, and 980 °C further demonstrate a strong correlation between creep behavior and aging state, reinforcing the applicability of the proposed classification for life cycle assessment methodologies.

The Text-to-SQL task is defined as “given a relational database and a natural language sentence that describes a question on the database, generate an SQL query over the database that expresses the question”. Text-to-SQL strategies based on Large Language Models (LLMs) achieve remarkable performance on well-known benchmarks, but their performance is significantly less for real-world databases. Some of the reasons for this performance decrease lie in the mismatch between the end user’s view of the data and the organization and naming conventions of the database schema, and in the differences between the end user’s data semantics and the encoding of such semantics in the database. This article then proposes an LLM-based strategy to compile natural language questions into SQL queries that uses a knowledge graph and incorporates a dynamic few-shot examples technique. The implementation of the strategy leverages a database keyword search tool and specific naming conventions to expose the knowledge graph to an LLM thereby simplifying the text-to-SQL task. The article includes experiments with a real-world, proprietary relational database and the Mondial database to assess the performance of the proposed strategy. The experiments suggest that the strategy achieves an accuracy on challenging relational databases that surpasses state-of-the-art approaches on the same databases.

Abstract This paper presents a real-time methodology for recognizing operational events in oil wells by integrating multivariate time-series segmentation with hybrid classification and novelty detection techniques. A 1D U-Net architecture is used to segment sensor data streams and isolate candidate event windows, which are then analyzed by an ensemble of binary classifiers trained to identify known production behaviors such as valve operations, spurious closures, severe slugging, and hydrate formation. To address the limitations of deep learning models, particularly their tendency to become overconfident on unseen events, and the noise sensitivity of statistical distance measures, the strategy incorporates Mahalanobis distance scoring computed in both the raw feature space and the low-dimensional latent space from an autoencoder. The resulting hybrid decision function increases robustness and generalization across a wide range of operating conditions. Experimental validation using the publicly available 3W dataset along with proprietary field data shows that latent-space Mahalanobis scoring improves novelty detection by an average of 17%, achieving more than 96% detection for critical events and raising the detection of challenging patterns such as Rapid Productivity Loss from near-zero accuracy with classifiers to 67.5%. The methodology also preserves strong performance on known-event classification, with binary classifiers exceeding 90% accuracy in most categories and U-Net segmentation delivering consistent event boundaries with 94% mean Intersection Over Union (IOU). These results demonstrate the suitability of the proposed approach for real-time operational monitoring, providing early warnings, reliable recognition of both known and novel events, and practical deployment potential in production environments.