Hasil untuk "Petroleum refining. Petroleum products"

Kai Zhu, Zhenyu Cui, Zehua Zang et al.

Recently, state space models have demonstrated efficient video segmentation through linear-complexity state space compression. However, Video Semantic Segmentation (VSS) requires pixel-level spatiotemporal modeling capabilities to maintain temporal consistency in segmentation of semantic objects. While state space models can preserve common semantic information during state space compression, the fixed-size state space inevitably forgets specific information, which limits the models' capability for pixel-level segmentation. To tackle the above issue, we proposed a Refining Specifics State Space Model approach (RS-SSM) for video semantic segmentation, which performs complementary refining of forgotten spatiotemporal specifics. Specifically, a Channel-wise Amplitude Perceptron (CwAP) is designed to extract and align the distribution characteristics of specific information in the state space. Besides, a Forgetting Gate Information Refiner (FGIR) is proposed to adaptively invert and refine the forgetting gate matrix in the state space model based on the specific information distribution. Consequently, our RS-SSM leverages the inverted forgetting gate to complementarily refine the specific information forgotten during state space compression, thereby enhancing the model's capability for spatiotemporal pixel-level segmentation. Extensive experiments on four VSS benchmarks demonstrate that our RS-SSM achieves state-of-the-art performance while maintaining high computational efficiency. The code is available at https://github.com/zhoujiahuan1991/CVPR2026-RS-SSM.

LI BING

Accelerating digital and intelligent transformation is a crucial measure for oil and gas enterprises to advance industrial transformation and upgrading and foster new productive forces. Sinopec’s upstream sector in China has thoroughly implemented the “Digital and Intelligent Sinopec” initiative, focusing on supporting corporate reform and management. By closely aligning with the development trends of digital and intelligent technologies and the demands of exploration and production operations, the digital and intelligent transformation has been steadily advanced. A group-level Exploration and Development Data Center (EPDC) has been established, aggregating 17.2 PB of various types of exploration and development data, which has enabled centralized data management and shared applications. An Internet of Things network covering oil and gas production sites has been nearly completed, with digital coverage rates for oil, gas, and water wells, and station facilities reaching 94.90% and 92.30%, respectively. This has fundamentally transformed the traditional manual management model of stationing personnel at wells and stations, effectively supporting the reform of production operation modes and labor organization under digital and intelligent conditions. The construction and deepened application of unified systems have been advanced coordinately, continuously improving the digital coverage across all exploration and development business operations. Sinopec has also actively promoted the construction of artificial intelligence (AI) scenarios and their pilot applications, achieving notable results in scenarios such as intelligent seismic processing and interpretation, intelligent rock thin-section identification and analysis, intelligent reservoir numerical simulation, intelligent drilling, intelligent fracturing, and intelligent well condition diagnosis. Looking ahead to the “15th Five-Year Plan”, Sinopec’s upstream sector in China aims to build intelligent oil and gas fields, accelerate the integration of data flow, business flow, value flow, and supervision flow (“four flows in one”), and promote the construction and application of high-value AI scenarios across the entire business chain. These efforts will support the deeper and more substantive integration of digitalization and intellectualization, enhancing the operational efficiency, economic benefits, and management capability of oil and gas exploration, development, and production.

Tianshu Yu, Chao Xiang, Mingchuan Yang et al.

Large language models (LLMs) have demonstrated remarkable evaluation and critique capabilities, providing insightful feedback and identifying flaws in various tasks. However, limited research has explored which types of critiques are most effective for improving model responses or how to generate such critiques. To address this gap, we introduce \textbf{R}efinement-oriented \textbf{C}ritique \textbf{O}ptimization (RCO), a novel framework designed to train critic models using refinement signals. RCO uses a feedback loop where critiques, generated by the critic model, guide the actor model in refining its responses. The critique utility (CU) quantifies the effectiveness of these refinements, serving as the reward signal for training the critic model. By focusing on critiques that lead to better refinements, RCO eliminates the need for direct critique preference assessment, ensuring that critiques driving meaningful improvements are rewarded. We evaluate RCO across five tasks, i.e., dialog generation, summarization, question answering, mathematical reasoning, and code generation, and show that it significantly outperforms traditional methods and open-source models in terms of critique quality and refinement outcomes. Our contributions include the introduction of RCO, a novel supervision scheme based on refined response preferences, and comprehensive experimental results that highlight the method's effectiveness in enhancing LLM critique-refinement loops.

Dmitry Brodsky, Benjamin Watson

As modeling and visualization applications proliferate, there arises a need to simplify large polygonal models at interactive rates. Unfortunately existing polygon mesh simplification algorithms are not well suited for this task because they are either too slow (requiring the simplified model to be pre-computed) or produce models that are too poor in quality. These shortcomings become particularly acute when models are extremely large. We present an algorithm suitable for simplification of large models at interactive speeds. The algorithm is fast and can guarantee displayable results within a given time limit. Results also have good quality. Inspired by splitting algorithms from vector quantization literature, we simplify models in reverse, beginning with an extremely coarse approximation and refining it. Approximations of surface curvature guide the simplification process. Previously produced simplifications can be further refined by using them as input to the algorithm.

JIANG Yong, LUO Xianbo, ZHANG Qixuan, WU Jintao, YANG Chenxu

The BZ condensate gas reservoir in the Bohai Sea, China, is a rare fractured buried hill condensate gas reservoir with high saturation and high content of condensate oil. The reservoir features high temperature, high pressure, ultra-low porosity, and ultra-low permeability. Due to the small difference between the fluid dew point and the pressure in the gas reservoir, it is prone to condensate oil precipitation, causing contamination in the near-wellbore zone. In the early development stage, the BZ gas reservoir pilot area was produced using natural energy. When the reservoir pressure drops below the dew point, retrograde condensation intensifies, leading to a rapid increase in the gas-oil ratio and an accelerated decline in production. Therefore, there is an urgent need for the evaluation of retrograde condensation damage and effective remediation methods. Core depletion experiments were conducted under high-temperature and high-pressure conditions using compound condensate gas to simulate retrograde condensate oil contamination. Gas-phase permeability was tested at different depletion pressure points to evaluate the degree of retrograde condensate contamination. Additionally, gas injection experiments were carried out to investigate the mechanisms of damage mitigation. Experimental results showed that as the reservoir pressure decreased, the amount of retrograde condensate in the core increased, and the effective gas-phase permeability decreased significantly. Ultimately, the resulting retrograde condensate damage to the reservoir reached 65.8% to 70.2%. Gas injection could reduce the viscosity of condensate oil, increase the volume expansion coefficient of reservoir fluids, and induce re-vaporization of retrograde condensate oil. This process reduced the amount and saturation of retrograde condensate liquid, relieved retrograde condensate blockage, and improved the effective gas-phase permeability of reservoir cores. The permeability recovery rates for N2, associated gas, and CO2 were 48.1%, 78.6%, and 81.7%, respectively. The final recovery rates for condensate oil reached 43.7%, 66.8%, and 69.2%, respectively. The research results provide technical support for gas injection development in the pilot zone of the BZ buried hill condensate gas reservoir. This approach effectively mitigates production decline and achieves good results, offering important guidance for the efficient large-scale gas injection development in the future.

QI HUAIYAN, YANG GUOBIN, ZHU YADI et al.

Imbibition plays a crucial role in waterflood development and the soaking stage after fracturing in tight oil reservoirs, serving as an effective method to enhance oil recovery. To investigate the effects of complex pore structures and rock-fluid interactions on imbibition mechanisms in tight reservoirs, this study combined nuclear magnetic resonance (NMR) technology with pore-scale imbibition numerical simulation techniques, conducting imbibition experiments and pore-scale imbibition numerical simulations on tight cores with different pore-throat characteristics. In the imbibition experiments, NMR <italic>T</italic>₂ spectra (transverse relaxation time) at different times were monitored in real time, which revealed the dynamic influencing patterns of pore structure on imbibition efficiency. In the pore-scale imbibition numerical simulations, realistic pore-scale physical models of tight sandstone were constructed based on thin sections, and the pore-scale imbibition process of tight sandstone was simulated by solving the Navier-Stokes equations combined with the phase field method. Based on the mutual verification of experimental and simulation results, the effects of contact angle, crude oil viscosity, and reservoir physical properties on imbibition efficiency were analyzed in detail. The results showed that the pore-scale imbibition numerical simulation results were in good agreement with the experimental data. The complexity of the pore structures significantly affected the oil displacement characteristics of imbibition, showing a relatively fast imbibition rate that gradually decreased with the extension of imbibition time. The aqueous phase preferentially entered smaller pores and then displaced the oil phase in larger pores. The smaller the contact angle resulting from rock-fluid interaction (i.e., the stronger the hydrophilicity of the rock), the greater the oil-water displacement driving force in the imbibition process and the higher the imbibition efficiency. In addition, a lower oil-water viscosity ratio and lower core permeability both generated stronger imbibition driving force. The research findings deepen the understanding of imbibition mechanisms in tight oil reservoirs at the microscopic level and provide theoretical foundation and experimental support for improving the development efficiency of tight oil reservoirs.

P. K. Kilpatrick

Jingtao Zhan, Qingyao Ai, Yiqun Liu et al.

For text-to-image generation, automatically refining user-provided natural language prompts into the keyword-enriched prompts favored by systems is essential for the user experience. Such a prompt refinement process is analogous to translating the prompt from "user languages" into "system languages". However, the scarcity of such parallel corpora makes it difficult to train a prompt refinement model. Inspired by zero-shot machine translation techniques, we introduce Prompt Refinement with Image Pivot (PRIP). PRIP innovatively uses the latent representation of a user-preferred image as an intermediary "pivot" between the user and system languages. It decomposes the refinement process into two data-rich tasks: inferring representations of user-preferred images from user languages and subsequently translating image representations into system languages. Thus, it can leverage abundant data for training. Extensive experiments show that PRIP substantially outperforms a wide range of baselines and effectively transfers to unseen systems in a zero-shot manner.

Xin Zhang, Teodor Boyadzhiev, Jinglei Shi et al.

In this paper, we leverage image complexity as a prior for refining segmentation features to achieve accurate real-time semantic segmentation. The design philosophy is based on the observation that different pixel regions within an image exhibit varying levels of complexity, with higher complexities posing a greater challenge for accurate segmentation. We thus introduce image complexity as prior guidance and propose the Image Complexity prior-guided Feature Refinement Network (ICFRNet). This network aggregates both complexity and segmentation features to produce an attention map for refining segmentation features within an Image Complexity Guided Attention (ICGA) module. We optimize the network in terms of both segmentation and image complexity prediction tasks with a combined loss function. Experimental results on the Cityscapes and CamViD datasets have shown that our ICFRNet achieves higher accuracy with a competitive efficiency for real-time segmentation.

Jalal M. Ahammad, Mohammad Azizur Rahman, Stephen D. Butt et al.

Abstract The occurrence of fluid flow near a wellhead is the major concern of the petroleum industry, as pressure drop, loss of formation, and other variables of interest are mostly affected in this region. The fluid flows from the hydrocarbon reservoir to the wellbore can be characterized as laminar to turbulent; thus, it is important to model this phenomenon with the integrated wellbore-reservoir model. Using 3D Navier–Stokes equations, an integrated wellbore-reservoir model is created in this study, and it incorporates the formation damage zone. For the porous-porous and porous-fluid interfaces, the General Grid Interface (GGI) approach is applied in conjunction with the conservative mass flux interface model. Model equations are solved using a velocity-pressure coupling solver that is pressure-based. For reliable and quick results, the system of equations is solved using an algebraic multigrid approach. The pressure diffusivity equation’s analytical solution under steady-state flow circumstances is used to validate the model. The integrated wellbore-reservoir model is applied to different reservoir scenarios, for example, different production rates, formation zones, and reservoir formation conditions. The results indicate that the present Computational Fluid Dynamics (CFD) model can be extended to simulate the real field scale model. integrated wellbore-reservoir modeling based on 3D Navier–Stokes equations with efficient computational techniques can lead the field of petroleum industries to advance current knowledge.

Pingping Sun, Vincenzo Cappello, A. Elgowainy et al.

In 2019, U.S. petroleum refineries emitted 196 million metric tons (MT) of CO2, while the well-to-gate and the full life cycle CO2 emissions were significantly higher, reaching 419 and 2843 million MT of CO2, respectively. This analysis examines decarbonization opportunities for U.S. refineries and the cost to achieve both refinery-level and complete life-cycle CO2 emission reductions. We used 2019 life-cycle CO2 emissions from U.S. refineries as a baseline and identified three categories of decarbonization opportunity: (1) switching refinery energy inputs from fossil to renewable sources (e.g., switch hydrogen source); (2) carbon capture and storage of CO2 from various refining units; and (3) changing the feedstock from petroleum crude to biocrude using various blending levels. While all three options can reduce CO2 emissions from refineries, only the third can reduce emissions throughout the life cycle of refinery products, including the combustion of fuels (e.g., gasoline and diesel) during end use applications. A decarbonization approach that combines strategies 1, 2, and 3 can achieve negative life-cycle CO2 emissions, with an average CO2 avoidance cost of $113-$477/MT CO2, or $54-$227/bbl of processed crude; these costs are driven primarily by the high cost of biocrude feedstock.

Yulan Hu, Zhirui Yang, Sheng Ouyang et al.

Self-Supervised Learning (SSL) has shown significant potential and has garnered increasing interest in graph learning. However, particularly for generative SSL methods, its potential in Heterogeneous Graph Learning (HGL) remains relatively underexplored. Generative SSL utilizes an encoder to map the input graph into a latent representation and a decoder to recover the input graph from the latent representation. Previous HGL SSL methods generally design complex strategies to capture graph heterogeneity, which heavily rely on contrastive view construction strategies that are often non-trivial. Yet, refining the latent representation in generative SSL can effectively improve graph learning results. In this study, we propose HGVAE, a generative SSL method specially designed for HGL. Instead of focusing on designing complex strategies to capture heterogeneity, HGVAE centers on refining the latent representation. Specifically, HGVAE innovatively develops a contrastive task based on the latent representation. To ensure the hardness of negative samples, we develop a progressive negative sample generation (PNSG) mechanism that leverages the ability of Variational Inference (VI) to generate high-quality negative samples. As a pioneer in applying generative SSL for HGL, HGVAE refines the latent representation, thereby compelling the model to learn high-quality representations. Compared with various state-of-the-art (SOTA) baselines, HGVAE achieves impressive results, thus validating its superiority.

Kumar Shridhar, Koustuv Sinha, Andrew Cohen et al.

In recent years, Large Language Models (LLMs) have demonstrated remarkable generative abilities, but can they judge the quality of their own generations? A popular concept, referred to as self-refinement, postulates that LLMs can detect and correct the errors in their generations when asked to do so. However, recent empirical evidence points in the opposite direction, suggesting that LLMs often struggle to accurately identify errors when reasoning is involved. To address this, we propose a reasoning with refinement objective called ART: Ask, Refine, and Trust, which asks necessary questions to decide when an LLM should refine its output, and either affirm or withhold trust in its refinement by ranking the refinement and the initial prediction. On two multistep reasoning tasks of mathematical word problems (GSM8K) and question answering (StrategyQA), ART achieves a performance gain of +5 points over self-refinement baselines, while using a much smaller model as the decision maker. We also demonstrate the benefit of using smaller models to make refinement decisions as a cost-effective alternative to fine-tuning a larger model.

Rui Cao, Tianrui Wang, Meng Ge et al.

Supervised speech enhancement has gained significantly from recent advancements in neural networks, especially due to their ability to non-linearly fit the diverse representations of target speech, such as waveform or spectrum. However, these direct-fitting solutions continue to face challenges with degraded speech and residual noise in hearing evaluations. By bridging the speech enhancement and the Information Bottleneck principle in this letter, we rethink a universal plug-and-play strategy and propose a Refining Underlying Information framework called RUI to rise to the challenges both in theory and practice. Specifically, we first transform the objective of speech enhancement into an incremental convergence problem of mutual information between comprehensive speech characteristics and individual speech characteristics, e.g., spectral and acoustic characteristics. By doing so, compared with the existing direct-fitting solutions, the underlying information stems from the conditional entropy of acoustic characteristic given spectral characteristics. Therefore, we design a dual-path multiple refinement iterator based on the chain rule of entropy to refine this underlying information for further approximating target speech. Experimental results on DNS-Challenge dataset show that our solution consistently improves 0.3+ PESQ score over baselines, with only additional 1.18 M parameters. The source code is available at https://github.com/caoruitju/RUI_SE.

Bin YUAN, Mingze ZHAO, Siwei MENG et al.

The existing approaches for identifying events in horizontal well fracturing are difficult, time-consuming, inaccurate, and incapable of real-time warning. Through improvement of data analysis and deep learning algorithm, together with the analysis on data and information of horizontal well fracturing in shale gas reservoirs, this paper presents a method for intelligent identification and real-time warning of diverse complex events in horizontal well fracturing. An identification model for “point” events in fracturing is established based on the Att-BiLSTM neural network, along with the broad learning system (BLS) and the BP neural network, and it realizes the intelligent identification of the start/end of fracturing, formation breakdown, instantaneous shut-in, and other events, with an accuracy of over 97%. An identification model for “phase” events in fracturing is established based on enhanced Unet++ network, and it realizes the intelligent identification of pump ball, pre-acid treatment, temporary plugging fracturing, sand plugging, and other events, with an error of less than 0.002. Moreover, a real-time prediction model for fracturing pressure is built based on the Att-BiLSTM neural network, and it realizes the real-time warning of diverse events in fracturing. The proposed method can provide an intelligent, efficient and accurate identification of events in fracturing to support the decision-making.

Kong Yongchao, Xia Hui, Luo Lei et al.

As the automated drilling rig is used widely,decentralized operation can no longer meet the requirements of the oilfield,and the integrated control system has become the mainstream configuration of automated drilling rig due to its high operational safety,remote centralized control,convenient operation and easy exPa·sion.With the rapid development of a new generation of information technology such as artificial intelligence,big data,internet of things and edge computing,the integrated control system of drilling rig is also continuously innovating.In order to promote the rapid development and technological progress of integrated control system of oil drilling rig in China,the current technical status of integrated control system of foreign drilling rig manufacturers such as NOV,Schlumberger and ACS was first investigated,and the system composition,characteristics and performance were discussed; then,technical investigation and analysis were conducted on the integrated control system of drilling rig manufacturers such as BOMCO,TSC-QD and HHG in China.The research results show that there is a large gap between domestic and foreign companies in terms of intelligence of integrated control system of drilling rig,mainly manifested in the relative lack of intelligent drilling software.Finally point out that the integrated control system of oil drilling rig in China will develop in three directions: data sharing,automated drilling,fault detection and diagnosis.The research results provide reference significance for the intelligent development of drilling rig technology in China.

Kamil Piwowarek, E. Lipińska, E. Hać-Szymańczuk et al.

Propionic acid bacteria (PAB) are a source of valuable metabolites, including propionic acid and vitamin B12. Propionic acid, a food preservative, is synthesized from petroleum refining by-products, giving rise to ecological concerns. Due to changing food trends, the demand for vitamin B12 has been expected to increase in the future. Therefore, it is necessary to look for new, alternative methods of obtaining these compounds. This study was conducted with an aim of optimizing the production of PAB metabolites using only residues (apple pomace, waste glycerine, and potato wastewater), without any enzymatic or chemical pretreatment and enrichment. Media consisting of one, two, or three industrial side-streams were used for the production of PAB metabolites. The highest production of propionic acid was observed in the medium containing all three residues (8.15 g/L, yield: 0.48 g/g). In the same medium, the highest production of acetic acid was found — 2.31 g/L (0.13 g/g). The presence of waste glycerine in the media had a positive effect on the efficiency of propionic acid production and P/A ratio. The concentration of vitamin B12 obtained in the wet biomass of Propionibacterium freudenreichii DSM 20271 ranged from 90 to 290 µg/100 g. The highest production of cobalamin was achieved in potato wastewater and apple pomace, which may be a source of the precursors of vitamin B12 — cobalt and riboflavin. The results obtained show both propionic acid and vitamin B12 can be produced in a more sustainable manner through the fermentation of residues which are often not properly managed. • The tested strain has been showed metabolic activity in the analyzed industrial side-streams (apple pomace, waste glycerine, potato wastewater). • All the side-streams were relevant for the production of propinic acid. • The addition of waste glycerine increases the propionic acid production efficiency and P/A ratio. • B12 was produced the most in the media containing potato wastewater and apple pomace as dominant ingredients.

Piergiuseppe Mallozzi, Inigo Incer, Pierluigi Nuzzo et al.

We address the problem of modeling, refining, and repairing formal specifications for robotic missions using assume-guarantee contracts. We show how to model mission specifications at various levels of abstraction and implement them using a library of pre-implemented specifications. Suppose the specification cannot be met using components from the library. In that case, we compute a proxy for the best approximation to the specification that can be generated using elements from the library. Afterward, we propose a systematic way to either 1) search for and refine the `missing part' of the specification that the library cannot meet or 2) repair the current specification such that the existing library can refine it. Our methodology for searching and repairing mission requirements leverages the quotient, separation, composition, and merging operations between contracts.

CHEN Qiang, LI Wenhou, SUN Jiaopeng et al.

The Caojiagou section in Qishan County, Shaanxi Province is located in the southern margin of the Ordos Basin. The well-exposed section deposited complete Ordovician strata and developed various sedimentary types, and its location has convenient transportation. In this paper, the stratigraphy and sedimentary characteristics of the Caojiagou section are introduced in detail. The lower Ordovician Yeli Formation is mainly composed of argillaceous dolomite and Mount Liangjia Formation is characterized by crystal dolomite intercalated with siliceous mass and develops stromatolites, all of which belong to tidal flat facies. The Middle Ordovician Majiagou Formation consists of medium-thin and medium-thick layers of silty-fine dolomite and limestone, belonging to open platform and platform front slope facies deposits. The Fengfeng Formation is mainly composed of medium-thin layer laminated dolomite with a great quantity of thin carbonate debris flow and turbidity current deposits, which represent the gravity flow deposit of the front slope of the platform in the deep-water environment. The Upper Ordovician Pingliang Formation mainly consists of a set of flyschoid rhythmic deposits of the continental slope facies. The Tangwangling Formation represents the debris flow environment of the continental slop and develops the ice-water deposits with typical ice-rafted dropstones.

Zhenzhen Wang, Carla Saoud, Sintawat Wangsiricharoen et al.

Annotating cancerous regions in whole-slide images (WSIs) of pathology samples plays a critical role in clinical diagnosis, biomedical research, and machine learning algorithms development. However, generating exhaustive and accurate annotations is labor-intensive, challenging, and costly. Drawing only coarse and approximate annotations is a much easier task, less costly, and it alleviates pathologists' workload. In this paper, we study the problem of refining these approximate annotations in digital pathology to obtain more accurate ones. Some previous works have explored obtaining machine learning models from these inaccurate annotations, but few of them tackle the refinement problem where the mislabeled regions should be explicitly identified and corrected, and all of them require a -- often very large -- number of training samples. We present a method, named Label Cleaning Multiple Instance Learning (LC-MIL), to refine coarse annotations on a single WSI without the need of external training data. Patches cropped from a WSI with inaccurate labels are processed jointly within a multiple instance learning framework, mitigating their impact on the predictive model and refining the segmentation. Our experiments on a heterogeneous WSI set with breast cancer lymph node metastasis, liver cancer, and colorectal cancer samples show that LC-MIL significantly refines the coarse annotations, outperforming state-of-the-art alternatives, even while learning from a single slide. Moreover, we demonstrate how real annotations drawn by pathologists can be efficiently refined and improved by the proposed approach. All these results demonstrate that LC-MIL is a promising, light-weight tool to provide fine-grained annotations from coarsely annotated pathology sets.