Seismic facies analysis from well logs based on supervised classification scheme with different machine learning techniques

Bagheri, Majid; Riahi, Mohammad Ali

doi:10.1007/s12517-014-1691-5

Cited by 27 publications

(7 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…SVM and ANN are typical methods of machine learning models, which are not only suitable for studying the relationship between the predicted value of civil aircraft spare parts and influencing parameters, but also for the realization of spare parts prediction through historical consumption data. [68][69][70] The advantage of SVM in spare parts prediction is that its optimal hyper plane can separate the projections according to different categories, thereby avoiding the risk of local prediction minimums. Li et al 71 applied SVM to predict spare parts demand, the input of SVM are the main factors that influenced spare parts consumption.…”

Section: Machine Learning Of Spare Parts Predictionmentioning

confidence: 99%

Civil Aircraft Spare Parts Prediction and Configuration Management Techniques: Review and Prospect

Feng

Chen

et al. 2021

Advances in Mechanical Engineering

View full text Add to dashboard Cite

Spare parts are treated as the basis to guarantee the safe and economic operation of civil aircraft, its scientific prediction and reasonable configuration play an important role in perfecting integrated logistics support (ILS) and achieving win-win situation of stakeholders (e.g. manufacturers, operators, maintenance providers). This paper studies the existing spare parts prediction and configuration methods of civil aircraft, and discusses future development trend of spare parts from the perspective of prediction and configuration, respectively. The current development status of civil aircraft spare parts prediction techniques are firstly introduced, according to demand characteristics of spare parts; the present research status of civil aircraft spare parts configuration methods are then elaborated, based on the traditional methods and machine learning methods; the development trend of civil aircraft spare parts prediction and configuration methods is finally analyzed, combined with the advantages and disadvantages of the existing methods of civil aircraft spare parts management methods. The efforts of this work provide a reference for the comprehensive management of civil aircraft spare parts and improve the perfection degree for the ILS of civil aircraft.

show abstract

Section: Machine Learning Of Spare Parts Predictionmentioning

confidence: 99%

Civil Aircraft Spare Parts Prediction and Configuration Management Techniques: Review and Prospect

Feng

Chen

et al. 2021

Advances in Mechanical Engineering

View full text Add to dashboard Cite

show abstract

“…Various studies have therefore attempted to show the benefits of a combined approach using both seismic and well data, in order to reduce the uncertainties of reservoir characterization (Toublanc et al, 2005;Fang et al, 2017;Albesher et al, 2020;Boersma et al, 2020;Méndez et al, 2020). Another aspect to consider is that manual interpretation of seismic data can be a very time-consuming task due to the high amount of data, which is why computational solutions, such as supervised and unsupervised neural networks have been increasingly used for seismic interpretation, pattern recognition, and lithology classification in recent years (Saggaf et al, 2003;Baaske et al, 2007;Bagheri & Riahi, 2015;Roden et al, 2015;Brcković et al, 2017;Zahmatkesh et al, 2021). Besides the long-time use for hydrocarbon reservoir investigation, seismic attribute analysis has also been increasingly used in geothermal exploration in recent years, especially for complex structured reservoirs (Pendrel, 2001;Chopra & Marfurt, 2007;Doyen, 2007;Abdel-Fattah et al, 2020), e.g., in Poland (Pussak et al, 2014) and Denmark (Bredesen et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

Advanced seismic characterization of a geothermal carbonate reservoir – Insight into the structure and diagenesis of a reservoir in the German Molasse Basin

Wadas

Bauer

Shipilin

et al. 2023

Preprint

View full text Add to dashboard Cite

Abstract. The quality of geothermal carbonate reservoirs is controlled by numerous factors and processes, such as the depositional environment, lithology, diagenesis, karstification, fracture networks, and tectonic deformation. Carbonatic rock formations are thus often extremely heterogeneous, and reservoir parameters and their spatial distribution are therefore difficult to predict. For the example of a 3D seismic dataset combined with well data from Munich, Germany, we demonstrate, how an advanced analysis can deliver an improved reservoir model concept and help to identify possible exploitation targets within the Upper Jurassic carbonates. To identify possible reservoir sections and to understand their above-mentioned controlling factors, we analyse different seismic single- and multi-attributes. Some of the seismic attributes, together with lithology logs from wells, are then used to identify parameter correlations between the seismic attributes and the different carbonate lithologies to obtain a supervised neural network based 3D lithology model of the geothermal reservoir. Furthermore, we compare the fracture orientations measured in seismic (ant-tracking analysis) and well scale (image log analysis), to address scalability. Our results show that, for example, acoustic impedance is well suited to identify reefs and karst-related dolines. Areas with strong karstification or fault- and fracture-related deformation, which are both associated with high permeabilities, are also indicated by e.g. strong frequency attenuation, variance anomalies, and/or morphological features like bowl-shaped structures derived from the shape index. Furthermore, by using sweetness we can reconstruct the reef development of two exemplary reefs, and regarding the lithology distribution, we show that the upper part of the reservoir is dominated by limestone and dolomitic limestone rather than dolomite. In addition, we observe spatial trends in the degree of dolomitization. With respect to the fracture orientations on seismic- and well scales, we point out that a general scalability is not possible due to a combination of methodological limitations and geological reasons. Nonetheless, we argue that the combination of both methods provides an improved overall impression of the fracture systems, and therefore possible fluid pathways. By taking all the results into account, we are able to improve and adapt recent reservoir concepts to outline the different phases of its structural and diagenetic evolution. Furthermore, our results help to identify high quality reservoir zones in the Munich area.

show abstract

“…Various studies have therefore attempted to show the benefits of a combined approach using both seismic and well data in order to reduce the uncertainties of reservoir characterization (Toublanc et al, 2005;Fang et al, 2017;Albesher et al, 2020;Boersma et al, 2020;Méndez et al, 2020). Another aspect to consider is that manual interpretation of seismic data can be a very time-consuming task due to the high amount of data, which is why computational solutions, such as supervised and unsupervised neural networks, have been increasingly used for seismic interpretation, pattern recognition, and lithology classification in recent years (Saggaf et al, 2003;Baaske et al, 2007;Bagheri and Riahi, 2015;Roden et al, 2015;Brcković et al, 2017;Zahmatkesh et al, 2021). Besides the long-time use for hydrocarbon reservoir investigation, seismic attribute analysis has also been increasingly used in geothermal exploration in recent years, especially for complex structured reservoirs (Pendrel, 2001;Chopra and Marfurt, 2007;Doyen, 2007;Abdel-Fattah et al, 2020), e.g.…”

Section: Introductionmentioning

confidence: 99%

Advanced seismic characterization of a geothermal carbonate reservoir – insight into the structure and diagenesis of a reservoir in the German Molasse Basin

Wadas,

Krumbholz,

Shipilin

et al. 2023

Solid Earth

View full text Add to dashboard Cite

Abstract. The quality of geothermal carbonate reservoirs is controlled by, for instance, depositional environment, lithology, diagenesis, karstification, fracture networks, and tectonic deformation. Carbonatic rock formations are thus often extremely heterogeneous, and reservoir parameters and their spatial distribution difficult to predict. Using a 3D seismic dataset combined with well data from Munich, Germany, we demonstrate how a comprehensive seismic attribute analysis can significantly improve the understanding of a complex carbonate reservoir. We deliver an improved reservoir model concept and identify possible exploitation targets within the Upper Jurassic carbonates. We use seismic attributes and different carbonate lithologies from well logs to identify parameter correlations. From this, we obtain a supervised neural-network-based 3D lithology model of the geothermal reservoir. Furthermore, we compare fracture orientations measured in seismic (ant-tracking analysis) and well scale (image log analysis) to address scalability. Our results show that, for example, acoustic impedance is suitable to identify reefs and karst-related dolines, and sweetness proves useful to analyse the internal reef architecture, whereas frequency- and phase-related attributes allow the detection of karst. In addition, reef edges, dolines, and fractures, associated with high permeabilities, are characterized by strong phase changes. Fractures are also identified using variance and ant tracking. Morphological characteristics, like dolines, are captured using the shape index. Regarding the diagenetic evolution of the reservoir and the corresponding lithology distribution, we show that the Upper Jurassic carbonate reservoir experienced a complex evolution, consisting of at least three dolomitization phases, two karstification phases, and a phase of tectonic deformation. We observe spatial trends in the degree of dolomitization and show that it is mainly facies-controlled and that karstification is facies- and fault-controlled. Karstification improves porosity and permeability, whereas dolomitization can either increase or decrease porosity. Therefore, reservoir zones should be exploited that experienced only weak diagenetic alteration, i.e. the dolomitic limestone in the upper part of the Upper Jurassic carbonates. Regarding the fracture scalability across seismic and well scales, we note that a general scalability is, due to a combination of methodological limitations and geological reasons, not possible. Nevertheless, both methods provide an improved understanding of the fracture system and possible fluid pathways. By integrating all the results, we are able to improve and adapt recent reservoir concepts, to outline the different phases of the reservoir's structural and diagenetic evolution, and to identify high-quality reservoir zones in the Munich area. These are located southeast at the Ottobrunn Fault and north of the Munich Fault close to the Nymphenburg Fault.

show abstract

Seismic facies analysis from well logs based on supervised classification scheme with different machine learning techniques

Cited by 27 publications

References 18 publications

Civil Aircraft Spare Parts Prediction and Configuration Management Techniques: Review and Prospect

Civil Aircraft Spare Parts Prediction and Configuration Management Techniques: Review and Prospect

Advanced seismic characterization of a geothermal carbonate reservoir – Insight into the structure and diagenesis of a reservoir in the German Molasse Basin

Advanced seismic characterization of a geothermal carbonate reservoir – insight into the structure and diagenesis of a reservoir in the German Molasse Basin

Contact Info

Product

Resources

About