Research on lithology identification method based on mechanical specific energy principle and machine learning theory
Introduction
Lithology identification is an important part of oil drilling engineering. Monitoring the working status of bottom hole drilling tools is directly related to the normal construction of drilling engineering. Traditionally, lithology identification is mostly through visual observation, relying on the experience of the staff to compare engineering data, usually using the cross-graph method to interpret the logging curve(Min, Pengbo, & Fengwei, 2020), but this method is cumbersome and requires a lot of materials and does not meet expectations (Saporetti, da Fonseca, & Pereira, 2019). In order to overcome this shortcoming, many research institutions have devoted themselves to the study of combining computer technology with lithology recognition technology. With the advancement of science and technology, lithology recognition technology has been inseparable from computer-aided design (Pour et al., 2019, Xie et al., 2018, Yang et al., 2016). The development prospect of logging technology is to develop high precision, multi information, integrated and intelligent comprehensive logging technology (Yang & Deng, 2020).
Commonly used lithology identification methods are mainly divided into lithology identification methods based on surface tools and lithology identification methods based on downhole tools. The lithology identification method based on surface instruments mainly relies on logging data (Liang, Chen, & Lu, 2019). For example, element logging is the cuttings returned by sampling the wellhead, and special methods are used to analyze the content of, or and other elements in the cuttings (Liang, Yun, Junaid Kan, & Gao, 2019), to determine the underground information. This method is simple to operate and easy to obtain data. However, the comprehensive logging is to monitor the entire drilling process. The data is more versatile and not targeted. This method is simple to operate and easy to obtain data, but the comprehensive logging is to monitor the whole drilling process, the data is more widely used and less targeted and is not only used to identify lithology, so the effectiveness of ground data for formation of lithology is weaker than the method based on the data of downhole measurement. The method of lithology identification based on downhole tools is mainly to use logging tools to collect downhole formation data and judge the downhole formation information (Singh, Ojha, & Sain, 2020). The lithology information collected in this way is more reliable. Drilling is usually stopped during logging. Therefore, the logging information can reflect the actual formation information. However, this method requires high downhole instrument performance, and obtains data slowly and is expensive (Gooneratne, Li, & Moellendick, 2017). Therefore, this research establishes a ground instrument lithology recognition method based on artificial intelligence algorithms to make up for ground instruments have the disadvantages of insufficient lithology recognition effect and insufficient real-time performance.
This paper has made the following contributions: applying specific mechanical energy to the field of lithology discrimination, broadening the application field of mechanical specific energy, and making up for the lack of theoretical basis for lithological discrimination. To research simulated annealing optimization (SA) and support vector machine (SVM) algorithms, designing a lithology recognition model based on SA optimized SVM algorithm (SA-SVM), and forming a set of lithology recognition methods based on machine learning. Combining machine learning and mechanical specific energy principles into the practice of lithology recognition engineering, establishing a lithology intelligent recognition model, realizing lithology prediction of unmarked samples, and providing theoretical references for the formation lithology recognition link of petroleum drilling engineering.
The structure of the rest of the article is arranged as follows: Chapter 2 introduces related work, including the study of lithology identification methods and the study of mechanical specific energy theory. Chapter 3 introduces SVM and SA algorithm theory. Chapter 4 establishes the lithology recognition model based on SA optimized SVM. Chapter 5 combines engineering data to realize the simulation research and analysis of the model. And finally puts forward a summary.
Section snippets
Related work
Olalere Oloruntobi and Stephen Butt et al. proposed a lithology identification method based on the LWD instrument, which correlated the mechanical specific energy with the lithology identification, and believed that the mechanical specific energy is a function of rock properties (Oloruntobi & Butt, 2020). Jian Sun and Qi Li et al. based on the measurement data of the LWD instrument, the machine learning algorithm is used to identify lithology, and the accuracy rate is greater than 90% (Sun, et
Support vector machines (SVM)
The two-class machine learning model is to use the labeled standard sample to establish a classification model through an artificial intelligence algorithm, where the standard sample label (Teimoorinia, Toyonaga, Fabbro, & Bottrell, 2020), input the unlabeled sample into the established classification model, and can predict the label of the unlabeled sample Attribute .
As shown in Fig. 2, it is an integrated learning and safe supervised learning method to
Sample data selection and processing
Select the actual logging data of a shale gas development well from 2000 m to 3075 m, and calculate the by eq. (3). The well mainly includes four rock formations: Lithology I (2000 m-2480 m), Lithology II (2480 m-2500 m), Lithology III (2500 m-2626 m), and Lithology IV (2626 m-3075 m).The main component of Lithology I are pyroxene and quartzite, which are common silicate rock-forming minerals, pyroxene has low hardness and quartzite has high hardness. Lithology II is Sandwich structure,
Conclusion
In oil drilling engineering, accurately identifying underground lithology and obtaining stratum information can provide guidance for the entire construction operation and ensure safe and orderly construction. The idea of SVM is to determine a hyperplane in a high-dimensional space under the condition of allowing a certain error rate to realize the soft division of classes and classes to the greatest extent. The SA algorithm is used to simulate the physical annealing process, the penalty
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgment
This work was supported by National Natural Science Foundation of China (NSFC), China (No.52074233). This work was supported by Science and Technology Cooperation Project of the CNPC-SWPU Innovation Alliance, China (No.2020CX040302).
References (33)
Real-time monitoring of mechanical specific energy and bit wear using control engineering systems
Journal of Petroleum Science and Engineering
(2017)- et al.
Geomechanical parameter estimation from mechanical specific energy using artificial intelligence
Journal of Petroleum Science and Engineering
(2019) - et al.
Estimating Proctor parameters in agricultural soils in the Ardabil plain of Iran using support vector machines, artificial neural networks and regression methods
Catena
(2020) - et al.
Real-time optimization of drilling parameters based on mechanical specific energy for rotating drilling with positive displacement motor in the hard formation
Journal of Natural Gas Science and Engineering
(2016) - et al.
Lithology identification using kernel Fisher discriminant analysis with well logs
Journal of Petroleum Science and Engineering
(2016) - et al.
Incremental on-line learning: A review and comparison of state of the art algorithms
Neurocomputing
(2018) - et al.
Lithology identification on well logs by fuzzy inference
Journal of Petroleum Science and Engineering
(2019) - et al.
Incremental and decremental fuzzy bounded twin support vector machine
Information Sciences
(2020) - et al.
Research and application of logging lithology identification for igneous reservoirs based on deep learning
Journal of Applied Geophysics
(2020) - et al.
Application of specific energy for lithology identification
Journal of Petroleum Science and Engineering
(2020)