Machine learning models for the well logging data processing automatization in geophysical surveys
Year: 2025
Type: article
Abstract: Purpose. Creation and integration of a machine learning model to automatize the process of processing well log data from geophysical surveys in order to achieve prompt and highly accurate processing of geological and geophysical information in the context of PJSC Ukrnafta's business processes.. Methods. In the process of realizing the identified goal of the proposed study, the following methods of scientific research were identified: theoretical, empirical and data processing methods. Theoretical methods include an analysis of previous studies on a particular problem, a method of system analysis (analysis of the logging data processing process as a complex system, search for interconnected elements and connections between them, search for the role and place of the designed model in a complex system), a method of comparative analysis (identification and comparison of existing methods of logging data processing, as well as methods of digitizing existing analog data), a method of generalization and systematization (analysis and systematization of the data obtained from industrial production, generation of general conclusions and principles preceding the development of a machine learning model for processing well log data from geophysical surveys), and a modeling method (construction of a theoretical model of an automated logging data processing system). In the process of analyzing the research problem, empirical methods of scientific research were also used, namely the method of data analysis (obtaining and analyzing analog logs and digital well log data with further analysis of the format, structure, characteristics, presence of noise and artifacts, determining the amount and quality of data required to train a machine learning model, the method of observation (observation of the work of specialists engaged in professional activities, which will help identify key stages, problems and needs that should be taken into account when developing a machine learning model for automatization of processing well log data process from geophysical surveys. The data processing method is also no less important in the scientific research process and involves the comparison and use of machine learning models (to build a model for digitizing, classifying, regressing or clustering well log data), image processing methods for digitizing data (using pattern recognition algorithms, line extraction, segmentation and other image processing methods for automatic digitization of analog well logs) and data visualization methods (using various graphical methods such as graphs, charts, maps, etc. to visually represent well log data, the results of their processing and interpretation, and to visualize the operation of the machine learning model). Results. Expected results from the proposed research based on the analysis of the data in line with the peculiarities of building a machine learning model for process automatization of well logging data processing of geophysical surveys of wells proposed solution for process automation of digitization of data of the well logging station of GDW PJSC «Ukrnafta» by designing an artificial intelligence model based on machine learning (ML) methods will be an advantage over conventional image processing methods (which, in any case, will not allow processing all types of well logging data) based on automatization and performance processes (machine learning will create a model that can automatically process well logs without the need for manual intervention, which will significantly reduce the time required to process each well log and allow for fast processing of large amounts of data); in the future, the finished product will be able to select a large number of well logs in the «. JPG» format and process them all in one approach, without human intervention, accuracy and consistency (the machine learning model can learn to recognize and process different types of well logs with high accuracy and consistency, avoiding possible errors that may occur during manual processing), but it should be noted that accuracy is achieved due to the large amount of input and output data used to train the AI model, difficulty of processing (manual processing of well logs can be very difficult due to the variety of quality of well logs and their image on different background substrates), where a machine learning model can be trained to recognize and correct these difficulties, scalability (once trained, the model can be used to process any number of well logs, making it scalable and efficient for use with large amounts of data), adaptation (It is thanks to machine learning capabilities that the AI model can be trained to adapt to new types of data or changes in the log format, which in turn makes the model more flexible and allows it to respond to changing requirements). Therefore, the use of artificial intelligence and machine learning methods, in particular using «TensorFlow», is a promising way to ensure prompt and highly accurate processing of geological and geophysical information. Scientific significance. The scientific novelty of the proposed study is the first proposed approach to building a machine learning model to automate the process of processing well log data from geophysical surveys, and a solution to the process of digitizing the data of the well logging station of GDW PJSC «Ukrnafta» by designing an artificial intelligence model based on machine learning methods. Practical significance. The development of a machine learning model to automate the processing of well logging data for GDW PJSC «Ukrnafta» will significantly speed up the digitization of well logging data, effectively deal with the complexity of logging data processing, which often arises due to the different quality of records and their display, as machine learning creates a tool that can independently process large amounts of logging data without the need for constant manual work, which, in turn, not only reduces the time spent on processing each individual well log, but also makes it possible to quickly analyze large amounts of accumulated data, which, in the future, will allow processing a large number of well log images, even in «.JPG» format, in one approach, without involving human resources, improving the accuracy and uniformity of processing various types of well logs. Therefore, the designed machine learning model can be specially trained to recognize and adapt to such complexities, which greatly expands the capabilities of automated processing, and, due to the properties of machine learning, such a model has the potential to adapt. It can be trained and adjusted to process new types of data or in case of changes in the log format, ensuring the flexibility and compliance of the system with changing requirements in the future. Thus, the implementation of a machine learning model to automatize the processing of well logs is a strategically important step for PJSC «Ukrnafta», which promises to significantly improve the efficiency, accuracy and speed of work with geophysical data. Keywords: model, artificial intelligence, machine learning, deep learning, automation, well logs, algorithm, image, picture.
Language: English
Subfield: Ocean Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4410924345