Literature review: Current trends and advances in the use of artificial intelligence for ensuring the safety and efficiency of gas pipeline operations

Abstract

• Key Research and Application Areas: • Gas leak and fault detection using sensors, IoT, and machine learning. • Predictive maintenance through analysis of historical and real-time data. • Optimization of gas transportation, especially with hydrogen blending (HBNG). • Use of digital twins, SCADA systems, generative networks (GANs), and deep learning (DL). • Most Common AI Methods: • Machine Learning (ML): SVM, KNN, Random Forest, Naive Bayes • Deep Learning (DL): CNN, LSTM, YOLOv5, RBFNN, GAN • Hybrid Models: combinations of RNN, ANFIS, fuzzy logic • Numerical simulations and analyses: CFD, FEM, EMAT, GIS • Model Results and Accuracy: • Model accuracy reaches up to 99-100% in some cases: • YOLOv5: 97-98% • Random Forest: 99.9% • LSTM + Adam: 100% • ANFIS, CNN, SVM, GAN models: above 90% • ML + DL combinations often lead to higher efficiency than traditional methods. • Challenges in AI Implementation: • Lack of high-quality datasets - major barrier to model development. • High costs of sensors, data collection, and processing. • Security and regulatory constraints in gas industry applications. • Need for skilled personnel to develop and manage AI systems. • Future Trends: • Increased use of generative models (GAN) for data synthesis. • AI integration into real-world operations via SCADA and digital twins. • Development of robust hybrid models with better generalization. • Focus on pipeline lifespan prediction and early-warning systems. • Confirmed Hypotheses: • AI significantly reduces the risk of gas leaks (97-99% accuracy, faster response). • AI-assisted maintenance lowers costs and extends pipeline life. The use of artificial intelligence (AI) in gas pipeline monitoring and maintenance represents a significant advancement in the energy industry. This article provides an overview of current trends and AI technologies applied in fault detection, failure prediction, and gas transportation optimization. Key methods include machine learning, deep neural networks, numerical simulations, and digital twins. Research highlights the importance of integrating AI with the physical properties of materials for localizing and assessing corrosion defects. A bibliometric analysis reveals that most studies focus on the application of neural networks, support vector machines, and Bayesian networks in predictive maintenance. Despite significant progress, challenges remain, such as the lack of high-quality datasets, high implementation costs, and regulatory barriers. Future research trends focus on the integration of AI with SCADA systems, improving predictive models, and the broader use of generative neural networks for data synthesis. This review of research trends from 2020 to 2025 underscores the importance of artificial intelligence in the transportation sector and highlights its potential for further development in enhancing the reliability and safety of energy infrastructures.

Topics & Concepts