Advances in Pipeline Leak Detection: A Review of Computational Fluid Dynamics, Machine Learning, and Hybrid Diagnostic Frameworks

Abstract

Pipeline leakage is one of the current economically important operational issues in the water distribution systems, petroleum and chemical processing and pharmaceutical fluid procedures networks in the world. Over the past two decades the technology of the leak detection has been transformed fundamentally- beginning with manual scrutiny and primitive pressure assessments to considerably more intricate sensor-based, signal-driven, and the most current concentrating on hybrid-calculation-machine learning diagnostic schemes. The given review critically reviews the current state of the field of pipeline leak detection and particularly how the combination of computational fluid dynamics (CFD) simulation and supervised and unsupervised machine learning can take place. The paper in question categorizes the previous detection methodologies in terms of hardware-based, software-based and hybrid ML-CFD and evaluates the capability, constraints and practicable restrictions of its application with regards to synthesis of over thirty representative publications published in the period between ninth year 1987 and 2025. Much attention is specifically paid to such emergent themes as: Science (i) used physics-consistency of generating synthetic training data via physics simulation through COMSOL Multiphysics and ANSYS based models; (ii) the relative usefulness of classical ML algorithms (Random Forest, Support Vector machine, Ridge Regression, Gradient Boosting) versus deep learning models ( CNN, LSTM, hybrid CNNLSTM); (iii) how pre-electra-steady-state single-audioid models can be applied to the problem of variability; variability (iv) the treatment of This review has highlighted the presence of such unanswered challenges as the pre-eminence of steady-state single-fluid research, lack of experimental validation, and lack of standardized benchmark datasets, and suggests a systematic research agenda to fully autonomous, multi-fluid, real-time pipeline integrity management systems.