| Abstract |
Urban drainage pipeline networks play a pivotal role in ensuring the sustainability of urban infrastructure, but conventional evaluation procedures are hampered by issues with accuracy, efficiency, and scalability. Current strategies, such as manual inspection, empirical models, and traditional machine learning methods, tend to perform poorly in delineating the complicated, nonlinear forms of degradation exhibited by drainage pipes, resulting in inefficient maintenance regimes and expensive pipe failures. To surpass these constraints, this research presents a state estimation model based on the Genetic Algorithm-Backpropagation (GA-BP) neural network. The GA addresses the BP network's weight initialization, speeding up convergence and avoiding local minima to improve predictive precision. The suggested model learns well from past pipeline condition data and adjusts to changing environmental factors, providing a more stable evaluation system. Comparative performance with conventional BP neural networks and other machine learning methods illustrates that GA-BP exhibits better performance in predictive accuracy, stability, and computational efficiency. Experimental findings prove that our method significantly enhances evaluation accuracy for timely maintenance interventions and minimizes long-term operation costs. The suggested GA-BP model presents an intelligent, data-based decision-making tool for the evaluation of urban drainage pipelines, enabling proactive infrastructure management and enhancing urban sustainability. Future studies can expand this model to real-time monitoring and incorporation into smart city frameworks. © 2025 IEEE. |
