Uncertainty analysis method for diagnosing multi-point defects in urban drainage systems

Area of Science:

• Environmental Engineering
• Computational Fluid Dynamics
• Geotechnical Engineering

Background:

• Urban drainage systems (UDS) are critical infrastructure, but pipe defects cause seepage, leading to urban flooding and environmental issues.
• Current defect detection methods, like inverse optimization, often yield a single solution, neglecting uncertainty and risking misdiagnosis.
• Diagnosing multiple defects simultaneously presents a significant computational challenge due to the high-dimensional parameter space.

Purpose of the Study:

• To develop a novel hybrid method for accurate and efficient multi-point defect diagnosis in UDS.
• To incorporate uncertainty analysis into defect localization to mitigate misdiagnosis.
• To reduce the computational burden associated with multi-point defect identification.

Main Methods:

• A hybrid approach combining a multi-population genetic algorithm (GA) for broad model space exploration and the adaptive Metropolis (AM) algorithm for posterior probability distribution (PPD) estimation.
• GA is utilized to identify probable defect locations, followed by AM to refine the PPD of defect parameters.
• Performance evaluation using accuracy (ACC), Matthews correlation coefficient (MCC), and mean absolute error (MAE) on synthetic UDS data.

Main Results:

• The proposed hybrid method demonstrated superior performance in multi-point defect diagnosis compared to the DiffeRential Evolution Adaptive Metropolis method, achieving higher ACC (0.91 vs. 0.78) and MCC (0.87 vs. 0.69).
• Diagnosis speed was enhanced by 32%, addressing the computational burden of traditional methods.
• The estimated PPD passed 90% confidence interval validation, confirming the reliability of the uncertainty analysis.

Conclusions:

• The hybrid GA-AM method offers an effective solution for diagnosing multi-point defects in UDS, providing robust uncertainty analysis.
• This approach significantly reduces the risk of misdiagnosis often associated with traditional single-solution methods.
• The enhanced speed and accuracy make this method suitable for practical application in urban infrastructure management.