Numerical Simulation and ANN Prediction of Crack Problems within Corrosion Defects

Area of Science:

• Civil Engineering
• Mechanical Engineering
• Materials Science

Background:

• Buried pipelines are critical infrastructure susceptible to fractures from corrosion and internal pressure.
• Existing research often focuses on failure pressure or residual strength, necessitating further analysis under seismic loads.

Purpose of the Study:

• To conduct a fracture analysis of buried pipelines with corrosion defects under seismic loads.
• To develop and validate a predictive model for crack tip stress fields in corroded pipelines.

Main Methods:

• Utilized the extended finite element method (XFEM) for modeling buried pipelines under seismic conditions.
• Developed a backpropagation (BP) neural network algorithm to predict stress fields at pipe crack tips.
• Optimized the neural network using a genetic algorithm (GA) to improve accuracy and convergence (GA-BPNN).

Main Results:

• Maximum stress at the crack tip occurred near a 5° circumferential angle in the corrosion area.
• The BP neural network achieved a prediction error of less than 10%.
• The optimized GA-BPNN model demonstrated a maximum error of 5.32% and superior adaptability.

Conclusions:

• Corrosion depth significantly influences the crack tip stress field, especially with increasing internal pressure.
• The GA-BPNN model provides accurate and adaptable predictions for buried pipeline fracture analysis.
• This research enhances understanding of buried pipeline integrity under combined internal pressure and seismic loading.