Development of an integrated deep learning-based remaining strength assessment model for pipelines with random corrosion defects subjected to internal pressures