Quality assessment method for metal magnetic memory pipe welds based on POA-CNN-BiGRU

Weld quality assessment is an indispensable part of pipeline maintenance. A metal magnetic memory (MMM) method was used to inspect pipeline welds and analyze the relationship between weld defects and MMM signals under different internal pressures. The magnitude of weld strength loss of defect-containing pipelines under different internal pressures was calculated, and a quantitative relationship between the strength loss value and the MMM signal was established. Subsequently, a method of weld strength loss inversion model is proposed, which employs a convolutional neural network (CNN) for local feature extraction of tangential and normal components of the MMM signal, further extracts the global sequence features of the signal sequence through a bidirectional gated recurrent network (BiGRU), and optimizes the parameters of the neural network model through the pelican optimization algorithm (POA). After experimental validation, the prediction accuracy of the model is improved by 26.57% and 15.15% compared with the CNN-BiGRU after grid optimization and sparrow search optimization, respectively, which provides a new idea for weld quality assessment.