A Multi-Sensor Detection Method Based on WGAN-GP and Attention-Bi-GRU for Well Control Pipeline Defects

With the rapid advances in the petroleum industry, the demand for defect detection of well control pipeline is increasing, and the detection accuracy of a single sensor can no longer meet the engineering requirements; therefore, In order to solve this problem, a multi-sensor joint detection method is established to improve the detection accuracy of well control pipeline defects. By using the combination of magnetic flux leakage, electromagnetic ultrasonic guided wave, and eddy current nondestructive testing, After collecting multi-sensor well control pipeline defect data, the WGAN-GP based enhanced model is used to expand the data set,. Taking GRU as the basis of multi-sensor well control pipeline defect data fusion model, two-way connection and attention mechanism are introduced into the original GRU network to complete the construction of multi-sensor well control pipeline defect data fusion model based on Attention-Bi-GRU, and the expanded defect data set is fused. The experimental results show that the relative error of the model in well control pipeline defect detection is less than 0.5% To be contribution and novelty (1) The WGAN-GP based enhancement model is used to expand the data set, which is used as the training set to enhance the generalization ability of the data fusion model. (2) The defect detection data of multiple sensors are fused based on Attention-Bi-GRU fusion model, which realizes comprehensive and accurate detection of well control pipeline defects.