Abstract: The long-gauge fiber Bragg grating (LG-FBG) sensing system enables accurate assessment of pipeline structural health.However,environmental complexities may lead to data loss,resulting in incomplete pipeline evaluation and reduced early warning capabilities for potential issues.To address this,the present study deployed an LG-FBG sensing system in pipeline engineering and analyzed the likelihood of monitoring data loss.Subsequently,an optimized bidirectional long short-term memory network (Bi-LSTM) combined with a generative adversarial network (GAN) was employed to capture the spatiotemporal correlations between available and missing data.The study further examined the impact of missing time proportions and the number of missing sensors on model recovery performance.Results indicate that when the missing time proportion is below 18/24,the model’s recovery performance exhibits minor degradation but remains stable overall.However,when the missing time proportion exceeds 18/24,recovery performance declines significantly.To ensure high recovery accuracy,it is recommended to maintain the missing time proportion within 18/24.Additionally,among various data recovery tasks,the Bi-LSTM-GAN model optimized with hunger game search demonstrated the best performance in evaluation metrics,effectively capturing the spatiotemporal correlations between available and missing data.In conclusion,this study integrates the LG-FBG sensing system with data-driven methods to systematically investigate the effectiveness of missing data recovery,providing a more comprehensive quantitative assessment for pipeline structural health monitoring.