Abstract: Pipeline corrosion damage assessment based on deterministic or probabilistic corrosion models often requires a large dataset. In-service subsea pipeline corrosion detection is challenging and expensive, leading to limited data and difficulties in selecting distribution models during the assessment process. A dynamic assessment method for subsea pipeline corrosion damage under small sample data conditions is thusly proposed. This method integrates the prior knowledge of subsea pipeline corrosion and employs Bayesian inference to dynamically update corrosion model parameters with actual inspection data, enabling the dynamic assessment of subsea pipeline corrosion damage. Taking the Shengli Oilfield subsea pipelines as an example, actual corrosion data from different years of service are obtained and a corrosion damage probability model is developed to validate the method. The assessment results indicate that the method proposed can continuously correct and update the corrosion model with limited data. As the sample size and the number of parameter update iterations increase, the dispersion of model parameters gradually decreases. The alignment between the posterior distribution of model parameters and the actual conditions improves with an increasing input sample dimension, resulting in enhanced model accuracy. Utilizing this method allows for planning the optimal detection sample size for pipeline corrosion damage assessment, considering both detection costs and model accuracy.