为了提高天然气集输管道内壁腐蚀深度的预测精度，提出了一种高斯残差过程管道内腐蚀预测方法。该方法通过粒子滤波器（particle filter, PF）推断残差的重要性，结合物理模型与历史数据，对数据驱动模型进行信任评估，并利用物理模型修正对应的物理模型和高斯过程模型的权重，从而实现内腐蚀深度的准确预测。方法的有效性通过某天然气管线的腐蚀监测数据进行了验证，结果表明，该方法的最大相对误差为1.276%，最小相对误差为0.045%。此外，该方法的均方根误差（E_RMSE）为0.92 μm，平均绝对百分比误差（E_MAPE）为0.69%。可见，所提方法具有很高的预测精度，能够有效且长期地对管道内壁腐蚀深度进行准确预测。

In order to improve the prediction accuracy of inner wall corrosion depth of natural gas gathering and transportation pipeline, a Gauss residual process corrosion prediction method was proposed. In this method, the importance of residual error was inferred by PF(particle filter), the trust of data-driven model was evaluated by combining physical model and historical data, and the weight of corresponding physical model and Gaussian process model was modified by physical model, so as to achieve accurate prediction of internal corrosion depth. The effectiveness of the method was verified by the corrosion monitoring data of a natural gas pipeline. The results show that the maximum relative error of this method is 1.276% and the minimum relative error is 0.045%. In addition, the root-mean-square error(E_RMSE) of the method is 0.92 μm and the mean absolute percentage error(E_MAPE) is 0.69%. It can be seen that the proposed method has high prediction accuracy and can accurately predict the corrosion depth of the inner wall of the pipeline effectively and for a long time.