为解决高温高压容器剩余寿命在线预测难题，提出一种基于数字孪生的高温高压容器剩余寿命预测模型构建方法。该方法基于实时工况条件，采用ANSYS仿真模型进行耦合仿真，获取高温高压容器一定时域物理场，通过多轴蠕变损伤模型建立高温高压容器剩余寿命预测样本数据集，利用Tent-SSA优化的BP（back propagation）神经网络算法进行训练预测，建立机理模型与机器学习融合驱动的数字孪生高温高压容器寿命预测模型。最后以某型钠冷快堆蒸汽发生器关键部件的管板作为对象，试验结果表明该预测模型总体均方误差由优化前的3.219 7 × 10^-2降低至7.744 9 × 10^-3，模型更稳定且鲁棒性强、收敛快。

In order to solve the difficult problem of online prediction of the remaining life of high-temperature and high-pressure vessel, a method of constructing the remaining life prediction model of high-temperature and high-pressure vessel based on digital twin was proposed. The method was based on real-time working conditions, using ANSYS simulation model for coupled simulation, obtaining a certain time-domain physical field of high-temperature and high-pressure vessel, establishing a sample dataset of remaining life prediction of high-temperature and high-pressure vessel through the multiaxial creep damage model, and using BP (back propagation) neural network algorithm optimized by Tent-SSA for training prediction, to establish a digital twin high-temperature and high-pressure vessel life prediction model driven by the fusion of the mechanism model and machine learning. Finally, the tube plate, which is a key component of a certain sodium-cooled fast reactor steam generator, was used as an object, and the experimental results show that the overall mean square error of the prediction model is reduced from 3.219 7 × 10^-2 before optimization to 7.744 9 × 10^-3, and the model is more stable, robust, and fast converging.