Background As the operating conditions of the heat pipe reactor vary, its key thermal parameters exhibit significant fluctuations, making it difficult for numerical models based on fixed parameter assumptions to accurately characterize its thermal behavior.

Purpose To address this, this work proposes an online correction method based on Bayesian networks for key thermal parameters that cannot be directly measured in the heat pipe reactor.

Methods Firstly, a three-dimensional numerical model of the heat pipe reactor is constructed and multi-condition simulations are carried out to build a simulation dataset. Subsequently, a Bayesian network model is trained to characterize the conditional probability relationships between directly measurable temperature and power, and indirectly measurable gap thermal resistance and heat loss rate. Then, based on measured temperature and power, the Bayesian network is used to estimate the mean values of gap thermal resistance and heat loss rate, which are then used to update the parameters of the numerical model, thereby improving the model's online prediction accuracy.

Results Finally, the method is validated on a hardware-in-the-loop experimental platform for the heat pipe reactor. The results show that the maximum and average errors after correction are 0.65% and 0.42%, respectively, which are reduced by 13.18% and 9.59% compared to those before correction. Moreover, the prediction interval of the numerical model can encompass 92% of the experimental data.

Conclusions This work provides a reference for the application of artificial intelligence algorithms in thermal calculation of heat pipe reactors.