复杂系统可靠性分析结果的准确性与输入参数精度密切相关。针对包含多源不确定信息的可靠性分析问题，提出了基于贝叶斯最大熵的随机模型修正与参数校准方法。该方法将类型统计信息（例如矩信息、可靠度）转换为约束条件，进而利用不确定性优化求解传统参数估计问题。考虑混合不确定性，引入Wasserstein距离构建似然函数，并利用近似算法提高计算效率。该方法通过增加“熵项”的方式拓展了经典贝叶斯推理的适用范围，可处理多源异构数据与混合不确定性问题。针对多部件航空发动机转子系统，建立了基于survival signature的多态系统可靠性模型，并应用上述方法开展可靠性分析，结果显示本方法较传统方法具有更高精度与更强的稳健性。

The accuracy of reliability analysis results of complex systems is closely related to the accuracy of input parameters. A stochastic model correction and parameter calibration method based on Bayesian maximum entropy is proposed to solve the reliability analysis problem containing multi-source uncertain information. By converting multi-source statistical information （such as moment information and reliability） into constraint conditions， this method transforms parameter estimation into uncertainty optimization problem. Further considering the mixed uncertainty， Wasserstein distance is introduced to construct the likelihood function， and the approximation algorithm is used to improve the computational efficiency. This method extends the application scope of classical Bayesian inference by adding "entropy term" and can deal with multi-source heterogeneous data and mixed uncertainty problems. A multi-state system reliability model based on survival signature was established for a multi-component aero-engine rotor system， and the reliability analysis was carried out by using the above method. Through comparative analysis， it was verified that the proposed method has higher accuracy and stronger robustness than the traditional method.