为了解决卫星这类复杂系统采用单一的智能故障诊断技术的诊断局限性问题,提出一种基于模糊综合评判的神经网络和专家系统相结合的诊断方法。选用改进的BP神经网络,添加动量因子和可变学习速率,并通过计算十个相同结构神经网络输出的标准差获得诊断结果的可信度。采用基于可信度的专家系统不确定性推理,给出了可信度计算方法。采用代数积与代数和模糊算子对两种诊断方法的结果进行综合评判。成功地应用于某卫星电源系统的故障诊断中。结果表明诊断的可信度和诊断的正确率得到提高。 In order to solve the diagnostic limitation of a single intelligent fault diagnosis technology for complex systems such as satellites, a diagnostic method combining neural network with expert system based on fuzzy comprehensive evaluation is proposed. The improved BP neural network was used to add the momentum factor and variable learning rate. The reliability of the diagnostic results was obtained by calculating the standard deviation of ten neural networks with the same structure. Adopting the uncertainty reasoning based on the expert system, the credibility calculation method is given. Using algebraic product and algebra and fuzzy operator to make a comprehensive evaluation of the results of the two diagnostic methods. Successfully applied to a satellite power system fault diagnosis. The results show that the credibility of the diagnosis and diagnostic accuracy is improved.