拉格朗日松弛算法在平台罗经多故障诊断技术中的应用研究

最优多故障诊断问题是一个NP-hard问题。针对平台罗经这一复杂系统，采用有向图描述元件与测试点间的因果依赖关系，并建立系统的多信号模型。在考虑元件发生故障的先验概率的前提下，提出一种基于拉格朗日算法（LRA）和子梯度优化算法（SOA）近最优多故障诊断算法，并在某型平台罗经的方位稳定系统多故障诊断中得到应用。结果表明：该方法相对于传统的单故障诊断方法诊断速度快，适用于平台罗经这类大型复杂系统的多故障诊断。

Multiple Fault Diagnosis Technology for Stabilized Gyrocompass Based on Lagrangian Relaxation Algorithm

The exact computation of conditional probabilities for the multiple fault diagnosis(MFD) is NP-hard.The multi-signal model for complex stabilized gyrocompass is presented which uses diagram to subscribe the causal dependency relationship among components and test nodes.A heuristic algorithm to find approximately the most likely candidate fault set was introduced which is based on Lagrangian relaxation algorithm(LRA) and Subgradient optimization algorithm(SOA) under the consideration for prior probability of the components failure of system and the test nodes.The proposed method is validated in one test for the azimuth-stabilized system of stabilized gyrocompass.The results show that the reasoning process is faster than the traditional method using single-fault diagnosis.This novel method can be applied in MFD for the stabilized gyrocompass system.