基于Rollout算法的测试不可靠条件下的诊断策略*

实际复杂装备系统测试时存在虚警和漏检，其导致的误诊费用往往超过了测试成本。针对复杂系统故障诊断中诊断精度低、虚警率高的问题，进行了不可靠测试条件下，基于Rollout算法的诊断策略优化方法研究。首先建立综合考虑测试点的故障检测能力、信息量、测试费用以及测试结果信任度的启发函数，然后利用Rollout算法生成诊断策略，最后用阿波罗检测系统实例验证了该算法的优越性。理论和实验表明，Rollout算法的平均测试费用低于贪婪算法，而且计算时间在可接受范围内，因此可用于测试不可靠条件下的诊断策略设计。

Strategies for fault diagnosis under unreliable test based on Rollout algorithm

The actual complex system has false and missing alarm during testing and diagnosis, which can cause higher misdiagnosis cost than detection cost. Considering the false and missing alarm of complex system diagnosis, this article proposes strategies for fault diagnosis under unreliable test based on the Rollout algorithm. Firstly, establish a heuristic evaluation function considering fault diagnosis abilities, information entropy ,test cost and the reliability of the test, and then use the Rollout algorithm getting the fault diagnosis strategy. The theory and experiments demonstrate that the method is much better than greed algorithm and its computing time is acceptable. So it can be used to design the strategy for fault diagnosis under unreliable test.