改进拉格朗日松弛算法在多故障诊断中的应用

Improved Lagrange relaxation algorithms and its application in multiple fault diagnostic strategy

为解决单故障策略可能出现的误诊和漏诊问题,从多故障的征兆出发,建立表征故障-测试依赖关系的多信号模型图,基于后验概率最大的思想,借助单故障诊断树的测试序列,采用拉格朗日松弛与次梯度优化方法求解多故障问题。进一步对次梯度算法的迭代步长提出改进,在设备研制初期传感器信息不足的情况下,对多故障进行准确定位。实例结果验证了该方法能够有效地排除隐含故障和伪故障,改进算法具有更好的诊断效率。

To solve the problems of missed diagnosis and misdiagnose in signal fault diagnostic strategy, multi-signal flow graph model descrihing relationship between fault and test was established based on signature of multiple-fault. Considering available test sequence provided by single-fault diagnostic strategy, Lagrange relaxation and subgradient optimization algorithm was pro posed to solve the problem of multiple-fault diagnosis based on the maximum posteriori probability. An improved iteration step on subgradient optimization algorithm to promote efficiency was proposed, which accurately located the fault on the condition of lacking of sensor information. An application example indicates that this algorithm can exactly remove hidden fault and masking false failures, and it has higher diagnostic efficiency.