马尔可夫过程融合贝叶斯网络的测试性建模研究

A Testability Modeling Study of Markov Process Incorporating Bayesian Network

在设备测试性建模中,通常将设备运行状态划分为正常和故障两态,忽略了设备故障演进特征,从而导致测试性指标虚高。该文提出一种马尔可夫过程融合贝叶斯网络的测试性优化建模方法,充分考虑设备退化过程。在故障与测试不确定性矩阵的基础上,基于马尔可夫随机过程建立了部件多状态退化模型(正常、过渡、故障),减少了两态条件下故障模式判定的不确定模糊区间,提高了系统测试性指标。利用贝叶斯网络推理确定多态测试分布函数,优化了虚警率和检测率。最后,以某型龙门式自动洗车机为对象进行了验证分析,与传统测试性方法相比,系统检测率提高7.71%,隔离率提高11.58%,虚警率下降9.50%,验证了该方法的有效性。

For the current testability modeling,the testability index is inflated by the simple division of equipment operation state into normal and fault states,which ignores the equipment fault evolution characteristics.In this paper,we propose a testability modeling method of Markov process incorporating Bayesian network,which establishes a multi-state degradation model of components(normal,transition,fault)based on Markov stochastic process on the basis of fault and test uncertainty matrix,reduces the fuzzy interval caused by the uncertainty of failure mode under the traditional two states,and improves the accuracy of the model.Bayesian network inference is used to determine the multi-state test distribution function and optimize the values of false alarm rate and detection rate.Finally,a gantry-type automatic car washer was analyzed as a case study,and the detection rate of the system increased by 7.71%,the isolation rate increased by 11.55%,and the false alarm rate decreased by 9.51%,thus verifying the effectiveness of the proposed method.