为了研究故障在复杂工程系统中的传播机制,根据关键节点的状态异常信息预测系统发生故障的概率,提出一种基于贝叶斯网络的故障预测方法.根据工程系统自身固有的网络拓扑结构,构建了多层贝叶斯网络模型,利用定性趋势分析法将时间信息融入网络节点中,使得网络具有处理时序信息的能力,便于进行故障传播机理分析和故障预测.提出了基于元器件健康度的根节点故障概率确定方法,针对完备数据集和非完备数据集,选择不同的参数学习方法确定贝叶斯网络的条件概率表,采用多树传播算法进行联合概率推理,由系统根节点运行状态推测其余节点的故障概率.算法在Quanser三自由度四旋翼直升机上进行了仿真应用,结果验证了该方法的可行性和有效性.
To find the fault propagation mechanism and predict systemlevel faults probability ac cording to the abnormalities of key components in a complex engineering system, a fault prognosis method via Bayesian network is proposed. According to the inherent topological structure of an engi neering system, a multilayer Bayesian network is developed firstly, which can handle the timede pendent information by incorporating the qualitative signal trend information into the nodes of the Bayesian network. Therefore, the developed network is suitable for failure propagation analysis and fault prediction. A method for identifying the failure probability of the root nodes in the network is also proposed based on the nodes' integrative health index. Different parameter learning algorithms are adopted to determine the conditional probability table of the Bayesian network for complete and incomplete data sets, respectively. The Pearl's poly tree propagation algorithm is used for joint prob ability reasoning. The proposed fault prognosis method can predict the probabilities of possible fail ures based on the current status of root nodes. The application results on a Quanser 3DOF hover simulation system can verify the effectiveness and feasibility of the proposed method.