基于贝叶斯网络的控制回路关联故障诊断

Cross-correlation fault diagnosis in control loop based on Bayesian network

在综合控制回路不同的监控算法组成一个控制回路监控系统时,系统出现的故障彼此之间可能会存在相互关联的情况,针对这一问题提出一种故障诊断方法.该方法利用贝叶斯网络强大的不确定性表达能力和学习推理能力,将所有监控器组成一个控制回路的性能监控诊断系统,当控制回路在不确定的环境下操作出现故障时,即使各个监控器工作性能低下,也能够及时准确地进行诊断.对该方法进行了应用仿真,仿真结果表明该方法能够对回路故障之间存在的交叉关联问题进行很好地辨识,精确地推理出回路各个组成部分出现故障的概率,验证了该方法具有较高的诊断精度.

When synthesizing different monitoring algorithms to form a control loop diagnostic system,the faults in the system may have cross-correlation. This paper proposes a method to solve this problem. The method takes the advantages of Bayesian network such as uncertain expression abilities and strong learning and reasoning abilities,and synthesizes different monitors to form a control loop performance monitoring and diagnosis system. When the cross-correlation faults occur in the control loop which is operated in an uncertain environment,it can give an appropriate inference timely and accurately even if the performance of each individual monitor is low. A simulation example of the method is given. Simulation results show that the method can solve the problem that the faults may have cross-correlation and can inference accurately the probability of failure for every part in the control loop. It shows that the proposed method is of highly diagnostic accuracy.