论文部分内容阅读
针对工业系统监控中存在的非高斯性、动态性以及缺失数据等问题,提出了基于动态贝叶斯网络的故障辨识方法.构建了混合高斯输出动态贝叶斯网络(DBNMG)模型,并基于期望最大化算法推导了DBNMG模型的参数学习策略.对于缺失数据问题,提出了一种非修补的DBNMG模型推理方法,利用部分的观测数据实现对故障的检测和辨识.以连续搅拌釜式反应器(CSTR)为对象,对本文提出的方法进行了仿真研究,仿真结果证明了本文所提方法的有效性.
Aiming at the problems of non-Gaussian, dynamic and missing data in industrial system monitoring, this paper proposes a fault identification method based on dynamic Bayesian network, and builds a hybrid Gaussian output dynamic Bayesian Network (DBNMG) model based on expectation The maximum learning algorithm derives the parameter learning strategy of DBNMG model.For the missing data problem, a non-patching DBNMG model inference method is proposed to detect and identify the fault using some of the observed data.Using continuous stirred tank reactor CSTR) as the object, the method proposed in this paper is simulated and the simulation results prove the effectiveness of the proposed method.