贝叶斯诊断网络平台的开发与实现

针对贝叶斯诊断网络在实际应用时需要对具体对象建立相应诊断网络的问题,开发了一个贝叶斯诊断网络平台.重点讨论了网络的数字化、网络拓扑顺序的确定和赌轮选择等平台实现的3个关键问题.采用数字形式来描述网络的拓扑模型,刻画了网络的全部信息.节点间的关系则用关系矩阵简洁、直观地予以表达,并据此确定网络的拓扑顺序.利用赌轮的选择功能,实现了网络节点状态的实例化.该平台简单易用,为网络的建立和推理提供了一个有效、便利、通用的运行环境.利用该平台为天津石化炼油厂的一台烟机建立了贝叶斯诊断网络,实例表明,该平台能够用于贝叶斯诊断网络的构建和推理,也为贝叶斯诊断网络的工程应用提供了一个有力的工具.

Development and Realization of Bayesian Diagnostic Network Platform

The fact that Bayesian diagnostic network must be constructed for a specific diagnosed object in the practical applications motivated to develop a new network platform. The number pattern was introduced to describe the topological model of network, and relation matrix was adopted to represent those relationships among nodes, thus the platform provided an effective and flexible operation environment for construction and reasoning of Bayesian diagnostic network. The three important problems for realization of the platform were discussed: digitization of network, identification of topological order and selection of roulette wheel. A Bayesian diagnostic network for a gas turbine was constructed under the proposed platform. The practical example confirms the efficiency of the platform to construct and reason Bayesian diagnostic network in the engineering applications.