Due to the complicated structure of conventional signed directed graph (SDG)model of complex thermal system,problems like combinatorial explosion in search paths and difficulty in distinguishing mul-tiple fault sources occur easily during the diagnosis.Thus,the hierarchical probability signed directed graph (SDG)model was proposed,which uses the layered method to simplify the search paths and describes the causal relationship between failures by applying a combination method of qualitative and quantitative meth-od.Moreover,a new fault diagnosis method based on the hierarchical probabilistic signed directed graph was put forward.This method takes advantage of the established SDG model to reduce the search space first,and then employs the compatibility and Bayesian inference to diagnose the faults,at last,it calculates the probability of failure and possible sequence of the source.Finally,this method was applied in failure di-agnosis of high pressure heater system of a 300 MW unit.The results show that,this method not only im-proves the fault source resolution but also reduces the search space,which avoids the"combinatorial explo-sion".%复杂热力系统的传统符号有向图(SDG)模型结构复杂，诊断时搜索路径易产生“组合爆炸”和难于区分多故障源等问题，对此提出了基于分层概率的 SDG 模型，利用分层简化搜索路径，采用定性和定量相结合的方法描述故障之间的因果关系；基于分层概率 SDG 的故障诊断方法，通过建立系统的分层有向图模型减少搜索空间，利用相容性和贝叶斯推理进行诊断，得到各故障源的概率及可能性排序。将该方法应用于某300 MW机组高压加热器给水系统的故障诊断，其在提高故障源分辨率的同时减少了搜索空间，避免了“组合爆炸”的发生。
展开▼
