中文摘要：

针对现有改进互信息爬山（MI＆HC）算法精度低、耗时长及简化爬山（SHC）算法产生大量冗余边的问题,提出一种新的结构学习算法,即改进爬山（IHC）算法。通过计算互信息链得到贝叶斯初始结构,利用条件独立性测试以及对孤立节点进行处理来加边补充贝叶斯初始结构得到完全结构,利用改进的爬山搜索算子对完全结构进行搜索直到得出最优结构。将该算法与爬山（HC）算法、MI＆HC算法、SHC算法进行比较,仿真结果表明,IHC算法能够得到较高准确率的模型,时间开销最小而且产生的冗余边数远远少于SHC算法产生的冗余边数。最后基于IHC算法,结合某回转窑数据进行训练,得到了回转窑工艺参数的故障诊断模型,对回转窑的烧成带温度实现了较为准确的故障诊断。

英文摘要：

The improved mutual information hill climbing algorithm （MIHC） had less accuracy and time consuming,and the simplified hill-climbing （SHC） algorithm generated lots of redundant edges.Aiming at these problems,this paper proposed a new Bayesian structure learning algorithm,improved hill-climbing （IHC）.Firstly,this algorithm calculated the mutual information chains to obtain the Bayesian initial structure.Secondly,this algorithm utilized conditional independence tests and dealt with isolated nodes to supplement the initial structures to obtain the completed structure.Finally,the algorithm utilized improved search operators of HC to search for the optimal structure based on completed structure.The simulation results show that：compared with hill-climbing （HC）,MIHC and SHC algorithm,the IHC algorithm may obtain a more accurate and more rapid model,the redundant edges generated by IHC are far less than SHC.In the end,based on IHC algorithm and combined with some rotary kiln operating data,the fault diagnosis model of processing parameters may be constructed and a precise fault diagnosis of burning zone temperature in the rotary kiln is realized.