基于全局局部保留投影与测地距离的气化炉故障检测方法  

Global-local Preserving Projection and Geodesic Distance Based Gasifier Fault Detection

作　　者：庄稼 韦炜 朱书奔 李扬 鲍涛 王伟 常雪丁 王村松 ZHUANG Jia;WEI Wei;ZHU Shuben;LI Yang;BAO Tao;WANG Wei;CHANG Xueding;WANG Cunsong(l.Energy Group Coal to Liquid Branch,Ningxia Coal Industry Co.Ltd.,Lingwu 750411,Ningxia,China;Zhejiang Supcon Software Co.Ltd.,Hangzhou 310053,China;SUPCON Technology Co.Ltd.,Hangzhou 310053,China;Institute of Intelligent Manufacturing,Nanjing Tech University,Nanjing 210009,China)

机构地区：[1]国家能源集团宁夏煤业有限责任公司煤制油分公司,宁夏灵武750411 [2]浙江中控软件技术有限公司,杭州310053 [3]中控技术股份有限公司,杭州310053 [4]南京工业大学智能制造研究院,南京210009

出　　处：《科技和产业》2024年第14期266-273,共8页Science Technology and Industry

基　　金：国家能源集团项目(煤制油分公司服务(咨询)﹝2023﹞001号)。

摘　　要：高温、高压、强腐蚀工作环境下的气化炉易发生仪表测量故障,进而影响煤制油、煤制甲醇等生产工艺,甚至危及人员安全。针对上述问题,提出了一种基于全局局部保留投影(global-local preserving projection,GLPP)算法与测地距离的气化炉故障检测方法。首先,采用GLPP算法提取样本邻域确定的数据局部特征;然后,提出一种基于测地距离度量样本的非近邻关系的数据全局特征提取方法;进一步,利用提取的全局特征构建相应的统计量来进行故障检测。最后,分别通过田纳西伊斯曼(Tennessee Eastman,TE)与真实气化炉2个案例验证所提出方法的有效性和可行性。Gasifiers operating in high-temperature,high-pressure,and highly corrosive working environments are prone to instrument measurement failures.The failure affects production processes such as coal-to-liquids and coal-to-methanol,and even endangers personnel safety.In order to solve the above problems,aglobal local preserving projection and geodesic distance gasifier based fault detection method were proposed in this paper.Firstly,the GLPP algorithm was adopted to extract the local features of the data determined by the sample neighborhood.Then,ageodesic distance measurement sample's non-neighbor relationship-based data global feature extraction method was proposed.Further,the extracted global features were used to construct corresponding statistics for fault detection.Finally,the effectiveness and feasibility of the proposed method were verified through two cases which were Tennessee Eastman(TE)and a real gasifier.

关 键 词：故障检测 气化炉 测地距离 全局局部保留投影 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]