欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
往复压缩机现有报警方式单一，多采用“单特征值报警”与“门限报警”的方式，经常导致设备盲目停车而影响生产，无法综合分析设备当前运行状态是否异常并提前预警。针对该问题，提出一种基于状态子空间的往复压缩机自动预警方法。该方法提取设备运行状态信号的特征参数，构造多维特征矩阵，利用核主元分析（kernel principal component analysis，KPCA）方法对多维特征矩阵进行降维，构建状态子空间，计算正常状态和当前状态子空间之间的差异度，并通过故障案例数据自学习得到差异度指标的报警阈值。经实际故障案例验证，该方法能大幅提前往复压缩机典型故障报警时间点，提高在线状态监测系统的故障预警能力。
英文摘要：
The use of ＂single feature alarm＂ and ＂ threshold alarm＂ makes the single alarm mode for the reciprocating compressor, which leads to blindly parking and affect the production. It is also unable to analyze whether the operating condition of equipment is abnormal or not comprehensively and to alarm in advance. An automatic alarm system for reciprocating compressor based on state subspaee is proposed in this paper. Features of operating condition signals of the equipment are extracted to construct a multidimensional feature matrix. The dimension is then reduced by KPCA （Kernel Principal Component Analysis, KPCA） method to construct the state subspace. The difference degree between the normal state subspace and the current state subspace is calculated and the alarm threshold of the difference degree is obtained by self-learning of data of fault cases. Verified by actual failure cases, the method can significantly advance the alarm time point of typical failure of reciprocating compressor and improve the fault alarm ability of online condition monitoring system.
同期刊论文项目
同项目期刊论文
期刊信息
位置：