摘要
针对催化裂化装置烟气轮机运行环境恶劣,故障频发等特点,开展基于PCA和MSET联合模型的振动监测预警研究。基于原始振动波形进行时频域特征提取计算,获得29个时频域特征参量,采用PCA方法将29个时频域特征参量重新组合成6个新主元,可以有效实现数据降维。烟气轮机正常运行状态下振动监测数据量的选择影响T2值变化,主元99%可信区间阈值随着数据量的增加降低至稳定值,采用时间点数据量2000满足分析要求。主元6是均方根值、方差和频率标准差的综合表征,在发生故障阶段波动较大,采用MSET模型进行残差分析,相对固定高报报警监测,可以超前预警8.5 h。
In view of the harsh operating environment and frequent failures of smoke turbine in catalytic cracking unit,vibration monitoring and early warning research based on PCA and MSET combined models was carried out.Based on the original vibration waveform,29 time-frequency feature parameters were obtained by the feature extraction calculation.Then,29 time-frequency feature parameters were recombined into 6 new principal components through the PCA method,which can effectively reduce the data dimensionality.The selection of vibration monitoring data quantity affected the change of T~2 value under normal operation of smoke turbine.The 99%confidence interval threshold of principal component decreased to a stable value with the increase of data volume,and the data volume of 2000 was adopted to meet the analysis.As a comprehensive representation of the root mean square value,variance and frequency standard deviation,principal component 6 fluctuated greatly at the failure stage.The residual analysis was calculated by the MSET model,and 8.5 h in advance was obtained by early warning compared with the fixed value warning.
作者
张伟亚
陈文武
韩磊
潘隆
Zhang Weiya;Chen Wenwu;Han Lei;Pan Long(State Key Laboratory of Safety and Control for Chemicals,SINOPEC Research Institute of Safety Engineering Co.,Ltd,Shandong,Qingdao,266104)
出处
《安全、健康和环境》
2023年第1期35-39,48,共6页
Safety Health & Environment
基金
中国石化重大专项(321123-3),催化裂化装置智能化安全绿色运行关键技术——关键设备失效识别与早期预警技术。
关键词
烟气轮机
振动监测
故障预警
PCA数据降维
MSET残差分析
smoke turbine
vibration monitoring
fault early warning
PCA data dimensionality reduction
MSET residual analysis
