摘要
将全局保持结构分析技术融入核慢特征分析中,实现基于改进核慢特征分析的间歇过程故障检测,可以有效提高对间歇过程的监控效率。文章采用两步多路数据展开策略将三维训练数据展开成两维矩阵,利用改进核慢特征分析提取间歇过程的局部动态数据,检测青霉素发酵过程中6种故障的仿真数据。结果表明:相比于传统的多路核主元分析、多路核独立元分析和核慢特征分析监控方法,基于改进核慢特征分析的监控方法具有最早的故障检测时刻,分别在第100、100、102、108、112和108个采样时刻检测到故障F1~F6;对间歇过程的6种故障具有最高的故障检测率,F1~F6的故障检测率分别为100%、100%、99%、96.67%、96.62%和97.97%;构造保留慢特征数目的准则是有效可行的。
To tackle the high nonlinearity and inherently time-varying dynamics of batch process,an improved KSFA(IKSFA)based fault detection approach is proposed by integrating global preserving structure analysis into KSFA model in order to improve the monitoring performance of batch process.In the proposed method,a two-step multiway data unfolding strategy is utilized to convert the three-way training dataset into a two-way matrix.The IKSFA approach is then used to explore the local dynamic data relationships as well as to mine the global structure information.A rule based on the cumulative slowness contribution is designed to determine the number of retained slow features.The case study on the six simulated faults of the fed-batch penicillin fermentation process demonstrates that compared with the traditional MKPCA,MKICA and KSFA methods,IKSFA based method gains the earliest fault detection time as well as the highest fault detection rate and that the rule of determining the number of retained slow features is proved to be effective.Faults were detected at 100,100,102,108,112 and 108 h respectively.The fault detection rates of F1-F6 are 100%,100%,99%,96.67%,96.62%and 97.97%respectively.
作者
张汉元
张汉营
梁泽宇
ZHANG Hanyuan;ZHANG Hanying;LIANG Zeyu(School of Information and Electrical Engineering, Shandong Jianzhu University, Jinan 250101, China;School of Thermal Engineering, Shandong Jianzhu University, Jinan 250101, China)
出处
《山东建筑大学学报》
2020年第1期42-49,共8页
Journal of Shandong Jianzhu University
基金
山东建筑大学博士科研基金项目(XNBS1821)
关键词
核慢特征分析
间歇过程
故障检测
全局结构分析
kernel slow feature analysis
batch process
fault detection
global preserving structure analysis
