【摘要】 随着工业过程的规模和复杂程度的增加,对于过程安全性和可靠性的要求进一步提高.为了准确及时地检测设备故障,提出了一种基于连续隐马尔可夫模型(CHMM)的在线故障检测方法.采用主元分析(PCA)方法对过程变量数据进行特征提取,利用变长度滑动窗口技术跟踪动态数据,并提出了一个新的实时统计量作为在线故障检测的量化指标,结合实时阈值实现了CHMM的在线故障检测.将该方法应用于田纳西-伊斯曼(TE)化工过程,并与基于PCA和动态主元分析(DPCA)方法的故障检测结果进行比较,能够较准确地检测到故障,验证了该方法的有效性.更多还原

【Abstract】 With the increasing of industrial process scale and complexity,the demand for safety and reliability of process improves further.In order to detect the equipment fault accurately and timely,an online fault detection method based on continuous hidden Markov model(CHMM)was proposed.The principal component analysis(PCA)approach was adopted to take feature extraction of the process variables,and the variable moving window technology was utilized to track dynamic data,then,a new real-time statistic was presented as a quantitative index of on-line fault detection,and combined with realtime threshold to implement CHMM-based on-line fault detection.Then the proposed method was carried out in Tennessee Eastman(TE)process.Also,the method could detect fault more accurately compared with PCA and dynamic principal component analysis(DPCA)based methods.The effectiveness of the proposed method was verified by the experimental results.更多还原