Abstract:

Data-driven statistical process monitoring approach, including principal component analysis(PCA) and partial least square(PLS), has been widely used in chemical process monitoring and fault detection.Sensor fault diagnosis algorithms based on signal reconstruction were explored and a general formulation of signal reconstruction algorithm was presented.The fault isolation criteria of model space and residual spaced were defined and the sufficient and necessary conditions to detect and isolate sensor faults were given.Different signal reconstruction approaches were compared in detail with a CSTR simulation application.

Key words:

故障诊断, 传感器故障, 可检测性

摘要：

主元分析、偏最小二层等数据驱动的多元统计监控方法由于不依赖于精确的数学模型，在化工过程监控与故障检测方面取得了广泛应用.通过研究基于统计信号重构的传感器故障诊断算法，给出了统计信号重构算法的一般形式，并推导了基于统计信号重构算法进行传感器故障诊断的可检测与可分离性条件，定义了模型空间和余差空间的故障识别指标.通过CSTR仿真对象的应用比较了不同统计信号重构算法间的差异，验证了故障诊断算法的有效性.

关键词:

故障诊断, 传感器故障, 可检测性