The independent component analysis (ICA ) was applied to a control loop diagnosis which boasting of strong correlation and non-Gaussian distribution of monitoring signals associated with its separate devices. Having inputs and outputs of the controller and transmitter, as well as the output of the actuator taken as the monitoring signal,and the historical data in normal conditions utilized to construct ICA model so as to get control limits of I2 and SPE. A fault happened in a control loop can be detected by means of changes of the statistics. In addition, mean values of contribution rates were used to locate the fault so as to reduce mistakes in diagnosis. Simulation results demonstrate the efficiency of this approach.%针对控制回路中各种设备的监测信号之间存在较强的关联性和非高斯分布的特点,首次将独立分量分析法( ICA)用于控制回路的故障诊断中.分别选取回路中控制器输入和输出、执行器输出、变送器输入和输出作为信号观测点,选择控制回路正常运行状况的数据进行ICA建模,获得I2和SPE统计限,当控制回路某个元件发生故障时可以通过统计量的变化对其进行监测.此外,还采用均值贡献率进行故障点定位,有效降低了故障的误诊率.
展开▼
