论文部分内容阅读
提出了一种基于动态模拟的精馏塔故障诊断方法。该方法从现场采集数据,同时利用动态模拟获得这些数据的模拟值,将二者之差作为状态监测目标。为减少监测变量个数,将这些残差针对模型误差进行校正,并经标准化后形成T~2统计量及其阈值。如果该统计量低于阈值,则说明装置运行正常;否则,说明装置出现了故障,此时启动故障诊断算法。故障诊断通过动态模型的在线参数估计完成,这些参数代表了设备内在故障原因和未知的外界干扰,从这些参数的变化过程就可以分析出故障的原因。参数估计算法基于参数对残差的灵敏度分析来构建,以残差的平方和为优化目标,通过非线性最小二乘法来实现。将该方法应用到了田纳西-伊斯曼仿真流程(TEP)中的汽提塔,分析了故障1时T~2统计量的变化,并给出了故障7时的塔底进料损失参数变化。应用结果表明,该方法具有较为灵敏的监测特征,并可以较为准确地给出故障的原因。
A dynamic simulation based rectification tower fault diagnosis method is proposed. The method collects data from the field and uses the dynamic simulation to obtain the simulated values of these data, taking the difference between the two as the state monitoring target. To reduce the number of monitoring variables, these residuals are corrected for model errors and standardized to form T ~ 2 statistics and their thresholds. If the statistic is lower than the threshold, the device is operating normally; otherwise, it indicates that the device has failed and the fault diagnosis algorithm is started. Fault diagnosis is accomplished through on-line parameter estimation of the dynamic model. These parameters represent the intrinsic cause of the fault in the equipment and unknown external disturbances. The reason for the fault can be analyzed from the changing process of these parameters. The parameter estimation algorithm is based on the sensitivity analysis of the parameters to the residuals. The square sum of the residuals is used as the optimization target and is achieved by the nonlinear least-squares method. The method was applied to the stripper in the Tennessee - Eastman simulation process (TEP). The T ~ 2 statistics of failure 1 were analyzed. The variation of feed loss parameters at the bottom of the failure was also given. The application results show that this method has more sensitive monitoring features and can give the reason of failure more accurately.