Abstract: Partial least squares (PLS) has been extensively researched and applied in industrial process monitoring. In order to improve the monitoring effect based on PLS process monitoring, aiming at the problem that the traditional PLS method uses a fixed threshold which generates a lot of false alarms and missed detections, an adaptive threshold PLS process monitoring method is proposed. Firstly, the PLS monitoring model is established according to the normal historical data of the process, and the corresponding adaptive threshold is calculated according to the exponentially weighted moving average of the statistics for the process monitoring. Finally, using the Tennessee Eastman (TE) process and large blast furnace iron-making process simulation experiment to test the performance of the method. The experimental results show that the process monitoring based on adaptive threshold PLS can reduce the false alarms rate and improve the process monitoring performance compared with the traditional PLS method.

Key words:  Partial least squares, process monitoring, adaptive threshold, Tennessee Eastman process, blast furnace iron-making

摘要： 偏最小二乘法(Partial Least Squares, PLS)在工业过程监测等方面得到了广泛研究与应用。为提高基于PLS过程监测的监测效果，针对传统PLS方法采用固定阈值产生大量误报与漏报的问题，提出一种自适应阈值PLS的过程监测方法。该方法首先根据过程正常历史数据建立PLS监测模型，并根据统计量的指数加权移动平均值，计算相应的自适应阈值，用于过程监测。最后，采用田纳西-伊斯曼(TE)过程和大型高炉炼铁过程的仿真实验测试方法的性能，实验结果表明，相对于传统PLS方法，基于自适应阈值PLS的过程监测能够降低误报率，提高过程监测性能。

关键词: 偏最小二乘, 过程监测, 自适应阈值, 田纳西-伊斯曼过程, 高炉炼铁