基于正交信号修正与高效偏最小二乘的质量相关故障检测方法

偏最小二乘(PLS)是一种广泛应用于多变量统计过程监控中的有效算法,高效偏最小二乘(EPLS)是近年提出的一种PLS改进算法,在质量相关故障检测中具有良好的检测效果,但当测试数据含有质量无关故障时,EPLS算法的误报率偏高,可能导致误报警,对工业过程中的故障检测有较大影响.为降低检测质量无关故障的误报率,将EPLS结合4种正交信号修正(OSC)方法提出4种OSC-EPLS算法.用质量无关故障样本建立OSC模型对在线监测数据进行预处理,将处理后的信息用EPLS算法进行故障检测,误报率明显降低.最后结合田纳西-伊斯曼工业过程,应用OSC-EPLS、PLS、EPLS算法进行故障检测,分别比较误报率和有效报警率的大小,体现所提出算法在故障检测中的优势.

Quality-related fault detection method based on orthogonal signal correction and efficient PLS

The partial least squares algorithm(PLS) is an effective method, which has been widely used in multivariate statistical processes. Efficient projection to latent structures(EPLS) is an improved PLS algorithm proposed in recent years and has a good detection effect in quality-related fault detection. However, when the test data has a quality-independent failure, the false alarm rate of the EPLS algorithm is high, which may lead to false alarms and has a great influence on fault detection in industrial processes. In order to reduce the false alarm rate of detection-independent faults, four OSC-EPLS algorithms are proposed in combination with four orthogonal signal corrections(OSC) and the EPLS. In this paper, we establish an OSC model with quality-independent fault samples to preprocess on-line inspection data, and use the EPLS algorithm for fault detection to reduce the false alarm rate. Finally, combined with Tennessee Eastman industrial process, four kinds of the OSC-EPLS, PLS, and EPLS algorithms are used for fault detection. The false alarm rate is compared with the effective alarm rate to reflect the advantages of the proposed algorithm in fault detection.