An on-line detection method for outliers of dynamic unstable measurement data

作者：Su, Weixing; Liu, Fang^*; Zhao, Jianjun; He, Maowei; Chen, Hanning

来源：Cluster Computing, 2019, 22(S4): S7831-S7839.

DOI：10.1007/s10586-017-1458-3

• Outlier detection
• Auto-regression
• Wavelet
• HMM
• Time series

Aiming at the characteristics of the vibration data collected by the regulation system during the unstable regulation process and the deficiency of the traditional wavelet anomaly detection method, an on-line anomaly detection method combining the autoregressive and the wavelet analysis is proposed to detect the abnormal data of the regulation system. By introducing the improved robustARmodel, thismethod can overcome the problem that the time and frequency of traditional anomaly detection using wavelet analysis method cannot be well balanced, and ensure the rationality of abnormal value detection of process data. Considering the general parameters of the regulation system is time-varying and has strong dynamic characteristics, the method proposed in this paper has the ability of online detection and real-time updating of parameters to ensure that the control parameters of time-varying control system; In order to avoid the problem that the traditional anomaly detection method needs to set the detection threshold in advance, HMM is introduced to analyze the wavelet coefficients and update the HMM parameters online, which can ensure that the HMM can well reflect the actual distribution of process data anomalies. It is proved that the method of anomaly data detection proposed in this paper is more suitable for the unstable regulation process data and has certain practicability through experiment and application.

• 出版日期2019-7
• 单位天津工业大学; 北京矿冶研究总院