An automatic data process line identification method for dam safety monitoring data outlier detection. Issue 7 (9th March 2022)

• Record Type: Journal Article
• Authors: Zheng, Sen; Shao, Chenfei; Gu, Chongshi; Xu, Yanxin
• Is Part Of: Structural control and health monitoring. Volume 29:Issue 7(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-03-09
• Subjects: CS algorithm -- dam safety monitoring -- data process line -- image processing -- outlier detection; Structural engineering -- Periodicals; Structural control (Engineering) -- Periodicals; Automatic data collection systems -- Periodicals; Detectors -- Periodicals; 624.17
• DOI: 10.1002/stc.2948
• Languages: English
• ISSNs: 1545-2255

Abstract

Summary: In order to discover anomalies of dam structure behaviors and evaluate the operation status timely, it is quite demanding to analyze the dam safety monitoring data that has been collected from the instruments. However, outliers in original monitoring data may affect the accuracy of dam performance assessment, which need to be detected before analyzing monitoring data. Model‐based methods have been applied in outlier detection as a kind of common method for a long time, but they generally rely heavily on model accuracy and easily lead to misjudgment of outliers once the data structure is complex. Considering the monitoring data of dam effect variables (e.g., deformation, cracking, or seepage) tend to show strong continuity, complex periodic and trending changes with the environment, valid monitoring data can reflect the variation trend by forming a data process line. Therefore, data that deviate from the process line can be detected as outliers. In this paper, an automatic process line identification method for dam safety monitoring data outlier detection is proposed. First, after drawing a scatter plot of the dam monitoring data, a binary image of the scatter plot is inputted into the computer program. Afterwards, the binary image would be processed by Gaussian blur and image binarization techniques, and then the continuous points could be identified. After constant adjustment of the vertical ordinate range and introducing Cuckoo Search (CS) algorithm, the optimal process line identification and outlier detection were finally completed. The case studies demonstrate the proposed method can enhance the efficiency of outlier detection.