A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Outlier Detection and Smoothing Process for Water Level Data Measured by Ultrasonic Sensor in Stream Flows
Water level data sets acquired by ultrasonic sensors in stream-scale channels exhibit relatively large numbers of outliers that are off the measurement range between the ultrasonic sensor and water surface, as well as data dispersion of approximately 2 cm due to random errors such as water waves. Therefore, this study develops a data processing algorithm for outlier removal and smoothing for water level data measured by ultrasonic sensors to consider these characteristics. The outlier removal process includes an initial cutoff process to remove outliers out of the measurement range and an outlier detection process using modified Z-scores based on the median absolute deviation (MAD) of a robust estimator. In addition, an exponentially weighted moving average (EWMA) method is applied to smooth the processed data. Sensitivity analyses are performed for factors that are subjectively set by the user, including the window size for the MAD outlier detection stage, the rejection criterion for the modified Z-score outlier removal stage, and the smoothing constant for the EWMA smoothing stage, based on four different water level data sets acquired by ultrasonic sensors in stream-scale experiments.
Outlier Detection and Smoothing Process for Water Level Data Measured by Ultrasonic Sensor in Stream Flows
Water level data sets acquired by ultrasonic sensors in stream-scale channels exhibit relatively large numbers of outliers that are off the measurement range between the ultrasonic sensor and water surface, as well as data dispersion of approximately 2 cm due to random errors such as water waves. Therefore, this study develops a data processing algorithm for outlier removal and smoothing for water level data measured by ultrasonic sensors to consider these characteristics. The outlier removal process includes an initial cutoff process to remove outliers out of the measurement range and an outlier detection process using modified Z-scores based on the median absolute deviation (MAD) of a robust estimator. In addition, an exponentially weighted moving average (EWMA) method is applied to smooth the processed data. Sensitivity analyses are performed for factors that are subjectively set by the user, including the window size for the MAD outlier detection stage, the rejection criterion for the modified Z-score outlier removal stage, and the smoothing constant for the EWMA smoothing stage, based on four different water level data sets acquired by ultrasonic sensors in stream-scale experiments.
Outlier Detection and Smoothing Process for Water Level Data Measured by Ultrasonic Sensor in Stream Flows
Inhyeok Bae (author) / Un Ji (author)
2019
Article (Journal)
Electronic Resource
Unknown
Metadata by DOAJ is licensed under CC BY-SA 1.0
Clustering and outlier detection for trajectory stream data
| TIBKAT | 2020
Sensor Activity Pattern (SAP) Matching Process and Outlier Detection
| Springer Verlag | 2015
Outlier Detection in Multivariate Hydrologic Data
| Online Contents | 2008
Outlier Detection in Multivariate Hydrologic Data
| British Library Online Contents | 2008