TY  - JOUR
EP  - 219
TI  - Integrating Prophet Forecasting with Gaussian Mixture Model-Hidden Markov Model (GMM-HMM) for Early Warning System in Dam Deformation Monitoring
KW  - Gaussian Mixture Model
KW  -  Hidden Markov Model
KW  -  Prophet Model
KW  -  Dam Deformation Forecasting
AV  - public
ID  - eprints1322
A1  - Tshireletso, Thalosang
A1  - Moyo, Pilate
VL  - 14
SP  - 212
UR  - http://dx.doi.org/10.54203/jceu.2024.22
N2  - Ensuring dam safety requires a monitoring system that can predict deformations and detect anomalies in real-time. This study combines the forecasting capabilities of the Prophet model with the real-time anomaly detection of a Gaussian Mixture Model-Hidden Markov Model (GMM-HMM) framework. The Prophet model analyses historical deformation data to forecast future deformations, enabling early issue identification. The GMM-HMM framework continuously monitors incoming data to detect deviations from predictions. Results shows that the GMM-HMM, with 10 components and a Mahalanobis distance threshold of 0.1, achieved a precision of 0.602, recall of 1.0, and F-1 score of 0.751, ensuring high sensitivity and accurate anomaly detection on. The GMM-HMM was then used to detect anomalies on Prophet forecasted radial deformations. Anomalies were detected on upper limit and lower limit deformations. This combined approach enhances dam safety by integrating predictive and real-time monitoring capabilities, offering a comprehensive early warning system for dam infrastructure.
SN  - 2252-0430
PB  - Scienceline Publication
Y1  - 2024/09/15/
IS  - 3s
JF  - Journal of Civil Engineering and Urbanism
ER  -