Research on singular value detection method of concrete dam deformation monitoring

Abstract

Stable and reliable deformation monitoring data of concrete dam is the information basis and powerful guarantee for comprehensively analyzing the evolution law of structural morphology and evaluating the safety of long-term service. Given the characteristics of concrete dam deformation monitoring data such as small amplitude variation between neighbors, complex distribution characteristics, prominent local singular features, and the defect of traditional singular value detection methods, such as the prior hypothesis of the normal distribution, low local singular values sensitivity, and poor detection performance, the singular value detection method of concrete dam deformation based on improved local outlier factor (LOF) is established in this paper. Based on LOF that is sensitive to the local singular characteristics, differential expansion theory is applied to improve the adaptability and sensitivity to deformation monitoring data of concrete dam. A scientific and reasonable judgment standard for singular values is built with the help of the principle of small probability meanwhile. The application of engineering examples can draw that the concrete dam deformation singular value detection method based on improved LOF has high detection accuracy and stability, the recall rate of the test result is higher than 90%, the weighted evaluation indicator is higher than 80%, and the detection performance is significantly better than traditional singular value detection methods repeatedly. The application results of engineering examples show that the concrete dam deformation singular value detection method based on improved local outlier factor is highly sensitive to the singular values of concrete dam deformation, can effectively mine the local singular features of data, and has better singular value detection performance and detection efficiency.