水力发电学报2025，Vol.44Issue(8)：105-118,14.DOI:10.11660/slfdxb.20250810
基于优化统计模型的混凝土坝变形异常值自适应识别
Adaptive identification of concrete dam deformation outliers based on optimized statistical model
摘要
Abstract
The development of a safety monitoring model utilizing dam deformation monitoring data is a crucial method for the quantitative analysis of deformation patterns,but previous deformation monitoring models often suffer from severe shortcomings in the selection of influencing factors and resisting outlier interference.This study develops an adaptive identification method for deformation outliers in concrete dams based on optimized statistical models.This method not only identifies outliers during regression modeling,but also prevents the monitoring model from distortion caused by erroneous data cleansing.First,we use a Bayesian model selection technique to reduce redundancy among the deformation's influencing factors,helping identify significant explanatory variables in the statistical modeling phase.Then,we use the least trimmed squares estimation for robust regression analysis of the deformation monitoring data,and construct a monitoring model that adaptively identifies various types of anomalies in the deformation data series.Finally,we design and implement a visualization strategy for different types of anomalies to generate an intuitive representation of their locations and potential impacts.A case study demonstrates that this new method identifies key deformation factors effectively,and adaptively reduces the interference of different anomaly types to regression analysis.It leads to improvement on the significance of regression results and the goodness of fit and prediction accuracy,manifesting satisfactory applicability for detecting anomalies in monitoring data and conducting quantitative analyses of dam safety behaviors.关键词
安全监测/混凝土坝/统计模型/异常值识别/最小截平方和估计/贝叶斯模型选择Key words
safety monitoring/concrete dams/statistical models/outlier detection/least trimmed squares/Bayesian model selection分类
建筑与水利引用本文复制引用
肖晟,杨杰,程琳,马春辉,徐笑颜..基于优化统计模型的混凝土坝变形异常值自适应识别[J].水力发电学报,2025,44(8):105-118,14.基金项目
国家自然科学基金面上项目(52279140) （52279140）
国家自然科学基金青年科学基金项目(52409173) （52409173）
西安理工大学优秀博士学位论文创新基金项目(104-252072408) （104-252072408）
