Page - 1392 - in Book of Full Papers - Symposium Hydro Engineering

Province of China, is recognized based on data mining. The raw measured temperature data always contains missing values and outliers. Processing of the missing values and outliers is required before analysis of the data. The density- based spatial clustering of applications with noise method is employed to detect and delete the outliers and the the k-nearest neighbor algorithm is used to estimate and substitute the missing data. The K-means clustering with dynamic time warping algorithm is applied to classified the temperature time series and to find their features. Four types of temperature series are recognized. Fig. 1 shows the four types of temperature history curves determined by the K-means clustering algorithm. There is a significant difference in the temperature variation between the different types. Type I exhibits a rapid increase due to hydration heat and a decrease as cooling water flows within 20 days after concrete pouring, followed by a gradually increase in the next 3 months. Comparing with Type I, Type II has a stable curve after its drop due to water-cooling. Type III is a approximately horizontal curve, and the hydration heat induced increase and the water-cooling induced drop of temperature are possibly missing. Fluctuation is observed in the later stage of the temperature curve of Type IV, while it is smooth in the other three types, implying that temperature of Type IV is measured near the surface of the dam where the environmental temperature has significant influence and the other three types monitored inside the mass concrete. This study presents a preliminary analysis of features of monitoring temperature of the RCC dam during construction. Further study based on the big data mining and numerical simulation is needed to understand the cause of the difference between the features and the effect of different features on the concrete dam behavior. (a) Type I (b) Type II (c) Type III (d) Type IV Fig. 1 Types of the monitoring temperatures time series 1392