Association Rules Mining And Prediction On Temperature Of Dam During Construction Period

Temperature control measures of mass concrete in the construction period have always been important concerns in the field of hydraulic engineering.With water conservancy information being widely used in engineer practice,immense quantities of data set have been stored.Observational data contain valuable knowledge and can be used to analysis the temperature control of the construction period.Based on the temperature monitoring data of the construction period,this paper establishes construction database and analysis temperature changing regularities,mining association rules with related factors,predicts temperature fluctuation procedure by adopting data-mining algorithm aims at providing decision support for temperature control during the construction.The main research results of this paper are as follows:1.Observational data in real life is not able to comply with the requirements of data mining.It is inconsistent,noisy,they have missing values,unsuitable format,etc.Data pre-processing aims at improving data quality,the accuracy and the efficiency of data mining tasks.Temperature control during construction period database is established,it is facility for data mining tasks to query and extraction data sets.2.Based on the time series clustering,the K-means algorithm is used to divide the concrete temperature time history curve of thermometers into four types of different changing rules,and summarize the temperature variation law of each type.The observed value of the class ? curve fluctuates greatly,which may be located near the surface of the dam.There is the strong possibility that the first three clusters are located in the dam body.Temperature duration curves of the first three clusters are smoothing,however their temperature variation processes are different,which can reflect different temperature control effect.According to the law of temperature change,judge the location of the thermometer and the temperature control effect near the thermometer.3.Based on the association rule mining algorithm,the relationship between the temperature curve of concrete and pouring information,water cooling information and concrete type is analyzed.The results show that the roller compacted concrete has a greater probability of secondary temperature rise than the normal concrete,and the normal concrete has a greater probability of steady change in the later stage.However,it is possible for the normal concrete to have a secondary temperature rise when the water cooling measures are inadequate or the concrete is poured in summer.And there is a lower possibility for the roller compacted concrete to have a secondary temperature rise when the water cooling temperature is low.4.The temperature time course prediction model is established based on Support Vector Regression and Principal Component Analysis.Through the comparison of the prediction results of several typical measuring points and the observational temperature time history curve,all the residuals fluctuate are less than 2?.Compared with the whole model,classification model predicts higher accuracy.