Prediction Of Cement Strength Based On Principal Components Wavelet Neural Network

Cement's 28 days compressive strength value is the main basis of cement label.About how to forecast this value,the traditional analysis method can not deal with highly redundant data between the multivariable and nonlinear which is produced in the production process,but the high-dimensional data inchudes a lot of important information of the raw data,making considerable difficulties when we analysis data.Therefore in the data of high dimensional data processing and analysis,we should reduce the dimension of data,but traditional methods did not provide effective means to extract the useful information, thus causing tremedous waste of resources.In recent years,some researchers apply artificially intelligent algorithm to cement strength prediction,obtaining some effect. At present,it is popular that directly employing neural network to predict the original data. Direct prediction model of neural network has some limitations due to the lack of the sample data pretreatment.This paper compress and optimize the raw data of the sample with the principal component analysis in the condition of trying not to lose important imformation,reducing the dimension of the sample raw data,using the extraction of principal component as new input variables ,combining wavelet neural network,and constructing the compressive strength of cement 28 days prediction model.This method not only can extract the original original data of the important information,cut the number of the input variables to simplify th netword strcture, but also optimize the network model performance.So this get a netword model whose structure is simpler ,convergence speed is faster,and prediction is more accurate.