Study On The Intelligent Algorithm For Prediction Of Compressive Strength Of Cemented Sand,Gravel And Rock

The Cemented Material Dam(CMD)is a new type of Dam that is based on the new idea of "suitable material structure ",which adjusts the structure of the dam to adapt the characteristics of the material.The type of dam can make full use of local material by adjusting the structure of the dam to fully adapt to the material characteristics.CMD can efficiently utilize the broken aggregate rock-fill materials,such as river bed gravel,foundation excavation material,slope waste material and the surrounding building material.Compared with gravity dam,CMD can greatly reduce the cost.However,the strength of the cemented material dam develops slowly.The compressive strength of the standard test block can not be measured until it is maintained for 180 days.Moreover,the materials in different places are different,and the mechanical properties of the raw materials are also different.The time consumption of the mixture ratio required by the design of the laboratory trial fitting engineering is too long,which seriously restricts the further popularization of the cemented material dam at home and abroad.Since there is no prediction model for the compressive strength of cemented sand gravel at present,this paper uses artificial neural network(ANN)to accurately predict the multi age compressive strength value of cemented sand gravel test block according to the material mix proportion and raw material performance,which can provide specific mix proportion for the project in advance.At the same time,it can reduce the damage to the overall structure of the dam caused by the core drilling method,which is a traditional dam type testing method,and avoid the potential safety hazard.The main research achievements are as follows:1.The single factor analysis method is adopted to analyze the influence of waterbinder ratio,unit cement dosage,unit fly ash content,unit water consumption,sand ratio and mud content on the compressive strength of consolidated sand and gravel,which provides theoretical support for the intelligent algorithm research on the prediction of the compressive strength of consolidated sand and gravel based on the compressive strength values of cemented sand and gravel at different ages of 28 d,70d,90 d,180d and 190 d in 342 groups.2.With unit values of 28d compressive strength of cement,unit water dosage of fly ash,sand ratio,dosage of water reducing agent,water-cement ratio,silt content,ages as influencing factors of the neural network input,we select cementation sand gravel compressive strength value as the output and establish cementation sand gravel compressive strength of the BP neural network prediction model,goodness of fit of the model is 0.91207,MSE is 2.7212,MAPE is 19.05%.3.In order to avoid BP neural network model falling into local optimum,I choose genetic algorithm to update the weight threshold of BP neural network prediction model,and then re-assigned after optimized screening.The goodness of fit of the optimized model is 0.94121,MSE is 1.8175,MAPE is 16.30%.4.In order to get a prediction model with higher accuracy,the emerging random forest algorithm model was selected for fitting.Unit cement dosage,28 d compressive strength value of cement,unit fly ash content,unit water consumption,sand rate,water reducing agent dosage,water-binder ratio,mud content and age were selected as input parameters of the random forest algorithm model,and the compressive strength of cemented sand and gravel was taken as output parameters.The goodness of fit after prediction is 0.98386,MSE is 0.3776,MAPE is 14.66%.5.The correlation coefficient of model prediction are gradually increasing.The compressive strength of cemented sand and gravel can be predicted more accurately.