Predicting Compressive Strength Of Cellulose Nanofibers Reinforced Cement Based Composites Using Machine Learning Models

Cellulose nanofibers(CNFs)are the newly introduced plant-based materials in the construction industry to ensure sustainable development.The application of artificial intelligence(AI)approaches particularly machine learning(ML)models has assisted to economized the construction sector.The way to obtain the compressive strength of CNFs is by physical experiments,which are expensive and time consuming to obtain results,so the working efficiency will be very low.That is why,technological advances allow solving engineering problems at a lower cost by other methods,such as empirical regression,numerical simulation and the use of machine learning methods.These methods allow predicting the compressive strength of CNFs with the proportion of the designed mixture of different components(cement,water,CNFs,superplasticizer,fine aggregate,coarse aggregate,and age).This research aims to determine the compressive strength of cellulose nanofibers reinforced cement based composites,concrete,cement paste,and cement mortar by using supervised regression machine learning techniques for analysis before adopting to utilize.To achieve this task,an experimental-based cement based composites dataset containing 695 data points was prepared and split into two categories(Training dataset=70%,Testing dataset=30%)for the evolution of ML models.This cement based composites dataset was further divided into 196,266,and233 observations of concrete,cement paste,and cement mortar datasets respectively.These four datasets of cement based composites,concrete,cement paste,and cement mortar were analyzed individually.There were seven independent variables:cement(kg/m~3),water(kg/m~3),CNFs(kg/m~3),superplasticizer(kg/m~3),fine aggregate(kg/m~3),coarse aggregate(kg/m~3),and age(Day)variables as an input and one dependent variable:compressive strength fc of CNFs reinforced concrete(MPa)as an output for cement based composites dataset.In the same way,there were seven independent variables for concrete dataset,five independent variables for cement paste dataset,and six independent variables for cement mortar dataset as an input variables with one dependent variable as an output variable.Firstly,the dataset analysis was executed in the Jupyter Notebook software by conducting the correlation coefficient analysis,descriptive analysis,outlier analysis,multivariate analysis,and density curve to investigate the properties and strength of designed datasets for cement based composites,concrete,cement paste,and cement mortar.The findings of designed datasets analysis indicated the less correlation between variables,very low number of outliers,and extremely less skewness which ensure the satisfaction and acceptability of datasets for the development of models.Secondly,the parameters of machine learning models:Random Forest(RF),Linear Regression(LR),Support Vector Regressor(SVR),Gradient Boosting Regressor(GBR),Ada Boosting Regressor(ABR),K-Neighbor Regressor(KNN),Bagging Regressor(BR),XG Boost Regressor(XGBR),Decision Tree(DT),and Pruned Decision Tree(PDT)were developed and implemented.Each of the models was analyzed for cement based composites,concrete,cement paste,and cement mortar datasets to compare the performance capability of the developed models.The following metrics were employed to gauge the ability and performance capability of the models:coefficient of determination(R~2),mean absolute percentage error(MAPE),mean absolute error(MAE),mean square error(MSE),and root mean square error(RMSE).The findings of metrics specified that seven out of ten models(RF,BR,XGBR,DT,GBR,ABR,and KNN)to predict the compressive strength of CNFs cement based composites had a firm capability(R~2>0.72,MAPE≤0.1,and MAE≤5)confirming to the standard of R~2 value greater than 0.60 and metrics values very less.Similarly,eight models(RF,BR,XGBR,DT,GBR,ABR,KNN,and SVR)for concrete,seven models(RF,BR,XGBR,DT,GBR,ABR,and KNN)for cement paste,and six models(RF,BR,XGBR,GBR,ABR,and KNN)for cement mortar had a strong ability to predict the compressive strength.The outcomes of cement based composites dataset showed that the RF(R~2=0.81,MSE=34.20,RMSE=5.84,MAPE=0.09,and MAE=4.10),BR(R~2=0.80,MSE=49.44,RMSE=10.09,MAPE=0.11,and MAE=4.80),and DT(R~2=0.79,MSE=37.96,RMSE=10.09,MAPE=0.10,and MAE=4.18)models presented the highest performance accuracy.The results of concrete dataset presented that the XGBR(R~2=0.93,MSE=9.77,RMSE=3.12,MAPE=0.051,and MAE=2.14),GBR(R~2=0.92,MSE=11.17,RMSE=3.34,MAPE=0.066,and MAE=2.18)and RF(R~2=0.913,MSE=12.44,RMSE=3.52,MAPE=0.078,and MAE=2.38)models presented the highest performance accuracy.The findings of cement paste dataset displayed that the XGBR(R~2=0.82,MSE=36.85,RMSE=6.07,MAPE=0.09,and MAE=4.10),DT(R~2=0.819,MSE=36.84,RMSE=6.05,MAPE=0.085,and MAE=2.17)and GBR(R~2=0.80,MSE=40.08,RMSE=6.33,MAPE=0.098,and MAE=4.63)models presented the highest performance accuracy.The results of cement mortar dataset showed that the KNN(R~2=0.73,MSE=19.87,RMSE=4.46,MAPE=0.095,and MAE=3.63),BR(R~2=0.73,MSE=19.87,RMSE=4.46,MAPE=0.095,and MAE=3.63)and RF(R~2=0.70,MSE=21.88,RMSE=4.47,MAPE=0.094,and MAE=3.62)models presented the highest performance accuracy.According to the sensitivity analysis of RF,DT,XGBR,GBR,and ABR models,water and cement were the factors with the biggest effects on the prediction of CNFs reinforced cement based composites,while the smallest effecting variable was coarse aggregate.There was a significant participation of coarse aggregate and CNFs independent parameters to output variable in concrete dataset.Also,cement and water in cement paste dataset;and water and age in the cement mortar dataset has placed noteworthy contribution than other independent variables.Finally,It can be concluded that the three RF,BR,and DT models for cement based composites,three XGBR,GBR,and RF models for concrete,three XGBR,DT,and GBR models for cement paste,and three KNN,BR,and RF models for cement mortar were the premier predicting models.Hence,RF was the one best predicting model of this research.