Research On The Influence Of Concrete/Cement Mix-proportion On Its Compressive Strength By Using SVR

Nowadays, it is more and more pay attention how to improve building safety,therefore concrete/cement has been greatly concerned by people. One of the mostserious problems is how to enhance the compressive strength of concrete/cement. Thecompressive strength of concrete/cement has been extensively investigated by scholarsbecause of its importance to the safety of the buildings. Preparation technique ofconcrete/cement material has been widely used in industry. Although large-scaleindustrial production has been achieved, it also faces many shortcomings anddeficiencies in the use of concrete/cement materials, thus it needs to be further studiedand improved. In the case of the concrete/cement possessing sufficient compressivestrength, how to improve the economic efficiency efficiently also plays very important.Because the material mix-proportion has a very closed relationship with its compressivestrength of concrete/cement, it is quite important for us to accurately model andoptimize the mix-proportion for concrete/cement. If there is no a feasible method forguidance, one needs a lot of manpower, material resources for a large number ofexperiments to find the suitable material mix-proportion. Therefore, how to design areasonable experiment, analysis of experimental data effectively, and reduce experimentcosts, becomes an unavoidable scientific problem.In this thesis, the support vector regression (SVR) theory combined with particleswarm optimization (PSO) was proposed to model the influence of the materialsmix-proportion on its compressive strength of concrete/cement. The main contents areas following:(1) According to an experimental dataset, the SVR was employed to regress andpredict the compressive strength of ordinary concrete via3factors including water,coarse aggregate, and cement. The result revealed that the modeling capacity of SVR issuperior to that of the spatial autocorrelation model. The optimal mix-proportion wasobtained based on the established SVR model, and the interactive influence of thefactors on the compressive strength of ordinary concrete was also found bygrid-screening.(2) The SVR was utilized to model and predict the compressive strength of carbonfiber reinforced concrete via7parameters (cement, silica fume, water, water reducer,sand, gravel and carbon fiber). The result demonstrated that SVR can efficiently dealwith the regression problems with small samples, and the modeling/predicting performance is also better than that of conventional multivariate nonlinear regression(MNR) approach. The interactive influence of the above7factors on the compressivestrength of carbon fiber reinforced concrete was also depicted by the constructed SVRmodel.(3) The SVR was also used to model and predict the compressive strength of flyashconcrete via5parameters, i.e., water to binder ratio (W/B), water content (W), fineaggregate ratio (s/a), fly ash replacement ratio (FA), air-entraining agent content (AE).The result is that the predicted results conducted by SVR model are much better thanthose of genetic algorithm reported in the literature. Based on the established SVRmodel, the optimal mix-proportion was found theoretically, and the theoreticalmaximum compressive strength (112.69MPa) of the flyash concrete was50.25%higherthan that (75MPa) obtained in the experiments.(4) Finally, the SVR was used to model/predict the compressive strength ofmagnesium oxychloride cement based on7factors (fly ash, MgO, water, MgCl2,sawdust, Fe2(SO4)3and H2SO4), the modeling/predicting results were compared withthose of ANN. It is found that both of the training and predicting abilities outperformthose of ANN. Compared with the experimental maximum index, the theoreticalmaximum compressive strength of the magnesium oxychloride cement was increasedby4.86%. In addition, the interactive influence among the above7factors on thecompressive strength of magnesium oxychloride cement was also illustrated by theconstructed SVR model.(5) In order to quantitatively illustrate the effect of various factors on thecompressive strength, sensitivity indices of the compressive strength of concrete/cementaffected by the changes of various factors via the established SVR model.The SVR can be efficiently employed to model/predict the compressive strength ofconcrete/cement. The quantitative analysis results about the influence of complex&multi variable factors on the compressive strength of concrete/cement can providescientific guidance for the construction project. The theoretical model is helpful to seekoptimal mix-proportion, effectively improve the compressive strength, thereby savetime and cost, reduce blindness in engineering operation. It has important practicalsignificance in improving the seismic performance of building facilities, reducingproject risk, guaranteeing people’s safety of life and properties.