Synthesis of geopolymer from fly ash and optimization of process parameters by neural network

Geopolymerization is an efficient approach to the utilization of the circulating fluidized bed boiler (CFBB) fly ash.The mechanical performances of synthesized geopolymers are largely dominated by the CFBB fly ash and process parameters,and have further impacts on the application of derived geopolymers used as building materials.In this study,the fly ash,from the CFBB of the Wangping power plant in Shuozhou,Shanxi Province,China was sampled and geopolymerized.The geopolymerization mechanism was studied with the aids of a high-resolution scanning electron microscope,Fourier transform infrared spectroscopy as well as solid-state nuclear magnetic resonance.Besides,the effects of primary process parameters on the mechanical properties of derived geopolymers were systematically investigated and optimized by the model of back-propagation artificial neural network (BPANN) to obtain geopolymer with excellent mechanical properties.The results showed that the amorphous silicon or/and aluminum compounds in the fly ash were activated and formed silicon or/and aluminum monomers in the process of geopolymerization.Then these monomers gradually became geopolymer gels by hy-drolysis,condensation,and gelation.Afterwards,these gels were combined,and the larger gels formed geopoly-mers through dehydration and solidification,and Si—O—Al bonds were the dominating chemical bonding patterns in the generated geopolymer.The mechanical properties of the geopolymer were improved with the piling up of the amount of gel resulted from the increase of the modulus and Na2O content of the activator,and a decrease in the ratio of liquid to solid and curing temperature.However,excessive curing and stirring time were detrimental to the mechanical properties of geopolymers.Based on the experimental data of the geopolymer cured at room tem-perature for 7 days,a BPANN model with the structure of 6-8-1 was established to predict the mechanical properties of the geopolymers.The training error and test error of the established model were 0.98% and 3.85% respectively,indicating that the model had good training accuracy and generalization ability in predicting the mechanical properties of the geopolymers.As a result,the optimized process parameters for synthesizing geopolymer with sufficient mechanical properties were obtained,where activator modulus was 1.6,the liquid-solid ratio was 0.8,Na2O content was 0.09,curing temperature was 20 ℃,curing time was 24 h,and stirring time was 20 min.