Ceramics-Silikáty

Graphene nanoplatelets (GrNs) are a significant enhancement for boosting the structural integrity and electrical conductivity of cement-based composites. Because of their unusually conductive properties, they can support the design of new composite materials and help improve the efficiency of non-destructive methods for monitoring structural integrity. About traditional tree-based models, however, they cannot accurately represent or describe all the complex nanoscale interactions that occur in these composites. This study utilized ML approaches, namely RFR and GBDT, to provide precise predicting methods for the compressive strength (f_c) of cementitious composites enhanced with graphene nanoplatelets. A thorough compilation of data from accessible sources was employed to develop these approaches. The dependability of the system is profoundly influenced by the choice of hyperparameters for RFR and GBDT, necessitating the use of metaheuristic optimization techniques for selection. A technique used for this purpose is the Artificial Protozoa (AP). A dataset including 172 points of data was gathered from the available publications to develop the model. The technique was trained on 75% (129) of the database, and it was assessed on 25% (43) of the database with new data to determine the model's performance. The data suggests that both RFR(AP) and GBR(AP) are capable of properly predicting f_c. The RMSE values for the GBR(AP) learning and testing steps were 2.5522 and 2.8178, respectively, with variance percentages of -15.94% and -13.6%, in contrast to the RFR (AP) technique. The RFR(AP) produced the lowest RMSE results for learning and assessment, at 2.1454 and 2.4346, correspondingly. Based on the data, the RFR(AP) design maintained high reliability while achieving the lowest values.