Determining the mechanical behavior of thin-walled cylindrical shells of cement composite containing graphene oxide and glass fibers with the help of fuzzy logic model and artificial neural network

Considering the great application and importance of thin-walled cylindrical shells (TWCS) in civil engineering, it is necessary to provide a suitable model as an important criterion in shell structures’ design. With the entry of new materials such as graphene oxide (GO) into the construction market, there is a need to evaluate the behavior and role of such materials in shell structures. In this research, an attempt has been made to evaluate the ability of the fuzzy logic (FL) model and artificial neural network (ANN) regarding the mechanical behavior of cement composite TWCS containing GO and reinforced with glass fibers (GF). In this regard, 36 small models of thin-walled shells containing 0%, 0.03% and 0.05% GO (replacement of cement weight) and 0%, 0.5%, 1%, 1.5%, 2% and 2.5% GF, with equal water to cement ratio were considered in all designs. After 28 days, they were tested in ideal laboratory conditions. Also, cubic and prismatic samples were made for assessment of the compressive and bending strengths of TWCS. And at the end, the results of the FL model and the results of the ANN model were measured and compared with the laboratory results of TWCS specimens. The results showed that the use of GO increased the compressive and bending strengths by 35% and 42%, respectively, compared to the control samples. Also, the compressive strength results indicate an increase in concrete samples with 0.05% GO compared to the samples containing 0.03% GO. The results of the compressive strength tests show the existence of a threshold limit in the consumption of GF. So that by increasing the amount of GF from these limits, the amount of increase in resistance decreases. In the TWCS, with the addition of GO, there was an increase of 20% and 12% in applied pressure (annular tensile strength) and strain, respectively. Also, with the addition of fibers, the amount of strain increased up to 7.9 times and the tensile strength of the ring or the amount of pressure on the shell increased by 59% compared to the control sample. On the other hand, by modeling and comparing the FL network and the ANN, it was observed that the ANN had a lower error of 4.24%. Therefore, this issue indicates that this model of the network will be able to predict and estimate the resistance of tensile stress and pressure on the TWCS specimens after conducting the training.