Prediction of Ultimate Bearing Capacity of Single-Layer Spherical Reticulated Shell Based on TensorFlow

In order to study the ultimate bearing capacity of the single-layer spherical reticulated shell structure, based on the BP neural network algorithm of TensorFlow, a neural network model was established to predict the ultimate bearing capacity of the K8 single-layer spherical reticulated shell structure by considering the complex mapping relationship in the nonlinear analysis. Moreover, another new neural network model was established to predict the ultimate bearing capacity of the Kn-type single-layer spherical reticulated shell structure. The prediction results were compared with the calculation results of the finite element and literature regression formulas. The results showed that the error mean between the predicted ultimate bearing capacity of the K8 single-layer spherical reticulated shell structure and the finite element calculation results was 1.666%, and that between the formula calculation results and the finite element calculation results was 3.994%; the error mean between the predicted ultimate bearing capacity of the Kn-type single-layer spherical reticulated shell structure and the finite element calculation was 4.774%, and that between the formula calculation results and the finite element calculation results was 5.163%. The feasibility of using neural network to predict the ultimate bearing capacity of single-layer reticulated shell structure is demonstrated.