Logic-guided neural network for predicting steel-concrete interfacial behaviors

The interfacial behaviors play significant roles in various composite materials and structures. This paper presents a logic-guided neural network to seamlessly integrate data-driven methods scientific knowledge predicting properties of steel-concrete composites. investigated include the bond strength, interface slip, bond-slip relationship. Three are proposed conform logic principles: (1) data generated supplement experimental data; (2) loss function is presented guide learning process; (3) unstructured incomplete utilized enlarge dataset. performance method compared with four representative machine methods, which artificial network, tree boosting, random forest, epsilon-support vector regression. results indicated that achieved highest accuracy. coefficients determination stress compressive strength higher than 0.95, predicted prior knowledge. useful for prediction