Abstract
Real-time monitoring of wave-induced loads supports ship masters’ scientific navigation decisions, where vertical bending moment is a core index representing hull longitudinal bending under waves. This paper combines a temporal convolutional network with the traditional section method to build a vertical bending moment identification model embedded with a section parameter correction mechanism. The main work includes the following: multiple wave condition strain–load datasets are generated via numerical simulations to train the model; the section method calibrates model parameters to boost prediction precision; and wave load tests are conducted to verify the model’s practicability. Results indicate the section method offers physical constraints that embed ship sectional features into the model, lifting identification accuracy, robustness and result rationality. This work applies a temporal convolutional network to the hull vertical bending moment identification with section parameter correction, offering technical references for hull structural safety evaluation and intelligent maritime decision-making.