Abstract: With the growing demand for the development of high-performance shipbuilding steel, the accumulation of industrial data and the rapid development of information and communication technologies, data-driven methods provide a new technical approach for the transformation of composition-process design of high-strength ship plate steel from empirical trial and error to efficient collaborative optimization. Aiming at the problems of long trial production cycle, difficulty in collaboratively meeting multiple performance indicators and insufficient consideration of manufacturing adaptability in traditional design, a data-driven optimization design method for composition-process of high-strength ship plate steel under multi-performance constraints based on industrial data is proposed in this paper. Based on the industrial production data of low-alloy steel from the medium and heavy plate production line of a large domestic iron and steel enterprise, 5 617 valid samples were obtained after data cleaning. An input system containing 24 core decision variables was constructed, and XGBoost algorithm was adopted to establish prediction models for yield strength, tensile strength, elongation after fracture and impact energy respectively. The coefficient of determination R² values of the test sets of the four performance models are 0.925, 0.949, 0.798 and 0.855, with the corresponding mean absolute errors (E_MA) of 12.147 MPa, 8.115 MPa, 1.379% and 15.065 J, respectively, which indicates that the models can well characterize the nonlinear mapping relationship among composition, process and performance. On this basis, taking the yield strength of 420 MPa, tensile strength of 520 MPa, elongation after fracture of 30% and impact energy of 120 J as the optimization targets, the performance errors and cold crack susceptibility index P_cm were incorporated into the NSGA-Ⅲ multi-objective optimization framework, and 495 groups of Pareto non-dominated solutions were acquired. A total of 68 groups of quasi-compliant candidate schemes were screened out and representative schemes including S418 were determined via the strategy of quasi-compliance screening and manufacturability priority ranking. For the S418 scheme, the predicted performance values are 420.48 MPa (yield strength), 520.67 MPa (tensile strength), 29.75% (elongation after fracture) and 182.36 J (impact energy), with a P_cm value of 0.137. The verification test with 10 kg-grade test steel showed that after hot rolling the 35 mm rolled piece to 7 mm, the measured yield strength, tensile strength, elongation after fracture and impact energy were 435 MPa, 516 MPa, 30.3% and 149.1 J, respectively. The results show that the proposed method can effectively narrow the composition-process design space of high-strength ship plate steel, and provides technical support for the digital development and multi-performance collaborative optimization of low-alloy high-strength ship plate steel.