A FEM-ML Hybrid Framework for Optimizing the Cooling Schedules of Roll-Bonded Clad Plates

In the production of clad rolled plates from asymmetric sandwich-type slab for pipeline applications, achieving both target mechanical properties and high geometric flatness remains a critical challenge due to differential thermal stresses between the dissimilar steel layers during accelerated cooling. This study aims to develop an optimal cooling schedule for a 25 mm thick clad plate, comprising a X70-grade steel base layer and an AISI 316L cladding, to ensure required strength and minimal bending. A comprehensive approach was employed, integrating a 3D finite element model (Ansys) for simulating thermoelastic stresses with a CatBoost machine learning model trained on industrial data to predict heat transfer coefficients accurately. A parametric analysis of cooling strategies was conducted. Results showed that a standard cooling strategy caused unacceptable bending of plate after cooling exceeding 130 mm. An optimized strategy featuring delayed activation of the lower cooling headers (on the cladding side) created a compensating thermoelastic moment, successfully reducing bending to approximately 20 mm while maintaining the base layer’s requisite mechanical properties. The findings validate the efficacy of the combined FEM-machine learning methodology and propose a viable, industrially implementable cooling strategy for high-quality clad plate production.