Session: 06-17-05 AI Technology for Ocean Engineering - V

Submission Number: 182086

Gradient-Based Topology and Shape Optimization of Offshore Jackets Using Differentiable Finite Element Analysis 

The structural optimization of deepwater jacket platforms faces growing challenges as water depth increases and geometric complexity grows exponentially.  Topology optimization of jacket platforms is one of the most challenging tasks. Previous studies have largely relied on spatial mesh representation and evolutionary or swarm intelligence methods, which are flexible but often computationally expensive, provide approximate sensitivities, and may require large numbers of structural evaluations to converge. To address these limitations, this study develops a differentiable finite element framework for the topology optimization of deepwater jackets, leveraging automatic differentiation (AD) in JAX. The structural topology is represented by a relation matrix of joints, defining the connectivity of tubular members and enabling efficient reconfiguration of jacket layouts during optimization. The finite element procedures are formulated as differentiable operations, allowing exact gradient computation of structural performance metrics such as compliance, stiffness, and dynamic response. Compared with heuristic optimization methods, the proposed approach provides higher efficiency, exact sensitivities, and implementation simplicity, enabling rapid convergence even for large-scale deepwater jacket models. Numerical experiments on representative jackets demonstrate that the framework preserves physical fidelity while achieving superior computational performance. The results highlight the potential of automatic differentiation and relation-matrix-based topology representation as a robust foundation for intelligent and scalable design optimization of offshore structures.

Presenting Author: Haiming Zhu Tianjin University

Presenting Author Biography: Haiming Zhu is a postdoc researcher at Tianjin Univeristy. His research interests include fluid-structure interation problems and differentiable fluid & structure dynamics in ocean engineering.

Authors:

Haiming Zhu Tianjin University
Zunfeng Du Tianjin University