Optimization method for jacket platforms using random forest surrogate model

The optimization of jacket platforms typically relies on computationally intensive finite element analysis (FEA), which is relatively time-consuming. Surrogate models are widely used in multi-domain optimization problems, while machine learning algorithms are employed to construct surrogate models, demonstrating promising applications. In this study, an optimization method that integrates machine learning surrogate models with an improved genetic algorithm (GA) is employed to enhance structural optimization efficiency. Then, the proposed method is applied to a typical jacket platform in the Bohai Sea using a seven-dimensional optimization space designed to minimize the total structural weight. A surrogate model is derived through a random forest algorithm to replace FEA. The surrogate model is combined with a real-coded GA featuring adaptive crossover and mutation for iterative optimization. Moreover, surrogate models are developed using Decision Tree and KNN methods. Evaluation results show that the random forest algorithm outperforms Decision Tree and KNN methods. The random forest surrogate model can save significant computation time while maintaining high accuracy, providing an efficient and effective alternative to the FEA optimization method. • Presents an optimization method using surrogate models to replace FEA. • Builds surrogate models using machine learning algorithms and evaluates. • Optimizes a Bohai Sea jacket platform by an improved genetic algorithm(GA). • Improved GA maintains population diversity, balancing exploration/exploitation. • Random forest model is more efficient and optimization results is very close to FEA.