Machine learning-based prediction of mechanical properties and optimization of controlled cooling process for steel pipes

To address the challenges of insufficient precision in the empirical design of controlled cooling process parameters for steel pipes and the difficulty in balancing high strength and high toughness, this paper proposes an intelligent method integrating machine learning and Bayesian multi-objective optimization. Firstly, Gaussian noise is utilized to augment the collected experimental data of controlled cooling parameters, mitigating the risk of overfitting caused by a limited sample size. Principal component analysis (PCA) is then applied to reduce the dimensionality of linearly redundant features, ensuring a cumulative variance contribution rate of 95%. A mechanical property prediction model for steel pipes is constructed based on the XGBoost algorithm. This model exhibits high-precision predictive capabilities, with a coefficient of determination of 0.96 and a mean absolute percentage error of less than 3% for key mechanical properties. To verify the accuracy of the prediction model, J55 oil casing pipe samples were trial-produced and subjected to tensile testing. The maximum relative error between the experimental mechanical property data and the model's predicted values is 3.14%. Subsequently, multiple sets of high-performance controlled cooling process parameters combinations were successfully identified through the Bayesian multi-objective optimization model. The tensile strength of the optimal solution reaches up to 927 MPa, and the mechanical property indicators of the optimization results all comply with the API SPEC 5CT standard. To examine the practical effectiveness of the Bayesian optimization algorithm, J55 oil casings were manufactured according to the optimized process parameters, and their mechanical properties were tested. The maximum relative error between the experimental results and the model's predicted values is 1.84%. The research results indicate that the proposed algorithm can achieve accurate prediction of steel pipe performance and efficient process optimization, providing a feasible solution for the intelligent manufacturing of steel pipes.