Prediction and Optimization of Flaw Detection Qualification Rate for 12Cr2Mo1V Large Cylindrical Forgings Using Random Forest and Particle Swarm Optimization

Large cylindrical forgings are extensively applied in petrochemical, nuclear, and other key industrial fields. Owing to their large dimensions and strict performance requirements, the manufacturing process is complicated with high production cost. Consequently, the accurate prediction of flaw detection (FD) performance and the optimization of manufacturing process are of great significance. In practice, the FD performance of large cylindrical forgings is affected by multiple processing parameters, including pouring temperature, pouring speed, preforging heating temperature, preforging heating time, forging temperature, and total forging ratio. In this paper, machine learning‐based (ML) data analysis was carried out to predict the flaw detection performance. To address the class imbalance issue, the ADASYN oversampling technique was strictly applied only to training subsets after stratified train‐test split and within each cross‐validation fold, to eliminate the risk of data leakage and ensure unbiased model evaluation. And Extreme Random Trees (ET), Support Vector Machines (SVMs), eXtreme Gradient Boosting (XGB), Logistic Regression (LR), and Random Forest (RF) algorithm were adopted to predict flaw detection qualification rate; results show that RF achieved the best prediction accuracy with an accuracy of 0.873 ± 0.044, an F1 score of 0.932 ± 0.025, a precision of 0.897 ± 0.034, and a recall of 0.971 ± 0.039 (mean ± standard deviation (SD) over 5‐fold cross‐validation). Aiming at improving the qualification rate, which is one of the critical objectives for industrial production, the particle swarm optimization (PSO) algorithm was employed to optimize the manufacturing process parameters. Experimental verification under the optimized process conditions indicates that the proposed machine learning method can effectively increase the qualification rate of large cylindrical forgings and realize the saving of production resources.