Prediction of Strip Width in Finishing-Mill Group Based on PCA-PSO-LightGBM

Aiming to improve the strip free width hitting rate in finishing rolling, this paper proposes a Light Gradient Boosting Machine (LightGBM) prediction algorithm based on particle Swarm Optimization (PSO) with Principal Component Analysis (PCA). First, the raw data are effectively dimensionally reduced using PCA after preprocessing, and input into LightGBM to establish a width prediction model, followed by optimization of the key parameters of LightGBM algorithm by PSO. Finally, data design experiments and simulations are carried out on the actual production line of a steel plant in Shanghai, and the results show that the PCA-PSO-LightGBM model proposed in this paper has good accuracy and robustness, and meets the requirements of practical applications.