strip crown prediction using mrf-based feature selection and hmm-based decision level fusion

Since the stability of the crown is an essential factor in determining strip crown quality, crown prediction can be a significant measure in hot-rolling mills. in this paper, a strip crown prediction model is proposed based on machine learning (ml) algorithms. for this purpose, influence factors of strip crown are identified then the most significant factors are selected and integrated based on proposed decision-level fusion-based feature selection using markov random field (mrf). ultimately, strip crown prediction is performed using the hidden markov model (hmm) whose parameters are optimized based on the leave-one-out (loo) cross-validation algorithm and training database. the proposed algorithm is evaluated using various error indicators, such as mean absolute percentage error (mape), root mean square error (rmse), and mean absolute error (mae) to verify its effectiveness. extensive experiments demonstrate that the proposed strip crown prediction performance is effective with determination coefficients greater than 0.97%. research results indicate the potential of the proposed method for strip crown prediction and the reliability of ml models in complex industrial problems.