Adaptive multivariate data fusion-based instability-state prediction framework for continuous casting process

In modern steel production, continuous casting is a crucial process that is evolving toward greater precision and integration. However, because of the large number of interrelated control variables in the continuous casting process, instability frequently occurs, adversely affecting product quality and production safety. Therefore, the development of effective methods for predicting instability states during continuous casting is crucial. Existing methods typically treat multivariate data as equally important and neglect the varying importance of different variables, thereby affecting the prediction performance of continuous casting instability states. To address this issue, this paper proposes an adaptive multivariate data fusion-based instability-state prediction framework for the continuous casting process. Specifically, a channel attention mechanism was employed to adaptively assign fusion weights to different variables according to their relevance to the process stability and various instability states. A weighted fusion was performed based on these adaptive weights. Subsequently, a one-dimensional convolutional neural network was utilized to extract features from the fused data, which were then used for instability-state prediction. Finally, two datasets were constructed using actual data from a multimode continuous casting and rolling production line, which were used to conduct comparative experiments. An interpretability analysis was performed on the proposed method. The experimental results demonstrated that the proposed method achieved prediction accuracies of 99.62% and 96.72% in the two cases, validating the effectiveness and superiority of the proposed framework.