A Deep Learning Approach for Rubber Product Quality Prediction with BiLSTM-Attention

The quality prediction of production in rubber manufacturing is crucial to ensure product quality and improve production efficiency. However, the strong coupling of process parameters and quality indicators in the rubber production process and the characteristics of time series data collected by high-frequency sensors make it difficult for traditional models to capture nonlinear interactions and long-term dependence, resulting in insufficient prediction accuracy. To this end, this paper proposes a rubber product quality prediction model that integrates bidirectional long short-term memory network and self-attention mechanism, which captures the two-way timing dependence of process parameters through BiLSTM, dynamically focuses on key time steps through the self-attention mechanism, and combines data preprocessing, BiLSTM layer, self-attention mechanism and fully connected layer to generate prediction values, so as to improve the prediction accuracy. Experimental results show that the model is significantly better than the baseline model in multiple evaluation indexes on different production lines, and has good robustness and generalization ability.