Prediction model for mechanism sand process parameters of vertical shaft impact crusher based on fuzzy-self-attention hybrid neural network

This study proposed a predictive model for process parameters of Vertical Shaft Impact Crusher for manufactured sand. It employs a fuzzy-self-attention hybrid neural network model to determine the optimal operational state of the crusher, effectively enhancing the accuracy of predicting the Fineness Modulus and particle size distribution of the manufactured sand. Extensive training of the model is conducted using a large dataset collected from three mining operation sites. The model is compared to two disintegration modules and a BP neural network. Simulation results reveal that this hybrid neural network model achieves convergence in fewer than 20 epochs during the training process. Its predicted particle size distribution curve closely matches actual sample data, and the Coefficient of Determination 𝑅ଶ is closest to 1, indicating superior predictive performance. This model exhibits the best predictive ability and can effectively forecast both the Fineness Modulus and particle size distribution of the sand.