PSO-DFNN: A particle swarm optimization enabled deep fuzzy neural network for predicting the pellet strength

In addressing the complexity, limited information, and dynamic spatiotemporal characteristics encountered in predicting pellet strength with traditional methods, this study proposes a novel prediction model for the strength of fusible pellets, developed on a Particle Swarm Optimization Deep Fuzzy Neural Network (PSO-DFNN). Initially, the model is constructed by observing and extracting fractal features of the microstructure of pellet ore. Subsequently, the fuzzy system is utilized to partition the spatiotemporal data and generate multi-layer fuzzy rules, thus constructing a deep fuzzy neural network. Lastly, the Particle Swarm Optimization algorithm is employed to optimize the fuzzy membership rule weights, achieving precise prediction of pellet strength. The results indicate a Mean Absolute Error (MAE) of 3.7218 and a Symmetric Mean Absolute Percentage Error (SMAPE) of 3.72 % when predicting pellet strength during the pellet roasting drying stage. The PSO-DFNN model exhibits high prediction accuracy, meeting the needs for pellet strength prediction and providing a more reliable basis for decision-making in the production process.