Parameter optimization of the classifying device of straw micro-crusher based on neural network and CFD

To improve the classification performance of the straw micro-crusher's classifying device, the characteristic parameters impacting classification performance are identified through an analysis of the forces acting on the straw particles between the blades of the rotor cage and the motion characteristics of the fluid. Taguchi's experimental design is used to ascertain the combinations of computational fluid dynamics (CFD) simulation parameters for cut size and classifying sharpness index, while a neural network (NN) model is developed using CFD data to predict the optimal feature parameter combination for classification performance. The results reveal that the rotor cage speed holds the most significant impact on both the cut size and classifying sharpness index. The optimal combination of feature parameters recommended by the neural network is v=8m/s, n=1200r/min, z=36, θ=40°, at which the cut size is 28.3 μm, a reduction of 9.0% compared to the Taguchi experiment. In addition, the relative error between the cut size predicted by the neural network model and the CFD simulation is less than 4.5%, indicating the reliability of the neural network model in predicting the optimal parameter combination of the classifying device.