Influence of Rotor Cage Structural Parameters on the Classification Performance of a Straw Micro-Crusher Classifying Device: CFD and Machine Learning Approach

Abstract

The rotor cage is a key component of the classifying device, and its structural parameters directly affect classification performance. To improve the classification performance of the straw micro-crusher classifying device, this paper proposes a CFD-ML-GA (Computational Fluid Dynamics-Machine Learning-Genetic Algorithm) method to quantitatively analyze the coupled effects of rotor cage structural parameters on classification performance. Firstly, CFD and orthogonal experimental methods are used to qualitatively investigate the effects of the number of blades, length of rotor blades, and blade installation angle on the classification performance. The conclusion obtained is that the blade installation angle exerts the greatest effect on classification performance, while the number of blades has the least effect. Subsequently, four machine learning algorithms are used to build a cut size prediction model, and, after comparison, the Random Forest Regression (RFR) model is selected. Finally, RFR is integrated with a Genetic Algorithm (GA) for quantitative parameter optimization. The quantitative analysis results of GA indicate that with 29 blades, a blade length of 232.8 mm, and a blade installation angle of 36.8°, the cut size decreases to 47.6 μm and the classifying sharpness index improves to 0.62. Compared with the optimal solution from the orthogonal experiment, the GA solution reduces the cut size by 9.33% and improves the classifying sharpness index by 9.68%. This validates the feasibility of the proposed method.