The Design and Optimization of a Multi-Channel Fertilizer Spreading System Based on EDEM Simulation and the CNN-LSTM–Attention Algorithm

To achieve the objectives of a variable spreading width and uniform fertilizer distribution in different operating environments, such as orchards and greenhouses, this study proposes a multi-channel fertilizer spreading method and designs a multi-channel fertilizer spreading system. The system mainly consists of a fertilizer delivery device and multi-channel spreading fertilizer disc. Based on the fertilizer delivery requirements, the fertilizer delivery device was designed using kinematic theories and the EDEM software. To address the optimization problem of the fin angle (λ), EDEM simulation is combined with machine learning methods, the multi-channel fertilizer spreading system was simulated using EDEM, and a spreading model was established based on the CNN-LSTM–Attention algorithm using simulation data. The model was employed to optimize the fin angle (λ), with spreading uniformity and width as the evaluation metrics. The optimization yielded a fin angle of λ = 34.15°, a coefficient of variation (Cv) of transverse distribution uniformity of 9.29%, and a spreading width of 3.0 m. The optimization results were validated through a combination of EDEM simulation and field tests. The EDEM simulations indicated a Cv of transverse distribution uniformity of 7.66% and an effective spreading width of 2.95 m. Field test results show a spreading width of 2.97 m and an average Cv of transverse distribution uniformity of 10.21%. The model optimization results align well with both the EDEM simulation and field test outcomes, providing new insights into researching fertilizer spreading methods.