Modeling and control methods of a multi-parameter system for threshing and cleaning in grain combine harvesters

The operating environment of grain combine harvesters is complex, and operators' experience is highly required. To meet the demand for high-quality and efficient harvesting in modern agriculture, harvesters must have the ability to control operational performance automatically. The core issue is to achieve multi-parameter cooperative control of the threshing and cleaning devices. The threshing and cleaning devices are located in the same rack space. The interaction mechanism between the internal agricultural material movement, mechanical structure, and airflow field is very complex, and it is a system with characteristics such as time-varying, hysteresis, and multi-parameter coupling. In response to the above challenges, real-time collection and adjustment methods for the main indicators and parameters of the threshing and cleaning devices were studied separately. A field experiment plan for identifying the threshing and cleaning system was designed. Based on the state space model and NARMAX (nonlinear auto-regressive moving average with exogenous input) model as the system structures, a fusion method of PSO (particle swarm optimization) and WNN (wavelet neural network) was proposed to optimize the structural parameters of the two models. Determine the optimal identification model for the threshing and cleaning system through testing. Then, adopt the MPC (model predictive control) method to regulate the multi-parameter of threshing and cleaning. The control process simulation was conducted in Simulink, and the software and hardware of the multi-parameter control system for the threshing and cleaning of the combine harvester were developed. The execution process of the control program was optimized, and field validation was conducted. The system could adjust the performance indicators to near the ideal set values in less than or equal to 3 s (indicator differences <= 0.23 %). In addition, the trigger times of the adjustment process were also few in the long-term continuous harvesting operation, which could make the performance indicators meet the requirements in the whole process. This indicated that the modeling and control methods proposed in this paper had good adaptability and stability, and laid the foundations for improving the intelligent level of harvesting machinery and achieving intelligent agriculture.