Collaborative optimization method of cleaning operational performance and multiparameter online control system for combine harvesters

• An optimization method for cleaning performance in combine harvest is proposed. • The PSO algorithm is used to optimize the SVR surrogate model of cleaning. • The MOPSO algorithm is used to decouple the inputs and outputs of the cleaning. • An online optimization system for cleaning performance is constructed. • Experiments show that the optimization system help in cleaning loss reduction. The cleaning process of a rapeseed combine harvester utilizes a highly complex multivariable system. Key operating parameters include the airflow velocity, vibrating sieve amplitude, frequency, and sieve opening. Strong coupling and nonlinear effects are observed between these variables and the cleaning operational performance, such as the loss and impurity rates. Furthermore, the relationship between these inputs and outputs significantly changes under diverse field conditions, resulting in an inconsistent harvest performance. Therefore, a multiparameter system optimization method that integrates the cleaning loss and impurity rates was explored, and an online regulation system for the operating parameters was developed. A support vector machine regression (SVR) methodology was proposed to establish a surrogate model of cleaning performance indicators. The fitting accuracy of the surrogate model was analyzed and compared after optimizing the key SVR parameters by the nonheuristic grid search (GS) algorithm and the heuristic particle swarm optimization (PSO) algorithm. The results showed that the surrogate model that employed PSO algorithm exhibited a 73.6% reduction in the mean squared error compared with the GS algorithm. This indicated that the PSO-SVR surrogate model had a higher prediction accuracy and enhanced generalizability. Based on the Pareto dominance principle, an improved multi-objective particle swarm optimization algorithm was adopted to solve the surrogate model and effectively decouple the cleaning performance indicators under multivariable coupling constraints. A nonlinear relationship between the two objectives was obtained. The optimal Pareto solution front showed the mapping relationships between independent variables and dependent variables and provided decision-makers with multiple solutions to achieve a lower loss or impurity rate. In addition, the accuracy of the optimization results was verified by field experiments. Upon implementation of the regulation system, the cleaning loss rate decreased from 4.84% to 3.05%, while the impurity rate was relatively unchanged. This study provides a reference for the intelligent decision-making design of the cleaning system of a combine harvester.