Optimal design of an omnidirectional adjustment mechanism for the chassis of a tracked combine harvester

Tracked combine harvesters have widely used in the hilly and mountainous terrains. It is highly demanded for the highly adaptive adjustment mechanism of the chassis, due to the complex operational environments. However, conventional systems can often suffer from irrational structural parameters, leading to large fluctuations in the track tension, excessive hydraulic loads, high risk of track derailment, and compromised stability during slope-crossing operations. Such challenges have severely limited the machine reliability, service life, and operational efficiency in real field conditions. The objective of this study was to design and then optimize an omnidirectional adjustable chassis. The coordinated lateral and longitudinal attitude adjustments were performed to maintain the stable track tension in order to reduce the requirements of the actuator thrust. The generalizable modeling and optimization framework was also constructed using terrain-adaptive platforms. A parametric model was then established for both the lateral and longitudinal adjustment mechanisms of the chassis. There were geometric relationships between critical link circumferences, hydraulic cylinder strokes, chassis attitude, the resulting track deformation, and tension states. Multiple working conditions were simulated to represent the real terrain scenarios, including the variations in the slope angle, chassis height, and combined attitude states. The structural parameters were also optimized under multi-constraint conditions using the AutoDesign module within the RecurDyn multibody dynamics platform. All components after optimization remained within the allowable stress and stroke limits, with the minimum hydraulic effort. As a result, the front and rear hydraulic cylinder strokes increased by 10.54% and 22.82%, respectively, leading to the reductions of 19.54% and 20.04% in the maximum thrust. The efficiency of the energy transfer was effectively improved to lower the mechanical impact on the support structure. In addition to the physical parameter optimization, a surrogate model was developed to rapidly predict the track circumference under different adjustment conditions. A multilayer perceptron neural network was employed to train the surrogate using adaptive moment estimation. The nonlinear mapping was approximately realized among tension wheel position, hydraulic stroke, and track circumference. The model was also trained on the simulation-generated data in the full adjustment range, and then validated on unseen test scenarios. The better performance was achieved in a root mean square error of 0.514 and a coefficient of determination of 0.997, indicating the excellent predictive accuracy and generalization suitable for engineering applications. The surrogate model was then embedded into the grey wolf optimizer. The global optimization was then performed on the tension wheel configuration. The optimal tension wheel was then positioned at 296.11 mm from the rearmost support roller. The track circumference remained consistently close to the actual value of 4590 mm under all attitude combinations and chassis height levels. Extensive validation experiments were performed on a full-scale physical prototype of the tracked combine harvester equipped with the optimal chassis. The prototype was tested under a series of lateral and longitudinal adjustment scenarios. Compared with the pre-optimization configuration, the optimal chassis reduced the maximum cylinder thrust by 2.5%-31.8% and the average thrust by 2.9%-19.7%, indicating the high effectiveness of the optimization. Furthermore, the track sag was consistently maintained within the effective tension range of 120-150 mm among the tested configurations. The smoothness, mechanical stability, and robustness of the tension control were substantially improved after adjustment under dynamic attitudes. The mechanical shock loads were reduced during adjustment in order to minimize the risk of slippage and derailment. Ground contact consistency and traction performance were enhanced during operation on uneven terrain. In summary, an integrated approach was combined with parametric modeling, neural network-based surrogate modeling, and optimization. The posture adjustment and track tension control were realized in the tracked agricultural machinery. The optimal chassis system significantly improved the mechanical efficiency to maintain the stable track tension. The practical applicability was obtained to reduce the actuator load in the combine harvesters. Moreover, the modeling and optimization framework demonstrated that the strong generalization was suitable for the construction equipment, autonomous ground vehicles, and mobile robots on complex terrains.