Application of the Discrete Element Method to Soil–Tool Interaction: A Critical Narrative Review of Contact Models, Calibration and Predictive Reliability

Abstract

Abstract

Numerical modelling of the interaction between soil-engaging tools and agricultural soils has become central to the design of tillage and seeding equipment, where draught reduction, controlled soil disturbance and residue management influence energy use and crop establishment. The discrete element method (DEM), which represents soil as an assembly of interacting particles rather than a continuum, has emerged as the dominant particle-scale approach for this problem. This review critically examines how DEM has been applied to soil–tool interaction, with attention to the choice of contact model, the calibration of microscopic parameters, the treatment of particle size and shape, and the reliability of predictions for draught force, vertical force, soil disturbance and soil translocation. The available evidence indicates that DEM can reproduce measured tillage forces and furrow profiles with useful accuracy across a range of tools, including sweeps, subsoilers, narrow openers, discs and mouldboard ploughs, provided that an appropriate plastic or cohesive-adhesive contact model is combined with carefully calibrated parameters. However, the field is characterised by persistent tensions: between computational tractability and physical fidelity in the selection of enlarged particles; between the many-to-one mapping of calibrated parameter sets and the physical properties they represent; and between force prediction, which is often robust, and soil-movement prediction, which remains more sensitive to particle scaling. Cohesive and adhesive wet soils, high-speed operations, heterogeneous layered profiles and tool wear are comparatively under-represented and less consistently validated. Coupling of DEM with multibody dynamics, computational fluid dynamics and finite element analysis extends its reach but introduces additional calibration and verification burdens. Machine-learning-assisted calibration and improved measurement of soil micro-properties are promising but not yet standardised. The review concludes that confidence in DEM is strongest for relative comparisons of tool geometry in well-characterised cohesionless and lightly cohesive soils, and weaker for absolute prediction in wet, adhesive or rapidly changing field conditions. Priorities include transparent and reproducible calibration, standardised validation metrics, and systematic evaluation across soil types and operating speeds.