Comparative study of CFD models of the air flow produced by an air-assisted sprayer adapted to the crop geometry

•A sprayer adapted to the crop geometry has been modeled in CFD in different ways.•Models based on air flow at the outlets does not yield satisfactory results.•Implementation of the geometry of the air ducts has been tested with good results.•Studying the effect of different air flows by varying the input flow is possible.•The instability of the model can be mitigated by calculating into two stages. A computational fluid dynamics (CFD) model for simulating the air flow produced by an air-assisted sprayer can be developed according to any one of several sets of criteria and for various specific applications. Thus far, most CFD models have focused on the characterization of the air flow generated by the sprayer. However, such models may not be the best when assessing the effectiveness of air-assisted sprayers adapted to crop geometry, such as those used in vineyards. In this study, the air flow produced by an air-assisted sprayer adapted to the geometry of a vineyard was simulated using four CFD models: Model 1 was designed to simulate air velocity measurements at the outlets; Model 2 was designed to simulate air velocity measurements at a certain distance from the outlets; Model 3 was dedicated to the modelling of the internal geometry of the air ducts rather than the characterization of the air flows generated by the sprayer; and Model 4, which was developed as a variant of Model 3, was created to perform the same calculation in several stages. The models were validated with actual measurements of the air velocity near the sprayer outlets. The results showed that although Models 1 and 2 (both of which have been used in most existing studies) simplified the calculation, they are impractical for simulating different air flows. By contrast, Models 3 and 4 provided complex meshes that complicate the convergence of the calculation and require a suitable treatment of both the viscosity and the flow near the walls. Model 3 showed the smallest error (16%) in the air velocity estimated in the treatment plane. Model 4 showed potential for future implementation of the dispersed phase and crop-air interaction because it mitigated the problems arising from the complexity of the meshes. It should also be noted that there are errors inherent to the implementation of the CFD model, errors related to inaccuracies in the geometry of the air ducts, inaccuracies related to the measurement of the air velocities, inaccuracies in the quantification of the air flow, and simplific