Research On Predicting The Droplet Deposition Distribution In Canopy During Air-Assisted Spraying

In the middle to late stages of field crop growth,the canopy becomes dense with poor ventilation and light penetration,resulting in uneven droplets deposition across the upper,middle,and lower layers of the plants and ineffective pest control.Air-assisted spraying helps achieve uniform droplet deposition in canopy and complete coverage of the plant,making it a widely adopted method of pesticide application.However,the existing understanding of droplet deposition mechanisms in air-assisted spraying lacks quantitative analysis,and the intrinsic correlation between operational parameters and spraying effects lacks digital transparency.Optimization of operational parameters often relies on empirical studies specific to crop morphology,making it challenging to reproduce and generalize results,resulting in unstable pest control efficacy,pesticide misuse,and low utilization rates.Currently,research on droplet deposition mechanisms is mostly qualitative or semi-quantitative,lacking quantitative prediction models for droplet deposition,which has become a technical bottleneck for achieving precise and environmentally friendly pesticide application.This paper takes the population of cotton plants in the middle to late stages as research object and aims to accurately predict the distribution of droplet deposition in canopy.To achieve this goal,a heterogeneous porous medium model considering the distribution pattern of leaves in canopy is constructed.A two-way fluid-structure interaction simulation method is proposed to analyze the attenuation pattern of airflow in canopy.Additionally,a prediction model for canopy airflow distribution is developed.Furthermore,a two-stage simulation method for droplet deposition under the deformation of canopy leaves induced by assisted airflow is studied.Finally,a digital twin prediction system for droplet deposition in canopy under the influence of airflow is established.This research aims to provide model references and technical support for rapidly predicting the efficacy of pesticide application.(1)Constructing a stratified sub-regional porous medium canopy airflow attenuation model based on a simplified virtual phenotype of cotton.Addressing the uneven spatial distribution of branches and leaves within the target population’s canopy structure,we propose a virtual canopy phenotype model driven by cotton agronomic planting parameters,3D point cloud scanning,and growth and development characteristics.This model extracts canopy structural parameters such as leaf area density and porosity to form a stratified sub-regional porous medium canopy model.Combining with application parameters,we study the attenuation distribution of airflow and vorticity in canopy and construct a simulation model for canopy airflow attenuation based on stratified sub-regionalporous medium.Spraying experiments show that the normalized mean absolute errors(NMAE)between simulated and measured values for the upper,middle,and lower canopy layers are 9.75%,21.35%,and 17.38%,respectively.This preliminary evidence demonstrates the effectiveness and feasibility of the model.Compared with traditional single and uniform porous medium models,this study is able to express the hindering effect of heterogeneous canopy structure on airflow and the attenuation law of airflow.(2)Building a simulation model for flexible leaf deformation and dynamic airflow changes based on two-way fluid-structure interaction.Leveraging the spatial distribution characteristics of multiple leaves in target canopy,we conducted mechanical tests on leaf deformation under wind loads.Using finite element and Lattice Boltzmann methods,we constructed a two-way fluid-structure interaction(airflow-leaf interaction)simulation model to analyze the dynamic distribution of airflow in canopy under action of assisted airflow and deformation of multiple flexible leaves.We also proposed a method for calculating the dynamic porosity of each layer in canopy by combining image processing techniques.Comparison results between simulation and empirical experiments show that the mean absolute error of prediction(MAEP)and normalized mean absolute error(NMAE)between simulated and measured airflow distribution values are 0.6317 and 11.38%,respectively.The maximum NMAE between simulated and measured values for leaf inclination change is 21.46%.These experimental results demonstrate the feasibility of simulating the interaction between flexible leaves and airflow,revealing the dynamic interaction patterns between multiple leaves and airflow at the leaf scale.(3)Predicting canopy airflow distribution by combining heterogeneous porous media and fluid-structure interaction simulation.To effectively predict airflow attenuation in canopy,we propose a machine learning-based method that integrates the analysis of airflow interaction with branches and leaves based on both heterogeneous porous media airflow simulation and fluid-structure interaction simulation datasets.This method utilizes algorithms such as artificial neural networks,random forest,and support vector regression to construct a model for predicting canopy airflow distribution.Comparative experimental results show that the random forest prediction results are in good agreement,with a coefficient of determination R~2 of 0.9860and a root mean square error(RMSE)of 0.3017,demonstrating that the model effectively predicts canopy airflow distribution and reduces the time and computational costs of computational fluid dynamics simulations.(4)Two-stage simulation method and deposition performance prediction of air-assisted droplet deposition behavior in canopy.To predict the deposition performance of droplets on branches and leaves during air-assisted spraying,a two-stage simulation method involving fluid-structure interaction(airflow-leaf interaction)and fluid-liquid-solid coupling(airflow-droplet-leaf interaction)for crop canopy droplet deposition is proposed.The deformation plant model is derived from the results of fluid-structure interaction simulation,and simulation of droplet deposition behavior on deformed plant canopy is performed using the discrete particle tracking method.Comparative experimental results show that the linear regression equation coefficient of determination(R~2)between simulation and measured values is 0.8684,with a root mean square error(RMSE)of 0.1463.There is no significant difference(P>0.05)between simulated and measured values of droplet deposition density in the upper,middle,and lower layers of canopy.Using the validated droplet deposition simulation model,analyze the effect of canopy droplet deposition distribution under different airflow velocities,spray flow rates,nozzle heights,and external natural wind conditions.Based on the simulated dataset of droplet deposition,propose a method for predicting canopy droplet deposition distribution using long short-term memory networks.(5)Construction of an airflow-induced droplets deposition prediction system based on digital twins in canopy.Drawing inspiration from the concept of digital twins,a system framework for predicting droplets deposition in canopy under airflow-induced based on digital twins.This framework encompasses system requirements analysis,theoretical models,overall system architecture design,data structure design,and key system technologies,laying the foundation for the digital twin system for canopy droplet deposition.Through application in specific scenarios,the feasibility of predicting droplet deposition distribution using the twin system is preliminarily validated,offering new insights into decision-making and regulation for droplet deposition prediction and digitized pesticide application in various air-assisted spraying scenarios.