Soil-cutting simulation and dual-objective optimization on tillage process parameters of micro-tiller by smoothed particle Galerkin modeling and genetic algorithm

Abstract

Precision tillage is a technology-based approach enabled by understanding the soil-tool interaction, which is of central importance to improving agricultural operations. Numerical simulation offers a valuable way to investigate soil behavior in tilling process which is typically difficult or impossible to obtain via physical experiments. Inspired by the recent development in the numerical analysis of processes involving large deformation (e.g., cutting of blood clots), smoothed particle Galerkin (SPG) method was used to simulate the interaction between rotary blade and soil in the tilling process. Motion of soil particles, torque on the rotary blade, and power consumption were obtained for performance analysis by simulation and soil bin experiments. It was showed that SPG method was an effective tool to simulate soil-cutting process which provided quite acceptable predictions compared with soil bin tests. Results also supported the presence of a direct relationship between process parameters (tillage depth, forward velocity, and rotation velocity) and micro-tiller’s working performance. Instead of conventional polynomial regression, extreme learning machine (ELM), a single hidden layer feedforward neural network, was used to model this relationship, which had better flexibility and scalability in this multi-input and multi-output problem. Using training data from SPG modeling, this physics-free fast prediction method provided satisfactory consistency with numerical simulation results but with considerably less time, which served as the basis for efficient parameter optimization in tillage. Reduction of torque on the rotary blade and power consumption of micro-tiller in mechanical tilling operation benefits the sustainability of the environment and farming systems. Taking these two as dual-minimization objectives, NSGA-II, a genetic algorithm-based iterative procedure together with ELM modeling was implemented to optimize operation parameters.