Optimization method of rotary tillage blade shape based on BP neural network particle swarm optimization and automated simulation control

Abstract

The present invention discloses a method for optimizing the shape of a rotary tillage scimitar based on BP neural network particle swarm optimization and automated simulation control, comprising the following steps: S1, using VBA to develop a plug-in for quickly generating a rotary tillage scimitar model in SolidWorks; S2, using batch processing to develop an EDEM control program for controlling scimitar soil-cutting simulation; S3, using Python to process scimitar soil-cutting data; S4, using a plug-in for quickly generating a scimitar model to complete scimitar modeling; S5, using the EDEM control program to complete scimitar soil-cutting simulation; S6, using a data processing program to complete soil-cutting simulation data collation; S7, using PyTorch to construct a regression model for the influence of the rotary tillage scimitar shape on the rotary tillage scimitar simulation data and train it; S8, establishing a fitness function with the soil-cutting effect of the rotary tillage scimitar as the optimization target, and optimizing the rotary tillage scimitar in combination with the regression model; the rotary tillage scimitar shape optimization method proposed by the present invention has the characteristics of high efficiency, strong operability, flexible design, good reliability and wide applicability.

Description

Technical Field

[0001] The invention belongs to the field of rotary tillage scimitar optimization and is an optimization method for optimizing the shape of a rotary tillage scimitar by driving soil cutting data of the scimitar. Background Art

[0002] The design and optimization of rotary tillage blades have always been the focus of research in the field of agricultural machinery. At present, there are two main methods for optimizing rotary tillage blades at home and abroad: one is to design several rotary tillage blades with different shaping parameters, and obtain the rotary tillage blade shape with the best soil cutting effect through physical soil cutting tests or simulated soil cutting tests. This method is simple to design and easy to operate, but the optimized rotary tillage blade obtained is not the optimal rotary tillage blade, but the better rotary tillage blade among several groups of rotary tillage blades in the cross-comparison; the other is to conduct a theoretical analysis of the soil cutting process of the rotary tillage blade, and obtain the optimal rotary tillage blade shaping parameters through analysis. This method is relatively complex to operate, and it is difficult to construct the corresponding mathematical relationship or obtain the optimal value. In response to the above problems, the present invention proposes a rotary tillage blade shape optimization method based on a combination of BP neural network, particle swarm optimization and automated simulation control. This method can find the optimal solution of the rotary tillage blade in a larger range, avoiding the defect of only obtaining the better solution among a few groups of rotary tillage blade shapes through cross-comparison. At the same time, the optimized rotary tillage blade shape is obtained by driving the rotary tillage effect data, avoiding the construction of complex mathematical relationships, making the optimization process simpler and more convenient both in terms of thinking and execution. Summary of the Invention

[0003] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a rotary tiller scimitar shape optimization method that combines the BP neural network regression model, the particle swarm optimization algorithm and the automated simulation control, which has the characteristics of high efficiency, strong operability, flexible design, good reliability and wide applicability.

[0004] The object of the present invention is achieved through the following technical solution: a rotary tillage blade shape optimization method based on BP neural network particle swarm optimization and automated simulation control, characterized in that it includes the following steps:

[0005] S1. Taking the soil cutting effect of rotary tillage blade as the optimization target, a simulation model of rotary tillage blade soil cutting was established in discrete element simulation software EDEM;

[0006] S2. Select the control points of the rotary tillage blade deformation, design and implement the rotary tillage blade automatic soil cutting simulation program, including: quickly generate the rotary tillage blade model program, the rotary tillage blade soil cutting simulation control program and the data processing program;

[0007] S3, constructing a BP neural network regression model of the rotary tillage scimitar shape and soil cutting effect, using a rotary tillage scimitar automatic soil cutting simulation program to obtain multiple sets of soil cutting data of rotary tillage scimitars with different shapes, and using this data to train the model;

[0008] S4. Use the particle swarm algorithm to optimize the trained regression model and find the rotary tillage blade model with the best rotary tillage effect.

[0009] The step S1 is specifically implemented as follows:

[0010] S11: Construct soil particle model and straw particle model in EDEM, and use the soil particle model and straw particle model to build the soil-straw simulation environment required for simulation.

[0011] The step S2 is specifically implemented as follows:

[0012] S21: Select the control point parameters of the rotary tiller blade, and use VBA language to build a program in SolidWorks that is driven by these parameters to quickly generate a rotary tiller blade model.

[0013] S22: With the goal of controlling the rotary tillage blade soil cutting simulation model built in EDEM, a control program for EDEM is constructed using batch commands. This program can control EDEM to automatically complete multiple sets of rotary tillage blade soil cutting simulation operations and data collection in sequence.

[0014] S23: Use Python to build a processing program for the rotary tillage blade cutting soil simulation data to process the data generated by the simulation.

[0015] The step S3 is specifically implemented as follows:

[0016] S31: Use Pytorch to build a BP neural network regression model about the rotary tillage blade shaping parameters and rotary tillage effect. It is mainly composed of input layer, hidden layer and output layer. The activation function is selected as ReLu function, the loss function is L1Loss function, and the optimizer is stochastic gradient descent function.

[0017] S32: A program for quickly generating rotary tillage scimitar models is used to generate a large number of rotary tillage scimitar models with different modeling parameters, a rotary tillage scimitar soil cutting simulation control program is used to control these rotary tillage scimitars to complete the rotary tillage soil cutting simulation, and a data processing program is used to collect and organize all simulation data.

[0018] S33: Use the simulated rotary tillage blade soil cutting data to train the BP neural network regression model.

[0019] The step S4 is specifically implemented as follows:

[0020] S41: Construct a fitness function for finding the rotary tillage blade shaping parameters corresponding to the optimal rotary tillage effect.

[0021] S42: Construct a particle swarm algorithm and initialize the particle swarm parameters to optimize the rotary tillage blade shaping parameters.

[0022] The present invention utilizes a BP neural network regression model and a particle swarm optimization algorithm to optimize the shape of rotary tiller blades. This method utilizes data from a large number of soil-cutting simulations of rotary tiller blades with different shapes, establishes a BP neural network regression model, and uses the simulation data to train the regression model. Finally, the particle swarm optimization algorithm optimizes the rotary tiller blades. This optimization method boasts a simple principle, strong operability, flexible design, high reliability, and wide applicability, providing valuable guidance and reference for research on rotary tiller blade optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Flowchart optimized for the rotary tiller blade of the present invention

[0024] Figure 2 Schematic diagram of rotary tillage blade control parameters

[0025] Figure 3 The simulation process of rotary tillage blade cutting soil

[0026] Figure 4 BP neural network regression model for the relationship between rotary tillage blade shape and rotary tillage effect

[0027] Figure 5 Flowchart for BP regression model training

[0028] Figure 6 This is a graph showing the change in loss function value during BP regression model training.

[0029] Figure 7 Particle swarm optimization process

[0030] Figure 8 The fitness change during particle swarm optimization

[0031] Figure 9 Secondary development process for simulation control of rotary tillage blade cutting soil

[0032] Figure 10 Design ideas for a program to quickly generate models for rotary tillage blades

[0033] Figure 11 Rapidly generate model program interface for rotary tillage blade

[0034] Figure 12 Design ideas for the automatic control program of rotary tillage blade cutting soil simulation

[0035] Figure 13 Design ideas for data processing programs

[0036] Figure 14 Abstract with attached diagram DETAILED DESCRIPTION

[0037] The technical solution of the present invention is further described below with reference to the accompanying drawings.

[0038] like Figure 1 As shown, the rotary tillage blade shaping optimization method proposed by the present invention based on the combination of BP neural network, particle swarm optimization and automated simulation control includes the following steps:

[0039] S1. Taking the soil-cutting effect of the rotary tillage blade as the optimization target, considering the influence of the shape of the rotary tillage blade on the soil-cutting performance of the rotary tillage blade, a series of discrete element soil-cutting simulation models of rotary tillage blades with different shape parameters are established;

[0040] In step S1, variable parameters of the rotary tiller scimitar are selected for secondary development to obtain a rapid generation model of the rotary tiller scimitar, and the parameters of the rotary tiller scimitar are modified to generate a series of rotary tiller scimitars with different shapes.

[0041] The parameters of the selected rotary tiller blade are as follows: Figure 2 The program design idea of ​​generating a scimitar model based on the rotary tillage scimitar model parameters is as follows: Figure 10 The interface of the program for quickly generating a scimitar model is as follows: Figure 11 , the program starts with Figure 2 The parameters in the figure are used as input to model the rotary tillage blade and generate a series of rotary tillage blade models. The values ​​of the rotary tillage blade modeling parameters are shown in Table 1.

[0042] Table 1

[0043]

[0044]

[0045]

[0046]

[0047]

[0048] S2, then use the rotary tillage blade soil cutting simulation automation program to control all rotary tillage blades to perform soil cutting simulation. The design idea of ​​the automation control program is as follows: Figure 12 As shown in the figure, the simulation process of rotary tillage blade cutting soil is as follows: Figure 3 shown.

[0049] After simulating the soil cutting of all rotary tillage blades, a data processing program is used to organize all the simulation data. The design idea of ​​the data processing program is as follows: Figure 13 The final data is shown in Table 2.

[0050] Table 2

[0051]

[0052]

[0053]

[0054]

[0055]

[0056] S3. After obtaining the soil cutting effect data of rotary tillage machetes with different shapes, a BP neural network regression model of rotary tillage machete shape parameters and soil cutting effect of rotary tillage machete is established. The established model is as follows: Figure 4 As shown. Then the simulation data is divided into training set and test set, and the regression model is trained using the training set data. The training process is as follows Figure 6 shown.

[0057] S4. After obtaining the regression model of the rotary tillage blade shape and rotary tillage effect, the particle swarm optimization algorithm is introduced to optimize the rotary tillage blade shape. The optimization idea is as follows: Figure 7 shown.

[0058] The optimization mathematical model of this example is as follows:

[0059]

[0060] Where X is the rotary tillage blade shaping parameter, G(X) is the BP regression model obtained by previous training, Z is the predicted value output by the regression model, which consists of the maximum soil throwing amount z1, the maximum stubble burying amount z2 and the maximum force torque z3, and the minF(Z) function outputs the fitness, z imax is the maximum value of each rotary tillage index in the previous simulation data, z imin It is the minimum value of each rotary tillage index in the previous simulation data.

[0061] The input parameter of this model is the rotary tillage blade shaping parameter X, and its constraints are:

[0062]

[0063] Since the values ​​of the maximum amount of soil thrown and the maximum stubble buried are as large as possible, and the maximum reaction torque on the rotary tillage blade is as small as possible, according to the requirements of the particle swarm algorithm, w1 and w2 are both negative values, and w3 is a positive value. Here, the present invention takes the absolute values ​​of the three as 0.3, that is, w1 = -0.3, w2 = -0.3, and w3 = 0.3.

[0064] When performing particle swarm optimization, it is necessary to initialize the particle swarm parameters. The values ​​of the particle swarm parameters initialized in the present invention are shown in Table 3.

[0065] Table 3

[0066] parameter Parameter value Objective function dimension X 8 Population size n 100 Maximum number of iterations m 100 Inertia weight ω 0.2 <![CDATA[Individual learning factor c1]]> 0.5 <![CDATA[Social learning factor c2]]> 0.5

[0067] Use the above parameters to initialize and start iteration. The iteration process is as follows: Figure 8 shown.

[0068] The optimized rotary tillage blade shaping parameter values ​​are shown in Table 4.

[0069] Table 4

[0070] Rotary tillage blade parameters Optimal value L1 51.9 L2 28.6 L3 112.6 L4 34.8 L5 68.8 L6 122.0 L7 23.2 L8 24.7

[0071] The data obtained by using the regression model to predict the optimized rotary tillage blade model, the data obtained by simulating the soil cutting of the optimized rotary tillage blade, and the simulated soil cutting data of the original design rotary tillage blade are sorted into the following Table 5. It can be seen from the table that the value of the rotary tillage blade regression prediction using the neural network is close to the simulated value of the rotary tillage blade soil cutting, indicating that it is feasible to use the neural network to regress the rotary tillage blade soil cutting.

[0072] Table 5

[0073]

[0074] From the analysis in Table 5, the soil cutting simulation comparison between the optimized rotary tillage blade and the original rotary tillage blade shows that the optimized rotary tillage blade has a 23.1% higher stubble burying ability and a 29.6% lower maximum reaction torque than the original rotary tillage blade. This shows that the method of using the neural network plus particle swarm algorithm to find the optimal parameters of the rotary tillage blade is feasible.

[0075] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A rotary tillage blade shape optimization method based on BP neural network particle swarm optimization and automated simulation control, characterized in that: The following steps are involved: S1. Selecting the desired controllable rotary tillage blade shaping parameters with the goal of optimizing the rotary tillage blade's soil throwing ability, stubble burying ability, and reaction torque when cutting soil; S2. Design and implement a secondary development program for rotary tillage blade soil cutting simulation, including: a rapid generation of a scimitar model program, an automatic control program for rotary tillage blade soil cutting simulation, and a simulation data collection and organization program; S3. Design multiple groups of rotary tillage scimitars with different shapes by changing the selected rotary tillage scimitar shaping parameter values; use a rapid scimitar model generation program and SolidWorks to establish a rotary tillage scimitar model with the designed shaping parameters; use a rotary tillage scimitar soil cutting simulation automatic control program and EDEM to complete multiple groups of rotary tillage scimitar soil cutting simulations; and use a simulation data collection and collation program to collect rotary tillage scimitar soil cutting simulation data; S4. Use PyTorch to establish a BP neural network regression model for rotary tillage blade shape optimization, and use rotary tillage blade soil cutting simulation data for model training; S5: Combine the particle swarm optimization algorithm with the regression model, and establish a fitness function for the soil cutting performance, stubble burying performance and reaction torque of the rotary tillage blade to optimize the rotary tillage blade.

2. The method for optimizing the shape of rotary tillage blades according to claim 1, characterized in that: The specific implementation method of step S1 is: selecting the height of the tangent end face of the rotary tiller, the length of the upper end face of the tangent, the radius of the back edge curve circle, the maximum angle of the tangent blade curve, the Y-direction distance of the back edge curve circle from the origin, the X-direction distance of the back edge curve circle from the origin, and the maximum angle of the side cutting blade curve as the rotary tiller shape control parameters, and selecting the rotary tillage reaction force, soil throwing ability, and stubble burying ability of the rotary tiller when cutting soil as optimization indicators.

3. The method for optimizing the shape of rotary tillage blades according to claim 1, characterized in that: The specific implementation method of step S2 is: using VBA to develop a plug-in in SolidWorks to quickly generate a rotary tillage machete model; using batch processing to develop an EDEM control program to automatically control the machete soil cutting simulation and data export; using Python to build a data processing program to process all simulated machete soil cutting data.

4. The method for optimizing the shape of rotary tillage blades according to claim 1, characterized in that: The specific implementation method of step S3 is: using SolidWorks and the rapid generation of scimitar model plug-in to complete the design and modeling of 177 sets of rotary tillage scimitars; using EDEM and EDEM control program to complete the soil cutting simulation test of all rotary tillage scimitars; and using a data processing program to complete the compilation of all soil cutting simulation data.

5. The method for optimizing the shape of rotary tillage blades according to claim 1, characterized in that: The specific implementation method of step S4 is: using the deep learning framework PyTorch to build a regression model about the influence of the rotary tillage scimitar shape on the rotary tillage scimitar simulation data, and using the rotary tillage scimitar soil cutting simulation data for training.

6. The method for optimizing the shape of rotary tillage blades according to claim 1, characterized in that: The specific implementation method of step S5 is: taking the soil cutting effect of the rotary tillage blade as the optimization target, establishing the fitness function of the rotary tillage blade's stubble burying ability, soil throwing ability and the reaction torque received by the rotary tillage blade during the soil cutting process, and combining the trained regression model with the particle swarm algorithm to optimize the rotary tillage blade shape.

Images