Maize harvester operation soil compaction degree judgment method based on machine tool real-time parameters

Abstract

The invention belongs to the technical field of agricultural mechanization and intelligent equipment, and particularly relates to a corn harvester operation soil compaction degree judgment method based on machine tool real-time parameters, in particular to a soil compaction degree judgment method. The invention aims to solve the problem of low soil compaction prediction accuracy caused by the fact that the existing soil compaction prediction method considers soil indexes and machine pressure and does not consider the influence of dynamic indexes which can be autonomously controlled by a driver on soil compaction. The method comprises the steps of 1, collecting a feature data training set; 2, obtaining a standardized feature data training set; 3, obtaining an optimal compaction degree prediction model; 4, constructing an optimal comprehensive compaction degree model; 5, constructing an optimal compaction early warning model; 6, constructing an optimal comprehensive compaction risk early warning model; 7, collecting a to-be-detected feature data set, obtaining an optimal comprehensive compaction degree model of the to-be-detected feature data set, and outputting a comprehensive compaction degree value and a risk early warning level; and 8, pushing on a screen end in real time.

Description

Technical Field

[0001] The invention belongs to the technical field of agricultural mechanization and intelligent equipment, and particularly relates to a method for determining soil compaction degree. Background Art

[0002] Black soil is one of the three major black soil belts in the world and is also the core area of ​​grain production in Northeast my country. With its high fertility and physical structure suitable for crop growth, black soil occupies an important position in agricultural production. However, its ecological environment is extremely fragile. Especially in the process of mechanized operation, the soil is easily compacted, resulting in a decrease in air permeability, water permeability and nutrient utilization efficiency. This change in soil structure not only weakens agricultural production efficiency, but also poses a serious threat to the sustainable use of black soil. In response to this problem, it is urgent to develop a technical method that integrates real-time monitoring, prediction and dynamic early warning to help scientifically guide agricultural mechanized operations and achieve effective protection of black soil resources. The present invention aims to protect the ecological environment of black soil, improve the operating efficiency and environmental protection performance of corn combine harvesters, and has wide agricultural application value. Summary of the invention

[0003] The purpose of the present invention is to solve the problem that the existing soil compaction prediction methods all consider soil indicators and the pressure of the machine itself, but do not consider the influence of dynamic indicators that the driver can independently control on soil compaction, resulting in low accuracy of soil compaction prediction, and propose a method for determining the soil compaction degree of a corn harvester based on real-time parameters of the machine.

[0004] A method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool is characterized in that the specific process of the method is as follows:

[0005] Step 1: Collect feature data training set; the specific process is:

[0006] The feature data training set includes the number of repeated rolling times X 1 , harvester weight X 2 , Track hardness X 3 , Track tension X 4 , unloading frequency X 5 、Total turning time ratio X 6 、Differential steering time ratio X 7 、Ratio of braking and steering time to X 8 、Ratio of time of turning in place X 9 ;

[0007] Step 2: Standardize the feature data collected in step 1 to obtain a standardized feature data training set;

[0008] Step 3: Based on the standardized feature data training set, an optimal compaction degree prediction model is obtained;

[0009] Step 4: Based on the optimal compaction degree prediction model, construct an optimal comprehensive compaction degree model;

[0010] Step 5: Based on the optimal comprehensive compaction degree model, construct an optimal compaction warning model;

[0011] Step 6: Based on the optimal comprehensive compaction degree model and the compaction warning model, an optimal comprehensive compaction risk warning model is constructed;

[0012] Step 7: Collect the feature data set to be tested, and based on Steps 2 to 6, obtain the optimal comprehensive compaction degree model of the feature data set to be tested, and output the comprehensive compaction degree value and risk warning level;

[0013] Step 8. Output the comprehensive compaction degree value and warning level according to the optimal comprehensive compaction degree model, and push the comprehensive compaction degree value f(C, g(C)), the data of each feature in the feature data set to be tested, and the warning level on the screen in real time. When the medium risk or high risk warning level is reached, a prompt alarm is issued.

[0014] Preferably, in step 2, the feature data collected in step 1 is standardized to obtain a standardized feature data training set; the specific process is:

[0015] Each feature data collected in step 1 is standardized to obtain each standardized feature data; the standardization formula is:

[0016]

[0017] Where: X j,i represents the i-th feature data of the j-th sample after standardization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0018] X' ji is the i-th feature data of the j-th sample;

[0019] μ i is the mean; σ i is the standard deviation.

[0020] Preferably, in step 3, an optimal compaction degree prediction model is obtained based on the standardized feature data training set; the specific process is:

[0021] Step 31: Based on the standardized feature data, a compaction degree prediction model is constructed; the specific process is as follows:

[0022] The compaction degree prediction model adopts the multivariate linear regression model:

[0023] Y B =B 1 X1 +B 2 X 2 +B 3 X 3 +B 4 X 4 +B 5 X 5 +B 6 X 6 +B 7 X 7 +B 8 X 8 +B 9 X 9 +b B

[0024] Y Z =Z 1 X 1 +Z 2 X 2 +Z 3 X 3 +Z 4 X 4 +Z 5 X 5 +Z 6 X 6 +Z 7 X 7 +Z 8 X 8 +Z 9 X 9 +b Z

[0025] Y S =S 1 X 1 +S 2 X 2 +S 3 X 3 +S 4 X 4 +S 5 X 5 +S 6 X 6 +S 7 X 7 +S 8 X 8 +S 9 X 9 +b S

[0026] in:

[0027] Y B It represents the surface soil compaction value predicted by the compaction degree prediction model;

[0028] Y Z It represents the compaction value of the middle layer soil predicted by the compaction degree prediction model;

[0029] Y S It represents the deep soil compaction value predicted by the compaction degree prediction model;

[0030] B 1 , B 2 , B 3 , B 4 , B 5 , B 6 , B 7 , B 8 , B 9 They represent the compaction degree prediction model for surface characteristic data X 1 , X 2 , X 3 , X 4 , X 5 , X 6 , X 7 , X 8 , X 9 The feature weights of

[0031] Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , Z 6 , Z 7 , Z 8 , Z 9 They represent the compaction degree prediction model for the middle layer feature data X 1 , X 2 , X 3 , X 4 , X 5 , X 6 , X 7 , X 8 , X 9 The feature weights of

[0032] S 1 , S 2 , S 3 , S 4 , S 5 , S 6 , S 7 , S 8 , S 9 They represent the compaction degree prediction model for deep feature data X 1 , X 2 , X 3 , X 4 , X 5, X 6 , X 7 , X 8 , X 9 The feature weights of

[0033] b B A bias term representing surface feature data;

[0034] b Z A bias term representing the middle-level feature data;

[0035] b S A bias term representing deep feature data;

[0036] Step 32: Construct the error loss function L of the compaction degree prediction model;

[0037] Step 3. Use the gradient descent method to optimize the feature weights of the compaction degree prediction model until the weights converge and the optimal feature weights are obtained;

[0038] The optimal compaction degree prediction model is obtained based on the optimal feature weights.

[0039] Preferably, the bias term b of the surface feature data B Initial setting b B =0; bias term b of middle-level feature data Z Initial setting b Z = 0; bias term b of deep feature data S Initial setting b S =0;

[0040] B 1 The initial value is set to B 1 =0.15; B 2 The initial value is set to B 2 =0.15; B 3 The initial value is set to B 3 =0.07; B 4 The initial value is set to B 4 =0.05; B 5 The initial value is set to B 5 =0.13; B 6 The initial value is set to B 6 =0.15; B 7 The initial value is set to B 7 =0.10; B 8 The initial value is set to B 8 =0.10; B 9 The initial value is set to B 9 =0.10;

[0041] Z 1The initial value is set to Z 1 =0.20; Z 2 The initial value is set to Z 2 =0.17; Z 3 The initial value is set to Z 3 =0.05; Z 4 The initial value is set to Z 4 =0.05; Z 5 The initial value is set to Z 5 =0.13; Z 6 The initial value is set to Z 6 =0.15; Z 7 The initial value is set to Z 7 =0.10; Z 8 The initial value is set to Z 8 =0.10; Z 9 The initial value is set to Z 9 =0.10;

[0042] S 1 The initial value is set to S 1 =0.25; S 2 The initial value is set to S 2 =0.17; S 3 The initial value is set to S 3 =0.05; S 4 The initial value is set to S 4 =0.05; S 5 The initial value is set to S 5 =0.13; S 6 The initial value is set to S 6 =0.15; S 7 The initial value is set to S 7 =0.10; S 8 The initial value is set to S 8 =0.10; S 9 The initial value is set to S 9 =0.10;

[0043] The 0cm≤surface soil depth≤10cm;

[0044] The 10cm<middle soil depth≤30cm;

[0045] The deep soil depth of 30cm < 50cm.

[0046] Preferably, in step 32, an error loss function L of a compaction degree prediction model is constructed; the specific process is:

[0047] The mean square error is used to measure the deviation between the predicted value and the true value:

[0048]

[0049] Where: Y j ′ represents the actual soil compaction value of the jth sample; Y j represents the predicted soil compaction value of the jth sample; N represents the number of training samples.

[0050] Preferably, in step 33, the feature weights of the compaction degree prediction model are optimized by using a gradient descent method until the weights converge to obtain the optimal feature weights;

[0051] The optimal compaction degree prediction model is obtained based on the optimal feature weights;

[0052] The specific process is:

[0053] The gradient descent method is used to optimize the feature weights. The formula for iterative update of feature weights is:

[0054]

[0055] Where: W represents the feature weight; W = B 1 , B 2 , B 3 , B 4 , B 5 , B 6 , B 7 , B 8 , B 9 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , Z 6 , Z 7 , Z 8 , Z 9 , S 1 , S 2 , S 3 , S 4 , S 5 , S 6 , S 7 , S 8 or S 9 ;

[0056] W j represents the feature weight of the jth sample;

[0057] α″ represents the learning rate;

[0058] Represents the loss function on weight W j The partial derivative of

[0059] Partial derivative calculation formula:

[0060]

[0061] Where: X j,i Represents the i-th feature data of the j-th sample after standardization, j=1,2,…,N, i=1,2,…,9.

[0062] Preferably, in step 4, an optimal comprehensive compaction degree model is constructed based on the optimal compaction degree prediction model; the specific process is:

[0063] Step 41: Based on the optimal compaction degree prediction model, a comprehensive compaction degree model is constructed; the specific process is as follows:

[0064] C=β B Y B +β Z Y Z +β S Y S

[0065] in:

[0066] C represents the comprehensive compaction value;

[0067] β B It represents the weight of surface compaction degree, and the initial value is set to 0.5;

[0068] β Z It represents the compaction degree weight of the middle layer, and its initial value is set to 0.3;

[0069] β S Represents the weight of deep compaction degree, with the initial value set to 0.2.

[0070] Step 4.2: Construct the error loss function of the comprehensive compaction degree model The specific process is:

[0071] The mean square error is used to measure the deviation between the predicted value and the true value:

[0072]

[0073] Where: C j ′ represents the actual soil comprehensive compaction value of the jth sample; C j represents the predicted soil comprehensive compaction value of the jth sample; N represents the number of training samples;

[0074] Step 4.3: Use the gradient descent method to optimize the weight of the comprehensive compaction degree model until the weight converges and obtains the optimal feature weight β B , β Z and β S ;

[0075] Based on the optimal feature weight β B , β Z and β S Obtain the optimal comprehensive compaction degree model;

[0076] The specific process is:

[0077] The gradient descent method is used to optimize the feature weights. The formula for iterative update of feature weights is:

[0078]

[0079] Where: β represents the feature weight; β = β B , β Z or β S ;

[0080] β j represents the weight of the jth sample;

[0081] represents the learning rate;

[0082] Represents the loss function for weight β j The partial derivative of

[0083] Partial derivative calculation formula:

[0084]

[0085] Where: X j,i Represents the i-th feature data of the j-th sample after standardization, j=1,2,…,N, i=1,2,…,9.

[0086] Until the weights converge, the optimal feature weight β is obtained B , β Z and β S ;

[0087] Based on the optimal feature weight β B , β Z and β S Obtain the optimal comprehensive compaction degree model.

[0088] Preferably, in step 5, an optimal compaction warning model is constructed based on the optimal comprehensive compaction degree model; the specific process is:

[0089] 1) When the optimal comprehensive compaction value obtained in step 4 is 0<C<a, the compaction area belongs to the low compaction area;

[0090] The optimal compaction warning model is expressed as:

[0091] g(C)=k 1C p

[0092] Where: k 1 It represents the risk increase parameter of low compaction area, which is set to 1;

[0093] p represents the degree of smoothness of risk change in the controlled low compaction area;

[0094] g(C) represents the compaction warning value corresponding to the optimal comprehensive compaction value C;

[0095] a represents the compaction degree dividing point;

[0096] 2) When the optimal comprehensive compaction value a≤C<b obtained in step 4, the compaction area belongs to the medium compaction area;

[0097] The optimal compaction warning model is expressed as:

[0098] g(C)=g(a)+k 2 (1-e -m(C-a) )

[0099] Among them: g(a) represents the compaction warning value corresponding to the compaction degree cutoff point a;

[0100] k 2 It represents the risk increase parameter of the medium compaction area;

[0101] b represents the compaction degree dividing point;

[0102] m represents the risk sensitivity adjustment parameter of the medium compaction zone, m>n;

[0103] n represents the risk sensitivity adjustment parameter of high-pressure compaction area, m>n;

[0104] 3) When the optimal comprehensive compaction value C≥b obtained in step 4, the compaction area belongs to the high compaction area;

[0105] The optimal compaction warning model is expressed as:

[0106] g(C)=g(b)+k 3 (1-e -n(C-b) )

[0107] Where: k 3 represents the risk increase parameter;

[0108] g(b) represents the compaction warning value corresponding to the compaction degree dividing point b.

[0109] Preferably, in step 6, an optimal comprehensive compaction risk warning model is constructed based on the optimal comprehensive compaction degree model and the compaction warning model; the specific process is:

[0110] Step 6.1: Based on the optimal comprehensive compaction degree model and the compaction warning model, a comprehensive compaction risk warning model is constructed; the expression is:

[0111] f(C, g(C)) = α 1 (β B Y B +β Z Y Z +β S Y S )+α 2 g(C)

[0112] Where: α 1 , α 2 They represent the weights of the optimal comprehensive compaction degree model and the compaction warning model respectively;

[0113] β B Y B +β Z Y Z +β S Y S represents the optimal comprehensive compaction degree model;

[0114] g(C) represents the compaction warning value corresponding to the optimal comprehensive compaction value C;

[0115] f(C, g(C)) represents the comprehensive compaction risk warning value;

[0116] Step 62: Construct the error loss function L″ of the comprehensive compaction risk warning model; the specific process is:

[0117] The mean square error is used to measure the deviation between the predicted value and the true value:

[0118]

[0119] Where: D j ′ represents the real comprehensive compaction risk warning value of the jth sample; D j represents the predicted comprehensive compaction risk warning of the jth sample; N represents the number of training samples;

[0120] Step 6. Use the gradient descent method to optimize the weight of the comprehensive compaction risk warning model until the weight converges to obtain the optimal comprehensive compaction degree model and compaction warning model weight α 1 , α 2 ;

[0121] Based on the optimal comprehensive compaction degree model and compaction warning model weight α 1 , α 2 , obtain the optimal comprehensive compaction risk warning model;

[0122] The specific process is:

[0123] The gradient descent method is used to optimize the weights of the comprehensive compaction risk warning model. The weight iterative update formula is:

[0124]

[0125] Where: α represents the feature weight; α = α 1 , α 2 ;

[0126] α j represents the weight of the jth sample;

[0127] represents the learning rate;

[0128] Represents the loss function for weight α j The partial derivative of

[0129] Partial derivative calculation formula:

[0130]

[0131] Where: X j,i represents the i-th feature data of the j-th sample after standardization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0132] Until the weights converge, the optimal comprehensive compaction degree model and compaction warning model weight α are obtained 1 , α 2 ;

[0133] Based on the optimal comprehensive compaction degree model and compaction warning model weight α 1 , α 2 , and obtain the optimal comprehensive compaction risk warning model.

[0134] Preferably, in step seven, a feature data set to be tested is collected to obtain a risk warning level of the feature data set to be tested;

[0135] The specific process is:

[0136] Collecting a feature data set to be tested, and standardizing the collected feature data set to be tested to obtain a standardized feature data set to be tested;

[0137] The feature data set to be tested includes the number of repeated rolling times X 1 , harvester weight X 2 , Track hardness X 3 , Track tension X 4 , unloading frequency X 5 、Total turning time ratio X 6 、Differential steering time ratio X7 、Ratio of braking and steering time to X 8 、Ratio of time of turning in place X 9 ;

[0138] The standardized feature data set to be tested is input into the optimal compaction degree prediction model, and the optimal compaction degree prediction model outputs the surface soil compaction value Y predicted by the compaction degree prediction model B , the compaction value Y of the middle layer soil predicted by the compaction degree prediction model Z , the deep soil compaction value Y predicted by the compaction degree prediction model S ;

[0139] The surface soil compaction value Y predicted by the compaction degree prediction model B , the compaction value Y of the middle layer soil predicted by the compaction degree prediction model Z , the deep soil compaction value Y predicted by the compaction degree prediction model S An optimal comprehensive compaction degree model is input, and the optimal comprehensive compaction degree model outputs a comprehensive compaction degree value C;

[0140] The comprehensive compaction degree value C output by the optimal comprehensive compaction degree model is input into the optimal compaction warning model, and the optimal compaction warning model outputs a compaction warning value g(C);

[0141] The optimal comprehensive compaction degree model C and the compaction warning model g(C) are input into the optimal comprehensive compaction risk warning model, and the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C));

[0142] When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C)) < 2, the warning level is low risk;

[0143] When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value 2≤f(C, g(C))<4, the warning level is medium risk;

[0144] When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C))≥4, the warning level is high risk.

[0145] The beneficial effects of the present invention are:

[0146] The present invention takes the protection of the ecological environment of black soil as its core goal, and proposes an innovative intelligent soil compaction prediction and early warning method for the impact of corn combine harvester operation on multi-dimensional soil compaction. By introducing multi-feature data, including machine deadweight, track tension, track hardness, and steering time ratio, and comprehensively considering the mechanical operation characteristics and soil response characteristics, a soil compaction prediction model combining multi-feature optimization is constructed, and a real-time early warning mechanism is embedded. In terms of technical implementation, this method utilizes real-time data collection and analysis, combined with a multivariate linear regression model and a gradient optimization algorithm, to achieve dynamic monitoring and prediction of the degree of soil compaction during operation. According to the different levels of compaction, the system can provide graded early warnings and operation adjustment suggestions, thereby significantly reducing the compaction impact of mechanical operations on the soil.

[0147] This invention not only improves the operating efficiency and environmental performance of corn combine harvesters, but also plays an important role in protecting black soil resources and slowing down soil degradation. At the same time, this technology is applicable to various types of farmland machinery operation scenarios and has a wide range of agricultural application value. It provides scientific technical support for the development of precision agriculture in the future and injects new impetus into the sustainable utilization and ecological protection of black soil.

[0148] The present invention proposes a soil compaction prediction and early warning technology based on multi-feature optimization, aiming to solve the problem of soil compaction at different depths. This technology combines the multi-dimensional characteristics of mechanical operations with soil reactions, and provides an accurate compaction degree prediction and risk early warning method. By real-time monitoring of soil compaction, especially the compaction degree of soil at different depths, the system can dynamically warn and promptly remind agricultural machinery operators to take adjustment measures to reduce the negative impact of mechanical operations on soil structure, protect black soil resources, and improve the environmental protection performance and efficiency of agricultural mechanization operations.

[0149] The early warning mechanism of the present invention can not only monitor the degree of soil compaction in real time, but also adjust the mechanical settings according to changes in track characteristics. By monitoring the real-time data of track tension and hardness, the system can automatically issue adjustment suggestions when the track status has a greater impact on soil compaction. For example, when the track tension is too high, the system will prompt to adjust the track tightness, or when the track hardness is high, it is recommended to reduce the operating speed or change the operating path. This dynamic adjustment can effectively slow down the occurrence of soil compaction and reduce damage to black soil.

[0150] The track is the key part of agricultural machinery in contact with the soil, especially in corn combine harvesters, and its influence on soil compaction cannot be ignored. Traditional soil compaction prediction methods usually focus on the overall weight of the machinery and the operating frequency, while ignoring the characteristics of the track itself. The present invention introduces key parameters such as track hardness (Shore hardness) and track tension (Newton / meter), combined with the contact characteristics of the track and the soil, to construct a more accurate soil compaction prediction model, thereby improving the accuracy of the prediction and the timeliness of the early warning, and solving the problem that the existing soil compaction prediction methods all consider the soil indicators and the pressure of the machine itself, but do not consider the influence of dynamic indicators that the driver can control independently on soil compaction, resulting in low accuracy of soil compaction prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0151] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0152] Specific implementation method 1: The specific process of this implementation method is as follows:

[0153] Step 1: Collect feature data training set; the specific process is:

[0154] The feature data training set includes the number of repeated rolling times X 1 , harvester weight X 2 , Track hardness X 3 , Track tension X 4 , unloading frequency X 5 、Total turning time ratio X 6 、Differential steering time ratio X 7 、Ratio of braking and steering time to X 8 、Ratio of time of turning in place X 9 ;

[0155] The unit of repeated rolling times is times / square meter; the unit of harvester deadweight is tons; the unit of track hardness is Shore hardness; the unit of track tension is Newton / meter; the unit of grain unloading frequency is times / hour; the unit of total steering time proportion is %; the unit of differential steering time proportion is %; the unit of brake steering time proportion is %; the unit of in-situ steering time proportion is %;

[0156] Step 2: Standardize the feature data collected in step 1 to obtain a standardized feature data training set;

[0157] Step 3: Based on the standardized feature data training set, an optimal compaction degree prediction model is obtained;

[0158] Step 4: Based on the optimal compaction degree prediction model, construct an optimal comprehensive compaction degree model;

[0159] Step 5: Based on the optimal comprehensive compaction degree model, construct an optimal compaction warning model;

[0160] Step 6: Based on the optimal comprehensive compaction degree model and the compaction warning model, an optimal comprehensive compaction risk warning model is constructed;

[0161] Step 7: Collect the feature data set to be tested, and based on Steps 2 to 6, obtain the optimal comprehensive compaction degree model of the feature data set to be tested, and output the comprehensive compaction degree value and risk warning level;

[0162] Step 8. Output the comprehensive compaction degree value and warning level according to the optimal comprehensive compaction degree model, and push the comprehensive compaction degree value f(C, g(C)), the data of each feature in the feature data set to be tested, and the warning level on the screen in real time. When the medium-risk or high-risk warning level is reached, a prompt alarm is issued and corresponding suggestions are provided.

[0163] Specific implementation method 2: This implementation method is different from the specific implementation method 1 in that: in step 2, the feature data collected in step 1 is standardized to obtain a standardized feature data training set; the specific process is:

[0164] Each characteristic data collected in step 1 is standardized to eliminate the dimension difference and obtain each standardized characteristic data; the standardization formula is:

[0165] Each feature data collected in step 1 is standardized to obtain each standardized feature data; the standardization formula is:

[0166]

[0167] Where: X j,i represents the i-th feature data of the j-th sample after standardization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0168] X' ji is the i-th feature data of the j-th sample;

[0169] μ i is the mean; σ i is the standard deviation (each feature corresponds to a mean and standard deviation).

[0170] The other steps and parameters are the same as those in the first embodiment.

[0171] Specific implementation method three: This implementation method is different from specific implementation methods one or two in that: in step three, the optimal compaction degree prediction model is obtained based on the standardized feature data training set; the specific process is:

[0172] Step 31: Based on the standardized feature data, a compaction degree prediction model is constructed; the specific process is as follows:

[0173] The compaction degree prediction model adopts the multivariate linear regression model:

[0174] Y B =B 1 X 1 +B 2 X 2 +B 3 X 3 +B 4 X 4 +B 5 X 5 +B 6 X 6 +B 7 X 7 +B 8 X 8 +B 9 X 9 +b B

[0175] Y Z =Z 1 X 1 +Z 2 X 2 +Z 3 X 3 +Z 4 X 4 +Z 5 X 5 +Z 6 X 6 +Z 7 X 7 +Z 8 X 8 +Z 9 X 9 +b Z

[0176] Y S =S 1 X 1 +S 2 X 2 +S 3 X 3 +S 4 X 4 +S 5 X 5 +S 6 X 6 +S 7 X 7 +S 8 X 8 +S 9 X9 +b S

[0177] in:

[0178] Y B It represents the surface soil compaction value (target value) predicted by the compaction degree prediction model;

[0179] Y Z It represents the middle soil compaction value (target value) predicted by the compaction degree prediction model;

[0180] Y S It represents the deep soil compaction value (target value) predicted by the compaction degree prediction model;

[0181] B 1 , B 2 , B 3 , B 4 , B 5 , B 6 , B 7 , B 8 , B 9 They represent the compaction degree prediction model for surface characteristic data X 1 , X 2 , X 3 , X 4 , X 5 , X 6 , X 7 , X 8 , X 9 The feature weights are in Table 1;

[0182] Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , Z 6 , Z 7 , Z 8 , Z 9 They represent the compaction degree prediction model for the middle layer feature data X 1 , X 2 , X 3 , X 4 , X 5 , X 6 , X 7 , X 8 , X 9 The feature weights are in Table 1;

[0183] S 1 , S 2 , S 3 , S 4 , S 5, S 6 , S 7 , S 8 , S 9 They represent the compaction degree prediction model for deep feature data X 1 , X 2 , X 3 , X 4 , X 5 , X 6 , X 7 , X 8 , X 9 The feature weights are in Table 1;

[0184] b B The bias term (intercept) representing the surface characteristic data;

[0185] b Z The bias term (intercept) representing the middle-level feature data;

[0186] b S The bias term (intercept) representing the deep feature data;

[0187] Step 32: Construct the error loss function L of the compaction degree prediction model;

[0188] Step 3: Use the gradient descent method to optimize the feature weights of the compaction degree prediction model until the weights converge and obtain the optimal feature weights B 1 , B 2 , B 3 , B 4 , B 5 , B 6 , B 7 , B 8 , B 9 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , Z 6 , Z 7 , Z 8 , Z 9 , S 1 , S 2 , S 3 , S 4 , S 5 , S 6 , S 7 , S 8 and S 9 ;

[0189] The optimal compaction degree prediction model is obtained based on the optimal feature weights.

[0190] The other steps and parameters are the same as those in the first or second embodiment.

[0191] Specific implementation method 4: This implementation method is different from any one of the specific implementation methods 1 to 3 in that: the bias term b of the surface feature data B Initial setting b B =0; bias term b of middle-level feature data Z Initial setting b Z = 0; bias term b of deep feature data S Initial setting b S =0;

[0192] B 1 The initial value is set to B 1 =0.15; B 2 The initial value is set to B 2 =0.15; B 3 The initial value is set to B 3 =0.07; B 4 The initial value is set to B 4 =0.05; B 5 The initial value is set to B 5 =0.13; B 6 The initial value is set to B 6 =0.15; B 7 The initial value is set to B 7 =0.10; B 8 The initial value is set to B 8 =0.10; B 9 The initial value is set to B 9 =0.10;

[0193] Z 1 The initial value is set to Z 1 =0.20; Z 2 The initial value is set to Z 2 =0.17; Z 3 The initial value is set to Z 3 =0.05; Z 4 The initial value is set to Z 4 =0.05; Z 5 The initial value is set to Z 5 =0.13; Z 6 The initial value is set to Z 6 =0.15; Z 7 The initial value is set to Z 7 =0.10; Z 8 The initial value is set to Z 8 =0.10; Z 9 The initial value is set to Z 9=0.10;

[0194] S 1 The initial value is set to S 1 =0.25; S 2 The initial value is set to S 2 =0.17; S 3 The initial value is set to S 3 =0.05; S 4 The initial value is set to S 4 =0.05; S 5 The initial value is set to S 5 =0.13; S 6 The initial value is set to S 6 =0.15; S 7 The initial value is set to S 7 =0.10; S 8 The initial value is set to S 8 =0.10; S 9 The initial value is set to S 9 =0.10;

[0195] The 0cm≤surface soil depth≤10cm;

[0196] The 10cm<middle soil depth≤30cm;

[0197] The deep soil depth of 30cm < 50cm.

[0198] The other steps and parameters are the same as those in Specific Embodiments 1 to 3.

[0199] Specific implementation method 5: This implementation method is different from the specific implementation methods 1 to 4 in that: in step 32, the error loss function L of the compaction degree prediction model is constructed; the specific process is:

[0200] The mean square error (MSE) is used to measure the deviation between the predicted value and the true value:

[0201]

[0202] Where: Y j ′ represents the actual soil compaction value of the jth sample; Y j represents the predicted soil compaction value of the jth sample; N represents the number of training samples.

[0203] The other steps and parameters are the same as those in Specific Embodiments 1 to 4.

[0204] Specific implementation method 6: This implementation method is different from the specific implementation methods 1 to 5 in that: in step 33, the gradient descent method is used to optimize the feature weights of the compaction degree prediction model until the weights converge to obtain the optimal feature weights B 1 , B 2 , B 3 , B 4 , B 5 , B 6 , B 7 , B 8 , B 9 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , Z 6 , Z 7 , Z 8 , Z 9 , S 1 , S 2 , S 3 , S 4 , S 5 , S 6 , S 7 , S 8 and S 9 ;

[0205] Based on the optimal feature weight B 1 , B 2 , B 3 , B 4 , B 5 , B 6 , B 7 , B 8 , B 9 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , Z 6 , Z 7 , Z 8 , Z 9 , S 1 , S 2 , S 3 , S 4 , S 5 , S 6 , S 7 , S 8 and S 9 Obtain the optimal compaction degree prediction model;

[0206] The specific process is:

[0207] The gradient descent method is used to optimize the feature weights. The formula for iterative update of feature weights is:

[0208]

[0209] Where: W represents the feature weight; W = B 1 , B 2 , B 3 , B 4 , B 5 , B 6 , B 7 , B 8 , B 9 , Z 1 , Z 2 , Z 3 , Z 4 , Z 5 , Z 6 , Z 7 , Z 8 , Z 9 , S 1 , S 2 , S 3 , S 4 , S 5 , S 6 , S 7 , S 8 or S 9 ;

[0210] W j represents the feature weight of the jth sample;

[0211] α″ represents the learning rate, which controls the update step (set α = 0.01);

[0212] Represents the loss function on weight W j The partial derivative of

[0213] Partial derivative calculation formula:

[0214]

[0215] Where: X j,i represents the i-th feature data of the j-th sample after standardization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0216] For example, X j,1 represents the first characteristic data of the jth sample after standardization (number of repeated rolling (times / square meter)), X j,2 Represents the second characteristic data of the j-th sample after standardization (harvester's own weight (tons)).

[0217] The other steps and parameters are the same as those in Specific Implementations 1 to 5.

[0218] Specific implementation method 7: This implementation method is different from any one of the specific implementation methods 1 to 6 in that: in step 4, an optimal comprehensive compaction degree model is constructed based on the optimal compaction degree prediction model; the specific process is:

[0219] Step 41: Based on the optimal compaction degree prediction model, a comprehensive compaction degree model is constructed; the specific process is as follows:

[0220] C=β B Y B +β Z Y Z +β S Y S

[0221] in:

[0222] C represents the comprehensive compaction value;

[0223] β B It represents the weight of surface compaction degree, with the initial value set to 0.5, and the weight is subsequently optimized by gradient descent method;

[0224] β Z It represents the compaction degree weight of the middle layer, with the initial value set to 0.3, and the subsequent gradient descent method optimizes the weight;

[0225] β S It represents the weight of deep compaction degree. The initial value is set to 0.2, and the weight is subsequently optimized by gradient descent method.

[0226] Step 4.2: Construct the error loss function of the comprehensive compaction degree model The specific process is:

[0227] The mean square error (MSE) is used to measure the deviation between the predicted value and the true value:

[0228]

[0229] Where: C j ′ represents the actual soil comprehensive compaction value of the jth sample; C j represents the predicted soil comprehensive compaction value of the jth sample; N represents the number of training samples;

[0230] Step 4.3: Use the gradient descent method to optimize the weight of the comprehensive compaction degree model until the weight converges and obtains the optimal feature weight β B , β Z and β S ;

[0231] Based on the optimal feature weight β B , βZ and β S Obtain the optimal comprehensive compaction degree model;

[0232] The specific process is:

[0233] The gradient descent method is used to optimize the feature weights. The formula for iterative update of feature weights is:

[0234]

[0235] Where: β represents the feature weight; β = β B , β Z or β S ;

[0236] β j represents the weight of the jth sample;

[0237] Represents the learning rate and controls the update step (set );

[0238] Represents the loss function for weight β j The partial derivative of

[0239] Partial derivative calculation formula:

[0240]

[0241] Where: X j,i represents the i-th feature data of the j-th sample after standardization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0242] Until the weights converge, the optimal feature weight β is obtained B , β Z and β S ;

[0243] Based on the optimal feature weight β B , β Z and β S Obtain the optimal comprehensive compaction degree model.

[0244] The other steps and parameters are the same as those in Specific Embodiments 1 to 6.

[0245] Specific implementation eight: This implementation differs from specific implementations one to seven in that: in step five, an optimal compaction warning model is constructed based on the optimal comprehensive compaction degree model; the specific process is:

[0246] 1) When the optimal comprehensive compaction value obtained in step 4 is 0<C<a, the compaction area belongs to the low compaction area;

[0247] The risk grows slowly, and the compaction warning model is represented by a linear or nonlinear incremental model;

[0248] The optimal compaction warning model is expressed as:

[0249] g(C)=k 1 C p

[0250] Where: k 1 It represents the risk increase parameter of low compaction area, which controls the growth rate of risk at each stage and is set to 1;

[0251] p represents the degree of smoothness of risk change in the low compaction area, usually p>1, and is set to 2;

[0252] g(C) represents the compaction warning value corresponding to the optimal comprehensive compaction value C;

[0253] a represents the compaction degree cutoff point, which is set to 0.4;

[0254] 2) When the optimal comprehensive compaction value a≤C<b obtained in step 4, the compaction area belongs to the medium compaction area;

[0255] The risk increases sharply, and the compaction warning model can be represented by an exponential or S-shaped curve;

[0256] The optimal compaction warning model is expressed as:

[0257] g(C)=g(a)+k 2 (1-e -m(C-a) )

[0258] Among them: g(a) represents the compaction warning value corresponding to the compaction degree cutoff point a;

[0259] k 2 It represents the risk increase parameter of the medium compaction area, which controls the growth rate of risk at each stage and is set to 5;

[0260] b represents the compaction degree cutoff point, which is set to 0.7;

[0261] m represents the risk sensitivity adjustment parameter of the medium compaction area. m>n ensures that the medium compaction area has a higher sensitivity and is set to 10;

[0262] n represents the risk sensitivity adjustment parameter of the high compaction area. m>n ensures that the medium compaction area has a higher sensitivity and is set to 2;

[0263] 3) When the optimal comprehensive compaction value C≥b obtained in step 4, the compaction area belongs to the high compaction area;

[0264] Risks level off or saturate.

[0265] The optimal compaction warning model is expressed as:

[0266] g(C)=g(b)+k 3 (1-e -n(C-b) )

[0267] Where: k 3 It represents the risk increase parameter, which controls the growth rate of risk at each stage and is set to 2;

[0268] g(b) represents the compaction warning value corresponding to the compaction degree dividing point b.

[0269] The other steps and parameters are the same as those in Specific Embodiments 1 to 7.

[0270] Specific implementation method 9: This implementation method is different from any one of specific implementation methods 1 to 8 in that: in step 6, an optimal comprehensive compaction risk warning model is constructed based on the optimal comprehensive compaction degree model and the compaction warning model; the specific process is:

[0271] Step 6.1: Based on the optimal comprehensive compaction degree model and the compaction warning model, a comprehensive compaction risk warning model is constructed; the expression is:

[0272] f(C, g(C)) = α 1 (β B Y B +β Z Y Z +β S Y S )+α 2 g(C)

[0273] Where: α 1 , α 2 They represent the weights of the optimal comprehensive compaction degree model and the compaction warning model, which are set to 0.8 and 0.2 respectively;

[0274] β B Y B +β Z Y Z +β S Y S represents the optimal comprehensive compaction degree model;

[0275] g(C) represents the compaction warning value corresponding to the optimal comprehensive compaction value C;

[0276] f(C, g(C)) represents the comprehensive compaction risk warning value;

[0277] Step 62: Construct the error loss function L″ of the comprehensive compaction risk warning model; the specific process is:

[0278] The mean square error (MSE) is used to measure the deviation between the predicted value and the true value:

[0279]

[0280] Where: D j ′ represents the real comprehensive compaction risk warning value of the jth sample; D j represents the predicted comprehensive compaction risk warning of the jth sample; N represents the number of training samples;

[0281] Step 6. Use the gradient descent method to optimize the weight of the comprehensive compaction risk warning model until the weight converges to obtain the optimal comprehensive compaction degree model and compaction warning model weight α 1 , α 2 ;

[0282] Based on the optimal comprehensive compaction degree model and compaction warning model weight α 1 , α 2 , obtain the optimal comprehensive compaction risk warning model;

[0283] The specific process is:

[0284] The gradient descent method is used to optimize the weights of the comprehensive compaction risk warning model. The weight iterative update formula is:

[0285]

[0286] Where: α represents the feature weight; α = α 1 , α 2 ;

[0287] α j represents the weight of the jth sample;

[0288] Represents the learning rate and controls the update step (set );

[0289] Represents the loss function for weight α j The partial derivative of

[0290] Partial derivative calculation formula:

[0291]

[0292] Where: X j,i represents the i-th feature data of the j-th sample after standardization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0293] Until the weights converge, the optimal comprehensive compaction degree model and compaction warning model weight α are obtained 1 , α 2 ;

[0294] Based on the optimal comprehensive compaction degree model and compaction warning model weight α 1 , α 2 , and obtain the optimal comprehensive compaction risk warning model.

[0295] The other steps and parameters are the same as those in Specific Embodiments 1 to 8.

[0296] Specific implementation method ten: This implementation method is different from any one of specific implementation methods one to nine in that: in step seven, a feature data set to be tested is collected to obtain a risk warning level of the feature data set to be tested;

[0297] The specific process is:

[0298] Collecting a feature data set to be tested, and standardizing the collected feature data set to be tested to obtain a standardized feature data set to be tested;

[0299] The feature data set to be tested includes the number of repeated rolling times X 1 , harvester weight X 2 , Track hardness X 3 , Track tension X 4 , unloading frequency X 5 、Total turning time ratio X 6 、Differential steering time ratio X 7 、Ratio of braking and steering time to X 8 、Ratio of time of turning in place X 9 ;

[0300] The standardized feature data set to be tested is input into the optimal compaction degree prediction model, and the optimal compaction degree prediction model outputs the surface soil compaction value Y predicted by the compaction degree prediction model B , the compaction value Y of the middle layer soil predicted by the compaction degree prediction model Z , the deep soil compaction value Y predicted by the compaction degree prediction model S ;

[0301] The surface soil compaction value Y predicted by the compaction degree prediction model B , the compaction value Y of the middle layer soil predicted by the compaction degree prediction model Z , the deep soil compaction value Y predicted by the compaction degree prediction model S An optimal comprehensive compaction degree model is input, and the optimal comprehensive compaction degree model outputs a comprehensive compaction degree value C;

[0302] The comprehensive compaction degree value C output by the optimal comprehensive compaction degree model is input into the optimal compaction warning model, and the optimal compaction warning model outputs a compaction warning value g(C);

[0303] The optimal comprehensive compaction degree model C and the compaction warning model g(C) are input into the optimal comprehensive compaction risk warning model, and the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C));

[0304] When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C)) < 2, the warning level is low risk;

[0305] When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value 2≤f(C, g(C))<4, the warning level is medium risk;

[0306] When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C))≥4, the warning level is high risk.

[0307] Enter the comprehensive compaction degree prediction value and determine the compaction warning level based on the prediction value:

[0308]

[0309] The other steps and parameters are the same as those in Specific Embodiments 1 to 9.

[0310] Table 1 Model input features and corresponding initial weights

[0311]

[0312] The present invention may also have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art may make various corresponding changes and modifications based on the present invention, but these corresponding changes and modifications should all fall within the scope of protection of the claims attached to the present invention.

Claims

1. A method for determining soil compaction degree of a corn harvester based on real-time parameters of the machine, characterized in that: The specific process of the method is: Step 1: Collect feature data training set; The specific process is: The feature data training set includes the number of repeated rolling times X1, harvester weight X2, track hardness X3, track tension X4, grain unloading frequency X5, total steering time proportion X6, differential steering time proportion X7, braking steering time proportion X8, and in-situ steering time proportion X9; Step 2: Standardize the feature data collected in step 1 to obtain a standardized feature data training set; Step 3: Based on the standardized feature data training set, an optimal compaction degree prediction model is obtained; Step 4: Based on the optimal compaction degree prediction model, construct an optimal comprehensive compaction degree model; Step 5: Based on the optimal comprehensive compaction degree model, construct an optimal compaction warning model; Step 6: Based on the optimal comprehensive compaction degree model and the compaction warning model, an optimal comprehensive compaction risk warning model is constructed; Step 7: Collect the feature data set to be tested, and based on Steps 2 to 6, obtain the optimal comprehensive compaction degree model of the feature data set to be tested, and output the comprehensive compaction degree value and risk warning level; Step 8. Output the comprehensive compaction degree value and warning level according to the optimal comprehensive compaction degree model, and push the comprehensive compaction degree value f(C, g(C)), the data of each feature in the feature data set to be tested, and the warning level on the screen in real time. When the medium risk or high risk warning level is reached, a prompt alarm is issued.

2. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 1, characterized in that: In the step 2, the feature data collected in the step 1 is standardized to obtain a standardized feature data training set; the specific process is: Each feature data collected in step 1 is standardized to obtain each standardized feature data; the standardization formula is: Where: X j,i represents the i-th feature data of the j-th sample after standardization, j = 1, 2, ..., N, i = 1, 2, ..., 9; X' ji is the i-th feature data of the j-th sample; μ i is the mean; σ i is the standard deviation.

3. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 2, characterized in that: In step 3, based on the standardized feature data training set, the optimal compaction degree prediction model is obtained; the specific process is: Step 31: Based on the standardized feature data, a compaction degree prediction model is constructed; the specific process is as follows: The compaction degree prediction model adopts the multivariate linear regression model: Y B =B1X1+B2X2+B3X3+B4X4+B5X5+B6X6+B7X7+B8X8+B9X9+b B Y Z =Z1X1+Z2X2+Z3X3+Z4X4+Z5X5+Z6X6+Z7X7+Z8X8+Z9X9+b Z Y S =S1X1+S2X2+S3X3+S4X4+S5X5+S6X6+S7X7+S8X8+S9X9+b S in: Y B It represents the surface soil compaction value predicted by the compaction degree prediction model; Y Z It represents the compaction value of the middle layer soil predicted by the compaction degree prediction model; Y S It represents the deep soil compaction value predicted by the compaction degree prediction model; B1, B2, B3, B4, B5, B6, B7, B8, and B9 represent the feature weights of the compaction degree prediction model for the surface feature data X1, X2, X3, X4, X5, X6, X7, X8, and X9, respectively; Z1, Z2, Z3, Z4, Z5, Z6, Z7, Z8, and Z9 represent the feature weights of the compaction degree prediction model for the middle-layer feature data X1, X2, X3, X4, X5, X6, X7, X8, and X9, respectively; S1, S2, S3, S4, S5, S6, S7, S8, and S9 represent the feature weights of the compaction degree prediction model for the deep feature data X1, X2, X3, X4, X5, X6, X7, X8, and X9, respectively; b B A bias term representing surface feature data; b Z A bias term representing the middle-level feature data; b S A bias term representing deep feature data; Step 32: Construct the error loss function L of the compaction degree prediction model; Step 3. Use the gradient descent method to optimize the feature weights of the compaction degree prediction model until the weights converge and the optimal feature weights are obtained; The optimal compaction degree prediction model is obtained based on the optimal feature weights.

4. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 3, characterized in that: The bias term b of the surface feature data B Initial setting b B =0; The bias term b of the middle-level feature data Z Initial setting b Z =0; The bias term b of the deep feature data S Initial setting b S =0; The initial value of B1 is set to B1=0.15; the initial value of B2 is set to B2=0.15; the initial value of B3 is set to B3=0.07; the initial value of B4 is set to B4=0.05; the initial value of B5 is set to B5=0.13; the initial value of B6 is set to B6=0.15; the initial value of B7 is set to B7=0.10; the initial value of B8 is set to B8=0.10; the initial value of B9 is set to B9=0.10; The initial value of Z1 is set to Z1=0.20; the initial value of Z2 is set to Z2=0.17; the initial value of Z3 is set to Z3=0.05; the initial value of Z4 is set to Z4=0.05; the initial value of Z5 is set to Z5=0.13; the initial value of Z6 is set to Z6=0.15; the initial value of Z7 is set to Z7=0.10; the initial value of Z8 is set to Z8=0.10; the initial value of Z9 is set to Z9=0.10; The initial value of S1 is set to S1=0.25; the initial value of S2 is set to S2=0.17; the initial value of S3 is set to S3=0.05; the initial value of S4 is set to S4=0.05; the initial value of S5 is set to S5=0.13; the initial value of S6 is set to S6=0.15; the initial value of S7 is set to S7=0.10; the initial value of S8 is set to S8=0.10; the initial value of S9 is set to S9=0.10; The 0cm≤surface soil depth≤10cm; The 10cm<middle soil depth≤30cm; The deep soil depth of 30cm < 50cm.

5. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 4, characterized in that: In step 32, the error loss function L of the compaction degree prediction model is constructed; the specific process is: The mean square error is used to measure the deviation between the predicted value and the true value: Where: Y j ′ represents the actual soil compaction value of the jth sample; Y j represents the predicted soil compaction value of the jth sample; N represents the number of training samples.

6. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 5, characterized in that: In the step 33, the gradient descent method is used to optimize the feature weights of the compaction degree prediction model until the weights converge to obtain the optimal feature weights; The optimal compaction degree prediction model is obtained based on the optimal feature weights; The specific process is: The gradient descent method is used to optimize the feature weights. The formula for iterative update of feature weights is: Where: W represents feature weight; W = B1, B2, B3, B4, B5, B6, B7, B8, B9, Z1, Z2, Z3, Z4, Z5, Z6, Z7, Z8, Z9, S1, S2, S3, S4, S5, S6, S7, S8 or S9; W j represents the feature weight of the jth sample; α″ represents the learning rate; Represents the loss function on weight W j The partial derivative of Partial derivative calculation formula: Where: X j,i Represents the i-th feature data of the j-th sample after standardization, j=1,2,…,N, i=1,2,…,9.

7. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 6, characterized in that: In step 4, an optimal comprehensive compaction degree model is constructed based on the optimal compaction degree prediction model; the specific process is: Step 41: Based on the optimal compaction degree prediction model, a comprehensive compaction degree model is constructed; the specific process is as follows: C=β B Y B +b Z Y Z +b S Y S in: C represents the comprehensive compaction value; β B It represents the weight of surface compaction degree, and the initial value is set to 0.5; β Z It represents the compaction degree weight of the middle layer, and its initial value is set to 0.3; β S Represents the weight of deep compaction degree, with the initial value set to 0.

2. Step 4.2: Construct the error loss function of the comprehensive compaction degree model The specific process is: The mean square error is used to measure the deviation between the predicted value and the true value: Where: C j ′ represents the actual soil comprehensive compaction value of the jth sample; C j represents the predicted soil comprehensive compaction value of the jth sample; N represents the number of training samples; Step 4.3: Use the gradient descent method to optimize the weight of the comprehensive compaction degree model until the weight converges and obtains the optimal feature weight β B , β Z and β S ; Based on the optimal feature weight β B , β Z and β S Obtain the optimal comprehensive compaction degree model; The specific process is: The gradient descent method is used to optimize the feature weights. The formula for iterative update of feature weights is: Where: β represents the feature weight; β = β B , β Z or β S ; β j represents the weight of the jth sample; represents the learning rate; Represents the loss function for weight β j The partial derivative of Partial derivative calculation formula: Where: X j,i Represents the i-th feature data of the j-th sample after standardization, j=1,2,…,N, i=1,2,…,9. Until the weights converge, the optimal feature weight β is obtained B , β Z and β S ; Based on the optimal feature weight β B , β Z and β S Obtain the optimal comprehensive compaction degree model.

8. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 7, characterized in that: In step 5, an optimal compaction warning model is constructed based on the optimal comprehensive compaction degree model; the specific process is: 1) When the optimal comprehensive compaction value obtained in step 4 is 0<C<a, the compaction area belongs to the low compaction area; The optimal compaction warning model is expressed as: g(C)=k1C p Where: k1 represents the risk increase parameter of low compaction area, which is set to 1; p represents the degree of smoothness of risk change in the controlled low compaction area; g(C) represents the compaction warning value corresponding to the optimal comprehensive compaction value C; a represents the compaction degree dividing point; 2) When the optimal comprehensive compaction value a≤C<b obtained in step 4, the compaction area belongs to the medium compaction area; The optimal compaction warning model is expressed as: g(C)=g(a)+k2(1-e -m(C-a) ) Where: g(a) represents the compaction warning value corresponding to the compaction degree cutoff point a; k2 represents the risk increase parameter of the medium compaction area; b represents the compaction degree dividing point; m represents the risk sensitivity adjustment parameter of the medium compaction zone, m>n; n represents the risk sensitivity adjustment parameter of high-pressure compaction area, m>n; 3) When the optimal comprehensive compaction value C ≥ b obtained in step 4, the compaction area belongs to the high compaction area; The optimal compaction warning model is expressed as: g(C)=g(b)+k3(1-e -n(C-b) ) Among them: k3 represents the risk increase parameter; g(b) represents the compaction warning value corresponding to the compaction degree dividing point b.

9. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 8, characterized in that: In the step 6, based on the optimal comprehensive compaction degree model and the compaction warning model, an optimal comprehensive compaction risk warning model is constructed; the specific process is: Step 6.1: Based on the optimal comprehensive compaction degree model and the compaction warning model, a comprehensive compaction risk warning model is constructed; the expression is: f(C,g(C))=α1(β B Y B +b Z Y Z +b S Y S )+α2g(C) Among them: α1 and α2 represent the weights of the optimal comprehensive compaction degree model and the compaction warning model respectively; β B Y B +β Z Y Z +β S Y S represents the optimal comprehensive compaction degree model; g(C) represents the compaction warning value corresponding to the optimal comprehensive compaction value C; f(C, g(C)) represents the comprehensive compaction risk warning value; Step 62: Construct the error loss function L″ of the comprehensive compaction risk warning model; the specific process is: The mean square error is used to measure the deviation between the predicted value and the true value: Where: D j ′ represents the real comprehensive compaction risk warning value of the jth sample; D j represents the predicted comprehensive compaction risk warning of the jth sample; N represents the number of training samples; Step 6.

3. Use the gradient descent method to optimize the weights of the comprehensive compaction risk warning model until the weights converge, and obtain the optimal comprehensive compaction degree model and compaction warning model weights α1 and α2; Based on the optimal comprehensive compaction degree model and the compaction warning model weights α1 and α2, the optimal comprehensive compaction risk warning model is obtained; The specific process is: The gradient descent method is used to optimize the weights of the comprehensive compaction risk warning model. The weight iterative update formula is: Where: α represents the feature weight; α = α1, α2; α j represents the weight of the jth sample; represents the learning rate; Represents the loss function for weight α j The partial derivative of Partial derivative calculation formula: Where: X j,i represents the i-th feature data of the j-th sample after standardization, j = 1, 2, ..., N, i = 1, 2, ..., 9; Until the weights converge, the optimal comprehensive compaction degree model and compaction warning model weights α1 and α2 are obtained; Based on the optimal comprehensive compaction degree model and the compaction warning model weights α1 and α2, the optimal comprehensive compaction risk warning model is obtained.

10. The method for determining soil compaction degree of a corn harvester based on real-time parameters of a machine tool according to claim 9, characterized in that: In the step 7, the feature data set to be tested is collected to obtain the risk warning level of the feature data set to be tested; The specific process is: Collecting a feature data set to be tested, and standardizing the collected feature data set to be tested to obtain a standardized feature data set to be tested; The feature data set to be measured includes the number of repeated rolling times X1, the weight of the harvester X2, the hardness of the track X3, the tension of the track X4, the grain unloading frequency X5, the total steering time proportion X6, the differential steering time proportion X7, the braking steering time proportion X8, and the in-situ steering time proportion X9; The standardized feature data set to be tested is input into the optimal compaction degree prediction model, and the optimal compaction degree prediction model outputs the surface soil compaction value Y predicted by the compaction degree prediction model B , the compaction value Y of the middle layer soil predicted by the compaction degree prediction model Z , the deep soil compaction value Y predicted by the compaction degree prediction model S ; The surface soil compaction value Y predicted by the compaction degree prediction model B , the compaction value Y of the middle layer soil predicted by the compaction degree prediction model Z , the deep soil compaction value Y predicted by the compaction degree prediction model S An optimal comprehensive compaction degree model is input, and the optimal comprehensive compaction degree model outputs a comprehensive compaction degree value C; The comprehensive compaction degree value C output by the optimal comprehensive compaction degree model is input into the optimal compaction warning model, and the optimal compaction warning model outputs a compaction warning value g(C); The optimal comprehensive compaction degree model C and the compaction warning model g(C) are input into the optimal comprehensive compaction risk warning model, and the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C)); When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C)) < 2, the warning level is low risk; When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value 2≤f(C, g(C))<4, the warning level is medium risk; When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value f(C, g(C))≥4, the warning level is high risk.

Images