Soil compaction determination method for corn harvester operation based on real-time parameters of the machine

Abstract

A method for determining soil compaction during corn harvester operation based on real-time parameters of the machine belongs to the technical field of agricultural mechanization and intelligent equipment, and specifically relates to a method for determining soil compaction. 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 in soil compaction prediction. The process is: 1. Collect feature data training set; 2. Obtain standardized feature data training set; 3. Obtain optimal compaction degree prediction model; 4. Construct optimal comprehensive compaction degree model; 5. Construct optimal compaction warning model; 6. Construct optimal comprehensive compaction risk warning model; 7. Collect the feature data set to be tested, obtain the optimal comprehensive compaction degree model of the feature data set to be tested, output the comprehensive compaction degree value and risk warning level; 8. Push in real time on the screen.

Description

Technical Field

[0001] The present invention belongs to the technical field of agricultural mechanization and intelligent equipment, and particularly relates to a method for determining soil compaction. 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 during mechanized operations, 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, there is an urgent need 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 broad agricultural application value. Summary of the Invention

[0003] The purpose of this invention is to solve the problem that the existing soil compaction prediction methods only 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. A method for determining the soil compaction degree of a corn harvester based on real-time parameters of the machine is proposed.

[0004] A method for determining soil compaction during corn harvester operation based on real-time parameters of the machine 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 (X1), harvester weight (X2), track hardness (X3), track tension (X4), grain unloading frequency (X5), total steering time (X6), differential steering time (X7), braking steering time (X8), and stationary steering time (X9).

[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, obtain the optimal compaction degree prediction model;

[0009] Step 4: Based on the optimal compaction degree prediction model, construct the 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 output the comprehensive compaction degree value and risk warning level based on the optimal comprehensive compaction degree model of the feature data set to be tested obtained from Steps 2 to 6;

[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 normalized to obtain each normalized feature data; the normalization formula is:

[0016]

[0017] Where: X j,i represents the i-th feature data of the j-th sample after normalization, 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 3.1: Based on the standardized feature data, build a compaction degree prediction model; the specific process is as follows:

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

[0023] Y B =B1X1+B2X2+B3X3+B4X4+B5X5+B6X6+B7X7+B8X8+B9X9+b B

[0024] Y Z=Z1X1+Z2X2+Z3X3+Z4X4+Z5X5+Z6X6+Z7X7+Z8X8+Z9X9+b Z

[0025] Y S =S1X1+S2X2+S3X3+S4X4+S5X5+S6X6+S7X7+S8X8+S9X9+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 middle soil compaction value 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] 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;

[0031] 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;

[0032] 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;

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

[0034] b Z Represents the bias term of the middle-level feature data;

[0035] b S 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] 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;

[0041] 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;

[0042] 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;

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

[0044] The depth of the middle soil layer is 10 cm < 30 cm;

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

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

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

[0048]

[0049] Where: Y j ′ represents the true 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 using a gradient descent method until the weights converge to obtain the optimal feature weights;

[0051] Obtaining the optimal compaction degree prediction model 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 = 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;

[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 normalization, 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 4.1: Based on the optimal compaction degree prediction model, construct a comprehensive compaction degree model; 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 Indicates the weight of surface compaction degree, with the initial value set to 0.5;

[0068] β Z Indicates the compaction degree weight of the middle layer, with the initial value set to 0.3;

[0069] β S Indicates 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 squared 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 on the 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 normalization, j = 1, 2, ..., N, i = 1, 2, ..., 9.

[0086] Until the weight converges, 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)=k1C p

[0092] Where: k1 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)+k2(1-e -m(C-a) )

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

[0100] k2 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 obtained in step 4 is greater than or equal to b, the compaction area belongs to the high compaction area;

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

[0106] g(C)=g(b)+k3(1-e -n(C-b) )

[0107] Among them: k3 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)) = α1(β B Y B +β Z Y Z +β S Y S )+α2g(C)

[0112] Among them: α1 and α2 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 squared 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 6.3: 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;

[0121] 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;

[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 = α1, α2;

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

[0127] represents the learning rate;

[0128] Represents the loss function on the 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 normalization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0132] Until the weights converge, the optimal comprehensive compaction degree model and compaction warning model weights α1 and α2 are obtained;

[0133] 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.

[0134] Preferably, in step seven, the feature data set to be tested is collected to obtain the 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 measured includes the number of repeated rolling times X1, the weight of the harvester X2, the track hardness X3, the track tension X4, the grain unloading frequency X5, the total steering time ratio X6, the differential steering time ratio X7, the braking steering time ratio X8, and the in-situ steering time ratio X9;

[0138] The standardized feature data set to be measured 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 middle soil compaction value Y predicted by the compaction degree prediction model Z , 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 middle soil compaction value Y predicted by the compaction degree prediction model Z , deep soil compaction value Y predicted by the compaction degree prediction model S Input the optimal comprehensive compaction degree model, and the optimal comprehensive compaction degree model outputs the 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 black soil ecological environment as its core goal, and proposes an innovative intelligent soil compaction prediction and early warning method for the impact of multi-dimensional soil compaction during the operation of corn combine harvesters. By introducing multi-feature data, including machine weight, track tension, track hardness, and steering time ratio, and comprehensively considering the mechanical operation characteristics and soil response characteristics, a soil compaction prediction model combined with multi-feature optimization is constructed, and a real-time early warning mechanism is embedded. In terms of technical implementation, this method uses 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 the operation process. According to the different levels of compaction, the system can provide graded early warnings and operation adjustment suggestions, thereby significantly reducing the impact of mechanical operations on soil compaction.

[0147] This invention not only improves the efficiency and environmental performance of corn combine harvesters, but also plays a significant role in protecting black soil resources and mitigating soil degradation. Furthermore, this technology is applicable to a wide range of agricultural applications, providing scientific and technological support for the future development of precision agriculture and injecting new impetus into the sustainable utilization and ecological protection of black soil.

[0148] This paper proposes a soil compaction prediction and early warning technology based on multi-feature optimization, aiming to address the problem of soil compaction at different depths. This technology combines the multi-dimensional characteristics of mechanical operations with soil reactions to provide an accurate method for predicting the degree of compaction and providing risk warnings. By monitoring soil compaction in real time, particularly at different depths, the system can provide dynamic early warnings and promptly alert agricultural machinery operators to take corrective measures, reducing the negative impact of mechanical operations on soil structure, protecting black soil resources, and improving the environmental performance and efficiency of agricultural mechanization.

[0149] The early warning mechanism of this invention not only monitors soil compaction in real time but also adjusts mechanical settings based on changes in track characteristics. By monitoring real-time data on track tension and hardness, the system automatically issues adjustment recommendations when the track condition significantly impacts soil compaction. For example, if track tension is too high, the system will prompt adjustments to the track tension. Alternatively, if track hardness is high, it may recommend reducing operating speed or changing the operating path. This dynamic adjustment can effectively mitigate 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 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. 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:

[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 (X1), harvester weight (X2), track hardness (X3), track tension (X4), grain unloading frequency (X5), total steering time (X6), differential steering time (X7), braking steering time (X8), and stationary steering time (X9).

[0155] The unit for repeated rolling is times / square meter; the unit for harvester deadweight is tons; the unit for track hardness is Shore hardness; the unit for track tension is Newton / meter; the unit for grain unloading frequency is times / hour; the unit for the proportion of total steering time is %; the unit for the proportion of differential steering time is %; the unit for the proportion of braking steering time is %; the unit for the proportion of stationary steering time 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, obtain the optimal compaction degree prediction model;

[0158] Step 4: Based on the optimal compaction degree prediction model, construct the 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 output the comprehensive compaction degree value and risk warning level based on the optimal comprehensive compaction degree model of the feature data set to be tested obtained from Steps 2 to 6;

[0162] Step 8. Output the comprehensive compaction degree value and warning level based on 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 embodiment 2: This embodiment differs from specific embodiment 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 feature data collected in step 1 is standardized to eliminate dimensional differences and obtain each standardized feature data; the standardization formula is:

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

[0166]

[0167] Where: X j,i represents the i-th feature data of the j-th sample after normalization, 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] Other steps and parameters are the same as those in the first embodiment.

[0171] Specific embodiment three: This embodiment differs from specific embodiment 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 3.1: Based on the standardized feature data, build a compaction degree prediction model; the specific process is as follows:

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

[0174] Y B=B1X1+B2X2+B3X3+B4X4+B5X5+B6X6+B7X7+B8X8+B9X9+b B

[0175] Y Z =Z1X1+Z2X2+Z3X3+Z4X4+Z5X5+Z6X6+Z7X7+Z8X8+Z9X9+b Z

[0176] Y S =S1X1+S2X2+S3X3+S4X4+S5X5+S6X6+S7X7+S8X8+S9X9+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] 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, in Table 1;

[0182] 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, in Table 1;

[0183] 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, 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 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 and S9;

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

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

[0191] Specific embodiment 4: This embodiment differs from any one of specific embodiments 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] 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;

[0193] 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;

[0194] 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;

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

[0196] The depth of the middle soil layer is 10 cm < 30 cm;

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

[0198] The other steps and parameters are the same as those in the first to third embodiments.

[0199] Specific embodiment 5: This embodiment differs from specific embodiments 1 to 4 in that the error loss function L of the compaction degree prediction model is constructed in step 32; the specific process is:

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

[0201]

[0202] Where: Y j ′ represents the true 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 the first to fourth embodiments.

[0204] Specific embodiment six: This embodiment differs from specific embodiments one to five in that: in step three, 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 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 and S9;

[0205] The optimal compaction degree prediction model is obtained based on the optimal feature weights 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 and S9;

[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 = 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;

[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 normalization, 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 the first to fifth embodiments.

[0218] Specific embodiment 7: This embodiment differs from any one of specific embodiments 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 4.1: Based on the optimal compaction degree prediction model, construct a comprehensive compaction degree model; 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 subsequent gradient descent method optimizes the weight;

[0224] β Z It represents the weight of the compaction degree of the middle layer. The initial value is set to 0.3, and the weight is optimized by the gradient descent method.

[0225] β S Indicates 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 squared 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 , β Z 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, controls the update step (set );

[0238] Represents the loss function on the 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 normalization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0242] Until the weight converges, 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 the first to sixth embodiments.

[0245] Specific embodiment eight: This embodiment differs from any one of specific embodiments 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 as follows:

[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 increases slowly, and the compaction warning model is represented by a linear or nonlinear increasing model;

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

[0249] g(C)=k1C p

[0250] Among them: k1 represents the risk increase parameter of low compaction area, which controls the growth rate of risk in 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)+k2(1-e -m(C-a) )

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

[0259] k2 represents the risk increase parameter of the medium compaction zone, 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 for 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 for high compaction areas. m>n ensures that the sensitivity of medium compaction areas is higher and is set to 2.

[0263] 3) When the optimal comprehensive compaction value C obtained in step 4 is greater than or equal to b, the compaction area belongs to the high compaction area;

[0264] Risks level off or become saturated.

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

[0266] g(C)=g(b)+k3(1-e -n(C-b) )

[0267] Among them: k3 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 the first to seventh embodiments.

[0270] Specific embodiment 9: This embodiment differs from any one of specific embodiments 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)) = α1(β B Y B +β Z Y Z +β S Y S )+α2g(C)

[0273] Where: α1 and α2 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 squared 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 6.3: 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;

[0282] 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;

[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 = α1, α2;

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

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

[0289] Represents the loss function on the 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 normalization, j = 1, 2, ..., N, i = 1, 2, ..., 9;

[0293] Until the weights converge, the optimal comprehensive compaction degree model and compaction warning model weights α1 and α2 are obtained;

[0294] 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.

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

[0296] Specific embodiment ten: This embodiment differs from any one of specific embodiments 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 measured includes the number of repeated rolling times X1, the weight of the harvester X2, the track hardness X3, the track tension X4, the grain unloading frequency X5, the total steering time ratio X6, the differential steering time ratio X7, the braking steering time ratio X8, and the in-situ steering time ratio X9;

[0300] The standardized feature data set to be measured 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 middle soil compaction value Y predicted by the compaction degree prediction modelZ , 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 middle soil compaction value Y predicted by the compaction degree prediction model Z , deep soil compaction value Y predicted by the compaction degree prediction model S Input the optimal comprehensive compaction degree model, and the optimal comprehensive compaction degree model outputs the 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 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 during corn harvester operation based on real-time parameters of the machine, characterized by: 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 crushing , Harvester weight , Track hardness , track tension , unloading frequency , Total turning time ratio , Differential steering time ratio , Braking and steering time ratio , the proportion of time spent turning in place ; 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, obtain the optimal compaction degree prediction model; the specific process is as follows: Step 3.1: Based on the standardized feature data, a compaction degree prediction model is constructed; The compaction degree prediction model is used to predict the surface soil compaction value , Middle soil compaction value , deep soil compaction value ; Step 3.2: Construct the error loss function 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; Obtaining the optimal compaction degree prediction model based on the optimal feature weights; Step 4: Based on the optimal compaction degree prediction model obtained in step 3, an optimal comprehensive compaction degree model is constructed; The optimal comprehensive compaction degree model expression is: in: represents the optimal comprehensive compaction value; represents the weight of the optimal surface compaction degree; represents the weight of the optimal middle layer compaction degree; represents the weight of the optimal deep compaction degree; Step 5: Based on the optimal comprehensive compaction degree model constructed in step 4, an optimal compaction warning model is constructed; Step 6: Based on the optimal comprehensive compaction degree model constructed in step 4 and the optimal compaction warning model constructed in step 5, an optimal comprehensive compaction risk warning model is constructed; The optimal comprehensive compaction risk warning model expression is: in: They represent the weights of the optimal comprehensive compaction degree model and the optimal compaction warning model respectively; Represents the optimal comprehensive compaction value The corresponding compaction warning value; represents the optimal comprehensive compaction risk warning value; Step 7: Collect the feature data set to be tested, output the optimal comprehensive compaction degree value based on the optimal comprehensive compaction degree model, and output the risk warning level based on the optimal comprehensive compaction risk warning model; Step 8: Push the optimal comprehensive compaction risk warning value on the screen in real time , the data of each feature in the feature data set to be tested and the warning level, and a prompt alarm will be issued when the medium risk or high risk warning level is reached.

2. The method for determining soil compaction during corn harvester operation 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 normalized to obtain each normalized feature data; the normalization formula is: in: Represents the standardized The first sample feature data, , ; Indicates the number of training samples; For the The first sample Characteristic data; is the mean; is the standard deviation.

3. The method for determining soil compaction during corn harvester operation based on real-time parameters of a machine tool according to claim 2, characterized in that: The compaction degree prediction model adopts a multiple linear regression model: in: 、 、 、 、 、 、 、 、 They represent the surface characteristic data of the compaction degree prediction model 、 、 、 、 、 、 、 、 The feature weights of 、 、 、 、 、 、 、 、 They represent the compaction degree prediction model for the middle layer characteristic data 、 、 、 、 、 、 、 、 The feature weights of 、 、 、 、 、 、 、 、 They represent the compaction degree prediction model for deep feature data 、 、 、 、 、 、 、 、 The feature weights of The bias term representing the surface feature data; Represents the bias term of the middle-level feature data; Represents the bias term of deep feature data.

4. The method for determining soil compaction during corn harvester operation based on real-time parameters of a machine tool according to claim 3, characterized in that: The bias term of the surface feature data Initial Settings ; Bias term for mid-level feature data Initial Settings ; Bias term for deep feature data Initial Settings ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; The initial value is set to ; ； ； 。 5. The method for determining soil compaction during corn harvester operation based on real-time parameters of the machine according to claim 4, characterized in that: The error loss function of the compaction degree prediction model constructed in step 32 is ; The specific process is: The mean squared error is used to measure the deviation between the predicted value and the true value: in: Indicates the The actual soil compaction value of each sample; Indicates the Predicted soil compaction values ​​for each sample.

6. The method for determining soil compaction during corn harvester operation based on real-time parameters of the machine according to claim 5, characterized 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; Obtaining the optimal compaction degree prediction model 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: in: Indicates the The feature weight of each sample; Indicates the updated The feature weight of each sample; represents the learning rate; Represents the loss function on weight The partial derivative of Partial derivative calculation formula: 。 7. The method for determining soil compaction during corn harvester operation based on real-time parameters of the machine according to claim 6, characterized in that: In step 4, based on the optimal compaction degree prediction model obtained in step 3, an optimal comprehensive compaction degree model is constructed; the specific process is: Step 4.1: Construct the error loss function of the comprehensive compaction degree model ; The specific process is: The mean squared error is used to measure the deviation between the predicted value and the true value: in: Indicates the The actual soil comprehensive compaction value of each sample; Indicates the The predicted soil comprehensive compaction value of each sample; Step 4.2: Use the gradient descent method to optimize the weight of the comprehensive compaction degree model until the weight converges to obtain the optimal surface compaction degree weight. , optimal middle layer compaction degree weight and optimal deep compaction weight ; Based on the optimal surface compaction degree weight , optimal middle layer compaction degree weight and optimal deep compaction weight 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: in: Indicates the updated The weight of the samples; Indicates the The weight of the samples; represents the learning rate; Represents the loss function on weight The partial derivative of Partial derivative calculation formula: Until the weight converges, the optimal surface compaction weight is obtained , optimal middle layer compaction degree weight and optimal deep compaction weight ; Based on the optimal surface compaction degree weight , optimal middle layer compaction degree weight and optimal deep compaction weight Obtain the optimal comprehensive compaction degree model.

8. The method for determining soil compaction during corn harvester operation based on real-time parameters of the machine according to claim 7, characterized in that: In step 5, based on the optimal comprehensive compaction degree model constructed in step 4, an optimal compaction warning model is constructed; the specific process is as follows: 1) When the optimal comprehensive compaction value obtained in step 4 is When , the compaction area belongs to the low compaction area; The optimal compaction warning model is expressed as: in: It represents the risk increase parameter of low compaction area, which is set to 1; Indicates the degree of smoothness of risk change in controlled low compaction areas; Represents the optimal comprehensive compaction value The corresponding compaction warning value; Indicates the compaction degree dividing point; 2) When the optimal comprehensive compaction value obtained in step 4 is When , the compaction area belongs to the medium compaction zone; The optimal compaction warning model is expressed as: in: Indicates the compaction degree dividing point The corresponding compaction warning value; It represents the risk increase parameter of the medium compaction area; Indicates the compaction degree dividing point; represents the risk sensitivity adjustment parameter of the medium compaction zone; 3) When the optimal comprehensive compaction value obtained in step 4 is When , the compacted area belongs to the high compacted area; The optimal compaction warning model is expressed as: in: represents the risk increase parameter; represents the risk sensitivity adjustment parameter of high-pressure solidification area, ; Indicates the compaction degree dividing point The corresponding compaction warning value.

9. The method for determining soil compaction during corn harvester operation based on real-time parameters of the machine according to claim 8, characterized in that: In step 6, based on the optimal comprehensive compaction degree model constructed in step 4 and the optimal compaction warning model constructed in step 5, an optimal comprehensive compaction risk warning model is constructed; The specific process is: Step 6.1: Construct the error loss function of the comprehensive compaction risk warning model ; The specific process is: The mean squared error is used to measure the deviation between the predicted value and the true value: in: Indicates the The actual comprehensive compaction risk warning value of each sample; Indicates the Comprehensive compaction risk warning for each sample; Step 6.2: Use the gradient descent method to optimize the weights of the comprehensive compaction risk warning model until the weights converge to obtain the optimal comprehensive compaction degree model and the optimal compaction warning model weights. ; Based on the weights of the optimal comprehensive compaction degree model and the optimal compaction warning model , obtain the optimal comprehensive compaction risk warning model; 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: in: Indicates the updated The weight of the samples; Indicates the The weight of the samples; represents the learning rate; Represents the loss function on weight The partial derivative of Partial derivative calculation formula: Until the weights converge, the weights of the optimal comprehensive compaction degree model and the optimal compaction warning model are obtained ; Weights based on the optimal comprehensive compaction degree model and the optimal compaction warning model , and obtain the optimal comprehensive compaction risk warning model.

10. The method for determining soil compaction during corn harvester operation based on real-time parameters of the machine according to claim 9, characterized in that: In step seven, a feature data set to be tested is collected, and an optimal comprehensive compaction degree value is output based on the optimal comprehensive compaction degree model, and a risk warning level is output based on the optimal comprehensive compaction risk warning model; 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 , Harvester weight , Track hardness , track tension , unloading frequency , Total turning time ratio , Differential steering time ratio , Braking and steering time ratio , the proportion of time spent turning in place ; The standardized feature data set to be measured is input into the optimal compaction degree prediction model, and the optimal compaction degree prediction model outputs the surface soil compaction value predicted by the compaction degree prediction model , the compaction value of the middle layer soil predicted by the compaction degree prediction model , deep soil compaction value predicted by compaction degree prediction model ; The surface soil compaction value predicted by the compaction degree prediction model , the compaction value of the middle layer soil predicted by the compaction degree prediction model , deep soil compaction value predicted by compaction degree prediction model Input the optimal comprehensive compaction degree model, and the optimal comprehensive compaction degree model outputs the optimal comprehensive compaction value ; The optimal comprehensive compaction value output by the optimal comprehensive compaction degree model Input the optimal compaction warning model, and the optimal compaction warning model outputs the compaction warning value ; The optimal comprehensive compaction value and compaction warning values Input the optimal comprehensive compaction risk warning model, and the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value ; When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value When the warning level is low risk; When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value When the warning level is medium risk; When the optimal comprehensive compaction risk warning model outputs the optimal comprehensive compaction risk warning value The warning level is high risk.

Images