Agricultural material dynamic regulation and cleaning method and system based on accumulation condition prediction

Abstract

The application discloses a kind of agricultural material dynamic regulation and control cleaning method and system based on accumulation condition prediction, method includes: obtaining one-dimensional air volume array data generated by all first air volume sensors, and two-dimensional air volume array data generated by all second air volume sensors;Based on one-dimensional air volume array data, the comprehensive non-uniformity index of air volume below screen surface is calculated, and whether it exceeds the first preset threshold is judged, yes, the data in two-dimensional air volume array data is compensated based on comprehensive non-uniformity index;According to two-dimensional air volume array data, the feature vector that reacts the non-uniformity of screen material is calculated;Based on feature vector and pre-trained prediction model, the accumulation height of material is predicted, to obtain predicted accumulation height, and the rotational speed of material dispersing roller is adjusted.The air volume data on the upper and lower sides of the cleaning screen can effectively predict the accumulation condition of the material on the screen surface, and the material is moderately dispersed accordingly, solving the material accumulation problem in the traditional cleaning process.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for material cleaning, and in particular to a method and system for dynamic regulation and cleaning of agricultural materials based on prediction of stacking conditions. Background Technology

[0002] Agricultural material cleaning is a core step in agricultural processing, directly affecting grain purity and storage quality. Screening equipment separates impurities through airflow, but localized accumulation on the screen surface disrupts the uniformity of airflow distribution: obstructed airflow in the accumulation area leads to impurity residue, while excessive airflow in adjacent areas causes excessive grain breakage or spillage. Traditional methods rely on fixed airflow or manual adjustment, making real-time monitoring of dynamic accumulation impossible, resulting in low cleaning efficiency and limited effectiveness.

[0003] In the prior art, the applicant's earlier application CN 115176603 A discloses a material distribution and guiding mechanism for the screen surface of a cleaning screen. Although it can distribute the material on the screen surface evenly, it requires manual control of the start-up, shutdown and rotation speed based on experience, and the control is not intelligent enough.

[0004] Patent CN 208912557 U discloses a wind-screen type grain cleaning test device, which has arrayed wind speed sensors installed on both the upper and lower sides of a vibrating screen to monitor the wind speed distribution at the fan outlet and near the screen surface in real time. A programmable controller analyzes the data and dynamically adjusts the cleaning parameters to ensure uniform airflow coverage of the screen surface. However, this technology lacks multi-dimensional data fusion, dynamic compensation, and intelligent prediction capabilities, resulting in insufficient adaptability and robustness in the cleaning process, and failing to effectively solve the material accumulation problem under complex working conditions. Summary of the Invention

[0005] Purpose of the invention: In order to overcome the shortcomings of the existing technology, the present invention provides a dynamic control cleaning method and system for agricultural materials based on the prediction of the accumulation of materials on the screen surface, which can effectively predict the material accumulation status on the screen surface based on sensor data and disperse the material as needed to improve the cleaning effect and efficiency.

[0006] Technical solution: To achieve the above objectives, the present invention provides a dynamic control and cleaning method for agricultural materials based on pile condition prediction, comprising:

[0007] Get the bottom of the cleaning screen r One-dimensional airflow array data generated by the first airflow sensor , For the first k Data generated by the first airflow sensor; and data acquired from above the cleaning screen. p OK q List all two-dimensional airflow array data generated by the second airflow sensor , For the first i Line 1 j Data generated by the second airflow sensor in the column;

[0008] Based on the one-dimensional air volume array data Calculate the comprehensive non-uniformity index Γ of the airflow distribution below the screen surface of the cleaning screen;

[0009] Determine whether Γ exceeds the first preset threshold Γ0; if so, then apply the comprehensive non-uniformity index to the two-dimensional airflow array data. The data in the image is compensated; otherwise, the two-dimensional airflow array data is maintained. The data remains unchanged;

[0010] Based on the two-dimensional air volume array data Calculate the eigenvector of material imbalance on the reaction sieve surface. ,in: The global standard deviation is... For the local maximum deviation, The maximum variance of the sub-region;

[0011] Based on the feature vectors and the pre-trained prediction model, the stacking height of the material is predicted, resulting in the predicted stacking height. The rotational speed of the material dispersing drum is adjusted using the following control strategy. n :

[0012] ;

[0013] in: This is the proportionality coefficient. This is the stacking height threshold.

[0014] Furthermore, the parameters in the feature vector are calculated based on the following formula:

[0015] Global standard deviation Among them: average wind speed ;

[0016] Local maximum deviation ;

[0017] Maximum variance of subregion ;

[0018] in, , to convert two-dimensional air volume array data Divided into a × b Sub-regions For the first M The number of sensors in each sub-region For the firstM The average air volume within each sub-region.

[0019] Furthermore, the comprehensive non-uniformity index Γ is calculated based on the following formula:

[0020]

[0021] in, This represents the average air volume.

[0022] Furthermore, the two-dimensional airflow array data is analyzed based on the comprehensive non-uniformity index. The data in the middle is compensated, including:

[0023] Calculate the normalized non-uniform weight at each position below the screen surface of the cleaning screen. ;

[0024] Two-dimensional airflow array data of the second airflow sensor To compensate, specifically: the formula ( The obtained value is assigned to In Perform data updates; among them, For compensation ratio coefficient, Let be the normalized non-uniformity weight corresponding to the i-th row. This refers to the air volume compensation range.

[0025] Furthermore, the prediction model is created based on the following method:

[0026] Obtaining experimental data ,in The actual height of the material accumulation on the screen surface , For feature vectors, N This represents the total number of experimental data sets.

[0027] The prediction model is obtained by training the experimental data and the SVR proxy model; the objective function of the SVR proxy model is:

[0028] ;

[0029] in, For the weight vector, For bias terms, As a penalty factor, and These are slack variables; The total number of trials;

[0030] Using radial basis function kernels ;in This is the kernel width parameter.

[0031] Furthermore, the rules for dividing the sub-regions include:

[0032] Determine the size of the sub-region c × d ,in c , d These represent the number of the second airflow sensors included in the horizontal and vertical directions, respectively.

[0033] Determine the number of overlapping rows and columns when selecting a sub-region e and f ;

[0034] Based on the sub-region size and the number of overlapping rows and columns, all combinations of the second air volume sensors that meet the conditions are traversed to form the sub-region that meets the conditions.

[0035] A dynamic control and cleaning system for agricultural materials based on accumulation condition prediction includes:

[0036] Cleaning sieves are used to clean agricultural materials to separate grains from impurities.

[0037] The one-dimensional linear array layout located below the cleaning screen r The first air volume sensor;

[0038] The button located above the cleaning screen p × q Multiple second airflow sensors arranged in a square array;

[0039] The material dispersing drum is used to disperse aggregated materials on the cleaning screen.

[0040] The controller is connected to the first air volume sensor, the second air volume sensor, and the material dispersing drum, and is capable of implementing the above-mentioned dynamic control and cleaning method for agricultural materials based on the prediction of the accumulation situation.

[0041] Beneficial Effects: The agricultural material dynamic control cleaning method and system based on accumulation condition prediction of the present invention has the following beneficial effects:

[0042] (1) By acquiring the air volume array data above and below the cleaning screen in real time, and combining the comprehensive non-uniformity index to dynamically compensate the two-dimensional air volume distribution, a multi-dimensional feature vector is constructed using the global standard deviation, local maximum deviation and sub-region maximum variance. Based on the surrogate model, the material accumulation height is predicted, and the adaptive adjustment of the drum speed is realized. This effectively solves the material accumulation problem caused by uneven air volume distribution in the traditional cleaning process, improves cleaning efficiency and quality, and reduces the need for manual intervention through data-driven closed-loop control, thereby enhancing the system's dynamic response capability and stability to complex working conditions.

[0043] (2) The three parameters in the feature vector obtained from the two-dimensional air volume array data can effectively reflect the distribution of air volume on the screen surface in three dimensions: overall, local and extreme differences. They are highly representative and can effectively reflect the distribution and accumulation of materials.

[0044] (3) The comprehensive non-uniformity index Γ comprehensively evaluates the overall dispersion and local mutation intensity of the air volume distribution below the screen by combining the global standard deviation and the absolute mean of the air volume difference between adjacent sensors. It can more scientifically reflect the true uniformity of airflow distribution, avoid one-sidedness, and thus accurately trigger dynamic compensation of the second sensor data, effectively correct the prediction error caused by the non-uniform airflow under the screen, and significantly improve the robustness of the accumulation height prediction and the adaptability of cleaning control.

[0045] (4) By dynamically adjusting the compensation parameters of the second air volume sensor based on the air volume at different positions below the screen, the data of the second air volume sensor can be compensated based on the data of the first air volume sensor, which can improve the accuracy of subsequent stacking height prediction.

[0046] (5) When dividing into sub-regions, the number of overlapping rows and columns is set to ensure the continuity of data between adjacent sub-regions, avoid the edge effect caused by division, enhance the ability to capture local airflow changes and abnormal accumulation, improve the comprehensiveness of feature extraction and model prediction accuracy, and adapt to different material distribution patterns to optimize the level of cleaning and control. Attached Figure Description

[0047] Figure 1 This is a structural diagram of an agricultural material dynamic control and cleaning system based on accumulation prediction.

[0048] Figure 2 This is a layout diagram of the second airflow sensor from a top-down perspective;

[0049] Figure 3 This is a flowchart illustrating a dynamic control and cleaning method for agricultural materials based on predictions of accumulation conditions.

[0050] In the diagram: 1- Cleaning screen; 2- First air volume sensor; 3- Second air volume sensor; 4- Material dispersing drum; 5- Fan. Detailed Implementation

[0051] The invention will now be further described with reference to the accompanying drawings.

[0052] like Figure 1The illustrated agricultural material dynamic control cleaning system based on accumulation prediction includes a cleaning sieve 1 for cleaning agricultural materials to separate grains from impurities, r first airflow sensors 2 arranged in a one-dimensional linear array below the cleaning sieve 1, and multiple second airflow sensors 3 arranged in a p×q square array above the cleaning sieve 1. Figure 2 As shown; in addition, a material dispersing roller 4 is installed above the cleaning screen to disperse the aggregated material on the cleaning screen 1; the cleaning system also includes a blower 5 for providing cleaning airflow.

[0053] The system also includes a controller, which is connected to the first air volume sensor 2, the second air volume sensor 3 and the material dispersing drum 4, and is capable of implementing the following dynamic control and cleaning method for agricultural materials based on the prediction of the accumulation situation.

[0054] like Figure 3 The method for dynamic control and cleaning of agricultural materials based on accumulation prediction, as shown, includes the following steps S101-S105:

[0055] Step S101, obtain the material below the cleaning sieve 1. r One-dimensional airflow array data generated by the first airflow sensor 2 , For the first k The data generated by the first airflow sensor 2; and the data obtained above the cleaning screen 1. p OK q List all two-dimensional airflow array data generated by the second airflow sensor 3 , For the first i Line 1 j The data generated by the second air volume sensor 3 in the column;

[0056] Step S102, based on the one-dimensional airflow array data Calculate the comprehensive non-uniformity index Γ of the airflow distribution below the screen surface of the cleaning screen 1;

[0057] Step S103: Determine whether Γ exceeds the first preset threshold Γ0. If yes, then apply the comprehensive non-uniformity index to the two-dimensional airflow array data. The data in the image is compensated; otherwise, the two-dimensional airflow array data is maintained. The data remains unchanged;

[0058] Step S104, based on the two-dimensional airflow array data Calculate the eigenvector of material imbalance on the reaction sieve surface. ,in: The global standard deviation is... For the local maximum deviation, The maximum variance of the sub-region;

[0059] Step S105: Based on the feature vector and the pre-trained prediction model, predict the stacking height of the material to obtain the predicted stacking height. The rotational speed of the material dispersing drum 4 is adjusted using the following control strategy:

[0060] ;

[0061] in: This is the proportionality coefficient. This is the stacking height threshold.

[0062] This method acquires real-time airflow array data above and below the cleaning screen 1, dynamically compensates for the two-dimensional airflow distribution by combining the comprehensive non-uniformity index, constructs a multi-dimensional feature vector using global standard deviation, local maximum deviation, and sub-region maximum variance, and predicts the material accumulation height based on a surrogate model to achieve adaptive adjustment of the drum speed. This effectively solves the material accumulation problem caused by uneven airflow distribution in the traditional cleaning process, improves cleaning efficiency and quality, and reduces the need for manual intervention through data-driven closed-loop control, enhancing the system's dynamic response capability and stability to complex working conditions.

[0063] Preferably, the parameters in the feature vector described in step S104 above are calculated based on the following formula:

[0064] Global standard deviation Among them: average wind speed ;

[0065] Local maximum deviation ;

[0066] Maximum variance of subregion ;

[0067] in, , to convert two-dimensional air volume array data Divided into a × b Sub-regions For the first M The number of sensors in each sub-region For the first M Average air volume in each sub-region For the sub-region, the first m Line 1 n The data generated by the second airflow sensor 3 in the column.

[0068] Of the three parameters mentioned above, the global standard deviation reflects the overall dispersion of airflow distribution across the entire screen surface, indicating the uniformity of the overall airflow. However, when material accumulates locally on the screen surface, the airflow in certain sub-regions may significantly deviate from the average. Therefore, by dividing the screen surface into sub-regions and calculating the variance of each sub-region, the maximum variance of the sub-regions can be selected to analyze the uniformity of local airflow distribution in detail. It highlights the areas with the most uneven airflow distribution, i.e., local areas where material accumulation may be severe. Even if the global standard deviation is low, the maximum variance of the sub-regions can detect these local anomalies. The local maximum deviation plays a crucial role in quickly and efficiently capturing extreme differences in airflow distribution throughout the method, providing important feature support for solving material accumulation problems.

[0069] It is evident that the three parameters in the feature vector obtained from the two-dimensional airflow array data can effectively reflect the airflow distribution on the screen surface in three dimensions: overall, local, and extreme differences. They are highly representative and can effectively reflect the distribution and accumulation of materials.

[0070] Preferably, the comprehensive non-uniformity index Γ in step S102 above is calculated based on the following formula:

[0071]

[0072] in, This represents the average airflow. In the formula above, the first term on the right side of the equals sign is the global standard deviation, and the second term is the absolute mean of the airflow differences between adjacent sensors.

[0073] The comprehensive non-uniformity index Γ comprehensively assesses the overall dispersion and local abrupt change intensity of the airflow distribution below the screen by combining the global standard deviation with the absolute mean of the airflow difference between adjacent sensors. It can more scientifically reflect the true uniformity of airflow distribution, avoid bias, and thus accurately trigger dynamic compensation of the second sensor data. It effectively corrects the prediction error caused by the non-uniform airflow under the screen, and significantly improves the robustness of the accumulation height prediction and the adaptability of the cleaning control.

[0074] Preferably, in step S103 above, the two-dimensional airflow array data is analyzed based on the comprehensive non-uniformity index. The data in the middle is compensated, including the following steps S201-S202:

[0075] Step S201: Calculate the normalized non-uniform weight at each position below the screen surface of the cleaning screen 1. ;

[0076] Step S202: Process the two-dimensional airflow array data of the second airflow sensor 3. To compensate, specifically: the formula ( The obtained value is assigned to In Perform data updates; among them, For compensation ratio coefficient, For the first i The corresponding normalized non-uniformity weights are arranged. This refers to the air volume compensation range.

[0077] In step S202, with Representing the formula ( The obtained value, during the compensation process, is subject to constraints. The variation range is determined to ensure that the compensated airflow data does not introduce new unevenness, i.e.:

[0078] ;

[0079] in, and The first Reasonable upper and lower limits for air volume data from the second air volume sensor.

[0080] By dynamically adjusting the compensation parameters of the second air volume sensor 3 based on the air volume at different positions below the screen surface, and by compensating the data of the second air volume sensor 3 based on the data of the first air volume sensor 2, the accuracy of subsequent stacking height prediction can be improved.

[0081] Preferably, the prediction model described in step S105 above is created based on the following method:

[0082] Step S301: Obtain experimental data ,in The actual height of the material accumulation on the screen surface , For feature vectors, N This represents the total number of experimental data sets; the feature vectors here are calculated using the algorithm described in steps S101-S104 above.

[0083] Step S302: Based on the experimental data and the SVR proxy model, perform data training to obtain the prediction model; the objective function of the SVR proxy model is:

[0084] ;

[0085] in, For the weight vector, For bias terms, As a penalty factor, and As slack variables, , ; This represents the total number of trials; all parameters here are local variables.

[0086] And it satisfies the following constraints:

[0087] For all , ,as well as, ;in, Define the tolerance range for prediction error as an insensitive parameter; It is a mathematical tool in machine learning that nonlinearly maps data to a high-dimensional space.

[0088] Using radial basis function kernels ;in This is the kernel width parameter.

[0089] This technology effectively balances model complexity and training error by combining a regularization term and a slack variable penalty mechanism with an SVR surrogate model. It uses a radial basis kernel function to map nonlinear features to a high-dimensional space, significantly improving the generalization ability and accuracy of stacking height prediction. Training based on experimental data ensures that the model fits the actual cleaning conditions, adapts to varying material distributions and airflow interference, provides a reliable decision-making basis for dynamic adjustment, and enhances the overall stability and intelligence level of the cleaning system.

[0090] Preferably, the sub-region division rule includes the following steps S401-S402:

[0091] Step S401: Determine the size of the sub-region. c × d ,in c , d These represent the number of the second airflow sensors 3 included in the horizontal and vertical directions, respectively.

[0092] Step S402: Determine the number of overlapping rows and columns when selecting a sub-region. e and f That is, when selecting a sub-region, two vertically adjacent sub-regions have... e The line sensors overlap, and two adjacent sub-regions in the lateral direction have f The rows of sensors are aligned. e Less than d , f Less than c .

[0093] Step S403: Based on the size of the sub-region and the number of overlapping rows and columns, traverse all combinations of the second air volume sensors 3 that meet the conditions to form the sub-region that meets the conditions.

[0094] By setting the number of overlapping rows and columns, the data continuity between adjacent sub-regions is ensured, the edge effect caused by division is avoided, the ability to capture local airflow changes and abnormal accumulation is enhanced, the comprehensiveness of feature extraction and the accuracy of model prediction are improved, and the fineness of cleaning and control is optimized to adapt to different material distribution patterns.

[0095] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for dynamic regulation and cleaning of agricultural materials based on prediction of accumulation conditions, characterized in that, The method includes: Acquire one-dimensional airflow array data generated by r first airflow sensors below the cleaning screen. , The data generated by the k-th first airflow sensor; and the two-dimensional airflow array data generated by all the second airflow sensors in rows p and columns q above the cleaning screen. , The data generated by the second airflow sensor in the i-th row and j-th column; Based on the one-dimensional air volume array data Calculate the comprehensive non-uniformity index Γ of the airflow distribution below the screen surface of the cleaning screen; Determine whether Γ exceeds the first preset threshold Γ0. If yes, then apply the comprehensive non-uniformity index to the two-dimensional airflow array data. The data in the data should be compensated; otherwise, the two-dimensional airflow array data should be maintained. The data remains unchanged; Based on the two-dimensional air volume array data Calculate the eigenvector of material imbalance on the reaction sieve surface. ,in: The global standard deviation is... For the local maximum deviation, The maximum variance of the sub-region; Based on the feature vectors and the pre-trained prediction model, the stacking height of the material is predicted, resulting in the predicted stacking height. The rotational speed n of the material dispersing drum is adjusted using the following control strategy: ； in: This is the proportionality coefficient. This is the stacking height threshold; The comprehensive heterogeneity index Γ is calculated based on the following formula: ； in, This represents the average air volume. The two-dimensional airflow array data is based on the comprehensive non-uniformity index. The data in the middle is compensated, including: Calculate the normalized non-uniform weight at each position below the screen surface of the cleaning screen. ; Two-dimensional airflow array data of the second airflow sensor To compensate, specifically: the formula The obtained value is assigned to In Perform data updates; among them, For compensation ratio coefficient, Let be the normalized non-uniformity weight corresponding to the i-th row. This refers to the air volume compensation range.

2. The method for dynamic regulation and cleaning of agricultural materials based on accumulation condition prediction according to claim 1, characterized in that, The parameters in the feature vector are calculated based on the following formula: Global standard deviation Among them: average wind speed ; Local maximum deviation ; Maximum variance of subregion ; in, , to convert two-dimensional air volume array data Divide into a×b sub-regions. Let M be the number of sensors in the Mth sub-region. Let be the average air volume in the Mth sub-region.

3. The method for dynamic regulation and cleaning of agricultural materials based on accumulation condition prediction according to claim 1, characterized in that, The prediction model was created based on the following method: Obtaining experimental data ,in The actual height of the material accumulation on the screen surface , Here, N represents the feature vector, and N is the total number of experimental data sets. The prediction model is obtained by training the experimental data and the SVR proxy model; the objective function of the SVR proxy model is: ； in, For the weight vector, For bias terms, As a penalty factor, and These are slack variables; The total number of trials; Using radial basis function kernels ;in This is the kernel width parameter.

4. The method for dynamic regulation and cleaning of agricultural materials based on accumulation condition prediction according to claim 2, characterized in that, The rules for dividing the sub-regions include: Determine the size of the sub-region as c×d, where c and d are the number of the second air volume sensors included in the horizontal and vertical directions, respectively; Determine the number of overlapping rows and columns, e and f, when selecting a sub-region; Based on the sub-region size and the number of overlapping rows and columns, all combinations of the second air volume sensors that meet the conditions are traversed to form the sub-region that meets the conditions.

5. A dynamic control and cleaning system for agricultural materials based on accumulation condition prediction, characterized in that, It includes: Cleaning sieves are used to clean agricultural materials to separate grains from impurities. r first airflow sensors are arranged in a one-dimensional linear array below the cleaning screen; Multiple second air volume sensors are located above the cleaning screen in a p×q square array. The material dispersing drum is used to disperse aggregated materials on the cleaning screen. The controller is connected to the first air volume sensor, the second air volume sensor and the material dispersing drum, and is capable of implementing the dynamic control and cleaning method for agricultural materials based on the prediction of the accumulation condition as described in any one of claims 1-4.

Images