Cotton picker pneumatic conveying system flow field characteristic prediction method and system based on digital twinning

Abstract

The invention provides a method and a system for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twinning. The method comprises the following steps of simulation data acquisition of the pneumatic conveying system of the cotton picker, data preprocessing, core sample screening, node flow velocity prediction model construction, rendering cloud picture generation and construction of a visual plate based on digital twinning. According to the method, a model is established for the input fan rotating speed and multiple groups of different types of output parameters through polynomial chaos expansion, and visualization of an air outlet plane flow field and an impeller circular ring expansion plane flow field is realized by utilizing a digital twin technology. According to the method, the flow field data of each node of the flow field of the pneumatic conveying system of the cotton picker is obtained through the numerical simulation model, key nodes in the flow field data are identified through the data clustering algorithm, the calculation cost is reduced, and the response speed is increased; a flow velocity prediction model of the key nodes is established by adopting polynomial chaos expansion, a flow field cloud picture is generated based on a node prediction result, and rapid prediction of the flow field characteristics of the pneumatic conveying system of the cotton picker is realized.

Description

Technical Field

[0001] The present invention belongs to the cross - technical field of agricultural engineering, big data and artificial intelligence, and particularly relates to a method and system for predicting the flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin. Background Art

[0002] The mechanization level of cotton production in China has developed rapidly, and the proportion of machine - picked cotton has gradually increased. Picking is a key step in cotton mechanized production. In the past, manual cotton picking was mainly used, which was inefficient and costly. To improve the picking efficiency and reduce costs, cotton pickers have become the main production method for cotton picking in China.

[0003] The cotton picker separates the seed cotton from the cotton plant through the picking head spindles, separates the seed cotton from the spindles through the doffer, and then transports the seed cotton to the cotton collecting box through the pneumatic conveying system to complete the cotton picking process.

[0004] In the pneumatic conveying system of the cotton picker, the centrifugal fan provides the air flow source for transporting the seed cotton. In the existing pneumatic conveying system, the fan speed, air distribution plate angle, etc. are fixed, resulting in fixed flow field characteristics of the pneumatic conveying system and unable to be adjusted in real - time according to the operating conditions of the cotton picker.

[0005] To achieve the adaptive regulation of the fan speed, air distribution plate angle, etc. with the operating conditions, it is necessary to obtain the flow field characteristics of the pneumatic conveying system under given parameter conditions such as the fan speed.

[0006] Existing flow field measurement methods mostly use Pitot tubes for measurement, which can only achieve single - point flow velocity measurement and cannot obtain parameters such as cross - section flow velocity, that is, the measurement of flow field characteristics.

[0007] Existing common flow field analysis methods include the analytical method and the numerical simulation method. Among them, the analytical method simplifies the scenario to a certain extent, and the solution accuracy is low; the numerical simulation method has high solution accuracy, but the solution time is long and cannot meet the requirement of rapid estimation of the flow field of the pneumatic conveying system.

[0008] Therefore, how to obtain the flow field characteristics of the pneumatic conveying system based on numerical simulation analysis data and achieve the rapid acquisition of flow field characteristics through technologies such as digital twin has become an urgent issue to be solved for obtaining the flow field characteristics of the pneumatic conveying system of the cotton picker. Summary of the Invention

[0009] To solve the above technical problems, the present invention provides a method and system for predicting the flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin. A model is established for the input fan speed and multiple groups of output parameters of different categories through polynomial chaos expansion (PCE), and the digital twin technology is used to visualize the flow field of the air outlet plane and the flow field of the impeller ring expansion plane. The present invention obtains the flow field data of each node of the pneumatic conveying system of the cotton picker through a numerical simulation model, uses a data clustering algorithm to identify the key nodes therein, reduces the calculation cost and improves the response speed; uses polynomial chaos expansion to establish a flow velocity prediction model for the key nodes, and then generates a flow field cloud map based on the node prediction results to achieve a rapid prediction of the flow field characteristics of the pneumatic conveying system of the cotton picker.

[0010] Note that the recording of these objectives does not prevent the existence of other objectives. One embodiment of the present invention does not need to achieve all of the above objectives. Objectives other than the above can be extracted from the descriptions of the specification, drawings, and claims.

[0011] The present invention achieves the above technical objectives through the following technical means.

[0012] A method for predicting the flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin includes the following steps:

[0013] Step S1, obtaining simulation data of the pneumatic conveying system of the cotton picker: Establish a high-fidelity numerical simulation model for predicting the flow field characteristic data of the pneumatic conveying system of the cotton picker, and obtain the flow field performance simulation data of the pneumatic conveying system of the cotton picker according to the simulation model;

[0014] Step S2, data preprocessing: Preprocess the flow field simulation data of the pneumatic conveying system of the cotton picker obtained in Step S1, identify and remove the outliers and anomalies inside the data, and form a training data set for each node;

[0015] Step S3, screening of core samples: Process the training data set obtained in Step S2. Considering problems such as too many data samples, cost, and noise, cluster the training data through a fuzzy clustering algorithm, and use the screened core data as the training data set;

[0016] Step S4, construction of a node flow velocity prediction model: Establish a flow velocity prediction model for each node based on polynomial chaos expansion, use the leave-one-out method to construct training samples and verification samples, use the training data set obtained in Step S3 to construct a flow velocity prediction model for each node of the flow field of the air outlet plane and the impeller ring expansion plane of the cotton picker, and verify the flow velocity prediction model using the verification set;

[0017] Step S5, Rendering Cloud Map Generation: Preprocess the prediction model obtained in step S4, optimize the number of prediction models, optimize the rendering cloud map generation rate, reduce the rendering time, speed up the display speed of the cloud map, and call the prediction model to calculate and generate a real-time rendered cloud map display based on the rendering software;

[0018] Step S6, Visualization: Build a visualization section based on digital twins, and establish a visualization interface display based on the software architecture with the prediction model obtained in step S5 as the basis.

[0019] In the above solution, in step S1, the flow field performance simulation data of the pneumatic conveying system of the cotton picker is obtained according to the simulation model, which specifically includes the following steps:

[0020] Step S1.1, Establish a high-fidelity numerical simulation model for predicting the flow field characteristic data of the pneumatic conveying system of the cotton picker for the overall structure of the fan pipeline of the cotton picker;

[0021] Step S1.2, Perform grid node division on the outlet plane flow field and the impeller ring expansion plane flow field of the cotton picker;

[0022] Step S1.3, Simulate and obtain all grid node performance data of the outlet plane flow field and the impeller ring expansion plane flow field under the corresponding fan operation data.

[0023] In the above solution, in step S1, the flow field simulation data of the pneumatic conveying system of the cotton picker is obtained by calculating the flow velocity data of each node through computational fluid dynamics, the finite element method, or the finite volume method.

[0024] In the above solution, in step S2, the isolation forest is used to identify and remove outliers, solving the problem of abnormal data in the training model data.

[0025] In the above solution, in step S3, the fuzzy clustering algorithm is used to delete samples, reduce the calculation cost, speed up the response speed of the prediction model, and use the screened core data as the training data set and save it to the database.

[0026] In the above solution, in step S4, a flow field characteristic prediction model is established based on polynomial chaos expansion, which specifically includes the following steps:

[0027] Step S4.1, Based on the processed data set, use the leave-one-out method to generate a training set and a validation set;

[0028] Step S4.2, Establish a flow field characteristic prediction model using polynomial chaos expansion;

[0029] Step S4.3, Use the validation set to verify the performance prediction effect of the flow field characteristic prediction model, and use the coefficient of determination R 2 to evaluate the accuracy and reliability of the model.

[0030] Further, the step of establishing a prediction model for the flow field characteristics based on polynomial chaos expansion in step S4.2 includes the following steps:

[0031] Construct a basis function y according to the input and output and the relationship conditions of their input and output;

[0032] Calculate the unknown polynomial chaos PC coefficients;

[0033] The relationship between the input x and the output y is expressed as follows:

[0034]

[0035] represents a polynomial chaos expansion model that attempts to approximate or fit the true system output response y;

[0036] is the symbol of the n-dimensional non-negative integer set;

[0037] p is the maximum order of the polynomial chaos expansion;

[0038] where y is the output;

[0039] x is the input;

[0040] ψ α (x) is the tensor product of univariate orthogonal polynomials and satisfies the following formula:

[0041]

[0042] x i is the i-th variable of the input;

[0043] β α is the polynomial chaos coefficient to be solved, that is, the PC coefficient;

[0044] where the total number of unknown PC coefficients P satisfies the following formula:

[0045]

[0046] P is the maximum order of the input variable;

[0047] n is the dimension of the input variable;

[0048] is a polynomial whose total order |α| does not exceed the given p order;

[0049] In order to calculate the PC value, it is necessary to find a polynomial chaos expansion PCE that satisfies the following formula:

[0050]

[0051] where N is the number of input-output samples;

[0052] The above formula can be rewritten as:

[0053] y = ψβ;

[0054] where

[0055] denotes a P-dimensional real vector

[0056] β is the PC coefficient vector to be solved;

[0057]

[0058] y is the output response vector, representing an N-dimensional real vector;

[0059]

[0060] ψ is the measurement matrix, and each column contains the evaluation values of the PC basis of N samples;

[0061] representing an N×P-dimensional real matrix;

[0062] The PC coefficients are calculated by minimizing the residual between the model response and the PCE approximation.

[0063] For N ≥ P, the unknown coefficients are calculated using least squares regression, and the formula is as follows:

[0064] β = (ψ 9 ψ) :" ψ 9 y;

[0065] When N < P, the problem of solving the PC coefficients is transformed into an l1 minimization problem:

[0066]

[0067] where the l2 norm constraint is used to consider the p-order truncation error ∈ of the PCE;

[0068] The constrained l1 minimization problem in the above formula can be reformulated as a regularized (unconstrained) optimization problem:

[0069]

[0070] ∥β∥ " is the l1 norm of the PC coefficients;

[0071] It is the degree of fitting with the true model response in the sense of the l2 norm.

[0072] In the above solution, in step S5, optimizing the rendering cloud map generation rate includes the following steps:

[0073] Step S5.1: Based on the numerical simulation model, obtain the flow field data of multiple nodes at different fan speeds, and retain the coordinate data of each node. Load the flow field data and coordinate data, and establish a surrogate model based on polynomial chaos expansion. Taking the coordinate data as the input and the node wind speed at a certain speed as the output, construct prediction models at different speeds based on polynomial chaos expansion;

[0074] Step S5.2: Load all the node flow velocity prediction models of step S5.1. According to the real-time response rate requirement of the cloud map, divide the horizontal and vertical coordinates into k equal parts within the cloud map boundary, determine the horizontal and vertical coordinate values at the division junctions, and combine them to generate (k + 1) × (k + 1) uniform node coordinate values. Taking the coordinates of all uniform nodes as the input, load the prediction model, and predict all data at each speed respectively; In the present invention, taking all the uniform node coordinate values as the input of the surrogate model established by polynomial chaos expansion, all data under each working condition are predicted respectively. Comparing the processed uniform node data with directly calling the simulation node data, based on the rendering software to calculate and render the real-time cloud map display, the former node data is more uniform, the calculation amount is smaller, and the response speed is faster;

[0075] Step S5.3: Load all the data of the uniform nodes. Taking each fan speed as the input and the corresponding flow velocity data of each node as the output, generate (k + 1) × (k + 1) uniform node flow velocity prediction models. According to the input fan speed, call all the prediction models through the rendering software and calculate the node wind speed values at each coordinate, and arrange the wind speed data in the cloud map plane according to the corresponding coordinate order. Finally, use the rendering software to interpolate and generate the cloud map according to the numerical relationship of the data in the plane.

[0076] In the above solution, the visualization interface display in step S6 includes the real-time display of the average wind speed in the central area of the flow field, the highest wind speed in the central area, the lowest wind speed in the central area, the outlet flow rate, the average working power of the fan, as well as the real-time display of the flow field in the outlet plane watershed and the impeller ring expansion plane.

[0077] Preferably, the average wind speed in the central area of the flow field, the highest wind speed in the central area, the lowest wind speed in the central area, the outlet flow rate, and the average working power of the fan are displayed in real time through line charts, and the flow field in the outlet plane watershed and the impeller ring expansion plane are displayed in real time through cloud maps.

[0078] A system for implementing the method for predicting the flow field characteristics of the pneumatic conveying system of a cotton picker based on digital twin, comprising a simulation data acquisition module, a data preprocessing module, a core sample screening module, a node flow velocity prediction model construction module, a rendered cloud map generation module, and a visualization module;

[0079] The simulation data acquisition module is used to establish a high-fidelity numerical simulation model for predicting the flow field characteristic data of the pneumatic conveying system of the cotton picker, and obtain the flow field performance simulation data of the pneumatic conveying system of the cotton picker according to the simulation model;

[0080] The data preprocessing module is used to preprocess the obtained flow field simulation data of the pneumatic conveying system of the cotton picker, identify and remove the outliers and anomalies inside the data, and form the training data set of each node;

[0081] The core sample screening module is used to process the obtained training data set. Considering problems such as excessive data samples, cost, and noise, the training data is clustered by a fuzzy clustering algorithm, and the screened core data is used as the training data set;

[0082] The node flow velocity prediction model construction module is used to establish a flow velocity prediction model for each node based on polynomial chaos expansion, construct training samples and verification samples using the leave-one-out method, construct a flow velocity prediction model for each node of the flow field of the cotton picker outlet plane and the impeller ring expansion plane using the training data set obtained in step S3, and verify the flow velocity prediction model using the verification set;

[0083] The rendered cloud map generation module is used to preprocess the obtained prediction model, optimize the number of prediction models, optimize the rendered cloud map generation rate, reduce the rendering time, speed up the cloud map display speed, and call the prediction model to calculate and generate a real-time rendered cloud map display based on the rendering software;

[0084] The visualization module is used to build a visualization section based on digital twin, and establish a visualization interface display based on the software architecture with the prediction model obtained in step S5 as the basis.

[0085] Compared with the prior art, the beneficial effects of the present invention are:

[0086] According to one aspect of the present invention, the method of the present invention first establishes a high-fidelity numerical simulation model for the flow field of the pneumatic conveying system of the cotton picker, divides the grid nodes of the flow field of the pneumatic conveying system such as the cotton picker outlet and the impeller ring, obtains the operation data and the flow field data of each node of the flow field of the pneumatic conveying system of the cotton picker, preprocesses the outliers and abnormal values of the flow field data, then screens the core data through a fuzzy clustering algorithm, and establishes a flow field node flow velocity prediction model through polynomial chaos expansion; calls the prediction model through the rendering software to generate prediction data, interpolates the data to generate a cloud map, and finally displays it through the visualization section.

[0087] According to one aspect of the present invention, the system of the present invention includes a simulation data acquisition module, a data preprocessing module, a core sample screening module, a node flow velocity prediction model construction module, a rendering cloud map generation module, and a visualization module. The entire device can clearly and real-time display the working state of the cotton picker and monitor the flow field performance data of the pneumatic conveying system of the cotton picker.

[0088] The present invention uses digital twin technology and polynomial chaos expansion to predict the flow field characteristics of the pneumatic conveying system of the cotton picker, which can realize the real-time monitoring of the flow field of the pneumatic conveying system of the cotton picker, solve the problem that the flow field characteristics of the pneumatic conveying system of the cotton picker cannot be quickly predicted and obtained, and provide a reference basis for the regulation of the pneumatic conveying system.

[0089] Note that the description of these effects does not prevent the existence of other effects. One aspect of the present invention does not necessarily have all the above effects. Effects other than the above can be obviously seen and extracted from the descriptions of the specification, drawings, claims, etc. Description of the Drawings

[0090] Figure 1 Schematic flowchart of the method for predicting the flow field characteristics of the pneumatic conveying system of the cotton picker according to an embodiment of the present invention;

[0091] Figure 2 High-fidelity numerical simulation result diagram of the cotton picker fan according to an embodiment of the present invention;

[0092] Figure 3 Schematic diagram of the case outlier recognition algorithm according to an embodiment of the present invention;

[0093] Figure 4 Schematic effect diagram of the case clustering algorithm according to an embodiment of the present invention;

[0094] Figure 5 Rendering cloud map of the outlet plane flow field of the case according to an embodiment of the present invention;

[0095] Figure 6 Rendering cloud map of the impeller ring expansion plane flow field of the case according to an embodiment of the present invention;

[0096] Figure 7 Real-time line chart of the fan speed, fan power, and minimum flow velocity of the central flow field of the case according to an embodiment of the present invention;

[0097] Figure 8 Real-time line chart of the comparison between the maximum flow velocity, minimum flow velocity, and average flow velocity of the central flow field and the real-time measured flow velocity of the case according to an embodiment of the present invention. Detailed Embodiments

[0098] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present invention, and should not be construed as a limitation of the present invention.

[0099] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by terms such as "center", "longitudinal", "lateral", "length", "width", "thickness", "front", "rear", "left", "right", "upper", "lower", "axial", "radial", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality" means two or more unless otherwise specifically defined.

[0100] In the present invention, unless otherwise clearly specified and limited, terms such as "installed", "connected", "connected to", "fixed" and the like should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0101] Embodiment 1

[0102] Figure 1 As shown, it is a preferred embodiment of the method for predicting the flow field characteristics of the pneumatic conveying system of a cotton picker based on digital twin. The method for predicting the flow field characteristics of the pneumatic conveying system of a cotton picker based on digital twin includes the following steps:

[0103] Step S1: Establish a high-fidelity numerical simulation model for predicting the flow field characteristic data of the pneumatic conveying system of the cotton picker. Divide the flow field of the air outlet plane of the pneumatic conveying system and the developed plane flow field of the impeller ring into 3,154 nodes, retain the coordinate positions corresponding to each node, and obtain 10 sets of flow field data of each node in the flow field of the air outlet plane of the pneumatic conveying system and the developed plane flow field of the impeller ring of the cotton picker in the rotational speed range of 1,000 rpm to 5,000 rpm according to the simulation model;

[0104] Step S2: Preprocess the flow field simulation performance data of the cotton picker pneumatic conveying system obtained in Step S1, and use the Isolation Forest to identify and remove outliers and anomalies within the data to form the training data set for each node;

[0105] Step S3: Process the training data set in Step S2. Considering problems such as excessive data samples, cost, and noise, cluster the training data through the fuzzy clustering algorithm, and use the selected core data as the training data set;

[0106] Step S4: Establish a flow velocity prediction model for each node based on the Polynomial Chaos Expansion. Use the leave-one-out method to construct training samples and validation samples, and use the training data set obtained in Step S3 to construct a flow velocity prediction model for each node in the flow field of the cotton picker outlet plane and the expanded plane of the impeller ring. Then, use the validation set to verify the flow velocity prediction model, and use the coefficient of determination R 2 to evaluate the accuracy and reliability of the model, and store the trained node flow velocity prediction model in the model storage section for the digital twin system to call;

[0107] Step S5: Preprocess the prediction model obtained in Step S4, optimize the number of prediction models, optimize the rendering cloud map generation rate, reduce the rendering time, speed up the cloud map display speed, and call the prediction model to calculate and generate a real-time rendered cloud map display based on the rendering software;

[0108] Step S6: Build a visualization section based on digital twin. Based on the prediction model obtained in Step S5, establish a visualization interface display based on the software architecture. The display of the visualization interface includes real-time line chart displays such as the average flow velocity in the central area of the flow field, the highest flow velocity in the central area, the lowest flow velocity in the central area, the outlet flow rate, and the average working power of the fan, as well as real-time cloud map displays of the outlet plane flow field and the expanded plane of the impeller ring.

[0109] In Step S1, the flow velocity data of each node is calculated by using numerical analysis such as, but not limited to, computational fluid dynamics, finite element method, and finite volume method to obtain the flow field simulation data of the cotton picker pneumatic conveying system. As Figure 2 shown in the high-fidelity numerical simulation result diagram of the cotton picker fan, grid nodes are divided based on the simulation model, and the flow velocity data of multiple nodes are simulated, providing data support for the subsequent establishment of a flow field characteristic prediction model based on the Polynomial Chaos Expansion. The specific steps of the Isolation Forest for outlier identification in Step S2 include:

[0110] Step S2.1: Use a random hyperplane to cut an arbitrary data space to obtain two data subspaces, and then use a random hyperplane to cut the data subspaces. Repeat this operation until there is only one data point in each subspace;

[0111] Step S2.2: Construct an isolation forest composed of a specified number of isolated binary trees;

[0112] Step S2.3: Randomly select a number of sample points from the dataset D as the sample set D for generating a single isolated binary tree N ;

[0113] Step S2.4: Randomly select a feature f and a cut-off value p from the sample set D N ;

[0114] Step S2.5: If the maximum and minimum values of all samples contained in node N under feature f are f_max and f_min respectively, then p ∈ [f_min, f_max];

[0115] Step S2.6: If the value of the sample with respect to feature f is less than the cut-off value p, then divide the sample into the left node of node N, otherwise divide it into the right node;

[0116] Step S2.7: Repeatedly split the left and right child nodes of node N to generate an isolated binary tree;

[0117] Step S2.8: When there are multiple identical data or only one data in the child nodes or the isolated binary tree has reached the set maximum height, stop generating the isolated binary tree;

[0118] Step S2.9: Calculate the outlier value of d according to the path length h(d) of the test sample point d in each isolated binary tree. The smaller the path length, the higher the outlier score of the sample point:

[0119]

[0120] where m is the total number of sample points in the sample set D N ;

[0121] E(h(d)) is the average value of all path lengths h(d);

[0122] is the normalization of the tree height, H = ln k + ξ, k is the input value, and ξ = 0.577 is the Euler constant.

[0123] As Figure 3 shown in the schematic diagram of the case outlier recognition algorithm, the outlier recognition by the isolation forest solves the problem of outliers in the training model data.

[0124] The specific steps of sample screening by fuzzy clustering in step S3 are as follows:

[0125] Step S3.1: Construct the following cost function J(U, V) to divide the training dataset into c datasets:

[0126]

[0127] Among them, J is the cost function, representing the sum of the weighted distances from data points to their cluster centers during the entire clustering process;

[0128] n is the total number of data points, and n is equal to 3154;

[0129] c is the number of clusters, and the nodes are divided into 300 clusters;

[0130] x n = [x "n , x 2n ,..., x Nn T is each data in the training dataset,

[0131] x "n represents the first attribute in data x n ;

[0132] x 2n represents the second attribute in data x n ;

[0133] x rn represents the s-th attribute in data x n , where s is a positive integer;

[0134] v s is the class prototype of each class of data;

[0135] m is the fuzzy coefficient;

[0136] ‖·‖ 2 represents the Euclidean distance;

[0137] u in is the membership degree, representing the possibility that data x n belongs to the i-th class and satisfies the following conditions:

[0138]

[0139] Step S3.2: Construct the membership matrix Obtain the necessary iterative conditions for minimizing the clustering cost function J(U, V):

[0140]

[0141] Repeat the above formula iteratively until the difference between the elements of the membership matrix in the previous generation and the next generation is less than a certain set threshold or the maximum number of iterations is reached, and the calculation terminates to obtain the optimal data classification result;

[0142] ​Step S3.4: For each subclass, select the class prototype as the representative sample. After sample screening, the rotational speed-corresponding flow velocity data under each node is 1,675, reducing the computational load for model construction;

[0143] As Figure 4 shown in the schematic effect diagram of the clustering algorithm, by using the fuzzy clustering algorithm to screen representative samples, the problem of excessive data volume in the training model is solved.

[0144] Regard the clustering prototype obtained by data clustering as the core data, so as to achieve the purpose of reducing the sample capacity of training data and reducing the calculation cost.

[0145] Represent the core data set as V = {v " , v 2 , …, v q};

[0146] The above-mentioned step S4 for establishing the node flow velocity prediction model of the air conveying outlet flow field and the impeller ring unfolded plane flow field of the cotton picker includes the following steps:

[0147] Step S4.1: Construct the basis function y according to the input and output and the relationship conditions of their input and output. In this embodiment, the basis function y is constructed according to the fan rotational speed and the node flow velocity data;

[0148] Step S4.2: Calculate the unknown PC coefficients;

[0149] Furthermore, according to the relationship condition (i.e., the relationship between the fan rotational speed of the cotton picker and the air outlet flow velocity of the fan) of the input x and output y in step S3.1, the formula is as follows:

[0150]

[0151] Among them, y is the output, that is, the flow velocity of the basin node;

[0152] x is the input, that is, the fan rotational speed;

[0153] ψ α (x) is the tensor product of univariate orthogonal polynomials, satisfying the following formula:

[0154]

[0155] β α is the polynomial chaos coefficient (PC coefficient) to be solved;

[0156] Among them, the total number of unknown PC coefficients P satisfies the following formula:

[0157]

[0158] P is the maximum order of the input variable;

[0159] n is the dimension of the input variable;

[0160] is a polynomial of total order |α| not exceeding the given p-th order.

[0161] To calculate the PC value, it is necessary to find the polynomial chaos expansion PCE that satisfies the following formula:

[0162]

[0163] where N is the number of input-output samples;

[0164] The above formula can be rewritten as:

[0165] y = ψβ;

[0166] where,

[0167] refers to a real vector representing a P-dimensional vector;

[0168] β is the PC coefficient vector to be solved;

[0169]

[0170] y is the output response vector, representing a real vector of N dimensions;

[0171]

[0172] ψ is the measurement matrix, each column of which contains the evaluation values of the PC basis of N samples;

[0173] representing a real matrix of N×P dimensions;

[0174] The PC coefficients are calculated by minimizing the residual between the model response and the PCE approximation. For N≥P, the unknown coefficients are calculated using least squares regression, and the formula is as follows:

[0175] β = (ψ 9 ψ) :" ψ 9 y;

[0176] When N<P, the problem of solving the PC coefficients is transformed into an l1 minimization problem:

[0177] where the l2 norm constraint is used to consider the p-th order truncation error ∈ of the PCE;

[0178] The constrained l1 minimization problem in the above formula can be reformulated as a regularized (unconstrained) optimization problem:

[0179]

[0180] ∥β∥ " is the l1 norm of the PC coefficients;

[0181] is the degree of fit to the true model response in the sense of the l2 norm;

[0182] In step S4, the leave-one-out method generates training samples and validation samples from the constructed data set. Specifically, the leave-one-out method is a commonly used cross-validation method for evaluating the performance of a model. For the flow velocity prediction model of each node in the prediction of flow field characteristics, the present invention uses each sample in the data set as the validation set, and the remaining samples as the training set for training and validation. The process is repeated n times (n is the number of samples in the data set). In this embodiment, since the sample set for each node is 10, the process is repeated 10 times (10 is the number of samples in the data set). For all the determination coefficients R 2 of the validation results of each flow velocity prediction model are averaged as the final evaluation result of the performance of each model.

[0183] The determination coefficient R 2 is defined as follows:

[0184]

[0185] y i is the true output value of the i-th sample in the validation set;

[0186] is the predicted output value of the i-th sample in the validation set;

[0187] m is the sample size of the validation set;

[0188] is the average value of the true output values;

[0189] R 2 The closer R is to 1, the higher the accuracy of the prediction model. The average accuracy R 2 of the leave-one-out cross-validation of the node models is all above 0.95, meeting the actual requirements;

[0190] In step S5, the real-time response rate of the flow field rendering cloud map of the pneumatic conveying system of the cotton picker is optimized. The specific optimization mechanism steps are as follows:

[0191] Step S5.1: Call the flow velocity data of multiple nodes of the pneumatic conveying system flow field at different fan speeds in step S3, and retain the coordinate data corresponding to each node;

[0192] Step S5.2: Load the flow rate data and coordinate data of Step S5.1. Using the planar coordinate data as the input and the coordinate data corresponding to the node flow rate at a certain rotational speed of the fan as the output, construct prediction models at different rotational speeds based on polynomial chaos expansion;

[0193] Step S5.3: Load all the prediction models of Step S5.2. Based on the real-time response rate requirement of the contour map, divide both the abscissa and ordinate into 20 equal parts within the boundary range of the contour map (on the basis of weighing the response rate and the accuracy of the contour map, divide the boundary range into 20 equal parts), determine the abscissa and ordinate data values at the intersection, and generate 441 uniform node coordinates through pairwise combination of the abscissa and ordinate values. Using the coordinates of all uniform nodes as the input, load the prediction models and predict all data at each rotational speed respectively.

[0194] Step S5.4: Load all the data of Step S5.3. Using the rotational speeds of each fan as the input and the flow rate data corresponding to each node as the output, generate flow rate prediction models for 441 uniform nodes. According to the input rotational speed of the fan, call all the prediction models through rendering software and calculate the node flow rate values at each coordinate, and arrange the flow rate data in the contour map plane according to the corresponding coordinate order. Finally, use the rendering software to interpolate and generate the contour map according to the data value relationship in the plane.

[0195] Figure 5 The figure shows the rendering cloud map display effect of the outlet plane flow field, revealing the details of the outlet plane flow field velocity. The cloud map can intuitively present the uniformity of the flow field and the distribution of the turbulent flow region, facilitating the analysis of the key performance of the system operation. The specific process corresponds to Step S5 in this embodiment. Figure 6 It is the rendering cloud map of the impeller ring expanded plane flow field, revealing the details of the impeller ring expanded plane flow field velocity distribution. The cloud map can intuitively present the uniformity of the flow field and the distribution of the turbulent flow region, facilitating the analysis of the key performance of the system operation.

[0196] Such as Figure 7 It is the real-time line graph of the rotational speed of the cotton picker fan, the fan power, and the minimum flow rate of the central flow field, Figure 8 It is the real-time line graph of the comparison between the maximum flow rate, the minimum flow rate, and the average flow rate of the central flow field and the real-time measured flow rate. Figure 7 and Figure 8 The real-time line graph display effect of the key performance data of the flow field is shown, which can dynamically plot the fluctuations of the core performance indicators during the system operation, ensuring the stable operation of the pneumatic conveying system of the cotton picker.

[0197] The present invention solves the problem that the flow field characteristics of the pneumatic conveying system of the cotton picker cannot be quickly predicted and obtained, providing a reference basis for the regulation of the pneumatic conveying system.

[0198] Example 2

[0199] A system for implementing the method for predicting the flow field characteristics of the pneumatic conveying system of a cotton picker based on digital twin described in Example 1, and thus has the beneficial effects of Example 1, which will not be elaborated here.

[0200] The system for predicting the flow field characteristics of the pneumatic conveying system of a cotton picker based on digital twin includes a simulation data acquisition module, a data preprocessing module, a core sample screening module, a node flow velocity prediction model construction module, a rendering cloud map generation module, and a visualization module;

[0201] The simulation data acquisition module is used to establish a high-fidelity numerical simulation model for predicting the flow field characteristic data of the pneumatic conveying system of a cotton picker, and perform grid node division on the outlet flow field. By using numerical simulation software such as computational fluid dynamics, finite element method, and finite volume method, the flow field performance simulation data of the pneumatic conveying system of a cotton picker is obtained according to the simulation model;

[0202] The data preprocessing module is used to preprocess the obtained flow field simulation data of the pneumatic conveying system of a cotton picker, identify and remove outliers and anomalies inside the data, and form a training data set for each node;

[0203] The core sample screening module is used to process the obtained training data set. Considering problems such as excessive data samples, cost, and noise, the training data is clustered by a fuzzy clustering algorithm, and the screened core data is used as the training data set;

[0204] The node flow velocity prediction model construction module is used to establish a flow velocity prediction model for each node based on polynomial chaos expansion, use the leave-one-out method to construct training samples and validation samples, use the training data set to construct a flow velocity prediction model for each node in the outlet plane flow field and the impeller ring expansion plane flow field of the cotton picker, and use the validation set to verify the flow velocity prediction model; The node flow velocity prediction model construction module, based on the processed core data set, uses polynomial chaos regression to establish a flow velocity prediction model for each node, which is the key to the prediction principle of this digital twin system;

[0205] The rendering cloud map generation module is used to preprocess the obtained prediction model, optimize the number of prediction models, optimize the rendering cloud map generation rate, reduce the rendering time, speed up the display speed of the cloud map, and call the prediction model to calculate and generate a real-time rendered cloud map display based on the rendering software; The rendering cloud map generation module performs deletion based on the prediction model to speed up the generation of cloud map data;

[0206] The visualization module is used to build a visualization section based on digital twin. Based on the obtained prediction model, a visualization interface is established and displayed based on the software architecture. The flow field performance visualization module calls the flow field characteristic prediction model through the software architecture to generate a visualization interface, realizing real-time monitoring of the flow field performance.

[0207] It should be understood that although this specification is described according to various embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0208] The series of detailed descriptions listed above are only specific descriptions of the feasible embodiments of the present invention, and they are not intended to limit the protection scope of the present invention. Any equivalent embodiments or changes made without departing from the technical spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twins, characterized in that: The following steps are involved: Step S1, acquiring simulation data of the pneumatic conveying system of the cotton picker: establishing a high-fidelity numerical simulation model for predicting flow field characteristic data of the pneumatic conveying system of the cotton picker, and obtaining simulation data of flow field performance of the pneumatic conveying system of the cotton picker according to the simulation model; Step S2, data preprocessing: preprocessing the flow field simulation data of the pneumatic conveying system of the cotton picker obtained in step S1, identifying and removing wild values ​​and anomalies in the data, and forming a training data set for each node; Step S3, core sample screening: processing the training data set obtained in step S2, clustering the training data using a fuzzy clustering algorithm, and using the screened core data as the training data set; Step S4, node velocity prediction model construction: establish a velocity prediction model for each node based on polynomial chaos expansion, use the leave-one-out method to construct training samples and verification samples, use the training data set obtained in step S3 to construct a velocity prediction model for each node of the plane flow field of the cotton picker outlet and the plane flow field of the impeller ring expansion, and use the verification set to verify the velocity prediction model; Step S5, rendering cloud map generation: preprocessing the prediction model obtained in step S4, optimizing the number of prediction models, optimizing the rendering cloud map generation rate, reducing the rendering time, speeding up the cloud map display speed, calling the prediction model to calculate and generate a real-time rendering cloud map display based on the rendering software; Step S6, visualization: Build a visualization module based on the digital twin, based on the prediction model obtained in step S5, and establish a visualization interface display based on the software architecture.

2. The method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin according to claim 1 is characterized in that: In the step S1, the flow field performance simulation data of the pneumatic conveying system of the cotton picker is obtained according to the simulation model, which specifically includes the following steps: Step S1.1, establishing a high-fidelity numerical simulation model for predicting flow field characteristic data of a pneumatic conveying system of a cotton picker for the overall structure of a fan duct of the cotton picker; Step S1.2, dividing the mesh nodes for the plane flow field of the cotton picker air outlet and the plane flow field of the impeller ring expansion; Step S1.3, simulate and obtain the performance data of all grid nodes of the outlet plane flow field and the impeller ring expanded plane flow field under the corresponding fan operation data.

3. The method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin according to claim 1 is characterized in that: The step S1 calculates the flow velocity data of each node by computational fluid dynamics, finite element method or finite volume method to obtain the flow field simulation data of the pneumatic conveying system of the cotton picker.

4. The method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin according to claim 1 is characterized in that: In step S2, the isolation forest method is used to identify and remove outliers.

5. The method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin according to claim 1 is characterized in that: In step S3, a fuzzy clustering algorithm is used to prune samples, and the filtered core data is used as a training data set and saved in a database.

6. The method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin according to claim 1 is characterized in that: In step S4, a flow field characteristic prediction model is established based on polynomial chaos expansion, which specifically includes the following steps: Step S4.1, based on the processed data set, generate a training set and a validation set using the leave-one-out method; Step S4.2, using polynomial chaos expansion to establish a flow field characteristic prediction model; Step S4.3: Use the validation set to verify the performance prediction effect of the flow field characteristic prediction model, and use the determination coefficient R 2 To evaluate the accuracy and reliability of the model.

7. The method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin according to claim 6 is characterized in that: The step S4.2 of establishing the flow field characteristic prediction model based on the polynomial chaos expansion includes the following steps: Construct the basis function y according to the input and output and the relationship between them; Calculate the unknown polynomial chaotic PC coefficients; The relationship between input x and output y is expressed as follows: Among them, y is the output; x is the input; ψ α (x) is the tensor product of univariate orthogonal polynomials, satisfying the following formula: x i is the i-th variable of the input; β α is the unknown PC coefficient; Among them, the total number of unknown PC coefficients P satisfies the following formula: P is the maximum number of inputs; n is the dimension of the input variable; is a polynomial whose total order |α| does not exceed the given order p; In order to calculate the PC value, it is necessary to find a polynomial chaos expansion PCE that satisfies the following formula: Where N is the number of input-output samples; The above formula can be rewritten as: y = ψβ; in, Refers to a real vector representing a P dimension; β is the PC coefficient vector; y is the output response vector, Represents an N-dimensional real vector; ψ is the measurement matrix, each column of which contains the estimated values ​​of the PC basis of N samples; represents an N×P dimensional real matrix; The PC coefficients are calculated by minimizing the residual between the model response and the PCE approximation; For N ≥ P, the unknown coefficients are calculated using least squares regression, and the formula is as follows: β=(ψ T (ψ) -1 ψ T y; When N < P, the problem of solving the PC coefficient is transformed into a minimization problem: in The norm constraint is used to consider the p-order truncation error ∈ of PCE; The constraints in the above formula The minimization problem can be reformulated as a regularized (unconstrained) optimization problem: ||β||1 is the PC coefficient norm; For The degree of fit to the true model response in the norm sense.

8. The method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin according to claim 1 is characterized in that: In step S5, optimizing the rendering cloud image generation rate includes the following steps: Step S5.1, obtain the flow field data of multiple nodes at different wind turbine speeds, and retain the coordinate data of each node, load the flow field data and coordinate data, establish an agent model based on polynomial chaos expansion, take the coordinate data as input, and the node wind speed at a certain speed as output, and build a prediction model at different speeds based on polynomial chaos expansion; Step S5.2, load all the node velocity prediction models of step S5.1, divide the horizontal and vertical coordinates into k equal parts within the boundary of the cloud map according to the real-time response rate requirements of the cloud map, determine the intersection of the divisions as the horizontal and vertical coordinate values, combine and generate (k+1)×(k+1) uniform node coordinate values, use the coordinates of all uniform nodes as input, load the prediction model, and predict all the data at each speed respectively; Step S5.3, load all the data of the uniform nodes, take the speed of each fan as input and the flow velocity data corresponding to each node as output, generate the flow velocity prediction model of (k+1)×(k+1) uniform nodes, call all the prediction models through the rendering software according to the input fan speed and calculate the node wind speed value under each coordinate, and arrange the wind speed data to the cloud map plane according to the corresponding coordinate order, and finally use the rendering software to interpolate and generate the cloud map according to the numerical relationship of the data in the plane.

9. The method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twin according to claim 1 is characterized in that: The visualization interface display in step S6 includes real-time display of the average wind speed in the center area of ​​the flow field, the maximum wind speed in the center area, the minimum wind speed in the center area, the outlet flow rate, the average working power of the fan, and the real-time display of the outlet plane flow domain and the impeller ring expanded plane flow field.

10. A system for implementing the method for predicting flow field characteristics of a pneumatic conveying system of a cotton picker based on digital twins as described in any one of claims 1 to 9, characterized in that: It includes simulation data acquisition module, data preprocessing module, core sample screening module, node velocity prediction model building module, rendering cloud map generation module and visualization module; The simulation data acquisition module is used to establish a high-fidelity numerical simulation model for predicting the flow field characteristic data of the pneumatic conveying system of the cotton picker, and obtain the flow field performance simulation data of the pneumatic conveying system of the cotton picker according to the simulation model; The data preprocessing module is used to preprocess the obtained flow field simulation data of the pneumatic conveying system of the cotton picker, identify and remove wild values ​​and anomalies in the data, and form a training data set for each node; The core sample screening module is used to process the obtained training data set, cluster the training data through the fuzzy clustering algorithm, and use the screened core data as the training data set; The node velocity prediction model construction module is used to establish the velocity prediction model of each node based on polynomial chaos expansion, use the leave-one-out method to construct training samples and verification samples, use the training data set to build the velocity prediction model of each node of the plane flow field of the cotton picker outlet and the plane flow field of the impeller ring expansion, and use the verification set to verify the velocity prediction model; The rendering cloud map generation module is used to pre-process the obtained prediction model, optimize the number of prediction models, optimize the rendering cloud map generation rate, reduce the rendering time, speed up the cloud map display speed, and call the prediction model to generate real-time rendering cloud map display based on the rendering software calculation; The visualization module is used to build a visualization section based on the digital twin, based on the prediction model and establishing a visualization interface display based on the software architecture.

Images