Method and system for obtaining and monitoring stress of key components of a combine harvester leveling mechanism

Abstract

The application discloses a stress acquisition and monitoring method and system for key components of a combine harvester leveling mechanism. The method comprises the following steps: through numerical simulation analysis, the mapping relationship between the measurement points and the overall stress is determined, the positions of the measurement points of the key components are determined, and the nonlinear fitting of the overall physical model of the key components is realized; based on the positions of the measurement points of the key components, the overall stress prediction model of the key components is constructed, the linear and nonlinear relationships between the experimental data and the simulation data are approximated through the neural network, the high experimental data and simulation data are replaced, and the data of the overall model of the key components are quickly solved; based on the digital twin monitoring system, the structural health dynamic monitoring of the key components is realized, and the data visualization is performed. According to the application, the overall stress data of the key components are predicted based on a small amount of actual data of the measurement points, the health state of the agricultural machinery equipment is monitored in real time, and guidance and reference are provided for the health state monitoring of the agricultural machinery which is difficult to install a complex surface sensor and the like.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of structural health monitoring in agricultural operation, and particularly relates to a stress acquisition and monitoring method and system for key components of a leveling mechanism of a combine harvester. BACKGROUND

[0002] The lifting leveling mechanism of a combine harvester is a key mechanism for the combine harvester to enter and exit a field, and the structural strength thereof is the basis for supporting the combine harvester to work stably in the field. The swing arm is one of the important components of the four-point lifting leveling mechanism, and is prone to breakage and other conditions when the combine harvester works in the field, and is an important object of health management of the combine harvester. In order to ensure safe and stable operation of the leveling mechanism, it is necessary to monitor the leveling mechanism under different working conditions. The traditional stress of the swing arm is measured by arranging strain gauges in the test stage, but is affected by the installation convenience and strain measurement reliability: the complex surface and spatial structure constraints and blockages make it difficult to install strain gauges, so it is difficult to measure the easy-to-break points. SUMMARY

[0003] To solve the problems in the background art, the present application provides a stress acquisition and monitoring method for key components of a leveling mechanism of a combine harvester, comprising the following steps:

[0004] A dynamic simulation model of the combine harvester is established to determine the load and constraint borne by the key components; a statics simulation model is constructed for the key components based on the load and constraint, and overall stress simulation data of the key components are obtained according to the statics simulation model;

[0005] Initial measurement points are determined according to the overall stress simulation data, and key measurement points and corresponding positions are obtained in the initial measurement points;

[0006] A key component overall stress prediction model is constructed based on the key measurement points and corresponding positions, and the key component overall stress prediction model is a test data and simulation data joint driving model;

[0007] Overall stress prediction data are obtained by using the key component overall stress prediction model, and the overall stress prediction data are visualized based on a digital twin system to realize monitoring and prediction of the overall stress data of the key components.

[0008] In further embodiments, the key components include a front swing arm, a rear swing arm, a walking beam, a connecting rod and a connecting arm.

[0009] In further embodiments, the determination process of the corresponding key measurement points and corresponding positions of the key components in the initial measurement points is as follows:

[0010] Step 101, a predetermined number of candidate measurement points are randomly selected in the initial measurement points;

[0011] Step 102, constructing a mapping relationship between the candidate measurement points and the overall stress simulation data by using a neural network, the mapping relationship being a relationship between the stress of the candidate measurement points and the overall stress;

[0012] Step 103, performing fitting effect and reliability evaluation on the mapping relationship constructed in step 102 to obtain an evaluation result ;

[0013] Step 104, if , then randomly extracting the same number of candidate measurement points again, repeating step 102 until , the corresponding candidate measurement points being the key measurement points corresponding to the key components, and the coordinates of the key measurement points being the corresponding positions; wherein is a pre-set difference value.

[0014] In further embodiments, the construction step of the key component overall stress prediction model is as follows:

[0015] Step 201, recording the number of determined key measurement points as , selecting one of the key measurement points as an output measurement point, and the remaining key measurement points as input measurement points;

[0016] Step 202, taking the test stress data of the input measurement points and the spatial coordinates of the key components as input features, and taking the overall stress simulation data as output, to construct a simulation data set;

[0017] Step 203, taking the test stress data of the input measurement points and the coordinates of the output measurement point as input features, and taking the test stress data of the output measurement point as output, to construct a test data set;

[0018] Step 204, based on the simulation data set and the test data set, constructing an MFNN multi-fidelity deep neural network from the aspects of network structure design and loss function optimization, each layer output calculation of the MFNN multi-fidelity deep neural network complying with a neural network algorithm output function.

[0019] In further embodiments, the acquisition process of the initial measurement points is as follows:

[0020] According to the spatial coordinates of the key components, sample points are extracted in the overall stress simulation data;

[0021] For regular boundaries, the following formula is used to obtain the sample points extracted in the dimension :

[0022] , wherein is the index of the sample point, For the preliminary random generated permutation, the interval of each dimension is divided into equal probability sub-intervals, is a random number sampled from a uniform distribution ;

[0023] For irregular boundaries, the sample points of the irregular boundary of the key component are obtained using the following formula:

[0024] ; Wherein, is the distance of the irregular point;

[0025] The distance between the extracted sample point and the reference point on the irregular boundary is calculated, and the extracted sample point with the smallest distance from the reference point on the irregular boundary is retained; the unique function is used to remove duplicate sample points.

[0026] In further embodiments, the flow of the visualization process is as follows:

[0027] A color rendering platform for the key component is built using Unity software, and the color function is used to realize color assignment of the overall stress prediction value of the key component.

[0028] In further embodiments, the structure design of the MFNN multi-fidelity deep neural network includes: a simulation data deep neural network and a test data deep neural network ;

[0029] Wherein, the test data deep neural network includes: a first fully connected deep neural network and a second fully connected deep neural network , is the input test data set, is the response value of the simulation data set, and are unknown parameters of the first fully connected deep neural network and the second fully connected deep neural network respectively;

[0030] The first fully connected deep neural network is used to approximate the linear mapping relationship between simulation data and test data , and the second fully connected deep neural network is used to approximate the nonlinear mapping relationship between simulation data and test data ;

[0031] Correspondingly, the MFNN multi-fidelity deep neural network outputs overall stress prediction data. :

[0032] .

[0033] In a further embodiment, the optimization method for the loss function is as follows:

[0034] Establish deep neural networks for simulation data respectively loss function and experimental data deep neural networks loss function ;

[0035] Based on loss function and loss function The loss function of the multi-fidelity deep neural network is calculated. .

[0036] A stress acquisition and monitoring system for key components of a combine harvester leveling mechanism, used to implement the stress acquisition and monitoring method for key components as described above, includes:

[0037] The data simulation unit is set up to establish a dynamic simulation model of the combine harvester, determine the loads and constraints borne by the key components, construct a static simulation model of the key components based on the loads and constraints, and obtain the overall stress simulation data of the key components based on the static simulation model.

[0038] The key measurement unit is configured to determine the initial measurement point based on the overall stress simulation data, and to obtain the key measurement point and its corresponding position from the initial measurement point.

[0039] The data processing and prediction unit is configured to construct an overall stress prediction model for key components based on key measurement points and corresponding locations. The overall stress prediction model for key components is a model jointly driven by experimental data and simulation data.

[0040] The visualization unit is configured to obtain overall stress prediction data using the overall stress prediction model of the key component, and to perform visualization processing on the overall stress prediction data based on the digital twin system, thereby realizing the monitoring and prediction of the overall stress data of the key component.

[0041] In a further embodiment, it also includes:

[0042] Physical entity units are physical entity objects in a digital twin system, used for acquiring actual stress data at measurement points and mapping objects to the digital twin virtual model;

[0043] A data acquisition unit is configured to acquire stress data and corresponding response data of the measurement points of the physical entity unit.

[0044] The present application provides a method and system for acquiring and monitoring stress of key components of a combine harvester leveling mechanism.

[0045] The present application acquires stress data of the whole leveling mechanism arm through numerical simulation technology, determines the positions of key measurement points using neural network algorithm (DNN), and reduces the test cost; a multi-fidelity neural network algorithm is used to establish a prediction model of measurement points and the whole arm stress data, and the whole arm stress data is quickly predicted based on the stress of the measurement points and the coordinate information of the arm.

[0046] The present application uses digital twin technology and multi-fidelity deep neural network to predict the health status of key components of the combine harvester leveling mechanism, which can realize real-time monitoring of the stress of key components of the leveling mechanism, and solve the problem that the strain gauge installation is difficult due to the constraint and block of the complex surface and spatial structure of the combine harvester leveling mechanism, and the stress data of the key components is difficult to measure. The present application uses multi-fidelity deep neural network algorithm, and only a small amount of measurement points can predict the stress value of the whole key component. The multi-fidelity technology uses a small amount of test sample points to realize the overall adjustment of the mapping relationship preliminarily fitted by simulation data, which reduces the test cost and achieves health monitoring of the key component strain gauge which cannot be pasted, and provides a reference for health status monitoring of key components of the combine harvester leveling mechanism. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 FIG. 1 is a flowchart of the method for acquiring and monitoring stress of key components of a combine harvester leveling mechanism according to embodiment 1.

[0048] FIG. 2 (a) is a projection of the simulation data set sample scatter points on the xy plane.

[0049] FIG. 2 (b) is a projection of the simulation data set sample scatter points on the xz plane.

[0050] FIG. 2 (c) is a projection of the simulation data set sample scatter points on the zy plane.

[0051] Figure 3 FIG. 3 is a schematic diagram of the measurement point position of the arm according to embodiment 1.

[0052] Figure 4 is a visualization processing diagram of the swing arm of Example 1.

[0053] Figure 5 is an architecture diagram of the stress acquisition and monitoring system of the key components of the combine harvester leveling mechanism of Example 2.

[0054] Figure 3 in which the annotations are: first measurement point 1, second measurement point 2, third measurement point 3, fourth measurement point 4, fifth measurement point 5, sixth measurement point 6, seventh measurement point 7, and eighth measurement point 8. DETAILED DESCRIPTION

[0055] The application will be further described below in conjunction with the drawings and examples of the specification.

[0056] Example 1

[0057] The tracked combine harvester has the characteristics of good terrain adaptability, superior ability to cross ditches and bumps, and strong field operation passability, and is widely used in agricultural harvesting work in hilly areas. The lifting and leveling mechanism of the combine harvester is the key mechanism for the combine harvester to enter and exit the field, and its structural strength is the basis for supporting the stable work of the entire harvester in the field. The leveling mechanism is mainly composed of front and rear swing arms, hydraulic cylinders, walking beams, connecting rods, connecting arms, etc. By adjusting the length of the hydraulic cylinder, the entire machine forms a certain angle with the ground to achieve the work requirements, such as realizing the three typical working abilities of overall lifting, front and rear tilting, and left and right tilting. Considering the actual working conditions, the front hydraulic cylinder length change range for overall lifting is set to -65mm~65mm, and the rear hydraulic cylinder length remains unchanged; the rear hydraulic cylinder length change range for front and rear tilting is set to -45mm~75mm, and the front hydraulic cylinder length remains unchanged; the front hydraulic cylinder length change range for left and right tilting is set to 0~65mm.

[0058] The present example relates to a stress acquisition and monitoring method for key components of a combine harvester leveling mechanism, aiming to build a multi-fidelity deep neural network to realize the prediction of the overall stress data of key components through a small amount of measurement point test data, and to monitor the health status of agricultural machinery equipment in real time. Therefore, the key components in the present example include: front swing arm, rear swing arm, walking beam, connecting rod and connecting arm.

[0059] In order to better understand the present example, the key components will be taken as an example of the front swing arm of the leveling mechanism. As shown in Figure 1 The stress acquisition and monitoring method for key components of the combine harvester leveling mechanism includes the following steps:

[0060] A dynamic simulation model of the combine harvester is established to determine the load and constraint borne by the key components; a static simulation model of the key components is constructed based on the load and constraint, and the overall stress simulation data of the key components is obtained based on the static simulation model. Taking the front swing arm as an example, the dynamic simulation model is used to determine the load and constraint borne by the front swing arm, and the static simulation model is used to obtain the overall stress response data of the front swing arm.

[0061] Considering the large range of stress variation and the region with large stress value, initial measurement points are determined based on the overall stress simulation data, and the key measurement points and corresponding positions of the key components are obtained in the initial measurement points.

[0062] Based on the key measurement points and corresponding positions, an overall stress prediction model of the key components is constructed, which is a joint driving model of test data and simulation data. The overall stress prediction data is obtained by using the overall stress prediction model of the key components, and the visualization processing of the overall stress prediction data is performed based on the digital twin system, so as to realize the monitoring and prediction of the overall stress data of the key components. In combination with the above example, the input test samples include the front hydraulic cylinder length, the rear hydraulic cylinder length, the overall attitude angle of the combine harvester, the front swing arm measurement point stress data and the corresponding coordinate information, and correspondingly, the output is the stress prediction data of the front swing arm.

[0063] In order to ensure the accuracy of the input data, the determination process of the key measurement points and corresponding positions of the key components in the initial measurement points is as follows:

[0064] Step 101, a predetermined number of candidate measurement points are randomly selected in the initial measurement points;

[0065] Step 102, a mapping relationship between the candidate measurement points and the overall stress simulation data is constructed by using a neural network, and the mapping relationship is the relationship between the stress of the candidate measurement points and the overall stress;

[0066] Step 103, the mapping relationship constructed in step 102 is evaluated in terms of fitting effect and reliability to obtain an evaluation result ;

[0067] Step 104, if , the same number of candidate measurement points are randomly selected again, and step 102 is repeated until , the corresponding candidate measurement points are the key measurement points of the key components, and the coordinates of the key measurement points are the corresponding positions; wherein, is a predetermined difference value.

[0068] It should be noted that the overall stress simulation data acquisition process in the embodiment is as follows: through dynamic simulation analysis, the constraints and loads of the front swing arm under typical working conditions, i.e., different hydraulic cylinder lengths, are obtained. They include the rotary pair, the moving pair, the position of the swing arm in space, and the load force received by the contact part of the front swing arm.

[0069] According to the position of the front swing arm in space obtained from the dynamic simulation results, considering that the spatial positions of the front swing arm at different times are not uniform in the global coordinate system, it is not convenient to perform statics simulation analysis on the same point. Therefore, by adding a local coordinate system of force, statics simulation analysis is performed on the front swing arm as a whole under different loads in the same three-dimensional model; after determining the position of the swing arm in space in the statics simulation, the corresponding load force is added, and statics simulation is performed to obtain the stress data of the swing arm under typical working conditions.

[0070] Further example, pre-set , according to the comparative analysis of step 104, that is, if the determination coefficient is far away from 1, it means that the above steps need to be re-executed, and if the determination coefficient is close to 1, it means that it can be used as a key measurement point.

[0071] Further example, the number of initial measurement points is 12, and 8 candidate measurement points are randomly selected from the 12 initial measurement points, The candidate measurement point combination with the value closest to 1 is used as the final result. As shown in Figure 3 , the coordinates of the measurement points are: the first measurement point 1 (22.783, -16.41, -28.782), the second measurement point 2 (-23.584, -15.026, -27.993), the third measurement point 3 (-29, -31.598, -1.6423), the fourth measurement point 4 (-29, -10.676, 9.2945), the fifth measurement point 5 (-21.75, -25.876, 16.288), the sixth measurement point 6 (21, 75, -25.876, 16.288), the seventh measurement point 7 (-9.5837, -30.675, 16.178), and the eighth measurement point 8 (5.768, -77.9, -13.957).

[0072] In this embodiment, the construction steps of the key component overall stress prediction model are as follows:

[0073] Step 201, the number of determined key measurement points is denoted as , one of the key measurement points is selected as an output measurement point, and the remaining key measurement points are used as input measurement points;

[0074] Step 202, input the test stress data of the input measurement points and the spatial coordinates of the key components as input features, and the overall stress simulation data as output, to construct a simulation data set;

[0075] Step 203, input the test stress data of the input measurement points and the coordinates of the output measurement points as input features, and the test stress data of the output measurement points as output, to construct a test data set;

[0076] Step 204, based on the simulation data set and the test data set, construct a MFNN multi-fidelity deep neural network from the network structure design and loss function optimization aspects, and each layer output calculation of the MFNN multi-fidelity deep neural network follows a neural network algorithm output function.

[0077] The simulation data deep neural network includes a first fully connected deep neural network and a second fully connected deep neural network. The neural network algorithm model has an output function as follows:

[0078] , wherein it is assumed that The network layer has a total of neurons, the network layer has a total of neurons, represents a matrix composed of linear system groups of the network layer, and the shape size is ; represents a column vector composed of the bias of the network layer; represents a column vector of the unactivated linear output of the network layer; represents a column vector composed of the neuron output of the network layer; represents a column vector composed of the neuron output of the network layer, represents an activation function.

[0079] Further, a proxy model of the front swing arm measurement point stress, spatial coordinate information and the overall stress data of the front swing arm is constructed using the multi-fidelity deep neural network, and the test sample point cost is reduced.

[0080] Therefore, the structure design of the MFNN multi-fidelity deep neural network includes: a simulation data deep neural network and a test data deep neural network. The test data deep neural network includes a first fully connected deep neural network and a second fully connected deep neural network , The input test data set is a response value of the simulation data set, and unknown parameters of a first fully connected deep neural network and a second fully connected deep neural network respectively;

[0081] The first fully connected deep neural network is used to approximate a linear mapping relationship between the simulation data and the test data , and the second fully connected deep neural network is used to approximate a nonlinear mapping relationship between the simulation data and the test data .

[0082] Correspondingly, the MFNN multi-fidelity deep neural network outputs overall stress prediction data :

[0083] .

[0084] In combination Figure 4 , the input value of the simulation data deep neural network is , and the output value of the simulation data deep neural network .

[0085] Further, the optimization method of the loss function optimization is as follows:

[0086] The loss function of the simulation data deep neural network and the loss function of the test data deep neural network are established respectively;

[0087] Based on the loss function and the loss function , the loss function of the multi-fidelity deep neural network is calculated.

[0088] In order to facilitate better understanding of the present embodiment, the calculation formula of the loss function is as follows:

[0089] ; wherein, is the number of sample points of the simulation data set, is the prediction value of the simulation data deep neural network , , is the gradient value , and is the gradient value of .

[0090] Correspondingly, the loss function The calculation formula is as follows:

[0091] ; wherein, is the number of test sample points, is the predicted value of the test data deep neural network, is the output value of the test data deep neural network, corresponding to the measurement point, the selection process is as follows:

[0092] First, obtain the initial sample points;

[0093] According to the spatial coordinates of the key components, sample points are extracted in the overall stress simulation data;

[0094] For regular boundaries, the following formula is used to obtain the sample points extracted in the first dimension :

[0095] , wherein, is the sample point index, is a preliminary randomly generated permutation, and the interval of each dimension is divided into equal probability subintervals, is a random number sampled from a uniform distribution ;

[0096] For irregular boundaries, the following formula is used to obtain the sample points of the irregular boundary of the key component:

[0097] ; wherein, is the coordinate of the reference point on the irregular boundary, is the coordinate of the sample point to be evaluated, is the distance from the reference point to the sample point to be evaluated;

[0098] Calculate the distance between the extracted sample points and the reference points on the irregular boundary, and retain the extracted sample points with the smallest distance from the reference points on the irregular boundary; use the unique function to remove duplicate sample points.

[0099] Secondly, obtain the key measurement points, the detailed steps are as follows:

[0100] Step 101, randomly select a predetermined number of candidate measurement points from the initial measurement points;

[0101] Step 102, use a neural network to construct a mapping relationship between the candidate measurement points and the overall stress simulation data, wherein the mapping relationship is the relationship between the stress of the candidate measurement points and the overall stress; ​​

[0102] Step 103, fitting effect and reliability evaluation is performed on the mapping relationship constructed in step 102, and an evaluation result is obtained ;

[0103] Step 104, if , a same number of candidate measurement points are randomly extracted again, and step 102 is repeated until , the corresponding candidate measurement point is the key measurement point corresponding to the key component, and the coordinates of the key measurement point are obtained as the corresponding position; wherein is a pre-set difference value.

[0104] A color rendering platform of the key component is built by using Unity software, and color assignment of the stress prediction value of the key component is realized by using a color function (Csharp language). For example, the rendering of the rotating arm can be referred to Figure 4 .

[0105] Embodiment 2

[0106] In combination with Figure 5 , the embodiment discloses a key component stress acquisition and monitoring system of a combine harvester leveling mechanism, which is used to realize the key component stress acquisition and monitoring method as described in Embodiment 1, and comprises:

[0107] A data simulation unit is configured to establish a dynamic simulation model of the combine harvester, determine the load and constraint borne by the key component, and construct a static simulation model of the key component based on the load and constraint, and obtain overall stress simulation data of the key component according to the static simulation model.

[0108] A key measurement unit is configured to determine initial measurement points according to the overall stress simulation data, and obtain key measurement points and corresponding positions in the initial measurement points.

[0109] A data processing and prediction unit is configured to construct a key component overall stress prediction model based on the key measurement points and the corresponding positions, wherein the key component overall stress prediction model is a joint driving model of test data and simulation data.

[0110] A visualization unit is configured to obtain overall stress prediction data by using the key component overall stress prediction model, and perform visualization processing on the overall stress prediction data based on a digital twin system, so as to realize monitoring and prediction of the overall stress data of the key component.

[0111] In further embodiments, it further comprises:

[0112] A physical entity unit is a physical entity object of the digital twin system, which is used for acquisition of actual stress data of the measurement point and mapping object of the digital twin virtual model.

[0113] a data acquisition unit configured to acquire stress data and corresponding response data of the measurement points of the physical entity unit.

Claims

1. A method for obtaining and monitoring stress in key components of a combine harvester leveling mechanism, characterized in that, Includes the following steps: Establish a dynamic simulation model of the combine harvester to determine the loads and constraints borne by key components; A static simulation model of the key component is constructed under load and constraint conditions, and the overall stress simulation data of the key component is obtained based on the static simulation model. Based on the overall stress simulation data, the initial measurement points are determined, and key measurement points and their corresponding positions are obtained from the initial measurement points. A comprehensive stress prediction model for key components is constructed based on key measurement points and corresponding locations. The comprehensive stress prediction model for key components is a model jointly driven by experimental data and simulation data. The overall stress prediction data of the key components is obtained by using the overall stress prediction model of the key components, and the overall stress prediction data is visualized based on the digital twin system to realize the monitoring and prediction of the overall stress data of the key components.

2. The method for obtaining and monitoring stress in key components of the leveling mechanism of a combine harvester according to claim 1, characterized in that, The key components include: front swing arm, rear swing arm, traveling beam, connecting rod, and connecting arm.

3. The method for obtaining and monitoring stress in key components of the leveling mechanism of a combine harvester according to claim 1, characterized in that, The process of obtaining key measurement points and their corresponding locations from the initial measurement points is as follows: Step 101: Randomly select a predetermined number of candidate measurement points from the initial measurement points; Step 102: Construct a mapping relationship between the candidate measurement points and the overall stress simulation data using a neural network. The mapping relationship is the relationship between the stress of the candidate measurement points and the overall stress. Step 103: Evaluate the fitting effect and reliability of the mapping relationship constructed in Step 102, and obtain the evaluation results. ; Step 104, if Then, randomly select the same number of candidate measurement points again, and repeat step 102 until... The corresponding candidate measurement points are the key measurement points corresponding to the key components, and obtaining the coordinates of the key measurement points gives the corresponding positions; among them... The difference is a pre-set value.

4. The method for obtaining and monitoring stress in key components of the leveling mechanism of a combine harvester according to claim 1, characterized in that, The steps for constructing the overall stress prediction model for the key component are as follows: Step 201: Record the number of key measurement points as follows: Select one of the key measurement points as the output measurement point, and the remaining... One key measurement point is used as the input measurement point; Step 202: Use the test stress data of the input measurement points and the spatial coordinates of key components as input features, and the overall stress simulation data as output to construct a simulation dataset; Step 203: Construct an experimental dataset by using the test stress data of the input measurement points and the coordinates of the output measurement points as input features, and the test stress data of the output measurement points as output. Step 204: Based on the simulation dataset and the experimental dataset, construct the MFNN multifidelity deep neural network from the aspects of network structure design and loss function optimization. The output calculation of each layer of the MFNN multifidelity deep neural network follows the neural network algorithm output function.

5. The method for obtaining and monitoring stress in key components of the leveling mechanism of a combine harvester according to claim 1, characterized in that, The process of obtaining the initial measurement point is as follows: Sample points are extracted from the overall stress simulation data based on the spatial coordinates of key components. For the rule boundary, the following formula is used to obtain the first... Dimension Extraction Sample Points : ,in, For sample point index, For the initially randomly generated permutation, the intervals in each dimension are divided into:

1. Equal probability subintervals From uniform distribution Random numbers obtained from sampling; For irregular boundaries, the following formula is used to obtain sample points of the irregular boundaries of key components: ;in, The coordinates of the reference point on the irregular boundary. The coordinates of the sample points to be evaluated. Distance from the reference point to the sample point to be evaluated; Calculate the distance between the extracted sample point and the reference point on the irregular boundary, and retain the extracted sample point with the smallest distance to the reference point on the irregular boundary; use the unique function to remove duplicate sample points.

6. The method for obtaining and monitoring stress in key components of the leveling mechanism of a combine harvester according to claim 1, characterized in that, The visualization process is as follows: A color rendering platform for key components was built using Unity software, and the color function was used to assign color values ​​to the overall stress prediction values ​​of the key components.

7. The method for obtaining and monitoring stress in key components of the leveling mechanism of a combine harvester according to claim 4, characterized in that, The structural design of the MFNN multifidelity deep neural network includes: simulated data deep neural network. and experimental data deep neural networks ; Among them, the experimental data deep neural network Including: the first fully connected deep neural network Second fully connected deep neural network , For the input test dataset, The response values ​​of the simulation dataset, and These are the first fully connected deep neural networks. Second fully connected deep neural network Unknown parameters; The first fully connected deep neural network Used to approximate the linear mapping relationship between simulation data and experimental data. The second fully connected deep neural network Used to approximate the nonlinear mapping relationship between simulation data and experimental data. ; Correspondingly, the MFNN multi-fidelity deep neural network outputs overall stress prediction data. : 。 8. The method for obtaining and monitoring stress in key components of the leveling mechanism of a combine harvester according to claim 4, characterized in that, The optimization method for the loss function is as follows: Establish deep neural networks for simulation data respectively loss function and experimental data deep neural networks loss function ; Based on loss function and loss function The loss function of the multifidelity deep neural network is calculated. .

9. A stress acquisition and monitoring system for key components of a combine harvester leveling mechanism, used to implement the stress acquisition and monitoring method for key components as described in any one of claims 1 to 8, characterized in that, include: The data simulation unit is set up to build a dynamic simulation model of the combine harvester and determine the loads and constraints borne by key components. A static simulation model of the key component is constructed under load and constraint conditions, and the overall stress simulation data of the key component is obtained based on the static simulation model. The key measurement unit is configured to determine the initial measurement point based on the overall stress simulation data, and to obtain the key measurement point and its corresponding position from the initial measurement point. The data processing and prediction unit is configured to construct an overall stress prediction model for key components based on key measurement points and corresponding locations. The overall stress prediction model for key components is a model jointly driven by experimental data and simulation data. The visualization unit is configured to obtain overall stress prediction data using the overall stress prediction model of the key component, and to perform visualization processing on the overall stress prediction data based on the digital twin system, thereby realizing the monitoring and prediction of the overall stress data of the key component.

10. The stress acquisition and monitoring system for key components of the leveling mechanism of a combine harvester according to claim 9, characterized in that, Also includes: Physical entity units are physical entity objects in a digital twin system, used for acquiring actual stress data at measurement points and mapping objects to the digital twin virtual model; The data acquisition unit is configured to acquire stress data and corresponding response data at the measurement points of the physical entity unit.

Images