Threshold optimized combine harvester load time domain extrapolation and load spectrum compilation method

Abstract

The invention provides a threshold-optimized combine harvester load time domain extrapolation and load spectrum compilation method, which comprises the following steps of: formulating a load spectrum acquisition scheme, and acquiring field operation time domain load of a combine harvester; carrying out preprocessing and stationary characteristic analysis on the time domain load; obtaining a stationary main load and a non-stationary trend load by using an empirical mode decomposition method; determining upper and lower thresholds in a load spectrum time domain extrapolation process by using a threshold optimization method; extracting a peak value and a valley value of a super-threshold part, and determining a super-threshold extreme value sample; and fitting an extreme point sample by using GPD to obtain an extrapolation main load, performing linear combination on the extrapolation main load and a trend load, and calculating to obtain a load spectrum of a target length. On the basis of obtaining the candidate threshold interval, the fitting test parameters and the residual square values of different candidate thresholds are analyzed, the optimal threshold is determined, and the extrapolation load spectrum is obtained; the method provides reference for load spectrum compilation, fatigue life prediction and reliability analysis of the combine harvester.

Description

Technical Field

[0001] This invention belongs to the field of agricultural machinery equipment load testing technology and experimental verification, specifically involving a method for threshold-optimized time-domain extrapolation of combine harvester load and load spectrum compilation. Background Technology

[0002] Accurately obtaining the field operation load spectrum of agricultural machinery is the key to predicting fatigue life. Theoretically, the load spectrum can be compiled by collecting load data throughout the entire life cycle of agricultural machinery. However, considering the test cost, the current method mainly measures the time-domain load data over a period of time and then obtains the full life load spectrum through extrapolation.

[0003] Currently, load spectrum extrapolation methods mainly include rainflow matrix extrapolation and time-domain extrapolation. Time-domain extrapolation has become a mainstream method for load spectrum compilation because it can preserve the temporal information of the load. The time-domain extrapolation method mainly uses the Peak Over Threshold (POT) model to extract extreme value samples from the measured load signal, and uses the Generalized Pareto Distribution (GPD) model to fit the extreme value samples to generate new extreme values ​​that meet the same distribution requirements. These new extreme values ​​replace the original extreme value samples, thus achieving load extrapolation to obtain the target length load spectrum.

[0004] However, in the current extrapolation process, when establishing extreme value samples, the threshold selection faces problems such as large differences in extrapolation results due to different candidate thresholds, lack of interpretability of the selected optimal threshold, and overly cumbersome threshold solution. Therefore, the threshold optimization method proposed in this invention provides a new approach for time-domain extrapolation and load spectrum compilation of combine harvesters, and provides support for fatigue life prediction and reliability analysis of combine harvesters. Summary of the Invention

[0005] To address the problems arising from the aforementioned threshold selection, the present invention aims to provide a method for time-domain extrapolation of combine harvester loads and load spectrum compilation based on threshold optimization, thereby solving the problems of inaccurate threshold selection, poor adaptability to non-stationary signals, and inaccurate load spectrum compilation in existing extrapolation techniques.

[0006] To address the above problems, the present invention provides the following technical solution:

[0007] A method for threshold-optimized time-domain extrapolation of combine harvester loads and compilation of load spectra includes the following steps:

[0008] S1. Develop a load spectrum acquisition scheme, select strain gauges as sensors for time-domain signal acquisition, preprocess and perform characteristic analysis on the acquired time-domain load, and analyze the temperature drift, power frequency interference, and stability characteristics of the test signal.

[0009] S1.1 Based on the characteristics of random alternating loads during the operation of the combine harvester, key parts such as the chassis, front and rear axles, and frame of the combine harvester are selected as measuring points. Strain signals are collected synchronously using strain gauges and strain rosettes, with the strain rosettes being attached at angles of 0°, 45°, and 90°.

[0010] S1.2 First, the load spectrum is preprocessed, including removing glitch and drift singular signal features.

[0011] Secondly, the Butterworth filtering method is used to filter the signal.

[0012] Next, calculate the maximum, minimum, mean, effective value, and standard deviation of the preprocessed time-domain load.

[0013] Finally, the statistical characteristics of the load signal are analyzed to determine its stationary properties.

[0014] S2. Use the empirical mode decomposition method to decompose the time-domain load and distinguish between the principal load and the trend load.

[0015] S2.1. The time-domain load signal is decomposed using the empirical mode decomposition method, as shown in Equation 1, to obtain n intrinsic mode components and the residual function:

[0016] Formula 1.

[0017] Where, X(t) For the payload time-domain signal, IMF i (t) is the th i Each intrinsic mode component, r( t ) is the residual function.

[0018] S2.2 Calculate the standard deviation, root mean square value, and kurtosis value of each intrinsic mode component, as shown in Formulas 2, 3, and 4:

[0019] Formula 2.

[0020] Where, t It is the standard deviation of each intrinsic mode component. x i It is the amplitude of each intrinsic mode component. α It is the mean of each intrinsic mode component.

[0021] Formula 3.

[0022] In the formula, RMS is the effective value of each intrinsic mode component. x i It is the amplitude of each intrinsic mode component.

[0023] Formula 4.

[0024] In the formula, K It is the kurtosis value of each intrinsic mode component. m It is the total number of peaks and troughs in the amplitude of each intrinsic mode component. n It is the number of intrinsic modal components. t It is the standard deviation of each intrinsic mode component. x i It is the amplitude of each intrinsic mode component. α It is the mean of each intrinsic mode component.

[0025] S2.3. Based on formulas 2, 3 and 4, analyze whether the standard deviation, root mean square value and kurtosis value of each intrinsic modal component change with time, and determine whether the load component contained in each intrinsic modal component is a stationary load or a non-stationary load.

[0026] Based on the determined number of stationary intrinsic mode components, the principal load X is calculated using Formula 5. main Based on the determined number of non-stationary eigenmode components, the eigenmode components of the non-stationary load are calculated using Formula 6, and used as the trend load X. trend :

[0027] Formula 5.

[0028] Formula 6.

[0029] In the formula, n is the number of each intrinsic mode component, X main ( t ) is the main load, X trend ( t ) is the trend load.

[0030] S3, Main load X obtained according to formula 5 main ( t The Mean Excess Function (MEF model) is used to analyze this data. e 1( m ), to obtain the interval where the candidate threshold is located.

[0031] Formula 7.

[0032] In the formula, D =( d 1, d 2, d 3,∙∙∙, d n Let be a random variable. m The threshold value is used.

[0033] S3.1, if D > m If the conditional distribution of is the GPD distribution, then the corresponding MEF function is: e 2( m ):

[0034] Formula 8.

[0035] In the formula, s It is a scale parameter. x It is a shape parameter.

[0036] S3.2, Define the threshold m Using Equation 8 as the independent variable, the graph of the excess mean function is calculated. The curve is approximately linear. If the curve slopes upward, the shape parameter of GPD is determined. x If the curve is positive, then the GPD data distribution is determined to be short-tailed. Based on the above method, the candidate threshold interval is defined as (μ1, μ2), and the interval containing the candidate threshold must satisfy two conditions:

[0037] (1) The candidate threshold does not belong to the short-tailed distribution interval.

[0038] (2) The selected threshold interval must satisfy the average excess approximate linear function.

[0039] S4. For the threshold interval (μ1, μ2) determined in S3.2, use the Least Squares Method (LSM) to fit all the threshold points included in the threshold interval.

[0040] Calculate the squared residuals at the fitted points, and select the threshold corresponding to the minimum squared residual as the optimal threshold μ. a。

[0041] The LSM method is used to fit all threshold points in the candidate threshold interval, and the residual squared value corresponding to each candidate threshold point is solved to obtain the residual squared equation, as shown in Equation 9.

[0042] Formula 9.

[0043] Solving Formula 9 yields the parameters. a 0 and a 1 The linear fitting equation is shown in Equation 10.

[0044] Formula 10.

[0045] S4.2 uses each candidate threshold as a threshold, extracts over-threshold samples using POT, fits the extreme value sample distribution using GPD, and uses R... 2 To evaluate the fit of the criteria.

[0046] By analyzing the relationship between the fitting effect and the squared residuals, the candidate threshold corresponding to the minimum squared residuals is identified as the optimal threshold, thus determining the best fitting effect and accurate extrapolation results.

[0047] Based on the above method, the candidate threshold corresponding to the minimum residual squared value is selected as the optimal upper limit threshold. m a1 and the optimal lower threshold m a2 。

[0048] S5, Optimal upper limit threshold obtained based on S4.2 m a1 and the optimal lower threshold m a2 POT (Point of Threshold) over-threshold sample extraction is performed on the time-domain signal to obtain the over-threshold distribution interval (μ). a1 ,μ max ) Sample and lower limit threshold distribution interval (μ min ,μ a2 )sample.

[0049] The GPD is selected as the extreme value sample distribution function, and the maximum likelihood estimation method is used to solve for the GPD shape parameters. x and scale parameters s This yields a fitted distribution function that matches the distribution of extreme value samples.

[0050] S6. Based on the fitted sample distribution function obtained in S5, extrapolate and reconstruct the extreme value samples exceeding the threshold to obtain the required extrapolated load; to determine the rationality of the extrapolated load, perform R-squared on the fitting results. 2 The fitting result is obtained by testing the samples exceeding the threshold obtained from the optimal threshold, and it needs to pass R. 2 The test result is greater than or equal to 0.9.

[0051] S6.1 When extrapolating and reconstructing the time-domain load, in order to preserve the time series information, when extracting the threshold in S4.2, the time corresponding to the extracted threshold is used, and the extrapolated extreme value and the corresponding time are replaced in the original load time series to obtain the extrapolated main load. Then, the extrapolated main load and the trend load are linearly superimposed to obtain the complete extrapolated load, and thus the target length load spectrum is obtained.

[0052] A load testing system for a combine harvester includes: strain gauges, a dynamic signal acquisition device, shielded wires, a signal conditioning module, a computer, a battery, and a combine harvester.

[0053] The dynamic signal acquisition device transmits signals through shielded wires to the strain gauges and strain rosettes, and is used to acquire, record, display and save strain data measured by the strain gauges and strain rosettes at different measurement points.

[0054] The battery powers the dynamic signal acquisition device, signal conditioning module, and host computer of the combine harvester during field operations.

[0055] The host computer communicates with the dynamic signal acquisition instrument via protocol, performs strain data preprocessing, data conversion, stationary signal characteristic statistics, and equivalent calculation of load spectrum, and outputs the load spectrum of the combine harvester.

[0056] Compared with existing technologies, the advantages of this invention are as follows:

[0057] 1. Since the measured load signal has non-stationary characteristics, the empirical mode decomposition method proposed in this invention can effectively solve the problem of distortion of time extrapolation results caused by non-stationary load signals.

[0058] 2. The MEF-LSM threshold optimization method proposed in this paper can effectively obtain the load spectrum of the target length of the combine harvester, providing an effective method for compiling the load spectrum and analyzing the fatigue life of the combine harvester under working conditions such as field harvesting and turning at the edge of the field.

[0059] The threshold optimization time-domain load extrapolation method proposed in this invention provides a reference for compiling the load spectrum of harvesting machinery, and provides a basis for accurately predicting the fatigue life and reliability analysis of agricultural machinery equipment. Attached Figure Description

[0060] Figure 1 A flowchart of a method for extrapolating the time domain load and compiling the load spectrum of a combine harvester with threshold optimization.

[0061] Figure 2 The diagram shows the empirical mode decomposition results of the measured load.

[0062] Figure 3 The graph shows the calculated standard deviation, root mean square, and kurtosis of each IMF signal component.

[0063] Figure 4 The figure shows the calculation results of the principal load and trend load based on empirical mode decomposition.

[0064] Figure 5 This is a schematic diagram of the preferred threshold.

[0065] Figure 6 This is a statistical characteristic diagram of IMF components.

[0066] Figure 7 The result of time-domain extrapolation for threshold optimization is shown in the figure.

[0067] Figure 8 The cumulative frequency of load cycles after extrapolation by 1 time for the MEF-LSM method.

[0068] Figure 9 The system diagram was developed for time-domain extrapolation of loads for combine harvesters, load spectrum testing, and compilation. Detailed Implementation

[0069] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0070] like Figure 1 As shown, a method for threshold-optimized time-domain extrapolation of combine harvester loads and compilation of load spectra includes the following steps:

[0071] S1. Develop a load spectrum acquisition scheme, select strain gauges as sensors for time-domain signal acquisition, preprocess and analyze the characteristics of the acquired time-domain load, analyze the temperature drift, power frequency interference, and stationary characteristics of the test signal, and obtain the stationary characteristics to further process the time-domain signal, providing a basis for the next step of time-domain signal extrapolation and reconstruction.

[0072] S1.1. Based on the complex environment and alternating load characteristics of the combine harvester during operation, the front and rear axles and chassis frame of the combine harvester are selected as measuring points. Strain gauges and strain rosettes are used to collect strain and collect strain signals. The strain rosettes are attached at angles of 0°, 45° and 90°.

[0073] S1.2 Preprocessing of the load spectrum includes not only deburring, de-drifting and filtering, but also calculating the maximum, minimum, effective value and standard deviation of the preprocessed load spectrum, analyzing the statistical characteristics of the load spectrum, and judging the quality and stationarity of the test signal.

[0074] S2. Use the empirical mode decomposition method to decompose non-stationary loads and distinguish between principal loads and trend load components.

[0075] S2.1, such as Figure 2 As shown, the time-domain load signal is decomposed using the empirical mode decomposition method to obtain n intrinsic mode components and a residual function.

[0076] Formula 1.

[0077] Where, X(t) For the payload time-domain signal, IMF i (t) is the th i Each intrinsic mode component, r( t ) is the residual function.

[0078] S2.2, such as Figure 3 As shown, calculate the standard deviation, root mean square, and kurtosis of each intrinsic mode component.

[0079] Formula 2.

[0080] Where, t It is the standard deviation of each intrinsic mode component. x i It is the amplitude of each intrinsic mode component. α It is the mean of each intrinsic mode component.

[0081] Formula 3.

[0082] In the formula, RMS is the root mean square value of each intrinsic modal component. x i It represents the amplitude of each intrinsic mode component.

[0083] Formula 4.

[0084] Where, K These are the kurtosis values ​​of each intrinsic mode component. m It is the total number of amplitude points for each intrinsic modal component. n It is the number of each intrinsic modal component. t It is the standard deviation of each intrinsic modal component. x i These are the amplitudes of each intrinsic modal component. α It is the mean of each intrinsic mode component.

[0085] S2.3, such as Figure 4 As shown, based on the standard deviation, root mean square, and kurtosis values ​​of each intrinsic modal component calculated using Formulas 2, 3, and 4, we analyze whether the load components contained in each intrinsic modal component are stationary, thereby determining the number of stationary intrinsic modal components. Then, we use Formula 5 to calculate the principal load X. main The remaining intrinsic mode components are calculated using Formula 6 and used as the trend load X. trend .

[0086] Formula 5.

[0087] Formula 6.

[0088] In the formula, n is the number of each intrinsic mode component, X main ( t ) is the main load, X trend ( t ) is the trend load.

[0089] S3, such as Figure 5 As shown, the main load X obtained according to Formula 5 main ( t ), and use the excess mean function on it. e 1( m ), to obtain the interval where the candidate threshold is located.

[0090] Formula 7.

[0091] Where, D = ( d 1, d 2, d 3,∙∙∙, d n Let be a random variable. m The threshold value is used.

[0092] S3.1, if D > m The conditional distribution of is a generalized Pareto distribution, then the corresponding average excess function is . e 2( m ).

[0093] Formula 8.

[0094] Where, s It is a scale parameter. x It is a shape parameter.

[0095] S3.2, For formula 8 m The independent variable analysis revealed that the curve of the excess mean function is approximately linear, and the candidate threshold interval is defined as (μ1, μ2). The curve slopes upwards, and the GPD shape parameter... x The result is positive; the curve slopes downwards, indicating that the data distribution originates from a short-tailed distribution.

[0096] Based on the above analysis, the interval containing the candidate threshold must satisfy two conditions:

[0097] (1) Thresholds that do not fall within the short-tailed distribution interval.

[0098] (2) The threshold that satisfies the average excess approximate linear function.

[0099] S4. For the threshold interval (μ1, μ2) determined in S3.2, use the least squares method to fit all the threshold points included in the threshold interval.

[0100] like Figure 6 The results are shown below. The squared residuals of the fitted points are calculated, and the threshold corresponding to the minimum squared residual is selected as the optimal threshold μ.a .

[0101] S4.1. Use the LSM method to fit all threshold points in the candidate threshold interval, solve for the residual squared value corresponding to each candidate threshold point, and obtain the residual squared equation, as shown in Formula 9.

[0102] Formula 9.

[0103] Solving Formula 9 yields the parameters. a 0 and a 1 The linear fitting equation is shown in Equation 10.

[0104] Formula 10.

[0105] S4.2 uses each candidate threshold as a threshold, extracts over-threshold samples using POT, fits the extreme value sample distribution using GPD, and uses R... 2 To evaluate the fit, the relationship between the fit and the squared residuals was analyzed. The candidate threshold corresponding to the minimum squared residual was identified as the optimal threshold, yielding the best fit and the most accurate extrapolation. Based on this principle, the candidate threshold corresponding to the minimum squared residual was selected as the optimal upper limit threshold. m a1 and the optimal lower threshold m a2 .

[0106] S5, Optimal upper limit threshold obtained based on S4.2 m a1 and the optimal lower threshold m a2 POT (Point of Threshold) over-threshold sample extraction is performed on the time-domain signal to obtain the over-threshold distribution interval (μ). a1 ,μ max ) Sample and lower limit threshold distribution interval (μ min ,μ a2 Sample; GPD is selected as the extreme value sample distribution function, and the maximum likelihood estimation method is used to solve for the GPD shape parameter. x and scale parameters s This yields a fitted distribution function that matches the distribution of extreme value samples.

[0107] S6. Based on the fitted sample distribution function obtained in S5, extrapolate and reconstruct the extreme value samples exceeding the threshold to obtain the required extrapolated load; to determine the rationality of the extrapolated load, perform R-squared on the fitting results. 2 The fitting results for the over-threshold samples obtained based on the optimal threshold passed the R-squared test. 2 The test result was 0.9987.

[0108] S6.1, such as Figure 7 As shown, when extrapolating and reconstructing the time-domain load, in order to preserve the time series information, when extracting the threshold in S4.2, the time corresponding to the extracted threshold is used, and the extrapolated extreme value and the corresponding time are replaced with the original load time series to obtain the extrapolated main load. Then, the extrapolated main load and the trend load are linearly superimposed to obtain the complete extrapolated load.

[0109] Based on S6.1, the target load spectrum length under various working conditions can be obtained by extrapolating the required length.

[0110] Analyzing the extrapolated load results, we obtain the following: Figure 8 The load cycle cumulative frequency diagram is shown.

[0111] A load time-domain extrapolation and load spectrum testing and compilation system for combine harvesters, such as Figure 9 The components shown include: strain gauges, strain rosettes, dynamic signal acquisition devices, shielded wires, signal conditioning modules, host computers, batteries, and combine harvesters.

[0112] The dynamic signal acquisition device transmits signals through shielded wires to the strain gauges and strain rosettes, and is used to acquire, record, display and save strain data measured by the strain gauges and strain rosettes at different measurement points.

[0113] The battery powers the dynamic signal acquisition device, signal conditioning module, and host computer of the combine harvester during field operations.

[0114] The host computer communicates with the dynamic signal acquisition instrument via protocol, performs strain data preprocessing, data conversion, stationary signal characteristic statistics, and equivalent calculation of load spectrum, and outputs the load spectrum of the combine harvester.

[0115] The above-described embodiments are merely one specific implementation of the present invention and should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, any modifications, equivalent substitutions, and improvements made within the scope of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for threshold-optimized time-domain extrapolation of combine harvester loads and compilation of load spectra, characterized in that, Includes the following steps: S1. Develop a load spectrum acquisition scheme, select strain gauges as sensors for time-domain signal acquisition, preprocess and perform characteristic analysis on the acquired time-domain load, and analyze the temperature drift, power frequency interference, and stability characteristics of the time-domain load of the test signal. S1.1 Based on the characteristics of random alternating loads during the operation of the combine harvester, key parts such as the chassis, front and rear axles, and frame of the combine harvester are selected as measuring points. Strain signals are collected synchronously using strain gauges and strain rosettes, with the strain rosettes being attached at angles of 0°, 45°, and 90°. S1.2 First, the load spectrum is preprocessed, including removing glitch and drift singular signal features; second, the signal is filtered using the Butterworth filtering method. Next, the maximum, minimum, mean, effective value and standard deviation of the preprocessed time-domain load are calculated, the statistical characteristics of the load signal are analyzed, and the stationary characteristics of the load signal are determined. S2. Decompose the time-domain load using the empirical mode decomposition method to distinguish between principal loads and trend loads; S2.

1. The time-domain load signal is decomposed using the empirical mode decomposition method, as shown in Equation 1, to obtain n intrinsic mode components and the residual function: Formula 1 In the formula, X(t) For the payload time-domain signal, IMF i (t) is the th i Each intrinsic mode component, r( t ) is the residual function; S2.2 Calculate the standard deviation, root mean square value, and kurtosis value of each intrinsic mode component, as shown in Formulas 2, 3, and 4: Formula 2 In the formula, τ It is the standard deviation of each intrinsic modal component. x i It is the amplitude of each intrinsic mode component. α It is the mean of each intrinsic mode component; Formula 3 In the formula, RMS is the effective value of each intrinsic mode component. x i It is the amplitude of each intrinsic mode component; Formula 4 In the formula, K It is the kurtosis value of each intrinsic mode component. m It is the total number of peaks and troughs in the amplitude of each intrinsic mode component. n It is the number of intrinsic modal components. τ It is the standard deviation of each intrinsic modal component. x i It is the amplitude of each intrinsic mode component. α It is the mean of each intrinsic mode component; S2.

3. Based on formulas 2, 3, and 4, analyze whether the standard deviation, root mean square value, and kurtosis value of each intrinsic modal component change with time, and determine whether the load components contained in each intrinsic modal component are stationary or non-stationary loads; based on the determined number of stationary intrinsic modal components, calculate the principal load X using formula 5. main Based on the determined number of non-stationary eigenmode components, the eigenmode components of the non-stationary load are calculated using Formula 6, and used as the trend load X. trend : Formula 5 Formula 6 In the formula, n is the number of intrinsic modal components, X main ( t ) is the main load, X trend ( t ) represents the trend load; S3, Main load X obtained according to formula 5 main ( t Using the mean function of the excess quantity e 1( μ ), calculate the interval of candidate thresholds required for the threshold method to extract samples exceeding the threshold; Formula 7 In the formula, D = ( d 1, d 2, d 3,∙··, d n Let be a random variable. μ For the threshold; S3.1, If ​​random variable D > μ The condition is that it conforms to a generalized Pareto distribution, and the corresponding average excess function is: e 2( μ ), Formula 8 In the formula, σ For scale parameters, ξ For shape parameters; S3.2, Define the threshold μ Using Equation 8 as the independent variable, the graph of the excess mean function is calculated. The curve is approximately linear. If the curve slopes upward, the shape parameter of GPD is determined. ξ If the curve is positive, then the GPD data distribution is determined to be short-tailed. Based on the above method, the candidate threshold interval is defined as (μ1, μ2), and the interval containing the candidate threshold must satisfy two conditions: 1) The candidate threshold does not belong to the short-tailed distribution interval; 2) The selected threshold interval must satisfy the approximate linear function of the average excess; S4. For the threshold interval (μ1, μ2) determined in S3.2, fit all threshold points in the threshold interval using the least squares method, calculate the squared residuals of the fitted threshold points, and select the smallest squared residual as the optimal threshold μ. a ; S4.

1. Use the LSM method to fit all threshold points in the candidate threshold interval, solve for the residual squared value corresponding to each candidate threshold point, and obtain the residual squared equation, as shown in Formula 9. Solving Formula 9 yields the parameters. a 0 and a 1 The linear fitting equation is shown in Equation 10. Formula 9 Formula 10 S4.2 uses each candidate threshold as a threshold, extracts over-threshold samples using the POT model, fits the extreme value sample distribution using the GPD function, and uses R... 2 As an evaluation criterion, the fitting results are tested; the relationship between the fitting results and the squared residuals is analyzed, and the evaluation criterion R is minimized when the squared residuals are minimized. 2 The candidate threshold corresponding to the maximum residual squared value is taken as the optimal threshold. Based on the above method, the candidate threshold corresponding to the minimum residual squared value is selected as the optimal upper limit threshold. μ a1 and the optimal lower threshold μ a2 ; S5, Optimal upper limit threshold obtained based on S4.2 μ a1 and the optimal lower threshold μ a2 The POT model is used to extract over-threshold samples of the time-domain signal, and the over-threshold distribution sample interval (μ) is obtained. a1 ,μ max ) and the sample interval of the lower limit threshold distribution (μ min ,μ a2 ); Select GPD as the extreme value sample distribution function, use the maximum likelihood estimation method to solve for the shape parameter and scale parameter of GPD, and calculate the fitting distribution function consistent with the distribution of samples exceeding the threshold; S6. Based on the fitted sample distribution function obtained in S5, extrapolate and reconstruct the extreme value samples exceeding the threshold to obtain the required extrapolated time-domain load; to determine the effectiveness of the extrapolated load, perform R-squared on the fitted extrapolated load results. 2 test; In step S6.1, when extrapolating and reconstructing the time-domain load, in order to preserve the time series information, when extracting the threshold in step S4.2, the time corresponding to the threshold is extracted, and the extrapolated extreme value and the corresponding time are replaced in the original load time series to obtain the extrapolated time-domain master load. Then, the extrapolated master load is linearly combined with the trend load to calculate the load spectrum of the target length that needs to be obtained.

2. A system for load time-domain extrapolation and load spectrum testing and compilation for a combine harvester, characterized in that, The method for threshold-optimized load extrapolation and load spectrum compilation of a combine harvester according to claim 1 includes a load spectrum testing and compilation system for the combine harvester comprising: strain gauges, strain rosettes, a dynamic signal acquisition device, shielded wires, a signal conditioning module, a host computer, a battery, and a combine harvester. The dynamic signal acquisition device transmits signals to the strain gauges and strain rosettes via shielded wires, and is used to acquire, record, display, and save strain data measured by the strain gauges and strain rosettes at different measuring points. The battery powers the dynamic signal acquisition device, signal conditioning module, and host computer during the combine harvester's field operation. The host computer communicates with the dynamic signal acquisition device via a protocol, performs strain data preprocessing, data conversion, stationary signal characteristic statistics, and equivalent load spectrum calculation, and outputs the load spectrum of the combine harvester.

Images