A method for compiling a load spectrum of a shovel working device fatigue life bench test procedure

Abstract

The present application relates to the field of engineering vehicles, in particular to a method for compiling a load spectrum of a shovel working device fatigue life bench test procedure, comprising: collecting load signals of key parts of a shovel in the field, and preprocessing the load signals; performing a fatigue test on the shovel, obtaining input loads and output loads of each measuring point, and constructing a training set; establishing a shovel load spectrum prediction model, which is based on an NARX model; identifying the training set through a FROLS algorithm, and training the shovel load spectrum prediction model using the identified training set; performing inverse virtual iteration using the preprocessed load signals to obtain a field input load spectrum; performing rainflow statistical counting on the field input load spectrum to obtain a two-dimensional load spectrum; and converting the two-dimensional load spectrum into a one-dimensional load spectrum based on the equivalent damage principle and the variable amplitude method. The present application realizes fast backstepping and spectrum compiling of a large engineering machinery overall working device fatigue test loading spectrum, and significantly improves efficiency and precision.

Description

Technical Field

[0001] This invention relates to the field of engineering vehicles, and specifically to a method for compiling a load spectrum for a fatigue life bench test procedure for an excavator working device. Background Technology

[0002] Hydraulic excavators are a mainstay of my country's construction machinery industry, widely used in mining, road construction, and disaster relief. As a common piece of engineering equipment used for soil excavation, hydraulic excavators experience complex and variable working loads during construction. Under long-term random loads, the main load-bearing components of the working device are prone to fatigue damage, affecting the excavator's normal operation. To analyze whether the working device under the current structure meets its design life, it is necessary to extract the load spectrum and perform fatigue life analysis.

[0003] Existing fatigue tests on excavator working devices mainly fall into two categories: one is a whole-machine fatigue test where the operator repeatedly digs, lifts, rotates, unloads, and slews until the working device fails due to fatigue; the other is a fatigue test conducted on separately designed fatigue test benches for different structures of the excavator working device. The former has a long test cycle, high investment costs, and is limited by environmental conditions, making it non-repeatable. The latter shortens the test time by increasing the rate of damage accumulation by increasing load amplitude, frequency, or test temperature. A prerequisite for bench fatigue testing is that the load in the test must reflect the actual load's impact on the structural fatigue life. However, when actually measuring the load on the excavator working device, it is often impossible to directly obtain the input load. The automotive industry commonly uses sensors placed in key parts such as wheels, engines, and bodies to collect vehicle driving signals on endurance roads through a fatigue data acquisition system. After filtering and deburring the measured signals, they are compiled into a program load spectrum that can be bench-loaded for bench testing, thereby analyzing and predicting the fatigue life of automotive components. However, this method involves dynamic loading, while excavator bench fatigue tests can only be performed with loading in a fixed posture. Therefore, this method is not suitable for excavators. Furthermore, most excavator working device fatigue life bench test programs use parametric extrapolation to develop load spectra. This requires the average amplitude distribution of the loads on various components of the working device to conform to certain rules, making it highly susceptible to subjective factors. Moreover, the parametric method only extrapolates the frequency of the load, failing to achieve bidirectional extrapolation of both load and frequency. Consequently, it may not accurately predict extreme loads that could significantly impact fatigue life throughout the entire lifespan. Summary of the Invention

[0004] To address the aforementioned problems of long testing cycles, high costs, and non-repeatability in existing excavator fatigue testing, the inability of bench tests to directly obtain input loads, the unsuitability of existing automotive industry methods for excavator fixed-post loading requirements, and the significant subjective influence of traditional parametric extrapolation methods, which cannot achieve bidirectional extrapolation of load and frequency and accurately predict extreme loads, this invention provides a method for compiling load spectra for excavator working device fatigue life bench tests. This invention primarily utilizes a NARX-based model to establish an excavator load spectrum prediction model, which is then identified and trained using the FROLS algorithm. The trained model is used to perform inverse virtual iteration on the preprocessed load signal to obtain the field input load spectrum. Rainflow load data is then obtained through rainflow statistical counting, and a two-dimensional load spectrum is obtained through extrapolation of the nonparametric rainflow matrix using two-dimensional kernel density estimation. Finally, based on the equivalent damage principle and the amplitude method, the two-dimensional load spectrum is converted into a one-dimensional load spectrum. This allows for the transfer of excavator working device fatigue life testing from the test field to the bench while ensuring damage equivalence between field and bench tests, facilitating fatigue life prediction of the working device. Furthermore, the new load spectrum compilation method solves the accuracy problem of the load spectrum in bench test programs.

[0005] The technical means employed in this invention are as follows:

[0006] A method for compiling a load spectrum for a fatigue life bench test procedure of an excavator's working device includes the following steps: The load signals of key parts of the excavator on site are collected and preprocessed. The excavator was subjected to fatigue tests to obtain the input load and the output load at each measuring point. A training set was then constructed based on the input load and the output load. An excavator load spectrum prediction model is established, which is based on the NARX model. The excavator load spectrum prediction model is identified using the FROLS algorithm, and the identified excavator load spectrum prediction model is trained using the training set. Based on the trained excavator load spectrum prediction model, the preprocessed load signal is used to perform inverse virtual iteration to obtain the field input load spectrum. Rainflow statistical counting is performed on the field input load spectrum to obtain rainflow load data. The nonparametric rainflow matrix extrapolation of the rainflow load data by two-dimensional kernel density estimation is then performed to obtain the two-dimensional load spectrum. Based on the principle of equivalent damage and the variable amplitude method, the two-dimensional load spectrum is converted into a one-dimensional load spectrum.

[0007] Further, the step of performing rainflow statistical counting on the field input load spectrum to obtain rainflow load data includes: Rainflow statistics and counting are performed on the field input load spectrum to obtain peak and valley values ​​and cycle number; The peak and valley values ​​and the number of cycles are statistically analyzed in a hierarchical manner to obtain the rainflow matrix.

[0008] Furthermore, the nonparametric rainflow matrix extrapolation of the two-dimensional kernel density estimation of the rainflow load data yields a two-dimensional load spectrum, including: The bandwidth of the rainflow matrix is ​​calculated using the following formula:

[0009] in, h For bandwidth, x i Peak value, y i It is the lowest value; The Epanechnikov kernel function is used to perform two-dimensional kernel density estimation on the rainflow matrix. The Epanechnikov kernel function is as follows:

[0010] in, This is the output of the Epanechnikov kernel function. This is the input to the Epanechnikov kernel function; Based on the bandwidth and the Epanechnikov kernel function, the rainflow matrix is ​​extrapolated nonparametrically to obtain a two-dimensional load spectrum.

[0011] Furthermore, the conversion of the two-dimensional load spectrum into a one-dimensional load spectrum based on the equivalent damage principle and the variable amplitude method includes: based on the equivalent damage principle and the variable amplitude method, with the total damage value as the objective, transferring loads with smaller damage contributions to loads with larger damage contributions to obtain a one-dimensional load spectrum.

[0012] Furthermore, the calculation process of the one-dimensional load spectrum includes: Based on the maximum value and minimum value at the start and inflection points of the two-dimensional load spectrum, the amplitude range and mean of the one-dimensional load spectrum are calculated. The formula for calculating the amplitude range is as follows:

[0013] in, R The range is defined as follows: max is the maximum value, min is the minimum value, and the mean is the average of the maximum and minimum values. The damage values ​​under different mean values ​​of the rainflow matrix amplitudes are calculated and summed to obtain the total damage value. The mean of the level with the greatest damage contribution is taken as the target mean. Based on the amplitude value method, the remaining damage is equivalent to the target mean to obtain the equivalent target mean and the corresponding amplitude levels. Divide the total damage value by the equivalent target mean to obtain the scaling factor for each level, and multiply the scaling factor by the number of cycles to obtain the equivalent number of cycles. Based on the target mean and the amplitudes corresponding to each amplitude level obtained after equivalent re-statistical analysis, the maximum and minimum values ​​of the one-dimensional load spectrum are calculated. The formula for calculating the maximum value of the one-dimensional load spectrum is as follows:

[0014] in, The maximum value of the one-dimensional load spectrum. The equivalent target mean, The formula for calculating the minimum value of the one-dimensional load spectrum, representing the equivalent amplitude range, is as follows:

[0015] in, This represents the minimum value of the one-dimensional load spectrum; A one-dimensional load spectrum is constructed based on the maximum value, minimum value, and equivalent number of cycles of the one-dimensional load spectrum.

[0016] Further, the preprocessing of the load signal includes: Small-amplitude load signals whose damage effect is negligible are removed from the load signals. The criteria for determining small-amplitude load signals are as follows: according to the stress-life curve of the material used in the excavator working device, when the stress amplitude corresponding to the load is lower than the preset damage threshold, it is determined to be a small-amplitude load signal that does not affect fatigue damage and is removed. A digital filter is used to filter the load signal after amplitude removal in order to overcome the error caused by random interference; The filtered load signal is then processed to remove glitches. These glitches are determined based on the signal standard deviation, which is calculated using the following formula:

[0017] in, The standard deviation of the signal. For the i-th numerical point, The mean of the signal. n For the total number of data points, when When the i-th numerical point is identified as a burr point, k’ This is the threshold coefficient.

[0018] Furthermore, the key components include the boom, stick, and bucket, and the load signal includes stress signal and strain signal.

[0019] Compared with the prior art, the present invention has the following advantages: (1) This invention addresses the problem that the load spectrum time-domain extrapolation compilation method is not applicable under the complex and multi-condition working conditions of excavators, and that the parameter extrapolation method is greatly affected by subjective factors, which restricts the accuracy of the extrapolation results. It proposes a program load spectrum compilation method applicable to the fatigue life bench test of excavator working device.

[0020] (2) The present invention uses data identification technology to deduce the bench loading load based on the stress and strain measured at the measuring points, which solves the problem that the virtual iteration method is not applicable to excavators, and thus can compile the loading spectrum of the fatigue bench test program for excavator working device.

[0021] (3) This invention solves the problem of reliability bench testing of excavator working device, and transfers the reliability test of excavator working device from the test field to the bench. It also solves the problem of complex stress and irregular load distribution of excavator during operation by using a new load spectrum compilation method.

[0022] Based on the above reasons, this invention can be widely promoted in fields such as engineering vehicles. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the method for compiling the load spectrum of a fatigue life bench test procedure for an excavator working device according to the present invention.

[0025] Figure 2 This is a flowchart illustrating the NARX model of the system based on the FROLS algorithm of this invention.

[0026] Figure 3 This is a flowchart of the rainflow counting process of the present invention.

[0027] Figure 4 This is a schematic diagram of the output load at each measuring point in this invention.

[0028] Figure 5 This is a diagram of stress measurement points.

[0029] Figure 6 A comparison diagram of the approximate load applied for the experiment and the inferred input load.

[0030] Figure 7 A comparison chart of the experimental output load and the target output load at each measuring point. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] like Figure 1 As shown, this invention provides a method for compiling the load spectrum of a fatigue life bench test procedure for an excavator's working device, comprising the following steps: S1. Collect load signals from key parts of the excavator on site and preprocess the load signals.

[0034] Key components included the boom, stick, and bucket, and load signals included stress and strain signals. Following the specifications for excavator whole-machine testing, strain gauges were placed at key components of the working device. Six stages of typical excavator operation—digging, lifting, rotating, unloading, slewing, and lowering—were tested, and stress and strain signals at these key components were obtained.

[0035] Preprocessing of load signals includes: Small-amplitude load signals with negligible damage impact are removed from the load signal. The criteria for determining small-amplitude load signals are: based on the stress-life curve (SN curve) of the material used in the excavator's working device, when the stress amplitude corresponding to the load is lower than a preset damage threshold, it is determined to be a small-amplitude load signal that does not affect fatigue damage and is therefore removed. A digital filter is used to filter the load signal after amplitude removal to overcome errors caused by random interference, removing burrs from the filtered load signal. Burrs are determined based on the signal standard deviation, which is calculated using the following formula:

[0036] in, The standard deviation of the signal. For the i-th numerical point, The mean of the signal. n For the total number of data points, when When the i-th numerical point is identified as a burr point, k’ This is the threshold coefficient.

[0037] S2. Conduct fatigue tests on the excavator to obtain the input load and the output load at each measuring point, and construct a training set based on the input load and the output load.

[0038] The excavator's working device was fixed on a fatigue testing bench for fatigue testing. Input load and data were acquired through a data acquisition system. Figure 5 The output loads corresponding to each measurement point are used as the training set for the data-driven model.

[0039] Specifically, random loads are applied to the bucket tip of the excavator on the test bench as input, and stress and strain signals measured at key parts of the working device are used as outputs, which serve as the training set for the data-driven model.

[0040] S3. Establish an excavator load spectrum prediction model, which is based on the NARX model.

[0041] Specifically, such as Figure 2 As shown, the NARX model, as a data-driven numerical model for identifying "black box" nonlinear systems, can establish an identification model of the system simply by using input and output data, and can be used to approximate any nonlinear dynamic system.

[0042] The NARX model can be implicitly defined as:

[0043] In the formula, d For a time delay, n u , n y The maximum time lag between the input and output sequences. It is an independent signal relative to the input and output signals.

[0044] Let x i (k) is:

[0045] When d=1, the expansion of the power polynomial of NARX can be expressed in the following form:

[0046] In the formula, t is the current sampling point, and θ i For each xi (k) Corresponding coefficients before the term.

[0047] The NARX model can be represented by the following linear regression form:

[0048] Furthermore, it can be written in the following matrix form:

[0049] in, y For the system output vector, P The regression matrix is ​​composed of candidate model terms. θ For X above i Correspondence between terms e It is an independent signal relative to the input and output signals.

[0050] In the formula, The system output vector is N, where N is the number of sampling points. Let be the model coefficient matrix, and P be the regression matrix composed of candidate model terms.

[0051]

[0052] S4. Identify the excavator load spectrum prediction model using the FROLS algorithm, and train the identified excavator load spectrum prediction model using the training set.

[0053] The parameter identification method introduces a model item reordering process, in which each step of the model item search process comprehensively filters the remaining model items, ultimately obtaining the simplest model structure.

[0054] Furthermore, such as Figure 3 As shown, FROLS, a standard algorithm for identifying the structure of nonlinear systems, has the following steps: Ignoring the effects of noise, the matrix form can be written in matrix form as follows:

[0055] In the formula, y is the system output vector. Representing model terms The sampled value at the sampling time.

[0056] The first step is to make ,calculate:

[0057]

[0058] The superscript (1) indicates the first step. This is the m-th orthogonality coefficient in the first step. Let m be the error reduction ratio of the m-th candidate.

[0059] make:

[0060] In the formula, This represents the value of the independent variable that maximizes the function, i.e. When the maximum value is obtained, the term in the above equation of the model is the l1th term, and it is set to be the first orthogonal vector of the orthogonal type. ,Right now ; Let there be a vector ,count . The coefficients of the first selected original model term. It is the first orthogonal vector.

[0061] Among them, the orthogonal type is:

[0062] The second step is to... For the column it belongs to, calculate:

[0063]

[0064]

[0065]

[0066] Let the above expression be the second orthogonal vector of the orthogonal form. ,Right now And count .

[0067] In the next step s, let Based on the already screened orthogonal vectors calculate:

[0068]

[0069]

[0070] make:

[0071] make Thus, the s-th orthogonal term is obtained. and calculate . Let m be the orthogonalized vector of the m-th candidate model term in step s. Let m be the orthogonality coefficient in step s. For the m-th element in step s, l s The index of the candidate with the largest ERR in step s. Let s be the s-th orthogonal vector. is the coefficient of the s-th selected original model term.

[0072] When the algorithm reaches step M0, it stops if the ESR meets the following condition.

[0073]

[0074] generally, . The proportion of unexplained error. M0 is the stopping threshold for the algorithm, and M0 is the number of model items finally selected.

[0075] The orthogonality of the model finally obtained by the FRLOS algorithm can be described by the following formula:

[0076] However, the above equation is not the final model expression we need. Therefore, the coefficients in the above equation need to be adjusted. Perform the inverse transform, i.e., the inverse Schmitt orthogonalization transform:

[0077]

[0078]

[0079] Transform into Then, the corresponding model term is . Let be the final coefficient of the m-th model term in the original NARX model. Let m be the m-th coefficient after orthogonalization. These are the transformation coefficients in the orthogonalization process.

[0080] Thus, after identification using the FROLS algorithm, the final expression of the NARX model is:

[0081] like Figure 4 As shown, taking the numerical model of measurement point CD_L1 as an example: the sampling frequency is... Based on the random input signal and the output signal of the measuring point, set the following parameters: maximum input time delay Output maximum time delay highest order The FROLS algorithm was used for system identification, and a multi-input single-output NARX model of the working device was established, as detailed below.

[0082]

[0083] Where y(t) is the output at time t, and u2(t), u3(t) and u4(t) are the load signals of different input channels, respectively.

[0084] S5. Based on the established NARX model, the inverse virtual iteration is performed using the preprocessed load signal to obtain the field input load spectrum.

[0085] S6. Perform rainflow statistical counting on the field input load spectrum to obtain rainflow load data. Extrapolate the nonparametric rainflow matrix of the rainflow load data by performing two-dimensional kernel density estimation to obtain the two-dimensional load spectrum.

[0086] Specifically, in this example, the "three-point method" is used for rainflow counting, and a 64×64 rainflow matrix is ​​used to store the statistical results. After performing rainflow statistical counting on the time-domain data of the loaded spectrum, peak and valley values ​​and cycle counts can be obtained. Hierarchical statistical analysis of the peak and valley values ​​and cycle counts yields a rainflow matrix in "From-to" format. S6 includes: The first step is to perform rainflow statistics and counting on the on-site input load spectrum to obtain peak and valley values ​​and the number of cycles.

[0087] The second step is to perform hierarchical statistics on peak and valley values ​​and the number of cycles to obtain the rainflow matrix.

[0088] The third step is to calculate the bandwidth of the rainflow matrix. The formula for calculating the bandwidth is:

[0089] in, h For bandwidth, x i Peak value, y i This is the lowest value.

[0090] Furthermore, assuming It is a random population Partial sample These are the experimentally measured values ​​for each sample. The measured values ​​are then sorted in ascending order to form a sequence. ,remember( x i , xi +1) is I i , f i Representative interval I i If the frequency of the measured value in the middle range accounts for the total number of measured values, then:

[0091] in, The empirical density function expression for the measured values. h i For bandwidth. Let:

[0092] but:

[0093] The simplified formula for the optimal bandwidth is:

[0094] Where A is a constant that depends only on the kernel function, and K is a constant factor determined by the kernel function K. This formula can be substituted for the formula for calculating bandwidth mentioned above.

[0095] Step 4: Perform two-dimensional kernel density estimation on the rainflow matrix using the Epanechnikov kernel function. The Epanechnikov kernel function is as follows:

[0096] in, This is the output of the Epanechnikov kernel function. This is the input to the Epanechnikov kernel function.

[0097] Step 5: Based on the bandwidth and Epanechnikov kernel function, perform nonparametric extrapolation of the rainflow matrix to obtain the two-dimensional load spectrum.

[0098] Furthermore, the rainflow matrix is ​​extrapolated using nonparametric kernel density estimation to obtain the two-dimensional load spectrum under typical operating conditions throughout the entire lifespan of the working device.

[0099] Specifically, by selecting an appropriate threshold (representing 20% ​​of the data range or 5% of the sample), the extreme values ​​in the domain are calculated using the statistical rainflow matrix, and the crossing intensity of these extreme values ​​is estimated to obtain the extrapolated load intensity above the threshold. Multiplying the limiting rainflow matrix by the corresponding scaling factor yields the two-dimensional load spectrum over the entire lifespan of the working device.

[0100] S7. Based on the principle of equivalent damage and the variable amplitude method, the two-dimensional load spectrum is converted into a one-dimensional load spectrum.

[0101] Based on the equivalent damage principle and the variable amplitude method, with the total damage value as the objective, loads with smaller damage contributions are transferred to loads with larger damage contributions, resulting in a one-dimensional load spectrum. The extrapolated "From-to" type rainwater basin load spectrum is then edited into a two-dimensional load spectrum, transforming it into a "Range-Mean" type rainwater basin load spectrum. Based on the damage equivalence principle, with the total damage value as the objective, damage with smaller contributions is transferred to loads with larger contributions, resulting in the bench test program load spectrum.

[0102] The calculation process of the one-dimensional load spectrum includes: Step 1: Based on the maximum and minimum values ​​at the start and inflection points of the two-dimensional load spectrum, calculate the amplitude range and mean of the one-dimensional load spectrum. The formula for calculating the amplitude range is:

[0103] in, R The range is defined by the amplitude, where max is the maximum value, min is the minimum value, and mean is the average of the maximum and minimum values.

[0104] The second step is to calculate the damage value under different mean values ​​of the rainflow matrix and sum them to obtain the total damage value.

[0105] The third step is to take the mean of the level with the greatest damage contribution as the target mean. Based on the variable amplitude method, the remaining damage is equivalent to the target mean to obtain the equivalent target mean and the corresponding amplitude levels.

[0106] Step 4: Divide the total damage value by the equivalent target mean to obtain the scaling factor for each level. Multiply the scaling factor by the number of cycles to obtain the equivalent number of cycles.

[0107] Step 5: Based on the target mean and the amplitudes corresponding to each amplitude level obtained after equivalent statistical analysis, calculate the maximum and minimum values ​​of the one-dimensional load spectrum. The formula for calculating the maximum value of the one-dimensional load spectrum is:

[0108] in, The maximum value of the one-dimensional load spectrum. The equivalent target mean, The formula for calculating the minimum value of the one-dimensional load spectrum, representing the equivalent amplitude range, is as follows:

[0109] in, This is the minimum value of the one-dimensional load spectrum.

[0110] Step 6: Construct a two-dimensional load spectrum based on the maximum and minimum values ​​and the equivalent number of cycles of the one-dimensional load spectrum.

[0111] During rainflow counting, the results of the "Range-Mean" type rainflow matrix with an 8×8 input load on the actuator long are directly extracted, as shown in Tables 1 and 2.

[0112] Table 1. Two-dimensional load spectrum of long-loaded actuators in the "Range-Mean" type.

[0113] Table 2. Actuator Long One-Dimensional Program Loading Spectrum

[0114] Through observation Figure 6 It can be seen that the approximate load captures the overall trend and general shape of the reverse-input load, but there are certain differences in detail and accuracy. The approximate load is applied to the actuator of the test bench, and the output load at each measuring point of the loader's working device is measured using a data acquisition device. Figure 6 As shown.

[0115] Through observation Figure 7 It can be clearly seen that there is a certain error between the experimental output load and the target output load at each measuring point. This is mainly because the input load derived in reverse cannot be directly applied to the actuator of the fatigue test bench. In order to further analyze the error between the experimental output load and the target output load, fatigue damage was calculated using nCode. During the calculation, the slope of the pseudo-damage SN curve was set to -3 and the intercept was 50000. The results are shown in Table 3.

[0116] Table 3 Comparison of output damage at each measuring point

[0117] from Figure 6 As can be seen from Table 3, the approximate load applied in the experiment captured the overall trend of the reverse-engineered input load and could accurately reproduce the strain of the loader during operation. This indicates that the method of reverse virtual iteration to reverse-engineer the input load of the loader's working device based on the NARX model and genetic algorithm is feasible.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for compiling a load spectrum for a bench test procedure of fatigue life of an excavator's working device, characterized in that, Includes the following steps: The load signals of key parts of the excavator on site are collected and preprocessed. The excavator was subjected to fatigue tests to obtain the input load and the output load at each measuring point. A training set was then constructed based on the input load and the output load. An excavator load spectrum prediction model is established, which is based on the NARX model. The excavator load spectrum prediction model is identified using the FROLS algorithm, and the identified excavator load spectrum prediction model is trained using the training set. Based on the trained excavator load spectrum prediction model, the preprocessed load signal is used to perform inverse virtual iteration to obtain the field input load spectrum. Rainflow statistical counting is performed on the field input load spectrum to obtain rainflow load data. The nonparametric rainflow matrix extrapolation of the rainflow load data by two-dimensional kernel density estimation is then performed to obtain the two-dimensional load spectrum. Based on the principle of equivalent damage and the variable amplitude method, the two-dimensional load spectrum is converted into a one-dimensional load spectrum.

2. The method for compiling the load spectrum of the fatigue life bench test procedure for excavator working device according to claim 1, characterized in that, The step of performing rainflow statistical counting on the field input load spectrum to obtain rainflow load data includes: Rainflow statistics and counting are performed on the field input load spectrum to obtain peak and valley values ​​and cycle number; The peak and valley values ​​and the number of cycles are statistically analyzed in a hierarchical manner to obtain the rainflow matrix.

3. The method for compiling the load spectrum of the fatigue life bench test procedure for excavator working device according to claim 1, characterized in that, The nonparametric rainflow matrix extrapolation of the two-dimensional kernel density estimation of the rainflow load data yields a two-dimensional load spectrum, including: The bandwidth of the rainflow matrix is ​​calculated using the following formula: in, h For bandwidth, x i Peak value, y i It is the lowest value; The Epanechnikov kernel function is used to perform two-dimensional kernel density estimation on the rainflow matrix. The Epanechnikov kernel function is as follows: in, This is the output of the Epanechnikov kernel function. x for Epanechnikov Input to the kernel function; Based on the bandwidth and the Epanechnikov kernel function, the rainflow matrix is ​​extrapolated nonparametrically to obtain a two-dimensional load spectrum.

4. The method for compiling the load spectrum of the fatigue life bench test procedure for excavator working device according to claim 1, characterized in that, The method of converting a two-dimensional load spectrum into a one-dimensional load spectrum based on the equivalent damage principle and the variable amplitude method includes: based on the equivalent damage principle and the variable amplitude method, with the total damage value as the objective, transferring loads with smaller damage contributions to loads with larger damage contributions to obtain a one-dimensional load spectrum.

5. The method for compiling the load spectrum of the excavator working device fatigue life bench test procedure according to claim 4, characterized in that, The calculation process of the one-dimensional load spectrum includes: Based on the maximum and minimum values ​​at the start and inflection points of the two-dimensional load spectrum, the amplitude range and mean of the one-dimensional load spectrum are calculated. The formula for calculating the amplitude range is as follows: in, R The range is defined as follows: max is the maximum value, min is the minimum value, and the mean is the average of the maximum and minimum values. The damage values ​​under different mean values ​​of the rainflow matrix amplitudes are calculated and summed to obtain the total damage value. The mean of the level with the greatest damage contribution is taken as the target mean. Based on the amplitude value method, the remaining damage is equivalent to the target mean to obtain the equivalent target mean and the corresponding amplitude levels. Divide the total damage value by the equivalent target mean to obtain the scaling factor for each level, and multiply the scaling factor by the number of cycles to obtain the equivalent number of cycles. Based on the target mean and the amplitudes corresponding to each amplitude level obtained after equivalent re-statistical analysis, the maximum and minimum values ​​of the one-dimensional load spectrum are calculated. The formula for calculating the maximum value of the one-dimensional load spectrum is as follows: in, The maximum value of the one-dimensional load spectrum. The equivalent target mean, The formula for calculating the minimum value of the one-dimensional load spectrum, representing the equivalent amplitude range, is as follows: in, This represents the minimum value of the one-dimensional load spectrum; A one-dimensional load spectrum is constructed based on the maximum value, minimum value, and equivalent number of cycles of the one-dimensional load spectrum.

6. The method for compiling the load spectrum of the fatigue life bench test procedure for excavator working device according to claim 1, characterized in that, The preprocessing of the load signal includes: Small-amplitude load signals whose damage effect is negligible are removed from the load signals. The criteria for determining small-amplitude load signals are as follows: according to the stress-life curve of the material used in the excavator working device, when the stress amplitude corresponding to the load is lower than the preset damage threshold, it is determined to be a small-amplitude load signal that does not affect fatigue damage and is removed. A digital filter is used to filter the load signal after amplitude removal in order to overcome the error caused by random interference; The filtered load signal is then processed to remove glitches. These glitches are determined based on the signal standard deviation, which is calculated using the following formula: in, The standard deviation of the signal. For the i-th numerical point, The mean of the signal. n For the total number of data points, when When the i-th numerical point is identified as a burr point, k’ This is the threshold coefficient.

7. The method for compiling the load spectrum of the fatigue life bench test procedure for excavator working device according to claim 1, characterized in that, The key components include the boom, stick, and bucket, and the load signals include stress signals and strain signals.

Images