High-fidelity method, system and storage medium for gearbox model based on digital twin

Through multi-channel adaptive fidelity evaluation and multi-objective optimization algorithm for gray wolf particle swarm, the problem of incomplete fidelity and low accuracy of gearboxes is solved, and more efficient diagnosis and calculation results are achieved.

CN117592332BActive Publication Date: 2024-11-01SPECIAL EQUIP SAFETY SUPERVISION INSPECTION INST OF JIANGSU PROVINCE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311563535.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-21
Publication Date
2024-11-01
Estimated Expiration
2043-11-21

AI Technical Summary

Technical Problem

In the prior art, the fidelity of the gearbox is incomplete and the accuracy is low, resulting in poor diagnostic accuracy, and traditional particle swarm optimization algorithms are prone to fall into local optimal solutions, and the accuracy of optimization results is low.

Method used

Using multi-channel adaptive fidelity evaluation method combined with the gray wolf particle swarm multi-objective optimization algorithm, the high-fidelity model of the gearbox is constructed through finite element simulation, vibration test data acquisition and algorithm integration, so as to avoid local optimal solutions and improve the model Diversity and accuracy.

Benefits of technology

It realizes more comprehensive and accurate fidelity optimization of the gearbox model, improving diagnostic accuracy and accuracy of calculation results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117592332B_ABST
    Figure CN117592332B_ABST
Patent Text Reader

Abstract

The present invention discloses a high-fidelity method, system and storage medium for a gearbox model based on digital twins. This solution first establishes a rigid-flexible coupling finite element dynamic twin model for the healthy and faulty gearboxes, completes the gearbox modal analysis and dynamic simulation, preliminarily verifies the accuracy of the twin model, and arranges sensor measurement points according to the modal. Then, a gearbox vibration test bench is used to collect multi-channel vibration information, and a noise-reduced vibration image is generated in combination with the Vmd noise reduction method. A fusion image recognition algorithm is used to calculate the fidelity of the single-channel simulation and test vibration images. Finally, the twin model parameters are optimized based on the Gray Wolf Particle Swarm Multi-Objective Optimization Algorithm to achieve high-fidelity optimization of the model, and a multi-sensor channel adaptive evaluation method is used for fidelity evaluation. The multi-channel adaptive fidelity evaluation method proposed in the scheme of the present invention can help screen the optimal fidelity solutions for multiple fault conditions and multiple sensor channels, making the model fidelity more comprehensive and accurate.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of equipment digital twins, and in particular to a high-fidelity method, system and storage medium for a gearbox model based on digital twins. Background Art

[0002] Port equipment plays an important role in logistics loading and unloading and transportation operations. Its harsh working environment often leads to degradation, failure and damage of the structure due to fatigue and wear of key components. In view of the shortcomings of the "planned maintenance" inspection method of gearboxes, such as high labor costs, serious lags and poor diagnostic accuracy, the new physical-driven simulation technology can use the high synchronization of digital models and physical models to achieve efficient and accurate "condition-based maintenance" for structural monitoring, in which the fidelity of the digital model determines the accuracy of the diagnosis.

[0003] The commonly used fidelity method is to optimize through single monitoring point data without performing multi-monitoring point optimization, resulting in poor comprehensiveness of model fidelity; the traditional particle swarm-based optimization algorithm is prone to fall into local optimal solutions, resulting in low fidelity accuracy after optimization.

[0004] The multi-channel adaptive fidelity evaluation method proposed in the present invention can help screen the optimal fidelity solutions for multiple fault conditions and multiple sensor channels, making the model fidelity more comprehensive and accurate. The gray wolf particle swarm multi-objective optimization algorithm adopted in the present invention can prevent the particle swarm search from falling into the local optimal solution, increase the diversity of particle search, improve the accuracy of the calculation results, and thus improve the fidelity of the twin model. Summary of the invention

[0005] The present invention provides a high-fidelity method, system and storage medium for a gearbox model based on digital twins, which aims to solve the problems of incomplete fidelity and low accuracy disclosed in the background technology.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A high-fidelity method for a gearbox model based on digital twins, comprising:

[0008] S01. Use finite element simulation software to establish a rigid-flexible coupling twin model of the gearbox, then perform modal analysis, obtain dynamic monitoring points, and perform dynamic simulation;

[0009] S02. Based on the parallel fixed-axis gearbox vibration test bench, collect the vibration physical data of the gearbox in the time domain and frequency domain corresponding to multiple channels when the gearbox is in normal or multiple fault states;

[0010] S03, calculate the fidelity of the single-channel fusion algorithm for vibration physical data based on the Sift algorithm and the PHash algorithm;

[0011] S04. Combining the single-channel fidelity and the preset dynamics comprehensive index, construct a multi-channel fidelity adaptive comprehensive evaluation index;

[0012] S05. Select the parameter variables to be corrected in the rigid-flexible coupling twin model of the gearbox, and construct a response surface function of the parameter to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox;

[0013] S06. Based on the Grey Wolf Particle Swarm Fusion Multi-Objective Optimization Algorithm, the non-decomposition of the response surface function of the parameters to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox is solved. Combined with the multi-channel fidelity adaptive comprehensive evaluation index, the optimal solution is screened out to complete the fidelity optimization of the rigid-flexible coupling twin model of the gearbox.

[0014] As a possible implementation method, further, this solution S01 includes: using finite element simulation software to establish a flexible finite element twin model of the gear and the housing, replacing the rigid housing wall and gear components with flexible components, setting boundary conditions and model parameters, and establishing a rigid-flexible coupling twin model of the parallel fixed-axis gearbox, and then performing modal analysis on the flexible housing, arranging 6 monitoring points at the maximum amplitude for vibration response monitoring, and performing dynamic simulation on the rigid-flexible coupling twin model of the gearbox to verify the accuracy of the twin model.

[0015] As a better implementation option, preferably, in this scheme S02, based on the parallel fixed-axis gearbox vibration test bench, the vibration physical data of the time domain and frequency domain corresponding to 6 channels when the gearbox is in normal or multiple fault states are collected, and then the data are denoised by the Vmd noise reduction algorithm, and then converted into frequency domain signals by fast Fourier transform, and the corresponding test images are generated.

[0016] As a better implementation option, preferably, in this solution S03, a single-channel fusion algorithm fidelity calculation is performed on the test image corresponding to the vibration physical data based on the Sift algorithm and the PHash algorithm, and the obtained comprehensive fidelity result is used as the calculation result of the final virtual-real signal image fidelity.

[0017] As a preferred implementation option, preferably, this solution S03 includes:

[0018] The formula for calculating the fidelity of the test image corresponding to the vibration physical data by the Sift algorithm is as follows:

[0019] ε′ d =sim1(B1,B2)×100% (1)

[0020] Among them, sim1(B1,B2) represents the similarity value between the vibration frequency domain response image B1 of the rigid-flexible coupling twin model and the vibration frequency domain response image B2 of the physical data when using the Sift algorithm, ε′d represents the fidelity of the two images of this channel under the Sift algorithm, d represents the sensor channel number, d = [1, 2, 3, 4, 5, 6];

[0021] The formula for calculating the fidelity of the test image corresponding to the vibration physical data by the PHash algorithm is as follows:

[0022] ε″ d =sim2(B1,B2)×100% (2)

[0023] Among them, sim2(B1,B2) represents the similarity value between the vibration frequency domain response image B1 of the rigid-flexible coupling model and the vibration frequency domain response image B2 of the physical data when using the PHash algorithm, ε″ d Represents the fidelity of the two images of this channel under the PHash algorithm;

[0024] The calculation formula for the fidelity of the single-channel fusion algorithm regarding the vibration response is as follows:

[0025]

[0026] As a preferred implementation option, preferably, in this solution S04, combining the single-channel fidelity and the preset dynamics comprehensive index to construct a multi-channel fidelity adaptive comprehensive evaluation index includes:

[0027] The Delphi method is used to calculate the importance coefficient of signal indicators under different fault conditions;

[0028] By using time domain characteristic indicators and combining them with the indicator importance coefficient, we can establish comprehensive sensor indicators to evaluate the sensor's ability to collect important information.

[0029] By using the fidelity and sensor comprehensive indicators, an adaptive comprehensive evaluation index of multi-channel fidelity is constructed to adaptively evaluate the multi-channel fidelity and provide a basis for the fidelity optimization of the gearbox multi-channel model.

[0030] As a preferred implementation option, preferably, in this solution S05, selecting the parameter variables to be corrected in the rigid-flexible coupling twin model of the gearbox, and constructing the response surface function of the parameters to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox includes:

[0031] The central composite design method was used for response surface experimental design; the second-order polynomial was used for response surface function fitting, and the fitting formula was as follows:

[0032]

[0033] Among them, a0 is a constant, a i 、b i 、c ijare the unknown coefficients of the first-order, second-order and second-order cross terms, n is the number of parameters to be corrected, and the minimum number of tests is

[0034] As a better implementation option, preferably, in this solution S06, the speed and position update formula of the gray wolf particle swarm fusion multi-objective optimization algorithm is as follows:

[0035] V d (k+1)=ω·[V d (k)+C1r1(X 1d (k)-X d (k)) +C2r2(X 2d (k)-X d (k))+C3r3(X 3d (k)-X d (k))] (5)

[0036]

[0037] Among them, r1, r2, r3 are random numbers in [0,1], ω is an adjustable inertia weight, and its range is [0.5,1]. C1, C2, C3 are calculated as follows:

[0038] C=2·n2 (7)

[0039] Due to the change in the above formula and the introduction of ω, the wolf pack position is updated as follows:

[0040]

[0041]

[0042] Among them, P1, P2, and P3 represent the position vectors of α wolf, β wolf, and δ wolf after update, and P(t+1) represents the position vector determined after optimization.

[0043] Based on the above, the present invention also provides a high-fidelity gearbox model system based on digital twin, which is applied with the above-mentioned high-fidelity gearbox model method based on digital twin.

[0044] Based on the above, the present invention also provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the high-fidelity method of the gearbox model based on digital twins as described above.

[0045] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art: the present invention proposes a multi-channel adaptive fidelity evaluation method, which can help screen the optimal fidelity solutions for multiple fault conditions and multiple sensor channels, making the twin model fidelity more comprehensive and accurate. In addition, the present invention adopts the gray wolf particle swarm multi-objective optimization algorithm, which can avoid the particle swarm search from falling into the local optimal solution, increase the diversity of particle search, improve the accuracy of the calculation results, and thus improve the fidelity of the twin model. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 This is a brief implementation flow chart of the high-fidelity method of the gearbox model based on digital twins in this solution;

[0048] Figure 2 This is the high-fidelity optimization flow chart of the gearbox model of this scheme;

[0049] Figure 3 A brief flowchart of the Sift algorithm mentioned in this solution;

[0050] Figure 4 A brief flowchart of the PHash algorithm mentioned in this solution;

[0051] Figure 5 It is the flow chart of determining the index importance coefficient by Delphi method;

[0052] Figure 6 This is the flow chart of the scheme through multi-channel adaptive high-fidelity evaluation;

[0053] Figure 7 This is the overall equipment diagram of the gearbox vibration test bench of this scheme;

[0054] Figure 8 This is the diagram of the data acquisition equipment for the gearbox vibration test in this scheme;

[0055] Fig. 9 This is the overall diagram of the gearbox equipment in this scheme;

[0056] Fig.10 This is the flow chart of the optimization fidelity response surface model using the Grey Wolf Particle Swarm Fusion Multi-Objective Optimization Algorithm in this scheme. DETAILED DESCRIPTION

[0057] The present invention will be further described in detail below in conjunction with the accompanying drawings and examples. It is particularly noted that the following examples are only used to illustrate the present invention, but are not intended to limit the scope of the present invention. Similarly, the following examples are only partial embodiments of the present invention rather than all embodiments, and all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0058] like Figure 1 , Figure 2 As shown, this embodiment provides a high-fidelity method for a gearbox model based on digital twins, which includes:

[0059] S01. Use finite element simulation software to establish a rigid-flexible coupling twin model of the gearbox, then perform modal analysis, obtain dynamic monitoring points, and perform dynamic simulation. In this step, the monitoring point positions are selected for sensor arrangement through modal analysis, and multi-point vibration responses are obtained by dynamic simulation.

[0060] S02. Based on the parallel fixed-axis gearbox vibration test bench, collect the vibration physical data of the gearbox in the time domain and frequency domain corresponding to multiple channels when the gearbox is in normal or multiple fault states;

[0061] S03, calculate the fidelity of the single-channel fusion algorithm for vibration physical data based on the Sift algorithm and the PHash algorithm;

[0062] S04. Combining the single-channel fidelity and the preset dynamics comprehensive index, construct a multi-channel fidelity adaptive comprehensive evaluation index;

[0063] S05. Select the parameter variables to be corrected in the rigid-flexible coupling twin model of the gearbox, and construct a response surface function of the parameter to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox;

[0064] S06. Based on the Grey Wolf Particle Swarm Fusion Multi-Objective Optimization Algorithm, the non-decomposition of the response surface function of the parameters to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox is solved. Combined with the multi-channel fidelity adaptive comprehensive evaluation index, the optimal solution is screened out to complete the fidelity optimization of the rigid-flexible coupling twin model of the gearbox.

[0065] This solution can also build twin models of healthy and faulty gearboxes, use the fidelity method to optimize, and compare the optimization results of the two models to verify the accuracy of model establishment and the feasibility of the fidelity optimization method.

[0066] In this scheme S01, finite element simulation software is used to establish a flexible finite element twin model of the gear and housing. The rigid housing wall and gear components are replaced with flexible components. The boundary conditions and model parameters are set to establish a rigid-flexible coupling twin model of the parallel fixed-axis gearbox. Then, a modal analysis is performed on the flexible housing. Six monitoring points are arranged at the point with the maximum amplitude to monitor the vibration response. At the same time, a dynamic simulation is performed on the rigid-flexible coupling twin model of the gearbox to verify the accuracy of the twin model.

[0067] In this solution S02, based on the parallel fixed-axis gearbox vibration test bench, the vibration physical data of the time domain and frequency domain corresponding to 6 channels when the gearbox is in normal or multiple fault states are collected, and then the data are denoised by the Vmd noise reduction algorithm, and then converted into frequency domain signals by fast Fourier transform, and the corresponding test images are generated.

[0068] In this solution S03, the fidelity of the single-channel fusion algorithm is calculated for the test image corresponding to the vibration physical data based on the Sift algorithm and the PHash algorithm, and the obtained comprehensive fidelity result is used as the calculation result of the final virtual-real signal image fidelity.

[0069] This solution S03 includes:

[0070] The formula for calculating the fidelity of the test image corresponding to the vibration physical data by the Sift algorithm is as follows:

[0071] ε′ d =sim1(B1,B2)×100% (1)

[0072] Among them, sim1(B1,B2) represents the similarity value between the vibration frequency domain response image B1 of the rigid-flexible coupling twin model and the vibration frequency domain response image B2 of the physical data when using the Sift algorithm, ε′ d represents the fidelity of the two images of this channel under the Sift algorithm, d represents the sensor channel number, d = [1, 2, 3, 4, 5, 6];

[0073] The formula for calculating the fidelity of the test image corresponding to the vibration physical data by the PHash algorithm is as follows:

[0074] ε″ d =sim2(B1,B2)×100% (2)

[0075] Among them, sim2(B1,B2) represents the similarity value between the vibration frequency domain response image B1 of the rigid-flexible coupling model and the vibration frequency domain response image B2 of the physical data when using the PHash algorithm, ε″ d Represents the fidelity of the two images of this channel under the PHash algorithm;

[0076] The calculation formula for the fidelity of the single-channel fusion algorithm regarding the vibration response is as follows:

[0077]

[0078] In this solution S04, the single-channel fidelity and the preset dynamics comprehensive index are combined to construct a multi-channel fidelity adaptive comprehensive evaluation index including:

[0079] The Delphi method is used to calculate the importance coefficient of signal indicators under different fault conditions;

[0080] By using time domain characteristic indicators and combining them with the indicator importance coefficient, we can establish comprehensive sensor indicators to evaluate the sensor's ability to collect important information.

[0081] By using the fidelity and sensor comprehensive indicators, an adaptive comprehensive evaluation index of multi-channel fidelity is constructed to adaptively evaluate the multi-channel fidelity and provide a basis for the fidelity optimization of the gearbox multi-channel model.

[0082] In this solution S05, the parameter variables to be corrected in the rigid-flexible coupling twin model of the gearbox are selected, and the response surface function of the parameters to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox is constructed, including:

[0083] The central composite design method was used for response surface experimental design; the second-order polynomial was used for response surface function fitting, and the fitting formula was as follows:

[0084]

[0085] Among them, a0 is a constant, a i , b i 、c ij are the unknown coefficients of the first-order, second-order and second-order cross terms, n is the number of parameters to be corrected, and the minimum number of tests is

[0086] In this solution S06, the speed and position update formulas of the gray wolf particle swarm fusion multi-objective optimization algorithm are as follows:

[0087] V d (k+1)=ω·[V d (k)+C1r1(X 1d (k)-X d (k))+C2r2(X 2d (k)-X d (k))+C3r3(X 3d (k)-X d (k))] (5)

[0088]

[0089] Among them, r1, r2, r3 are random numbers in [0,1], ω is an adjustable inertia weight, and its range is [0.5,1]. C1, C2, C3 are calculated as follows:

[0090] C=2·n2 (7)

[0091] Due to the change in the above formula and the introduction of ω, the wolf pack position is updated as follows:

[0092]

[0093]

[0094] Among them, P1, P2, and P3 represent the position vectors of α wolf, β wolf, and δ wolf after update, and P(t+1) represents the position vector determined after optimization.

[0095] As an implementation example, this solution is further explained with the following example:

[0096] like Figure 1 The schematic diagram and attached diagram of a high-fidelity method for a gearbox model based on digital twins are shown in Figure 1. Figure 2 The high-fidelity optimization flow chart of the gearbox model shown in the figure, this embodiment includes:

[0097] Based on finite element dynamics simulation software, a rigid-flexible coupling twin model of the gearbox is jointly established to perform modal analysis and dynamics simulation. The monitoring point locations are selected through modal analysis for sensor layout, and dynamics simulation is used to obtain multi-point vibration responses.

[0098] Based on the parallel fixed-axis gearbox vibration test bench, collect multi-channel time domain and frequency domain physical data under various fault conditions of the gearbox;

[0099] The single-channel fidelity is calculated based on the fusion of Sift algorithm and PHash algorithm; the multi-channel comprehensive fidelity adaptive evaluation index is constructed by combining the single-channel fidelity and the preset sensor comprehensive index;

[0100] Construct a response surface model of the twin model parameter variables to be corrected and the fidelity of the vibration response;

[0101] Based on the Gray Wolf Particle Swarm Fusion Multi-Objective Optimization Algorithm, the non-fragmentation of the response surface function of the twin finite element dynamics model is solved, and the optimal solution is screened out by combining the comprehensive fidelity adaptive evaluation index to complete the fidelity optimization of the rigid-flexible coupling model;

[0102] By constructing twin models of healthy and faulty gearboxes, optimizing them using the fidelity method, and comparing the optimization results of the two models, the accuracy of model establishment and the feasibility of the fidelity optimization method can be mutually verified.

[0103] As attached Figure 3 As shown in the Sift algorithm flow chart, this embodiment includes using the Sift algorithm to calculate the difference between the local features of the single-channel simulation and experimental vibration images, constructing a scale space based on a scale-invariant feature conversion algorithm, detecting spatial extreme points, generating key feature point positions, determining key feature point directions, generating descriptors, and matching and comparing key feature points.

[0104] Among them, the scale space construction can use Gaussian kernel function and image convolution, as shown in the formula:

[0105] L(x,y,σ)=G(x,y,σ)×I(x,y) (10)

[0106] Where σ represents the spatial scale parameter, and G(x, y, σ) is the Gaussian kernel function, as shown below:

[0107]

[0108] In order to make the key points rotationally invariant, the main direction of the key points is specified by the gradient direction characteristics of the domain pixels, where the gradient is represented by the direction in which the grayscale of the point changes fastest. The gradient modulus and direction of each pixel are calculated as follows:

[0109]

[0110]

[0111] The Euclidean distance is used to define the similarity of feature points. The Euclidean distance calculation formula and similarity calculation formula are as follows:

[0112]

[0113]

[0114] The number of matched feature points returned is m, the number of unmatched feature points is n, and sim1(A1, A2) indicates that the similarity values ​​of the two images A1 and A2 are calculated using the Sift algorithm.

[0115] As attached Figure 4 As shown in the PHash algorithm flow chart, the scheme of this embodiment includes using the PHash algorithm to calculate the difference between the overall characteristics of the single-channel simulation and experimental vibration images, compressing and graying the vibration image after input, calculating the mean of the DCT coefficients using DCT transformation, connecting the perceptual hash values ​​of the generated images, calculating the Hamming distance of the PHash values ​​of the two images, and obtaining the image similarity.

[0116] Among them, the Hamming distance calculation is obtained by formula (16):

[0117]

[0118] The image similarity is obtained by formula (17):

[0119]

[0120] Among them, the number of identical values ​​in the same position of the hash string sequence values ​​of the two images is p, and the number of different values ​​in the same position is q. sim2(A1, A2) indicates that the similarity value of the two images A1 and A2 is calculated using the PHash algorithm.

[0121] As attached Figure 5 The Delphi method shown in the figure determines the flow chart of the index importance coefficient. In the present embodiment, the Delphi method is a systematic data collection method, which collects the judgment information of experts to negotiate the specific issues in a certain field into a unified opinion, and it is necessary to investigate and analyze the opinions of different experts. If there are large differences in opinions, the questionnaire needs to be re-decentralized until the opinions are more unified. In order to ensure the reliability and consistency of the data, multiple surveys and feedback are required. The present invention distributes the questionnaire to multiple experts and conducts a total of 3 rounds of surveys.

[0122] As attached Figure 6 The multi-channel adaptive high-fidelity evaluation flow chart is shown in the figure. The specific method is as follows: for a kind of fault state, each state is composed of b kinds of indicators, where b = [1, 2, 3, 4, 5, 6]. Each expert gives a weight under a kind of fault state, where V a,b,t represents the weight given by the t-th expert, where t = [1, 2, ..., n], and the sum of the weights of different indicators under the same fault state is 1, as shown in the following formula:

[0123]

[0124] The Delphi method is used to take the average of the expert weights as the final importance coefficient, and the corresponding fault state is compared with X rms , C q , C e , C f ,I p , C w The relevant importance coefficient expression is as follows:

[0125]

[0126] In order to determine the ability of different sensors to collect different important information, the vibration signal is first collected and processed with a digital filter for noise reduction. Then, the characteristic values ​​of the same characteristic index collected by sensors of different channels are calculated, and the characteristic index coefficient is solved as follows:

[0127]

[0128] Among them, p represents the characteristic index, p = [1, 2, 3, 4, 5, 6], R p,q Represents the characteristic index coefficient, G p,q Represents the characteristic value of the indicator. The importance of each channel can be measured by the sensor comprehensive indicator, calculated by the following formula:

[0129]

[0130] Among them, T q Represents the comprehensive index of each sensor. The larger the index value, the more important information the sensor collects, and the smaller the index value, the less important information the sensor collects.

[0131] After calculating the comprehensive index of each sensor, the single-channel comprehensive fidelity is integrated with the comprehensive index to obtain the adaptive comprehensive fidelity evaluation index M of the channel. q , calculated by the following formula:

[0132] M q =ε q ·T q (twenty two)

[0133] Sum the adaptive fidelity evaluation indicators of multiple channels and find their maximum value. The calculation method is as follows:

[0134]

[0135] The maximum value of the adaptive comprehensive fidelity evaluation index is used as the screening basis to screen the Pareto optimal solution, and the screened correction parameter value and fidelity are taken as the optimal solution. If there are still multiple groups of solutions after screening, the relevant corresponding parameters are back-substituted into the rigid-flexible coupling simulation model for verification, and the model with the smallest difference between simulation fidelity and calculation fidelity is selected as the optimal solution.

[0136] In this embodiment, the gear box vibration test bench is as shown in the attached Figure 7-9 As shown, including

[0137] Attached Figure 7 In the overall equipment diagram of the gearbox vibration test bench shown, 1 represents the test bench control equipment, 2 represents the signal transmission equipment, 3 represents the data conversion and acquisition system, 4 represents the high-performance computer processing equipment, and 5 represents the vibration test bench.

[0138] Attached Figure 8 In the gearbox vibration test data acquisition equipment diagram shown, 1 represents signal vibration data acquisition and analysis software, 2 represents data converter, 3 represents high-performance computer, 4 represents load controller, 5 represents motor controller, 6 represents vibration data collector, 7 represents vibration data transmission line, 8 represents router, 9 represents switch, and 10 represents signal network cable.

[0139] Attached Fig. 9 In the overall diagram of the gearbox equipment shown, 1 represents the gearbox, 2 represents the laser tachometer, 3 represents the coupling, 4 represents the motor, 5 represents the protective cover, 6 represents the magnetic powder brake, which is used as a load, 7 represents the vibration acceleration sensor, and 8 represents the test bench frame and base.

[0140] Among them, the process of collecting vibration signals is as follows: the sensor is arranged on the surface of the gear box body, connected to the acquisition card through a signal connection line, the acquisition card is connected to the switch and router through a network cable, one end of the photoelectric speed sensor is connected to the signal converter, the signal converter is connected to the acquisition card with a signal connection line, and the voltage signal is converted into a square wave signal. The high-performance computer end is also connected to the switch through a network cable. Thus, the acquisition card, vibration acceleration sensor, photoelectric speed sensor, router, switch, and high-performance computer together form a local area network connection for information sharing. The motor control cabinet and load controller are connected to the motor and magnetic powder brake respectively. After covering the protective cover, the motor control cabinet and load controller switches are turned on in turn, and the actual speed of the motor control cabinet is set to 1500rpm. The load controller applies a torque of 2Nm to the magnetic powder brake. Start the vibration test bench, open the vibration monitoring software for data acquisition, and the acquisition card has a built-in digital filter noise reduction module to reduce the noise of the collected signal. The single acquisition time is 5s, the sampling frequency is 2kHz, and the time domain is collected and converted into a frequency domain signal. After the acquisition is completed, close the motor control cabinet and the load controller, open the gear box to replace the faulty pinion, and perform the above operations again to collect the time domain and frequency domain signals under the fault state.

[0141] The gray wolf particle swarm fusion multi-objective optimization algorithm optimizes the fidelity response surface model process as shown in the attached Fig.10 As shown, the specific process includes:

[0142] 1) Set the initial correction parameter value and set the range, take the maximum fidelity as the goal, select the appropriate test method according to the characteristics of the correction parameter and the fidelity objective function, and fit and verify the response surface model to obtain the response surface function.

[0143] 2) Set the population dimension dim, the number of iterations K, the size popsize, and the initial values ​​of the parameters α, A, C, and the inertia weight ω.

[0144] 3) Randomly generate the speed and position of the population in the initial state, and iteratively optimize the algorithm to achieve population initialization.

[0145] 4) Sort the gray wolves according to their individual fitness values, take the top three individuals as α, β and δ, and set X α (k), X β (k), X δ(k) is defined as their location information.

[0146] 5) Update the position of the individual gray wolf to obtain new values ​​of α, A, C, and ω, and return to step (3) to recalculate and obtain new α, β, and δ.

[0147] 6) If the maximum number of iterations is reached, it is considered that the gray wolf has found the optimal solution and the process goes to step (17). If not, the current position result is used as a potential solution and the process goes back to step (14) to continue looking for a better solution.

[0148] 7) Output the optimal solution group that meets the end conditions, update the correction parameters of the rigid-flexible coupling model, and return to the response surface model for verification.

[0149] The above descriptions are only some embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any equivalent device or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A high-fidelity method for gearbox model based on digital twin, characterized in that: It includes: S01. Use finite element simulation software to establish a rigid-flexible coupling twin model of the gearbox, then perform modal analysis, obtain dynamic monitoring points, and perform dynamic simulation; S02. Based on the parallel fixed-axis gearbox vibration test bench, collect the vibration physical data of the gearbox in the time domain and frequency domain corresponding to multiple channels when the gearbox is in normal or multiple fault states; S03, calculate the fidelity of the single-channel fusion algorithm for vibration physical data based on the Sift algorithm and the PHash algorithm; S04. Combining the single-channel fidelity and the preset dynamics comprehensive index, construct a multi-channel fidelity adaptive comprehensive evaluation index; S05. Select the parameter variables to be corrected in the rigid-flexible coupling twin model of the gearbox, and construct a response surface function of the parameter to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox; S06. Based on the Grey Wolf Particle Swarm Fusion Multi-Objective Optimization Algorithm, the non-decomposition of the response surface function of the parameters to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox is solved. Combined with the multi-channel fidelity adaptive comprehensive evaluation index, the optimal solution is screened out to complete the fidelity optimization of the rigid-flexible coupling twin model of the gearbox.

2. The high-fidelity method for gearbox model based on digital twin as claimed in claim 1, characterized in that: S01 includes: using finite element simulation software to establish a flexible finite element twin model of the gear and housing, replacing the rigid housing wall and gear components with flexible components, setting boundary conditions and model parameters, and establishing a rigid-flexible coupling twin model of the parallel fixed-axis gearbox. Then, modal analysis is performed on the flexible housing, and 6 monitoring points are arranged at the maximum amplitude to monitor the vibration response. At the same time, dynamic simulation is performed on the rigid-flexible coupling twin model of the gearbox to verify the accuracy of the twin model.

3. The high-fidelity method for gearbox model based on digital twin as claimed in claim 2, characterized in that: In S02, based on the parallel fixed-axis gearbox vibration test bench, the vibration physical data of the time domain and frequency domain corresponding to 6 channels when the gearbox is in normal or multiple fault states are collected, and then the data are denoised by the Vmd noise reduction algorithm, and then converted into frequency domain signals by fast Fourier transform, and the corresponding test images are generated.

4. The high-fidelity method for gearbox model based on digital twin as claimed in claim 3, characterized in that: In S03, the fidelity of the single-channel fusion algorithm is calculated for the test image corresponding to the vibration physical data based on the Sift algorithm and the PHash algorithm, and the obtained comprehensive fidelity result is used as the calculation result of the final virtual-real signal image fidelity.

5. The high-fidelity method for gearbox model based on digital twin as claimed in claim 4, characterized in that: S03 includes: The formula for calculating the fidelity of the test image corresponding to the vibration physical data by the Sift algorithm is as follows: <h2 style=";text-align:left;direction:ltr">ε′<h2 style=";text-align:left;direction:ltr"> d <h2 style=";text-align:left;direction:ltr"> = sim1(B1,B2)×100% (1) Among them, sim1(B1,B2) represents the similarity value between the vibration frequency domain response image B1 of the rigid-flexible coupling twin model and the vibration frequency domain response image B2 of the physical data when using the Sift algorithm, ε′ d represents the fidelity of the two images of this channel under the Sift algorithm, d represents the sensor channel number, d = [1, 2, 3, 4, 5, 6]; The formula for calculating the fidelity of the test image corresponding to the vibration physical data by the PHash algorithm is as follows: ε″ d =sim2(B1,B2)×100% (2) Among them, sim2(B1,B2) represents the similarity value between the vibration frequency domain response image B1 of the rigid-flexible coupling twin model and the vibration frequency domain response image B2 of the physical data when using the PHash algorithm, ε″ d Represents the fidelity of the two images of this channel under the PHash algorithm; The calculation formula for the fidelity of the single-channel fusion algorithm regarding the vibration response is as follows:

6. The high-fidelity method for gearbox model based on digital twin as claimed in claim 5, characterized in that: In S04, the single-channel fidelity and the preset dynamics comprehensive index are combined to construct the multi-channel fidelity adaptive comprehensive evaluation index, including: The Delphi method is used to calculate the importance coefficient of signal indicators under different fault conditions; By using time domain characteristic indicators and combining them with the indicator importance coefficient, we can establish comprehensive sensor indicators to evaluate the sensor's ability to collect important information. By using the fidelity and sensor comprehensive indicators, an adaptive comprehensive evaluation index of multi-channel fidelity is constructed to adaptively evaluate the multi-channel fidelity and provide a basis for the fidelity optimization of the gearbox multi-channel model.

7. The high-fidelity method for gearbox model based on digital twin as claimed in claim 1, characterized in that: In S05, the parameter variables to be corrected in the rigid-flexible coupling twin model of the gearbox are selected, and the response surface function of the parameters to be corrected and the fidelity of the rigid-flexible coupling twin model of the gearbox is constructed, including: The central composite design method was used for response surface experimental design; the second-order polynomial was used for response surface function fitting, and the fitting formula was as follows: Among them, a0 is a constant, a i 、b i 、c ij are the unknown coefficients of the first-order, second-order and second-order cross terms, n is the number of parameters to be corrected, and the minimum number of tests is 8. A high-fidelity gearbox model system based on digital twin, characterized in that: The invention applies the high-fidelity method of gearbox model based on digital twin as claimed in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor to implement the high-fidelity method of gearbox model based on digital twin as described in one of claims 1 to 7.