Digital twinborn system construction method of fatigue test gear transmission system

By constructing a dynamic simulation model of rigid-flexible coupling and CycleGAN network mapping, the problem of obtaining vibration signals of the fatigue test gear transmission system in the existing technology is solved, and efficient and accurate signal generation is achieved, fault diagnosis and intelligent operation and maintenance are supported, and fault diagnosis and intelligent operation and maintenance are reduced. Experimental cost.

CN119026266APending Publication Date: 2024-11-26CHONGQING UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411106592.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

When the existing technology studies the dynamic response of parallel axis displacement spur gear transmission system, the experimental method consumes a lot of manpower, material resources and time, while the dynamic simulation method has a different accuracy from the measured signal, which cannot completely replace the experimental data and lacks a A method that can accurately and efficiently obtain vibration signals of the gear transmission system of fatigue test.

Method used

The dynamic simulation model of the fatigue test gear transmission system with rigid-flexible coupling is constructed by multi-body dynamics method, and the mapping of virtual space data to physical space data is completed through the optimized CycleGAN network, and a vibration signal close to the actual measured signal is generated to support the fault. Diagnosis and intelligent operation and maintenance.

Benefits of technology

It realizes the generation of high-quality spur gear transmission vibration signals under different working conditions and faults, reduces experimental costs, improves analysis efficiency, and supports fault diagnosis and intelligent operation and maintenance of spur gear transmission systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026266A_ABST
    Figure CN119026266A_ABST
Patent Text Reader

Abstract

The invention relates to a digital twin system construction method of a fatigue test gear transmission system, and belongs to the technical field of mechanical equipment fault detection. The method specifically comprises the following steps: S1, constructing a rigid-flexible coupled fatigue test gear transmission system dynamics simulation model by using a multi-body dynamics method, and obtaining a dynamics simulation signal; s2, constructing twin data of the fatigue test gear transmission system, and completing mapping of virtual space data to physical space data through the optimized CycleGAN to generate a mapping signal closer to an actually measured signal; and S3, high-quality straight gear transmission system vibration signals under different working conditions and faults are generated by using the dynamic simulation signals and the mapping signals, and are used for supporting fault diagnosis and intelligent operation and maintenance of the straight gear transmission system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of mechanical equipment fault detection and relates to a method for constructing a digital twin system of a fatigue test gear transmission system. Background Art

[0002] As one of the most commonly used transmission methods in modern machinery, gear transmission systems are indispensable in mechanical equipment in various fields. Therefore, the research on fault detection and intelligent operation and maintenance of gear transmission systems has become a very important task. At present, the fault diagnosis research on mechanical equipment is generally completed by analyzing vibration signals. For the two methods of obtaining vibration signals, although the experimental method can obtain accurate test signals, it requires a lot of manpower, material resources and time costs; although the dynamic simulation method can obtain high-precision simulation signals in a short time, it always differs from the measured signals in terms of accuracy and other characteristics, and cannot completely replace the test data.

[0003] Therefore, in order to study the dynamic response of the parallel shaft displacement spur gear transmission system, a method that can accurately and efficiently obtain the vibration signal of the fatigue test gear transmission system is urgently needed. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a method for constructing a digital twin system of a fatigue test gear transmission system. The system can realize motion state detection and multi-body dynamics simulation of the fatigue test gear transmission system, and generate high-quality spur gear transmission system vibration signals under different working conditions and faults through CycleGAN mapping, making it possible to replace experiments with simulation.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for constructing a digital twin system of a fatigue test gear transmission system specifically comprises the following steps:

[0007] S1: Use multi-body dynamics method to build a dynamic simulation model of the rigid-flexible coupling fatigue test gear transmission system and obtain dynamic simulation signals;

[0008] S2: Construct twin data of fatigue test gear transmission system, and complete the mapping of virtual space data to physical space data through optimized CycleGAN, so as to generate mapping signals closer to measured signals; CycleGAN stands for Cycle Generative Adversarial Network;

[0009] S3: Generate high-quality spur gear transmission system vibration signals under different working conditions and faults using dynamic simulation signals and mapping signals to support fault diagnosis and intelligent operation and maintenance of spur gear transmission systems;

[0010] Further, in step S1, a dynamic simulation model of a rigid-flexible coupled fatigue test gear transmission system is constructed, which specifically includes the following steps:

[0011] S11: Construct rigid-flexible coupling model;

[0012] The three-dimensional model of the fatigue test gear transmission system is established using Soildworks software, where the gear transmission system includes a test gearbox, a drive gearbox, a pair of spur gears (drive gears), a pair of displacement spur gears (test gears), two transmission shafts and eight deep groove ball bearings; specifically, first, the test gears are strictly parametrically modeled, and other components are also modeled according to size requirements; then, based on the established multi-body dynamics model, the gearbox housing is made flexible to form a rigid-flexible coupling multi-body dynamics model; finally, a series of operations such as meshing, analysis step creation and modal analysis of the housing model are completed in ABAQUS, and finally the modal neutral file of the model can be obtained, which is opened in ABAQUS and assembled to the corresponding position of the original model to complete the establishment of a rigid-flexible coupling model in ADAMS; where ABAQUS represents finite element analysis software, and ADAMS represents mechanical system dynamics simulation software;

[0013] S12: Setting constraints and contact parameters, specifically including: adding constraints, loads and drives to the fatigue test gear transmission system model; setting contacts between meshing gears and at load-bearing bearings.

[0014] Further, in step S12, constraints, loads and drives are added to the fatigue test gear transmission system model, specifically including:

[0015] (1) Add fixing pairs between the bolt holes on the four legs of the gearbox and the ground;

[0016] (2) Add a fixed pair on the contact surface between the gear and the transmission shaft;

[0017] (3) Add a fixed pair on the contact surface between the inner ring of the bearing and the transmission shaft;

[0018] (4) Add a fixed pair on the contact surface between the outer ring of the bearing and the housing;

[0019] (5) Add a revolute pair between the transmission shaft and the ground;

[0020] (6) Apply a rotational drive to the revolving pair of the input shaft, use the STEP function to define the speed, select Velocity as the drive type, and the unit is ° / s;

[0021] (7) Apply a load force to the output shaft and use the STEP function to define the size in N.

[0022] Further, in step S12, contact is set between meshing gears, specifically including: among the parameters of the gear contact force, the meshing stiffness K of the gear needs to be determined, and the expression is:

[0023]

[0024] Where, E is the elastic modulus; μ is the Poisson's ratio; R is the radius of curvature of the gear, and the subscripts 1 and 2 represent the driving wheel and the driven wheel respectively; E * represents the comprehensive elastic modulus; R is the comprehensive curvature radius, and its calculation formula is:

[0025]

[0026] Among them, β n is the base circle helix angle, r1 is the driving wheel pitch circle radius, α n is the end pressure angle, i is the transmission ratio; β n and α n The calculation formulas are:

[0027]

[0028] β n =arctan(tanβcosα t )

[0029] Among them, α t is the end face engagement angle, and β is the helix angle.

[0030] Further, in step S12, contact is set at the load-bearing bearing, specifically including: in the setting of the contact force inside the bearing, the most important parameter to be determined is also the contact stiffness K', which is expressed as:

[0031]

[0032] Among them, K' is the contact stiffness of the two objects; R is the comprehensive curvature radius, R'1 and R'2 are the curvature radii of the contact parts; E' is the comprehensive elastic modulus, E1 and E2 are the elastic moduli of the contact parts; μ1 and μ2 are the Poisson's ratios of the contact parts.

[0033] Further, in step S2, the fatigue test gear transmission system twin data includes test bench parameter data, sensor acquisition working condition data and dynamic model simulation data.

[0034] Furthermore, in step S2, the optimized CycleGAN proposes a smooth cycle consistency loss based on the smooth L1 loss function, that is, the loss function of the optimized CycleGAN Loss CycleGAN is the loss of the two GANs and the cycle consistency loss Loss cycle (GX ,G Y ), the expression is:

[0035] Loss CycleGAN =Loss X-GAN (G X ,D Y )+Loss Y-GAN (G Y ,D X )+λLoss cycle (G X ,G Y )

[0036] Loss cycle (G X ,G Y )=E[G Y (G X (x))-x]+E[G X (G Y (y))-y]

[0037] Among them, λ is the weight coefficient, Loss X-GAN (G X ,D Y )、Loss Y-GAN (G Y ,D X ) represents the adversarial loss function, E[.] represents the expectation; x and y represent the input values, G Y (G X (x))、G X (G Y (y)) represents forward prediction;

[0038] By maximum or minimum Loss CycleGAN , which enables CycleGAN to learn the mapping relationship between non-parallel domains X and Y in an unsupervised manner, thus making the mapping of twin data possible.

[0039] The beneficial effect of the present invention is that the system constructed by the present invention utilizes dynamic simulation signals and measured signals to generate high-quality spur gear transmission system vibration signals under different working conditions and faults, thereby supporting fault diagnosis and intelligent operation and maintenance of the spur gear transmission system, greatly saving experimental costs and improving analysis efficiency.

[0040] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0042] Figure 1 It is a simplified structural diagram of the parallel shaft spur gear transmission system;

[0043] Figure 2 This is the modeling diagram of the test gear;

[0044] Figure 3 Assembly drawing of gear transmission system for fatigue test;

[0045] Figure 4 Rigid-flexible coupling model of gear transmission system for fatigue test;

[0046] Figure 5 It is the vibration acceleration diagram of the simulation signal;

[0047] Figure 6 This is a comparison chart of vibration acceleration;

[0048] Figure 7 Twin data structure for fatigue test gear transmission system;

[0049] Figure 8 Schematic diagram of the CycleGAN structure for smoothing cycle consistency loss;

[0050] Fig. 9 It is a comparison diagram between the mapped signal and the measured signal;

[0051] Fig.10 This is a test diagram of the loss function and accuracy of CycleGAN training;

[0052] Fig.11 is the multi-body dynamics simulation result;

[0053] Fig.12 Generate results for the mapped signal;

[0054] Fig.13 This is a comparison chart of the mapped signal and the measured signal. DETAILED DESCRIPTION

[0055] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0056] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0057] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0058] See also Figures 1 to 13 The present invention takes the fatigue test gear transmission system as the object, establishes a rigid-flexible coupled multi-body dynamic model and performs analysis. Based on the idea of ​​digital twin, a digital twin mapping model of the fatigue test gear transmission system is established, and on this basis, a digital twin system of the fatigue test gear transmission system driven by digital and analog co-operation is developed. The system can realize the motion state detection and multi-body dynamic simulation of the fatigue test gear transmission system, and generate high-quality spur gear transmission system vibration signals under different working conditions and faults through CycleGAN mapping, making it possible to replace experiments with simulation.

[0059] 1. Dynamics simulation of gear transmission system based on automatic dynamics analysis of mechanical systems (ADAMS) for fatigue test

[0060] 1) Establish a rigid-flexible coupling model

[0061] The three-dimensional model of the fatigue test gear transmission system is established using Soildworks software. The gear transmission system of the test bench includes a test gearbox, a drive gearbox, a pair of spur gears (drive gears), a pair of shifted spur gears (test gears), two transmission shafts, and eight deep groove ball bearings. The structural diagram is shown in the figure. Figure 1 shown.

[0062] As the most important component in the gear transmission system, the vibration characteristic frequency of the gear is closely related to its own parameters, so it is necessary to perform strict parametric modeling on the test gear. The main parameters of the test gear are shown in Table 1. The final test gear modeling is as follows: Figure 2 shown.

[0063] Table 1 Test gear parameters

[0064]

[0065] Other parts are also modeled according to the size requirements, including 2 gearboxes, 2 transmission shafts, and 8 Japanese NSK deep groove ball bearings. In order to simplify the calculation process, the model can be lightened by deleting some bolt holes, undercuts and other structures. Assemble the parts together according to the positional relationship, and check whether there is interference after setting the gear fit between the meshing gears. The final assembly is as follows Figure 3 shown.

[0066] In order to obtain the vibration signal of the test gearbox in the fatigue test gear transmission system, the present invention makes the gearbox flexible on the basis of the established multi-body dynamics model to form a rigid-flexible coupling multi-body dynamics model. A series of operations such as meshing, analysis step creation, and modal analysis of the box model are completed in ABAQUS, and finally the modal neutral file of the model can be obtained. The file is opened in ADAMS and assembled to the corresponding position of the original model to complete the establishment of the rigid-flexible coupling model in ADAMS. The final rigid-flexible coupling model is as follows Figure 4 shown.

[0067] 2) Set constraints and contact parameters

[0068] In ADAMS, kinematic pairs are used to simulate the real connection or motion relationship between components or local positions. In ADAMS, kinematic pairs are added to components such as gearboxes, transmission shafts, bearings, and gears according to the motion relationship and contact relationship between the components, and then loads and drives are applied according to the actual working conditions. The entire model completes the basic settings before simulation. In order to simulate the actual operation of the fatigue test gear transmission system, the present invention also adds constraints, loads, and drives to the fatigue test gear transmission system model, as follows:

[0069] (1) Add fixing pairs between the bolt holes on the four legs of the gearbox and the ground;

[0070] (2) Add a fixed pair on the contact surface between the gear and the transmission shaft;

[0071] (3) Add a fixed pair on the contact surface between the inner ring of the bearing and the transmission shaft;

[0072] (4) Add a fixed pair on the contact surface between the outer ring of the bearing and the housing;

[0073] (5) Add a revolute pair between the transmission shaft and the ground;

[0074] (6) Apply a rotary drive to the revolving pair of the input shaft, use the STEP function to define the speed, select Velocity as the drive type, and the unit is ° / s;

[0075] (7) Apply a load force to the output shaft and use the STEP function to define the size in N.

[0076] In ADAMS, contact is an important concept used to describe the interactions between objects such as collision and friction. ADAMS provides a series of contact models and parameter settings to simulate different contact behaviors. In the fatigue test gear transmission system model established by the present invention, it is necessary to set contact between meshing gears and at the load-bearing bearings.

[0077] Among the parameters of gear contact force, we mainly need to determine the gear meshing stiffness K. Gear tooth meshing can be equivalent to the collision of two cylinders with variable curvature radius, which can be solved by Hertz static elastic contact theory. According to Hertz contact theory, when the contact area is circular, we have:

[0078]

[0079] From the above formula, the relationship between the normal contact force P and deformation δ during collision is:

[0080]

[0081] Where K depends on the material and structural shape of the collision object:

[0082]

[0083] Where, E is the elastic modulus; μ is the Poisson's ratio; R is the radius of curvature of the gear, and the subscripts 1 and 2 represent the driving wheel and the driven wheel respectively; E * represents the comprehensive elastic modulus; R is the comprehensive curvature radius, and its calculation formula is:

[0084]

[0085] Among them, β n is the base circle helix angle; r1 is the driving wheel pitch circle radius; αn is the end pressure angle; i is the transmission ratio. n and α n The calculation formulas are:

[0086]

[0087] β n =arctan(tanβcosα t ) (8)

[0088] In the setting of the internal contact force of the bearing, the most important parameter is also the contact stiffness K'. During the operation of the gear transmission system, the deep groove ball bearing plays a role of transmission and support. The internal rollers and cages, as well as the inner and outer rings of the bearing are in contact. The actual contact stiffness is related to the surface shape and material properties of the two contacting objects. The specific contact stiffness can be calculated using the following formula:

[0089]

[0090] Among them, K' is the contact stiffness of the two objects; R is the comprehensive curvature radius, R'1 and R'2 are the curvature radii of the contact parts; E' is the comprehensive elastic modulus, E1 and E2 are the elastic moduli of the contact parts; μ1 and μ2 are the Poisson's ratios of the contact parts.

[0091] Substituting the relevant parameters of the test gear and deep groove ball bearing into formulas (3) to (9), the contact stiffness K or K' can be calculated. Other contact force parameters are selected based on experience. The final contact force parameter settings are shown in Table 2.

[0092] Table 2 Contact force parameter settings

[0093]

[0094] 3) Simulation results and analysis

[0095] In order to evaluate the results of multi-body dynamics simulation, this experiment collected the measured vibration signals of the fatigue test gear transmission system under the same working conditions and compared them with the dynamic simulation signals. The experimental working condition parameters and characteristic frequency theoretical values ​​of the fatigue test gear transmission system are shown in Table 3. In order to enhance the characteristics of the vibration signal, more realistically reflect the signal characteristics, and prevent the impact of the starting speed from being too large, thereby affecting the vibration signal collection, the STEP function definition is used on the application of speed and load. Set the input speed to STEP (time, 0, 0d, 0.1, 4236d), the load to STEP (time, 0, 0, 0.1, 500), the simulation time t to 0.5s, and the step to 12500. Select GSTIFF of the SI2 integral equation as the integrator of the dynamic model, and the integral error is set to 0.001.

[0096] Table 3 Experimental parameters and characteristic frequency theoretical values

[0097]

[0098] After the rigid-flexible coupling model of the fatigue test gear transmission system is established according to the above, the selection of simulation measurement points should be consistent with the actual sensor measurement points. The actual acceleration sensor of the fatigue test gear transmission system is located directly above the test gearbox. The layout and principle of the sensor measurement points can refer to the construction of the twin data of the fatigue test gear transmission system in Section 2 below. Therefore, a marker is set at the center point above the test gearbox model in ADAMS, and the vibration acceleration signal of this point is output for analysis.

[0099] The vibration acceleration signal output by the marker point is transformed into the frequency domain by fast Fourier transform. Figure 5 shown. Figure 5 (a) is the vibration acceleration time domain diagram. The red arrow in the figure indicates the meshing impact. The impact interval can be seen from the readings in the figure, which is approximately, which is consistent with the theoretical meshing frequency of 188 Hz; Figure 5 (b) is the acceleration spectrum, the meshing frequency and its multiples are clearly visible, with a single component; Figure 5 (c) is the acceleration envelope spectrum, which also has a single component. However, compared with the amplitude of the first three times of the frequency spectrum, the amplitude of each frequency in the envelope spectrum is relatively obvious.

[0100] In order to more intuitively compare the differences between dynamic simulation and experimental data, the vibration acceleration simulation signal and the measured signal at the same measuring point were compared. The results are as follows: Figure 6 As shown, Figure 6 (a) is the time domain comparison of vibration acceleration, Figure 6 (b) is the frequency domain comparison of vibration acceleration. Figure 6 (c) is the comparison of vibration acceleration envelope spectrum. It can be seen that the meshing frequency of the simulation signal is consistent with the measured signal, which is 188Hz. The trend is basically consistent with the measured signal, but there is still a slight difference in the accuracy of the characteristic frequency. This proves that the multi-body dynamics simulation of the fatigue test gear transmission system is accurate, but it cannot completely replace the experimental signal.

[0101] 2. Construct twin data of fatigue test gear transmission system

[0102] 1) Collect fatigue test gear transmission system data

[0103] Twin data is a multi-dimensional, multi-angle data obtained by integrating complex and huge information data during the operation of fatigue test gear transmission system. It is the data basis for driving the digital twin system of fatigue test gear transmission system. The twin data of fatigue test gear transmission system includes test bench parameter data, sensor acquisition condition data, dynamic model simulation data, and completes the data mapping from virtual space data to physical space through the integrated cycle generative adversarial network (CycleGAN), which can generate simulation signals that are closer to the measured signals. The specific composition is as follows: Figure 7 shown.

[0104] The actual measured signals of the gear's full life vibration are all from the fatigue test gear transmission system, which is mainly composed of torque loading system, water cooling circulation system, temperature control and monitoring system, oil lubrication system, gear transmission system and other parts. The main engine part uses a 380V three-phase motor as the power source, and the torque loading device is installed on the driven shaft. The load is applied by hanging the weight on the loading rod. The test gearbox and the drive gearbox are both cast in gray iron HT250. The test bench includes a drive motor, a drive gearbox, a torque loading clutch, a torque sensor (used to measure the load size and prevent sudden changes in torque) and a test gearbox, and the test gear is installed in the test gearbox.

[0105] The present invention mainly studies the dynamic response of the gear transmission system, so a vibration sensor is used to collect the vibration acceleration signal on the housing when the fatigue test gear transmission system is running. The quality of the actual vibration signal collection is inseparable from the selection of the sensor, so the selection of the vibration sensor is crucial. Commonly used vibration sensors include displacement sensors, velocity sensors, and acceleration sensors. The present invention chooses to use a contact acceleration sensor with a data acquisition card to collect the vibration acceleration signal of the test gear box housing. The model of the acceleration sensor is PCB352A60, which is matched with a Ni9234 acquisition card. The various parameters of the sensor are shown in Table 4.

[0106] Table 4 PCB352A60 sensor parameters

[0107]

[0108]

[0109] The location of the sensor measuring point also has a great impact on the accuracy and sensitivity of the vibration signal. Considering the economy, information complexity and structural operation status, it is necessary to optimize the number and location of the measuring points to ensure that as few sensors as possible are used to collect sufficient vibration signals. According to the effectiveness and feasibility of sensor layout, the sensor is easy to install and as close to the vibration source as possible. Acceleration sensors are arranged just above and on the side of the test gearbox.

[0110] 2) CycleGAN

[0111] In the process of constructing twin data of fatigue test gear transmission system, it is also necessary to find a "bridge" between the simulation data in the virtual space and the measured data in the physical space. The distribution of the dynamic simulation signal and the measured signal of the gear transmission system is very similar, but because the established dynamic model has errors and the measured signal also includes some external interference, it is also necessary to correct the simulation signal output by the dynamic model to make it more consistent with the measured signal.

[0112] With the development of deep learning, we can establish the mapping relationship between virtual space simulation data and physical space measured data through mapping networks. Artificial neural networks (ANN), recurrent neural networks (RNN) and convolutional neural networks (CNN) are currently commonly used neural networks in deep learning research. They are more used in pattern recognition, that is, extracting certain features from different objects, while the data mapping from virtual space to physical space belongs to the domain transformation problem. In order to solve the above problems, some scholars have proposed Generative Adversarial Network (GAN) and Cycle Generative Adversarial Network (CycleGAN). CycleGAN, as an upgraded version of GAN, introduces cycle-consistency loss to prevent overfitting, solves the problem of GAN network mode collapse, and realizes the migration between two non-parallel domains X and Y.

[0113] CycleGAN integrates two GANs, each of which has a generator for converting data and a corresponding discriminator for distinguishing actual data from generated data. Each adversarial loss function is used to train the corresponding generator. X-GAN (G X ,D Y ) and Loss Y-GAN (G Y ,D X ) are defined as:

[0114] Loss X-GAN (G X ,D Y )=E[log(D Y (y))]+E[log(1-D Y (G X (x)))] (10)

[0115] Loss Y-GAN (GY ,D X )=E[log(D X (x))]+E[log(1-D X (G Y (y)))] (11)

[0116] In order to perform unsupervised learning in non-parallel situations, a cycle consistency loss is also introduced in CycleGAN, which is defined as the loss between the input value x and the forward prediction G Y (G X (x)) and the input value y and the forward prediction G X (G Y (y)), the larger the difference, the farther the prediction is from the original input. The cycle consistency loss is defined as follows:

[0117] Loss cycle (G X ,G Y )=E[G Y (G X (x))-x]+E[|G X (G Y (y))-y] (12)

[0118] In summary, the loss function used to train the complete CycleGAN is the sum of the losses of the two GANs and the cycle consistency loss:

[0119] Loss CycleGAN =Loss X-GAN (G X ,D Y )+Loss Y-GAN (G Y ,D X )+λLoss cycle (G X ,G Y ) (13)

[0120] Among them, the weight coefficient λ is used to control the weight of the cycle consistency loss in the total loss. The higher the weight, the more meaningful it is to reduce the cycle consistency loss compared with other losses. CycleGAN , which enables CycleGAN to learn the mapping relationship between non-parallel domains X and Y in an unsupervised manner, thus making the mapping of twin data possible.

[0121] In summary, an optimized CycleGAN is introduced into the twin data of fatigue test gear transmission system. In this network, a smooth cycle consistency loss is proposed based on the smooth L1 loss function, which can improve the robustness and stability of CycleGAN for outlier data in the two domains. Its specific structure is as follows: Figure 8 By taking a small amount of dynamic simulation signals and measured signals as two domains and using the network for training, the mapping between the data can be completed, and mapping data that better fits the measured data can be generated, which saves various experimental costs and improves analysis efficiency.

[0122] 3) Analyze the mapping signal

[0123] The optimized CycleGAN can be used to map the simulation data to the measured data under any working condition. To achieve this function, CycleGAN must be pre-trained. In order to improve the generalization ability of the network, multiple sets of multi-body dynamic simulation signals of fatigue test gear transmission systems under different working conditions and the corresponding measured signals are paired into CycleGAN for training. 200,000 data points are intercepted for each set of data. The detailed information of the working conditions involved in the training is shown in Table 5. In terms of hyperparameter setting, the learning rates of the training discriminator and generator are 0.0001 and 0.0002 respectively, the weight coefficient λ is 10, and CycleGAN uses the Adam optimizer to train 1000 Epochs, the Batchsize is set to 8, and the momentum is set to 0.5.

[0124] Table 5 Working conditions involved in training

[0125]

[0126] After the network training is completed, a set of simulation data with a speed of 450rpm and a load of 120Nm is used to verify the mapping ability of the optimized CycleGAN. Finally, the network generates a set of mapping signals with the characteristics of the measured signals under this working condition. In order to better reflect the fitting effect of the mapping signal and the measured signal, the mapping signal is compared with the measured signal as follows Fig. 9 As shown:

[0127] Depend on Fig. 9 (a), 9(b), and 9(c) show that compared with the dynamic simulation signal, the mapping signal generated by the neural network after learning the characteristics of the measured signal is indeed more consistent with the measured signal. The time domain waveform, spectrum, envelope spectrum, and meshing frequency f m In order to further quantify the error between the mapped signal and the measured signal, the PRD and PMSE indicators of the spectrum are also calculated. The calculation formula is as follows:

[0128]

[0129] In the above formula, and x(n) represent the generated signal and the measured signal respectively, and N represents the signal length. It can be seen that the larger the amount of measured signal or simulated signal data involved in the training, the smaller the error between the generated mapping signal and the measured signal, and the more accurate the mapping signal. After calculation, the PRD of the mapping signal is 9.9 and the RMSE is 0.1309, which proves that the mapping signal generated by CycleGAN is not much different from the measured signal.

[0130] The quality of model training can be measured by the two indicators of loss function and accuracy. By adjusting the hyperparameters according to the different characteristics of the curve, the quality of network training can be continuously optimized. Fig.10 As shown, Fig.10 (a) is the loss comparison between the generator and the discriminator. It can be seen that the loss fluctuations of both are relatively stable, indicating that the adversarial effect is stable and the training effect is good; Fig.10 (b) is the accuracy comparison between the generated signal and the measured signal. It can be found that the spectrum line rises in the first 200 epochs. This is because the discriminator training speed is faster. Then, after the generator training speed catches up, the spectrum line gradually decreases and finally stabilizes at around 0.5. This also proves that the CycleGAN with smoothed cycle consistency loss has a better training effect.

[0131] 3. Development and application cases of digital twin system for fatigue test gear transmission system

[0132] 1) System development tools and operating environment

[0133] The digital twin system of the fatigue test gear transmission system of the digital-analog combination was developed using the Unity3D engine. With the help of Unity3D's superior three-dimensional model visualization function and physical collision system, the model was controlled through C# scripts, thereby completing the test bench model operation display function. With the help of ECharts' customizable data visualization function, the chart development work required for the interface was completed, and the front-end development of the digital twin system was completed by combining the above two software. The dynamic analysis model and CycleGAN are integrated into the system in the form of a dynamic link library to achieve the corresponding functions. The operating environment required by the system is detailed in Table 6.

[0134] Table 6 Operating environment of digital twin system of fatigue test gear transmission system

[0135]

[0136] 2) Application case of digital twin system for fatigue test gear transmission system

[0137] The following is a practical example to demonstrate the practical application of the digital twin system of the fatigue test gear transmission system. The system is used to obtain the vibration acceleration mapping signal of the healthy test gear under the conditions of a speed of 706rpm and a load of 500Nm:

[0138] (1) Perform multi-body dynamics simulation

[0139] The developed digital twin system is used for dynamic simulation. The simulation parameters under this condition are speed 706rpm, load 500Nm, sampling frequency 25kHz, and simulation time is set to 0.5s. The results are as follows: Fig.11 shown.

[0140] (2) Generate mapping signal

[0141] After generating the dynamic simulation signal under the working condition of 706rpm and 500Nm, the digital twin system is used to perform CycleGAN mapping to generate the mapping signal under the working condition of 706rpm and 500Nm. The mapping signal results are as follows: Fig.12 As shown in the figure, users can also judge the quality of model training based on Loss and Acc, thereby adjusting hyperparameters to regenerate mapping signals.

[0142] In order to more intuitively reflect the fitting degree between the mapping signal and the measured signal, Fig.13 The time domain, frequency domain and envelope spectrum of the mapped signal under 706rpm and 500Nm working conditions are compared with the measured signal under the same working conditions. It can be seen that the meshing frequency and its frequency multiples of the mapped signal are almost completely consistent with the measured signal. After calculation, the PRD of the mapped signal spectrum is 2.9 and the PMSE is 0.0293, which further verifies the mapping effect of the optimized CycleGAN.

[0143] 4. Summary

[0144] In order to study the dynamic response of the parallel axis displacement spur gear transmission system, the present invention uses a multi-body dynamics method to construct a rigid-flexible coupled fatigue test gear transmission system dynamic model based on the flat fatigue test gear transmission system. The vibration acceleration simulation signal is compared with the vibration acceleration measured signal under the same working condition, and the characteristic frequencies in the time domain, frequency domain and envelope spectrum are compared respectively, and the advantages and disadvantages of the dynamic simulation are analyzed. Subsequently, the twin data of the fatigue test gear transmission system is built, and the mapping of virtual space data to physical space data is completed through the CycleGAN network to generate a mapping signal that is closer to the real experimental signal. Finally, based on software such as Unity3D, functional modules such as operation status detection, simulation signal visualization, and neural network generation of mapping signals are developed and integrated together to establish a digital twin system of the fatigue test gear transmission system. The system uses dynamic simulation signals and measured signals to generate high-quality spur gear transmission system vibration signals under different working conditions and faults, thereby supporting fault diagnosis and intelligent operation and maintenance of spur gear transmission systems, greatly saving experimental costs and improving analysis efficiency.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A method for constructing a digital twin system of a fatigue test gear transmission system, characterized in that: The method specifically comprises the following steps: S1: Use multi-body dynamics method to build a dynamic simulation model of the rigid-flexible coupling fatigue test gear transmission system and obtain dynamic simulation signals; S2: Construct twin data of fatigue test gear transmission system, and complete the mapping of virtual space data to physical space data through optimized CycleGAN, and generate mapping signals close to measured signals; CycleGAN stands for Cycle Generative Adversarial Network; S3: Using dynamic simulation signals and mapping signals, vibration signals of the spur gear transmission system under different working conditions and faults are generated to support fault diagnosis and intelligent operation and maintenance of the spur gear transmission system.

2. The method for constructing a digital twin system of a fatigue test gear transmission system according to claim 1, characterized in that: In step S1, a dynamic simulation model of a rigid-flexible coupled fatigue test gear transmission system is constructed, which specifically includes the following steps: S11: Construct rigid-flexible coupling model; The software is used to establish a three-dimensional model of the fatigue test gear transmission system, wherein the gear transmission system includes a test gearbox, a drive gearbox, a pair of drive gears, a pair of test gears, two transmission shafts and eight deep groove ball bearings; specifically, the following steps are used: first, the test gears are strictly parametrically modeled, and other components are also modeled according to the size requirements; then, the gearbox housing is made flexible based on the established multi-body dynamics model to form a rigid-flexible coupling multi-body dynamics model; finally, the meshing, analysis step creation and modal analysis of the housing model are completed in ABAQUS, and finally the modal neutral file of the model is obtained, which is opened in ADAMS and assembled to the corresponding position of the original model to complete the establishment of the rigid-flexible coupling model in ADAMS; wherein ABAQUS represents finite element analysis software, and ADAMS represents mechanical system dynamics simulation software; S12: Setting constraints and contact parameters, specifically including: adding constraints, loads and drives to the fatigue test gear transmission system model; setting contacts between meshing gears and at load-bearing bearings.

3. The method for constructing a digital twin system of a fatigue test gear transmission system according to claim 2, characterized in that: In step S12, constraints, loads and drives are added to the fatigue test gear transmission system model, specifically including: (1) Add fixing pairs between the bolt holes on the four legs of the gearbox and the ground; (2) Add a fixed pair on the contact surface between the gear and the transmission shaft; (3) Add a fixed pair on the contact surface between the inner ring of the bearing and the transmission shaft; (4) Add a fixed pair on the contact surface between the outer ring of the bearing and the housing; (5) Add a revolute pair between the transmission shaft and the ground; (6) Apply a rotary drive to the revolving pair of the input shaft, use the STEP function to define the speed, and select Velocity as the drive type; (7) Apply a load force to the output shaft and define its magnitude using the STEP function.

4. The method for constructing a digital twin system of a fatigue test gear transmission system according to claim 2, characterized in that: In step S12, contact is set between meshing gears, specifically including: among the parameters of the gear contact force, the meshing stiffness K of the gear needs to be determined, and the expression is: Where, E is the elastic modulus; μ is the Poisson's ratio; R is the radius of curvature of the gear, and the subscripts 1 and 2 represent the driving wheel and the driven wheel respectively; E * represents the comprehensive elastic modulus; R is the comprehensive curvature radius, and its calculation formula is: Among them, β n is the base circle helix angle, r1 is the driving wheel pitch circle radius, α n is the end pressure angle, i is the transmission ratio; β n and α n The calculation formulas are: β n =arctan(tanβcosα t ) Among them, α t is the end face engagement angle, and β is the helix angle.

5. The method for constructing a digital twin system of a fatigue test gear transmission system according to claim 2, characterized in that: In step S12, contact is set at the load-bearing bearing, specifically including: in the setting of the contact force inside the bearing, it is necessary to determine the contact stiffness K', which is expressed as: Among them, K' is the contact stiffness of the two objects; R is the comprehensive curvature radius, R'1 and R'2 are the curvature radii of the contact parts; E' is the comprehensive elastic modulus, E1 and E2 are the elastic moduli of the contact parts; μ1 and μ2 are the Poisson's ratios of the contact parts.

6. The method for constructing a digital twin system of a fatigue test gear transmission system according to claim 1, characterized in that: In step S2, the fatigue test gear transmission system twin data includes test bench parameter data, sensor acquisition working condition data and dynamic model simulation data.

7. The method for constructing a digital twin system of a fatigue test gear transmission system according to claim 1, characterized in that: In step S2, the optimized CycleGAN proposes a smooth cycle consistency loss based on the smooth L1 loss function, that is, the loss function of the optimized CycleGAN Loss CycleGAN is the loss of the two GANs and the cycle consistency loss Loss cycle (G X ,G Y ), the expression is: Loss CycleGAN =Loss X-GAN (G X ,D Y )+Loss Y-GAN (G Y ,D X )+λLoss cycle (G X ,G Y ) Loss cycle (G X ,G Y )=E[G Y (G X (x))-x]+E[G X (G Y (y))-y] Among them, λ is the weight coefficient, Loss X-GAN (G X ,D Y )、Loss Y-GAN (G Y ,D X ) represents the adversarial loss function, E[.] represents the expectation; x and y represent the input values, G Y (G X (x))、G X (G Y (y)) represents forward prediction; By maximum or minimum Loss CycleGAN , enabling CycleGAN to learn the mapping relationship between non-parallel domains X and Y in an unsupervised manner, thus making the mapping of twin data possible.

Citation Information

Cited By

  • Digital twin tooth surface wear degradation morphology prediction method

    CN121980954A