Construction of rolling bearing digital twin model and model parameter optimization method

Abstract

The present invention relates to a method for constructing a digital twin model of a rolling bearing and optimizing model parameters, and belongs to the field of digital twins. The method comprises the following steps: S1: establishing a four-degree-of-freedom digital twin model of a bearing by combining radial force action, an outer ring, an inner ring, and a dynamic relationship between rolling elements, and inputting geometric parameters and operating parameters for measuring the bearing; S2: initializing and optimizing key parameters that cannot be measured inside the bearing based on a locust algorithm, and outputting a virtual twin signal from the four-degree-of-freedom digital twin model of the bearing; S3: calculating the root mean square difference between the virtual twin signal and the actual vibration signal, and the locust algorithm searches for key parameters in the direction of reducing the difference; S4: the locust algorithm iteration ends, and the optimal twin signal is the one with the smallest error between the output record and the measured vibration signal, and the bearing under the corresponding parameters is the optimal digital twin, so as to provide rich data for training anomaly detection, fault classification, and remaining service life prediction models, and monitor the health status of the bearing in real time.

Description

Technical Field

[0001] The present invention belongs to the field of digital twins and relates to a method for constructing a digital twin model of a rolling bearing and optimizing model parameters. Background Art

[0002] Rolling bearings are called "industrial joints". Due to the harsh working environment of rolling bearings, their operating performance and health status will gradually deteriorate over time. Therefore, it is of great significance to study the intelligent maintenance and health management of bearings.

[0003] Digital twin is a concept that originated from product lifecycle management. Digital twin performs real-time intelligent monitoring and lifecycle management through real-time data perception and high-precision models of dynamic environments. Vibration signals can effectively reflect the degradation information of rolling bearings, so they are used to build digital twins of bearings. Before building a digital twin of a bearing, it is necessary to understand the dynamic response of the bearing. In theory, many scholars have established different fault models to describe the impact of fault-induced shock on the vibration acceleration signal of the rolling bearing in order to obtain the vibration response of the rolling bearing under fault shock. At present, there are three main modeling methods:

[0004] Phenomenological modeling simplifies the fault into an impact signal and performs signal simulation. This method can qualitatively and intuitively simulate the characteristics of bearing fault signals, but it is too simplistic to directly regard the bearing as a one-dimensional single-degree-of-freedom system and ignore the contact relationship between the inner and outer rings and various components.

[0005] Dynamic modeling, based on Hertz theory, considers the contact relationship and fault model between the various components of the bearing, simplifies each component into a series of concentrated masses, and performs dynamic modeling of the bearing.

[0006] Finite element and multi-body dynamics related simulation software have achieved more complex and detailed bearing modeling. However, the simulation is time-consuming and cannot meet the needs of real-time information interaction of digital twins. Summary of the invention

[0007] In view of this, in order to solve the problems existing in the above-mentioned modeling method, the present invention provides a method for constructing a digital twin model of a rolling bearing and optimizing model parameters. The dynamic response of the bearing is simulated by establishing a four-degree-of-freedom dynamic model of the bearing, and a virtual signal is generated using the four-degree-of-freedom dynamic model in a virtual space. Then, the locust algorithm is used to search for the twin signal with the smallest error with the physical signal. High-fidelity life cycle twin signals can be generated, providing rich data for training anomaly detection, fault classification and remaining service life prediction models.

[0008] In order to achieve the above object, the present invention provides the following technical solutions:

[0009] The method for constructing a rolling bearing digital twin model and optimizing model parameters includes the following steps:

[0010] S1: Combine the radial force, the dynamic relationship between the outer ring, the inner ring and the rolling elements to establish a four-degree-of-freedom digital twin model of the bearing, and input the geometric parameters and operating parameters used to measure the bearing;

[0011] S2: Initialize and optimize the key parameters that cannot be measured inside the bearing based on the locust algorithm, and output the virtual twin signal from the four-degree-of-freedom digital twin model of the bearing;

[0012] S3: Calculate the root mean square difference between the virtual twin signal and the actual vibration signal, and the locust algorithm searches for key parameters in the direction of reducing the difference;

[0013] S4: The iteration of the locust algorithm ends, and the output signal with the smallest error between the output record and the measured vibration signal is the optimal twin signal, and the bearing under the corresponding parameters is the optimal digital twin.

[0014] Further, S1 specifically includes the following steps:

[0015] S11: Based on the bearing vibration response mechanism, a four-degree-of-freedom dynamic model of the bearing is established, combining the radial force action, the dynamic relationship between the outer ring, the inner ring, and the rolling elements;

[0016] S12: When modeling the bearing fault, the defect shape is simplified to a square; by modeling the defect as a square, the defect is introduced into the digital bearing. At the same time, considering that the defect height has little effect on the dynamic response, the defect height is simplified to a constant;

[0017] S13: Input the measurable bearing geometric parameters and operating parameters to calculate the theoretical failure frequency of the bearing; the geometric parameters include pitch diameter size, rolling element size, rolling element number and mass; the operating parameters include rotation speed and time.

[0018] Further, S2 specifically includes the following steps:

[0019] S21: Use the locust optimization algorithm to initialize and optimize the key parameters that cannot be measured inside the bearing;

[0020] S22: Select the parameters to be optimized according to the task; if d parameters are optimized, the search space is d-dimensional, and the search upper and lower limits ub of the d parameters are set. d and lb d ;

[0021] S23: The key parameters determined by the locust algorithm are input into the digital twin model, and the virtual twin signal is obtained by using the Runge-Kutta algorithm.

[0022] Further, S3 specifically includes the following steps:

[0023] S31: The measured vibration signal and the virtual twin signal interact in the locust algorithm; the virtual twin signal output by the specific bearing digital twin model is input into the locust algorithm, and the vibration signal of the measured real bearing is also input into the locust algorithm;

[0024] S32: The optimization goal of the locust algorithm is to minimize the root mean square difference between the measured vibration signal and the virtual twin signal;

[0025] S33: The locust algorithm continuously searches for key parameters that reduce the error between the measured vibration signal and the virtual twin signal. Specifically, the locust algorithm continuously iterates in the direction of reducing RMSD, searches for the virtual vibration response with the minimum error with the measured vibration signal, and continuously updates the key parameters during the search process.

[0026] S34: Record the minimum root mean square error and the corresponding key parameters, update the key parameters of the bearing digital twin model, and continue to iterate until the maximum number of iterations is reached. The locust algorithm records the minimum root mean square error and the corresponding optimal parameters during the iteration process. In a new round of iteration, continue to search for new parameters near the recorded optimal parameters, bring the new parameters into the bearing digital twin model, generate new virtual twin signals, and continue to iterate until the maximum number of iterations is reached.

[0027] Further, S4 specifically includes the following steps:

[0028] S41: The maximum number of iterations is reached, and the locust algorithm iteration ends; at the beginning of the locust algorithm, the number of locusts is set to 100 and the number of iterations is set to 50;

[0029] S42: Output the optimal twin signal with the smallest error between the recorded and measured vibration signals during the iteration process; the locust algorithm stops iterating after reaching the set maximum number of iterations. The locust algorithm outputs the minimum RMSD value during the search process, as well as the corresponding virtual twin signal at this time. This virtual twin signal with the smallest error between the measured vibration signal and the measured vibration signal can be considered as the optimal twin signal;

[0030] S43: Output the parameters corresponding to the optimal twin signal. The digital bearing under the optimal parameters is the optimal digital twin of the bearing. The locust algorithm stops iterating after reaching the set maximum number of iterations. The locust algorithm outputs the minimum RMSD value during the search process and the corresponding key parameter values ​​at this time. The digital twin model with these values ​​set can be considered as the optimal digital twin of the bearing.

[0031] Furthermore, the four-degree-of-freedom dynamic equation is calculated using the following formula:

[0032]

[0033] In the formula, m in is the mass of the inner ring and shaft, c in is the shaft connection damping, k in is the equivalent connection stiffness of the shaft, x in is the displacement of the inner circle along the x direction, y in is the displacement of the inner circle along the y direction, m out is the mass of the outer ring and bearing base, c out is the bearing base damping, k out is the equivalent connection stiffness of the bearing base, x out is the displacement of the outer ring in the x direction, y out is the displacement of the outer ring along the y direction, W x is the radial force in the x direction, W y is the radial force in the y direction, F HX and F HY are the contact force components between the inner and outer rings and the rolling elements in the x and y directions, e is the eccentricity, ω is the angular frequency of the inner ring rotation, t is the time, and g is the acceleration due to gravity.

[0034] Furthermore, the theoretical failure frequency of the bearing is calculated using the following formula:

[0035]

[0036] In the formula, f o is the outer race fault frequency, f i is the inner race fault frequency, f c is the cage failure frequency, f r is the rolling element failure frequency, z is the number of rolling elements, f is the rotational frequency, d is the rolling element diameter, D is the bearing pitch diameter, and α represents the bearing contact angle.

[0037] Furthermore, the locust algorithm uses the following calculation formula:

[0038]

[0039] Where N is the number of locusts, |x j -x i | is the distance from the i-th locust to the j-th locust, (x j -x i ) / d ij is the unit vector from the i-th locust to the j-th locust, s is the social force, ub d and lb d are the upper and lower boundaries of the d-dimensional search space, T d is the optimal solution found so far in the target direction. As the number of iterations increases, the outer c reduces the search area around the target, and the inner c reduces the interaction of locusts.

[0040] Furthermore, the calculation formula of the Runge-Kutta algorithm is as follows:

[0041] K1=h*f(x n ,y n )

[0042] K2=h*f(x n +h / 2,y n +K1 / 2)

[0043] K3=h*f(x n +h / 2,y n +K2 / 2)

[0044] K4=h*f(x n +h,y n +K3)

[0045] y n+1 =y n +(K1+2*K2+2*K3+K4) / 6

[0046] x n+1 =x n +h

[0047] In the formula, x n is the x coordinate of the current point, y n is the y coordinate of the current point, f is the differential, h is the step size, K1 is the initial slope of the current time step, K2 is the slope of the middle point of the current time step, K3 is the slope of the other middle point of the current time step, and K4 is the initial slope of the next time step.

[0048] Furthermore, the optimization objective of the locust algorithm is calculated using the following formula:

[0049] minRMSD|=min|rms p -rms v |

[0050] In the formula, rms p is the root mean square of the measured bearing vibration signal, rms v is the root mean square obtained by solving the bearing digital twin model, and RMSD is the error between the two.

[0051] The beneficial effects of the present invention are:

[0052] 1. The method for constructing a digital twin model of rolling bearings and optimizing model parameters disclosed in the present invention is that the digital twin of bearings can generate high-fidelity life cycle twin signals, which can be used to generate twin signals of bearings at different fault degrees and different speeds, and provide rich data for training anomaly detection, fault classification and remaining service life prediction models, thereby solving the problem that bearing intelligent anomaly detection, fault classification and remaining service life prediction models based on deep learning are limited by lack of data in industrial applications.

[0053] 2. The method for constructing a digital twin model of a rolling bearing and optimizing model parameters disclosed in the present invention can also use a twin model to reflect the internal information of the bearing that cannot be measured, such as fault size, stiffness, damping, etc., based on the measured signal. Constructing a digital twin model of the bearing is a real-time and high-fidelity reflection of the real bearing. Based on the digital twin and life cycle physical signals of the bearing, the growth trend of defects can be obtained, which is used to monitor the health status of the bearing in real time, which helps to monitor the health status of the bearing in real time and can also perform predictive maintenance on the bearing.

[0054] 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

[0055] 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:

[0056] Figure 1 It is a flowchart of the method for building a digital twin model of a rolling bearing and optimizing model parameters of the present invention;

[0057] Figure 2 This is a schematic diagram of a digital bearing in this embodiment;

[0058] Figure 3 This is a schematic diagram of a digital bearing defect in this embodiment;

[0059] Figure 4 The search history graph of the locust algorithm for the 234 data set in this embodiment;

[0060] Figure 5 This is a comparison diagram of the twin signal and the physical signal obtained when the data set is 234 in this embodiment;

[0061] Figure 6 This is the normal distribution fitting diagram for the 234 data sets in this embodiment;

[0062] Figure 7The search history graph of the locust algorithm for the 130 data sets in this embodiment;

[0063] Figure 8 This is a comparison diagram of the twin signal and the physical signal obtained when the data set is 130 in this embodiment;

[0064] Fig. 9 This is the normal distribution fitting diagram for the 130 data sets in this embodiment. DETAILED DESCRIPTION

[0065] 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.

[0066] 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.

[0067] 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.

[0068] like Figure 1-Figure 9 The method for constructing a rolling bearing digital twin model and optimizing model parameters shown in the figure includes the following steps:

[0069] S1: Based on the radial force, the dynamic relationship between the outer ring, inner ring and rolling elements, a four-degree-of-freedom digital twin model of the bearing is established, and the geometric parameters and operating parameters used to measure the bearing are input.

[0070] S11: Based on the bearing vibration response mechanism, a four-degree-of-freedom dynamic model of the bearing is established, combining the radial force action, the dynamic relationship between the outer ring, the inner ring, and the rolling elements;

[0071] Digital bearing reference Figure 2 Specifically, the four-degree-of-freedom dynamic equation uses the following calculation formula. It is worth noting that this equation takes into account the influence of radial force:

[0072]

[0073] In the formula, m in is the mass of the inner ring and shaft, c in is the shaft connection damping, k in is the equivalent connection stiffness of the shaft, x in is the displacement of the inner circle along the x direction, y in is the displacement of the inner circle along the y direction, m out is the mass of the outer ring and bearing base, c out is the bearing base damping, k out is the equivalent connection stiffness of the bearing base, x out is the displacement of the outer ring in the x direction, y out is the displacement of the outer ring along the y direction, W x is the radial force in the x direction, W y is the radial force in the y direction, F HX and F HY are the contact force components between the inner and outer rings and the rolling elements in the x and y directions, e is the eccentricity, ω is the angular frequency of the inner ring rotation, t is the time, and g is the acceleration due to gravity.

[0074] S12: When modeling bearing faults, the defect shape is simplified to a square;

[0075] The defect is introduced into the digital bearing by modeling it as a square. At the same time, considering that the defect height has little effect on the dynamic response, the defect height is simplified to a constant. Figure 3 .

[0076] S13: Input the measurable bearing geometric parameters and operating parameters to calculate the theoretical failure frequency of the bearing; the geometric parameters include pitch diameter size, rolling element size, rolling element number and mass; the operating parameters include rotation speed and time.

[0077] Specifically, the theoretical failure frequency of the bearing is calculated using the following formula:

[0078]

[0079] In the formula, f o is the outer race fault frequency, f iis the inner race fault frequency, f c is the cage failure frequency, f r is the rolling element failure frequency, z is the number of rolling elements, f is the rotational frequency, d is the rolling element diameter, D is the bearing pitch diameter, and α represents the bearing contact angle.

[0080] S2: Based on the locust algorithm, the key parameters that cannot be measured inside the bearing are initialized and optimized, and the virtual twin signal is output by the four-degree-of-freedom digital twin model of the bearing.

[0081] S21: Use the locust optimization algorithm to initialize and optimize the key parameters that cannot be measured inside the bearing.

[0082] Specifically, the locust algorithm uses the following calculation formula:

[0083]

[0084] Where N is the number of locusts, |x j -x i | is the distance from the i-th locust to the j-th locust, (x j -x i ) / d ij is the unit vector from the i-th locust to the j-th locust, s is the social force, ub d and lb d are the upper and lower boundaries of the d-dimensional search space, T d is the optimal solution found so far in the target direction. As the number of iterations increases, the outer c reduces the search area around the target, and the inner c reduces the interaction of locusts.

[0085] S22: Use the locust optimization algorithm to initialize and optimize the key parameters that cannot be measured inside the bearing. Specifically, select the parameters to be optimized according to the task. For example, if d parameters are optimized, the search space is d-dimensional, and the search upper and lower limits of d parameters are set ub d and lb d .

[0086] S23: The key parameters determined by the locust algorithm are input into the digital twin model, and the virtual twin signal is obtained by using the Runge-Kutta algorithm.

[0087] Specifically, the calculation formula of the Runge-Kutta algorithm is as follows:

[0088] K1=h*f(x n ,y n )

[0089] K2=h*f(x n +h / 2,y n +K1 / 2)

[0090] K3=h*f(x n +h / 2,y n +K2 / 2)

[0091] K4=h*f(x n +h,y n +K3)

[0092] y n+1 =y n +(K1+2*K2+2*K3+K4) / 6

[0093] x n+1 =x n +h

[0094] In the formula, x n is the x coordinate of the current point, y n is the y coordinate of the current point, f is the differential, h is the step size, K1 is the initial slope of the current time step, K2 is the slope of the middle point of the current time step, K3 is the slope of the other middle point of the current time step, and K4 is the initial slope of the next time step.

[0095] S3: Calculate the root mean square difference between the virtual twin signal and the actual vibration signal, and the locust algorithm searches for key parameters in the direction of reducing the difference.

[0096] S31: The measured vibration signal and the virtual twin signal interact in the locust algorithm; the virtual twin signal output by the specific bearing digital twin model is input into the locust algorithm, and the measured vibration signal of the real bearing is also input into the locust algorithm.

[0097] S32: The optimization goal of the locust algorithm is to minimize the root mean square difference between the measured vibration signal and the virtual twin signal.

[0098] Specifically, the optimization objective of the locust algorithm is calculated using the following formula:

[0099] minRMSD|=min|rms p -rms v |

[0100] In the formula, rms p is the root mean square of the measured bearing vibration signal, rms v is the root mean square obtained by solving the bearing digital twin model, and RMSD is the error between the two.

[0101] S33: The locust algorithm continuously searches for key parameters that reduce the error between the measured vibration signal and the virtual twin signal.

[0102] The specific locust algorithm continuously iterates in the direction of reducing RMSD, searching for the virtual vibration response with the minimum error with the measured vibration signal, and continuously updates key parameters during the search process.

[0103] S34: Record the minimum root mean square error and the corresponding key parameters, update the key parameters of the bearing digital twin model, and continue to iterate until the maximum number of iterations is reached.

[0104] Specifically, the locust algorithm records the minimum root mean square error and the corresponding optimal parameters during the iteration process. In a new round of iteration, it continues to search for new parameters near the recorded optimal parameters, brings the new parameters into the bearing digital twin model, generates new virtual twin signals, and continues to iterate until the maximum number of iterations is reached.

[0105] S4: The iteration of the locust algorithm ends, and the output signal with the smallest error between the output record and the measured vibration signal is the optimal twin signal, and the bearing under the corresponding parameters is the optimal digital twin.

[0106] S41: The maximum number of iterations is reached and the locust algorithm iteration ends.

[0107] At the beginning of the locust algorithm, the number of locusts was set to 100 and the number of iterations was set to 50.

[0108] S42: The optimal twin signal is the one with the smallest error between the recorded and measured vibration signals during the output iteration process.

[0109] Specifically, the locust algorithm stops iterating after reaching the maximum number of iterations. The locust algorithm outputs the minimum RMSD value during the search process, as well as the corresponding virtual twin signal at this time. This virtual twin signal with the smallest error with the measured vibration signal can be considered as the optimal twin signal.

[0110] S43: Output the parameters corresponding to the optimal twin signal. The digital bearing under the optimal parameters is the optimal digital twin of the bearing. The locust algorithm stops iterating after reaching the set maximum number of iterations.

[0111] Specifically, the locust algorithm outputs the minimum RMSD value during the search process, as well as the corresponding key parameter values ​​at this time. The digital twin model with these values ​​set can be considered as the optimal digital twin of the bearing.

[0112] In order to verify the high fidelity of the digital twin model constructed by the present invention and the high precision of the generated twin signal, experimental verification is carried out. The experimental data uses the CWRU data set, and the data sets numbered 130 (defect size is 0.1778mm) and 234 (defect size is 0.5334mm) are selected for testing. The speed is 1797rpm and the sampling frequency is 12kHz. The bearing model is SKF-6205, and the specific parameters are shown in Table 1 below.

[0113] Table 1 Rolling bearing parameters

[0114]

[0115] In this embodiment, 234 data sets are first used for testing. The geometric parameters and operating parameters of the bearing are input into the four-degree-of-freedom digital twin model of the bearing established by Simulink. The 234 data sets actually measured are loaded into the locust algorithm for standby use. The digital twin model of the bearing is solved using the Runge-Kutta algorithm, and the obtained virtual twin signal is input into the locust algorithm. In the locust algorithm, the root mean square of the measured signal and the root mean square of the virtual twin signal and the difference between the two are calculated, with the goal of minimizing this difference, updating the parameters, and continuously iterating. The optimization parameter set for this group of data is radial force. When the maximum number of iterations is reached, the locust algorithm stops iterating. The locust algorithm outputs the minimum error value between the two during the entire search process, as well as the corresponding optimization parameters and dynamic response at this time.

[0116] The locust algorithm finally obtained the minimum error between the two to be 1.1731×10 -6 The whole search process of the locust algorithm is as follows: Figure 4 As shown, the dynamic response of the output is Figure 5 As shown. Figure 4 It can be seen that after 50 rounds of searching, the optimal radial force finally determined is 116.1552N. Figure 5 It can be seen that the comparison between the measured vibration signal, the virtual twin signal and the virtual vibration response without locust algorithm optimization, the time domain signal is shown on the left and the envelope spectrum is shown on the right.

[0117] According to the bearing parameters and speed, the theoretical fault frequency of 107.4 Hz is calculated. Figure 5 As shown, most of the spectral lines of the virtual twin signal and the measured vibration signal match, but there are slight differences in amplitude. The cosine similarity between the twin signal and the physical signal is 0.62, which is greater than 0.6, and the consistency is good. In the envelope spectrum on the right, there are obvious spectral lines at the fault frequency and its multiples, as well as at the rotation frequency and sidebands. The key spectral lines all appear in the envelope spectrum, with only slight differences in amplitude. This is because the digital twin cannot fully reproduce the actual operating conditions of the bearing. The fault frequency error between the twin signal and the measured signal is less than 0.1%. The variance of the measured signal is 0.3406, and the variance of the twin signal is 0.3409, with an error of only 0.09%. It can be seen that this digital twin model of the bearing is high-fidelity. It can be seen that the virtual vibration response that has not been optimized by the locust algorithm has a small amplitude both in the time domain and in the envelope spectrum, which is far from the measured physical signal. The three signals are fitted with a normal distribution, and the fitting results are shown as follows. Figure 6As shown in the figure, it can be seen that the twin signal is closer to the measured signal. The distribution difference is quantified by the Vasastan distance. The Vasastan distance between the twin signal and the physical signal is 0.1092, and the Vasastan distance between the unoptimized virtual vibration response and the physical signal is 0.1876. Obviously, the distribution difference between the twin signal and the physical signal is smaller.

[0118] Using this high-fidelity model, we then used 130 data sets to conduct experiments. The optimization parameter set this time was the defect size. Previously, it was almost impossible to monitor the defect size of a real-time running bearing in industrial scenarios. By establishing a high-fidelity digital twin model, the size of the bearing fault can be estimated based on the measured signal of the bearing. This shows that the present invention is of great significance for monitoring the operating status of bearings in industrial scenarios, and for active maintenance and decision-making. When the locust algorithm completes the iteration, the minimum error maintained is 3.4665×10 -7 The whole optimization process of the locust algorithm is as follows: Figure 7 As shown in the figure, the optimal result of the search is 0.1767mm, while the actual defect size of the bearing is 0.1778mm, with an error of only 0.62%. The high-fidelity bearing digital twin model established by the present invention can accurately estimate the size of the bearing fault based on the measured signal. The comparison between the output twin signal and the real signal is as follows: Figure 8 As shown. It can be seen from the figure that most of the spectral lines of the virtual response coincide with those of the physical signal, but there are differences in amplitude. The cosine similarity between the twin signal and the physical signal is 0.66, which is greater than 0.6, and the consistency is good. From the envelope spectrum on the right side of the figure, obvious spectral lines can be seen at the rotation frequency and sidebands. Clear spectral lines also appear at the fault frequency and its multiples. The variance of the physical signal is 0.4475, and the variance of the twin signal is 0.4472, with an error of only 0.07%. This once again demonstrates the high fidelity of the data. From Figure 8 It can be seen that the virtual vibration response without optimization is very different from the physical signal in both time domain and envelope spectrum. The normal fitting results are as follows: Fig. 9 As shown in the figure, it can be seen that the distribution of the twin signal is closer to that of the measured signal. Using the Vasastan distance to quantify the distribution difference, the Vasastan distance between the twin signal and the physical signal is 0.106, and the Vasastan distance between the unoptimized virtual vibration response and the physical signal is 0.344. Obviously, the distribution difference between the twin signal and the physical signal is smaller.

[0119] Finally, the size of the bearing is estimated by using the established high-fidelity twin model and the measured bearing data of the 234 data set. The estimated result is 0.53449mm, and the actual fault size corresponding to the 234 data set is 0.5334mm, with an error of only 0.21%. 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 is described in detail with reference to the preferred embodiments, ordinary technicians in this field 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 covered by the scope of the claims of the present invention.

Claims

1. Construction of rolling bearing digital twin model and model parameter optimization method, characterized in that: The following steps are involved: S1: Combined with the radial force, the dynamic relationship between the outer ring, the inner ring and the rolling element, a four-degree-of-freedom digital twin model of the bearing is established. Based on the bearing vibration response mechanism, a four-degree-of-freedom dynamic model of the bearing is established, and the geometric parameters and operating parameters used to measure the bearing are input; the four-degree-of-freedom dynamic equation uses the following calculation formula: In the formula, m in is the mass of the inner ring and shaft, c in is the shaft connection damping, k in is the equivalent connection stiffness of the shaft, x in is the displacement of the inner circle along the x direction, y in is the displacement of the inner circle along the y direction, m out is the mass of the outer ring and bearing base, c out is the bearing base damping, k out is the equivalent connection stiffness of the bearing base, x out is the displacement of the outer ring in the x direction, y out is the displacement of the outer ring along the y direction, W x is the radial force in the x direction, W y is the radial force in the y direction, F HX and F HY is the contact force component between the inner and outer rings and the rolling element in the x and y directions, e is the eccentricity, ω is the angular frequency of the inner ring rotation, t is the time, and g is the acceleration due to gravity; S2: Initialize and optimize the key parameters that cannot be measured inside the bearing based on the locust algorithm, and output the virtual twin signal from the four-degree-of-freedom digital twin model of the bearing; it specifically includes the following steps: S21: Use the locust optimization algorithm to initialize and optimize the key parameters that cannot be measured inside the bearing; S22: Select the parameters to be optimized according to the task; optimize d parameters, the search space is d-dimensional, and set the search upper and lower limits ub of d parameters d and lb d ; S23: Input the key parameters determined by the locust algorithm into the digital twin model, and use the Runge-Kutta algorithm to solve and obtain the virtual twin signal; S3: Calculate the root mean square difference between the virtual twin signal and the actual vibration signal, and the locust algorithm searches for key parameters in the direction of reducing the difference; S4: The iteration of the locust algorithm ends, and the output signal with the smallest error between the output record and the measured vibration signal is the optimal twin signal, and the bearing under the corresponding parameters is the optimal digital twin.

2. The method for constructing a rolling bearing digital twin model and optimizing model parameters according to claim 1, characterized in that: S1 specifically includes the following steps: S11: Based on the bearing vibration response mechanism, a four-degree-of-freedom dynamic model of the bearing is established, combining the radial force action, the dynamic relationship between the outer ring, the inner ring, and the rolling elements; S12: When modeling the bearing fault, the defect shape is simplified to a square; by modeling the defect as a square, the defect is introduced into the digital bearing. At the same time, considering that the defect height has little effect on the dynamic response, the defect height is simplified to a constant; S13: Input the measurable bearing geometric parameters and operating parameters to calculate the theoretical failure frequency of the bearing; the geometric parameters include pitch diameter size, rolling element size, rolling element number and mass; the operating parameters include rotation speed and time.

3. The method for constructing a rolling bearing digital twin model and optimizing model parameters according to claim 1, characterized in that: S3 specifically includes the following steps: S31: The measured vibration signal and the virtual twin signal interact in the locust algorithm; the virtual twin signal output by the bearing digital twin model is input into the locust algorithm, and the measured vibration signal of the real bearing is also input into the locust algorithm; S32: The optimization goal of the locust algorithm is to minimize the root mean square difference between the measured vibration signal and the virtual twin signal; S33: The locust algorithm continuously searches for key parameters that reduce the error between the measured vibration signal and the virtual twin signal. The locust algorithm continuously iterates in the direction of reducing RMSD, searching for the virtual vibration response with the minimum error with the measured vibration signal, and continuously updates the key parameters during the search process. S34: Record the minimum root mean square error and the corresponding key parameters, update the key parameters of the bearing digital twin model, and continue to iterate until the maximum number of iterations is reached. The locust algorithm records the minimum root mean square error and the corresponding optimal parameters during the iteration process. In a new round of iteration, continue to search for new parameters near the recorded optimal parameters, bring the new parameters into the bearing digital twin model, generate new virtual twin signals, and continue to iterate until the maximum number of iterations is reached.

4. The method for constructing a rolling bearing digital twin model and optimizing model parameters according to claim 1, characterized in that: S4 specifically includes the following steps: S41: The maximum number of iterations is reached, and the locust algorithm iteration ends; at the beginning of the locust algorithm, the number of locusts is set to 100 and the number of iterations is set to 50; S42: Output the optimal twin signal with the smallest error between the recorded and measured vibration signals during the iteration process; after the locust algorithm reaches the set maximum number of iterations, it stops iterating, and the locust algorithm outputs the minimum RMSD value during the search process, as well as the corresponding virtual twin signal at this time. This virtual twin signal with the smallest error between the recorded and measured vibration signals is the optimal twin signal; S43: Output the parameters corresponding to the optimal twin signal. The digital bearing under this parameter is the optimal digital twin of the bearing. The locust algorithm stops iterating after reaching the set maximum number of iterations. The locust algorithm outputs the minimum RMSD value during the search process and the corresponding key parameter values ​​at this time. The digital twin model with these values ​​set is the optimal digital twin of the bearing.

5. The method for constructing a rolling bearing digital twin model and optimizing model parameters according to claim 2, characterized in that: The theoretical failure frequency of the bearing is calculated in S13 using the following formula: In the formula, f o is the outer race fault frequency, f i is the inner race fault frequency, f c is the cage failure frequency, f r is the rolling element failure frequency, z is the number of rolling elements, f is the rotational frequency, d is the rolling element diameter, D is the bearing pitch diameter, and α represents the bearing contact angle.

6. The method for constructing a rolling bearing digital twin model and optimizing model parameters according to claim 1, characterized in that: The locust algorithm in S21 uses the following calculation formula: Where N is the number of locusts, |x j -x i | is the distance from the i-th locust to the j-th locust, (x j -x i ) / d ij is the unit vector from the i-th locust to the j-th locust, s is the social force, ub d and lb d are the upper and lower boundaries of the d-dimensional search space, T d is the optimal solution found so far in the target direction. As the number of iterations increases, the outer c reduces the search area around the target, and the inner c reduces the interaction of locusts.

7. The method for constructing a rolling bearing digital twin model and optimizing model parameters according to claim 1, characterized in that: The calculation formula of Runge-Kutta algorithm in S23 is as follows: K1=h*f(x n ,y n ) K2=h*f(x n +h / 2,y n +K1 / 2) K3=h*f(x n +h / 2,y n +K2 / 2) K4=h*f(x n +h,y n +K3) y n+1 =y n +(K1+2*K2+2*K3+K4) / 6 x n+1 =x n +h In the formula, x n is the x coordinate of the current point, y n is the y coordinate of the current point, f is the differential, h is the step size, K1 is the initial slope of the current time step, K2 is the slope of the middle point of the current time step, K3 is the slope of the other middle point of the current time step, and K4 is the initial slope of the next time step.

8. The method for constructing a rolling bearing digital twin model and optimizing model parameters according to claim 3, characterized in that: The optimization objective of the locust algorithm in S32 is calculated using the following formula: min|RMSD|=min|rms p -rms v | In the formula, rms p is the root mean square of the measured bearing vibration signal, rms v is the root mean square obtained by solving the bearing digital twin model, and RMSD is the error between the two.

Images