Digital twinning-assisted bearing weak fault diagnosis method

Abstract

The invention belongs to the technical field of fault detection, and discloses a digital twin-assisted bearing weak fault diagnosis method, which comprises the following steps of: 1, establishing a dynamical model of three faults of a rolling bearing; and 2, setting parameters of the kinetic model according to an actual test bed, and updating according to actually acquired vibration data to obtain a simulated vibration signal. And step 3, performing sparse feature optimization by taking the simulated vibration signal as a sample, and obtaining each column of the weight matrix as a filter. And 4, filtering the original signal by using the optimized filter to obtain a filtered signal, and observing the fault characteristics of the filtered signal through a Hilbert envelope spectrum. According to the method, the influence of sample noise on the result in the sparse optimization process is reduced, and noise interference can be more effectively suppressed, so that fault features are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault detection, and particularly relates to a digital twin-assisted weak fault diagnosis method for bearings. Background Technique

[0002] As a core component of mechanical equipment, bearings are widely used in fields such as ships, wind power generation, and aviation. Their working conditions directly affect the normal operation of mechanical equipment. In actual work, bearings often operate in high-load and high-speed working environments. As the operating time of mechanical equipment increases, bearings are extremely prone to damage or failure, resulting in a decrease in power transmission efficiency or even equipment damage, thereby inducing safety accidents.

[0003] Due to the continuous development of mechanical systems towards precision and high speed, the complexity of the operation of mechanical equipment has made traditional fault diagnosis methods unable to handle quickly and accurately. Therefore, the safety and reliability of mechanical equipment have received more attention. In the initial stage of a fault, the energy of the weak fault characteristics of rolling bearings is low, and fault detection methods such as sparse feature optimization methods are greatly affected by sample noise. Therefore, it is difficult to form effective detection in the initial stage of a fault. Summary of the Invention

[0004] The purpose of the present invention is to provide a digital twin-assisted weak fault diagnosis method for bearings, which is used to solve the technical problems in the prior art that the energy of the weak fault characteristics of rolling bearings is low and the sparse feature optimization method is greatly affected by sample noise.

[0005] The described digital twin-assisted weak fault diagnosis method for bearings includes the following steps.

[0006] Step 1: Establish a dynamic model of three faults of a rolling bearing.

[0007] Step 2: Set the parameters of the dynamic model according to the actual test bench and update them according to the actually collected vibration data to obtain a simulated vibration signal.

[0008] Step 3: Optimize the sparse features using the simulated vibration signal as a sample, and each column of the weight matrix is used as a filter.

[0009] Step 4: Filter the original signal using the optimized filter to obtain a filtered signal, and observe the fault characteristics of the filtered signal through the Hilbert envelope spectrum.

[0010] Preferably, in Step 1, a dynamic model of three faults of a rolling bearing is established in the visualization simulation tool Simulink, including the following specific steps.

[0011] Step 1-1: Deduce the dynamic characteristics of three types of faults of rolling bearings for establishing the dynamic model of rolling bearings in Simulink; the dynamic model of the rolling bearing is simplified to a 5-degree-of-freedom dynamic model.

[0012] Step 1-2: Conduct dynamic modeling in the visualization simulation tool Simulink according to the deduced dynamic characteristics.

[0013] Preferably, in Step 1-1, the inner ring of the simplified bearing has 2 degrees of freedom, namely the horizontal displacement x i and the vertical displacement y i ; the outer ring has 2 degrees of freedom, namely the horizontal displacement x o and the vertical displacement y o ; the unit resonator has 1 degree of freedom of vertical displacement y r , representing the simplified rolling elements and cage; the inner ring of the bearing rotates together with the transmission shaft driven by the motor at a speed of ω s ; M, K, and C respectively represent the mass, stiffness, and damping coefficient of the simplified bearing, M i , k i , C i respectively represent the mass, stiffness, and damping coefficient of the inner ring of the simplified bearing, M o , K o , C o respectively represent the mass, stiffness, and damping coefficient of the outer ring of the simplified bearing, M r , K r , C r respectively represent the mass, stiffness, and damping coefficient of the rolling elements and cage of the simplified bearing; in this simplified model, the outer ring is fixed on the bearing seat, the inner ring rotates together with the rotating shaft, and the surface contact between the rolling elements and the bearing raceway satisfies the Hertz contact model; the dynamic differential equation of this simplified bearing is:

[0014]

[0015] In the formula, W is the radial load applied to the inner ring, F x , F y are respectively the contact forces of the bearing in the x and y directions, and g is the acceleration due to gravity; through this simplified model, the dynamic models of outer ring faults, inner ring faults, and rolling element faults of the bearing are established respectively.

[0016] Preferably, in step 1-2, five main 'Add' modules are used to represent five equations for simplifying the dynamic differential equations of the bearing; 'acc0' is the output of the model, simulating the acceleration vibration signal; the 'fcn' module can control the parameters of the simulated roller signal; the input parameter 'u' is used to define the type of bearing damage. When u = 0, the model simulates the healthy state of the bearing; when u = 1, the model simulates the outer raceway damage of the bearing; when u = 2, the model simulates the inner raceway damage of the bearing; when u = 3, the model simulates the rolling element damage of the bearing.

[0017] Preferably, step 2 includes the following specific steps.

[0018] Step 2-1: Set the structural parameters, geometric dimensions, bearing fault types, and operating conditions in the bearing model according to the actual test bearing.

[0019] Step 2-2: Set the sampling frequency in the acquisition system to be consistent with the test.

[0020] Step 2-3: Update the parameters of the simulation model through the actually collected vibration signals.

[0021] Step 2-4: Obtain the vibration acceleration signals of three fault types of the test bearing through simulation.

[0022] Preferably, step 3 includes the following specific steps.

[0023] Step 3-1: Construct the Hankel matrix X from the collected simulation vibration signals H :

[0024]

[0025] In the formula, N-l+1 represents the sample length, and l represents the number of samples.

[0026] Step 3-2: Set the number of columns of the filter according to the number of rows of the Hankel matrix constructed from the signal, and set the number of rows of the filter according to the requirements.

[0027] Step 3-3: Initialize the weight matrix with the size of the filter, and perform sparse feature optimization using these simulation vibration signals and the weight matrix respectively to obtain filters for different fault features.

[0028] Preferably, in step 3-3, these simulation vibration signals are respectively subjected to sparse optimization according to the following formula:

[0029]

[0030] Where represents the sample feature, that is, the jth feature of the ith sample, and F represents the feature matrix; in the above formula, H is composed of XH Replace to obtain filters for different fault characteristics:

[0031]

[0032] In the formula, N f represents the number of filters, and the filter length is the same as the sample length.

[0033] Preferably, step 4 includes the following specific steps:

[0034] Step 4-1: Filter the vibration signals collected in the actual test using the optimized filters;

[0035] Step 4-2: Perform Hilbert envelope spectrum on the filtered signals;

[0036] Step 4-3: Observe the frequency characteristics of the signals through the Hilbert envelope spectrum.

[0037] The advantages of the present invention are as follows: The present invention can simulate three types of faults of the bearing outer ring, inner ring, and rolling elements. The model established by Simulink is more intuitive than the models established by codes in other ways, and the parameter update is more convenient. By updating the parameters of the bearing in the model, the dynamic behavior of rolling bearings on different test benches can be simulated. The simulation signals generated by the model are used as samples for sparse optimization to obtain filters, and then these filters are used to filter the signals collected in the test to obtain enhanced fault characteristics, reducing the influence of sample noise on the results during the sparse optimization process. The present invention can more effectively suppress noise interference and thus enhance fault characteristics. Description of the Drawings

[0038] Figure 1 is a schematic diagram of the principle framework of a digital twin-assisted bearing weak fault diagnosis method of the present invention.

[0039] Figure 2 is a 5-degree-of-freedom dynamic model of the rolling bearing in the present invention.

[0040] Figure 3 is a simulation model of the rolling bearing obtained by performing dynamic modeling in the visualization simulation tool Simulink in the present invention.

[0041] Figure 4 is the envelope spectrum of the original vibration signal in the present invention.

[0042] Figure 5 is the envelope spectrum of the filtered signal obtained by using the present invention. Detailed Embodiments

[0043] The following will, with reference to the accompanying drawings, further elaborate on the specific embodiments of the present invention through the description of the embodiments, so as to help those skilled in the art have a more complete, accurate and in-depth understanding of the inventive concept and technical solution of the present invention.

[0044] As Figures 1 to 5 shown, the present invention provides a digital twin-assisted weak fault diagnosis method for bearings, including the following steps.

[0045] Step 1: Establish a dynamic model for three types of faults of a rolling bearing.

[0046] Establish a dynamic model for three types of faults of a rolling bearing in the visualization simulation tool Simulink, including the following specific steps.

[0047] Step 1-1: Deduce the dynamic characteristics of three types of faults of a rolling bearing for establishing a dynamic model of the rolling bearing in Simulink. The dynamic model of the rolling bearing is simplified to a 5-degree-of-freedom dynamic model, as Figure 2 shown.

[0048] This 5-degree-of-freedom dynamic model of the rolling bearing combines the contact deformation theory to simulate the dynamic behaviors of inner race faults, outer race faults and rolling element faults. Among them, the inner race of the simplified bearing has 2 degrees of freedom, namely the horizontal displacement x i and the vertical displacement y i ; the outer race has 2 degrees of freedom, namely the horizontal displacement x o and the vertical displacement y o ; the unit resonator has 1 degree of freedom of vertical displacement y r , representing the simplified rolling elements and cage; the inner race of the bearing rotates together with the transmission shaft driven by the motor at a speed of ω s ; M, K, and C respectively represent the mass, stiffness and damping coefficient of the simplified bearing, M i , k i , C i respectively represent the mass, stiffness and damping coefficient of the inner race of the simplified bearing, M o , K o , C o respectively represent the mass, stiffness and damping coefficient of the outer race of the simplified bearing, M r , K r , C r respectively represent the mass, stiffness and damping coefficient of the rolling elements and cage of the simplified bearing; in this simplified model, the outer race is fixed on the bearing housing, the inner race rotates with the rotating shaft, and the surface contact between the rolling elements and the bearing raceway satisfies the Hertz contact model; the dynamic differential equation of this simplified bearing is:

[0049]

[0050] where W is the radial load applied to the inner ring, and F x , F y are the contact forces of the bearing in the x and y directions respectively, and g is the acceleration due to gravity. Through this simplified model, the dynamic models of the outer ring fault, inner ring fault, and rolling element fault of the bearing are established respectively.

[0051] Step 1-2: Perform dynamic modeling in the visualization simulation tool Simulink according to the derived dynamic characteristics.

[0052] A dynamic model is established in MATLAB / Simulink, and this model is as Figure 3 shown. Five main 'Add' modules are used to represent the five equations of the dynamic differential equation of the simplified bearing; 'acc0' is the output of the model, simulating the acceleration vibration signal; the 'fcn' module can control the parameters of the simulation roller signal; the input parameter 'u' is used to define the type of bearing damage. When u = 0, the model simulates the healthy state of the bearing; when u = 1, the model simulates the outer ring raceway damage of the bearing; when u = 2, the model simulates the inner ring raceway damage of the bearing; when u = 3, the model simulates the rolling element damage of the bearing. The model established by MATLAB / Simulink is more intuitive and easy to understand, and the parameter adjustment of the model is also very convenient.

[0053] Step 2: Set the parameters of the dynamic model according to the actual test bench and update them according to the actually collected vibration data to obtain the simulated vibration signal. This step includes the following specific steps.

[0054] Step 2-1: Set the structural parameters, geometric dimensions, bearing fault type, operating conditions, etc. in the bearing model according to the actual test bearing.

[0055] Step 2-2: Set the sampling frequency in the acquisition system to be consistent with the test.

[0056] Step 2-3: Update the parameters of the simulation model through the actually collected vibration signal.

[0057] Step 2-4: Obtain the vibration acceleration signals of the three fault types of the test bearing through simulation.

[0058] Step 3: Use the simulated vibration signal as a sample for sparse feature optimization to obtain a filter for each column of the weight matrix. This step includes the following specific steps.

[0059] Step 3-1: Construct the collected simulated vibration signal into a Hankel matrix X H .

[0060]

[0061] In the formula, N-l+1 represents the sample length, and l represents the number of samples.

[0062] Step 3-2: Construct the number of rows of the Hankel matrix according to the signal, set the number of columns of the filter, and set the number of rows of the filter as required.

[0063] Step 3-3: Initialize the weight matrix with the size of the filter, use these simulated vibration signals and the weight matrix to perform sparse feature optimization respectively, and obtain filters for different fault characteristics.

[0064] These simulated vibration signals are used for sparse optimization according to the following formulas:

[0065]

[0066] in, represents the sample feature, that is, the jth feature of the i-th sample, and F represents the feature matrix. In the above formula, the matrix H is composed of X H Replace and get filters for different fault characteristics:

[0067]

[0068] Where N f Indicates the number of filters, and the filter length is consistent with the sample length.

[0069] These filters are used to filter the vibration signals collected in the actual test, so as to extract the fault characteristics in the actual vibration signals. The relationship between the vibration signal and the filter collected in the test is expressed by Z transform as follows:

[0070]

[0071] Where i represents the optimized i-th filter, that is, the i-th column of the filter matrix W.

[0072] Step 4: Filter the original signal with the optimized filter to obtain a filtered signal, and observe the fault characteristics of the filtered signal through the Hilbert envelope spectrum.

[0073] Step 4-1: Use the optimized filter to filter the vibration signal collected in the actual test.

[0074] Step 4-2: Calculate the Hilbert envelope spectrum of the filtered signal.

[0075] Step 4-3: Observe the frequency characteristics of the signal through the Hilbert envelope spectrum.

[0076] Figure 4is the envelope spectrum of the original vibration signal, and the envelope spectrum of the filtered vibration signal obtained by the method of the present invention is as Figure 5 shown. The circles in the figure represent the fault characteristic frequencies of the bearing inner ring and their multiples. In Figure 4 , the amplitudes of the characteristic frequencies and their multiples expressing the fault characteristics of the bearing inner ring are not prominent. In Figure 5 , the amplitudes of the characteristic frequencies and their multiples of the fault characteristics of the bearing inner ring are clearly visible.

[0077] The present invention has been described exemplarily above in conjunction with the accompanying drawings. Obviously, the specific implementation of the present invention is not limited by the above-mentioned manner. As long as various non-substantive improvements are made by adopting the inventive concept and technical solution of the present invention, or the inventive concept and technical solution of the present invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. A digital twin-assisted bearing weak fault diagnosis method, characterized by: The following steps are involved: Step 1: Establish the dynamic models of three types of rolling bearing faults; Step 2: Set the parameters of the dynamic model according to the actual test bench and update it according to the actual collected vibration data to obtain the simulated vibration signal; Step 3: Use the simulated vibration signal as a sample to perform sparse feature optimization, and obtain each column of the weight matrix as a filter; Step 4: Filter the original signal with the optimized filter to obtain a filtered signal, and observe the fault characteristics of the filtered signal through the Hilbert envelope spectrum.

2. A digital twin-assisted bearing weak fault diagnosis method according to claim 1, characterized in that: The step 1 establishes the dynamic models of three kinds of rolling bearing faults in the visual simulation tool Simulink, including the following specific steps: Step 1-1: Derive the dynamic characteristics of three types of rolling bearing faults to establish a dynamic model of the rolling bearing in Simulink; the dynamic model of the rolling bearing is simplified to a 5-DOF dynamic model; Step 1-2: Perform dynamic modeling based on the derived dynamic characteristics in the visual simulation tool Simulink.

3. A digital twin-assisted bearing weak fault diagnosis method according to claim 2, characterized in that: In step 1-1, the inner ring of the simplified bearing has two degrees of freedom, namely horizontal displacement x i and vertical displacement y i ; The outer ring has two degrees of freedom, namely horizontal displacement x o and vertical displacement y o ; The unit resonator has 1 degree of freedom of vertical displacement y r , represents a simplified rolling element and cage; the inner ring of the bearing is driven by the motor along the transmission shaft at ω s The bearing rotates with the speed of the shaft; M, K, and C represent the mass, stiffness, and damping coefficient of the simplified bearing, respectively. i , k i , C i are the mass, stiffness and damping coefficient of the inner ring of the simplified bearing, M o , K o , C o They represent the mass, stiffness and damping coefficient of the outer ring of the simplified bearing, M r , K r , C r Respectively represent the mass, stiffness and damping coefficient of the rolling element and cage of the simplified bearing; in this simplified model, the outer ring is fixed on the bearing seat, the inner ring rotates with the shaft, and the surface contact between the rolling element and the bearing raceway satisfies the Hertz contact model; the dynamic differential equation of the simplified bearing is: Where W is the radial load applied to the inner ring, F x , F y are the contact forces of the bearing in the x and y directions respectively, and g is the acceleration of gravity. Through this simplified model, the dynamic models of bearing outer ring fault, inner ring fault and rolling element fault are established respectively.

4. A digital twin-assisted bearing weak fault diagnosis method according to claim 2, characterized in that: In steps 1-2, the five main 'Add' modules are used to represent the five equations of the simplified bearing dynamic differential equations; 'acc0' is the output of the model, simulating the acceleration vibration signal; the 'fcn' module can control the parameters of the simulated roller signal; the input parameter 'u' is used to define the type of bearing damage, when u=0, the model simulates the healthy state of the bearing; when u=1, the model simulates the outer ring raceway damage of the bearing; when u=2, the model simulates the inner ring raceway damage of the bearing; when u=3, the model simulates the rolling element damage of the bearing.

5. The method for diagnosing weak bearing faults assisted by digital twins according to claim 1, characterized in that: The step 2 comprises the following specific steps: Step 2-1: Set the structural parameters, geometric dimensions, bearing fault types, and operating conditions in the bearing model according to the actual test bearing; Step 2-2: Set the sampling frequency in the acquisition system to be consistent with the experiment; Step 2-3: Update the parameters of the simulation model through the collected actual vibration signal; Step 2-4: Obtain the vibration acceleration signals of the three fault types of the test bearing through simulation.

6. A digital twin-assisted bearing weak fault diagnosis method according to claim 1, characterized in that: The step 3 comprises the following specific steps: Step 3-1: Construct the collected simulated vibration signal into the Hankel matrix X H : In the formula, N-l+1 represents the sample length, l represents the number of samples; Step 3-2: Construct the number of rows of the Hankel matrix according to the signal, set the number of columns of the filter, and set the number of rows of the filter according to the requirements; Step 3-3: Initialize the weight matrix with the size of the filter, use these simulated vibration signals and the weight matrix to perform sparse feature optimization respectively, and obtain filters for different fault characteristics.

7. A digital twin-assisted bearing weak fault diagnosis method according to claim 6, characterized in that: In step 3-3, these simulated vibration signals are used to perform sparse optimization according to the following formulas: in, represents the sample feature, that is, the jth feature of the i-th sample, and F represents the feature matrix; in the above formula, H is composed of X H Replace and get filters for different fault characteristics: Where N f Indicates the number of filters, and the filter length is consistent with the sample length.

8. A digital twin-assisted bearing weak fault diagnosis method according to claim 1, characterized in that: The step 4 comprises the following specific steps: Step 4-1: Use the optimized filter to filter the vibration signal collected in the actual test; Step 4-2: Calculate the Hilbert envelope spectrum of the filtered signal; Step 4-3: Observe the frequency characteristics of the signal through the Hilbert envelope spectrum.

Images