Modeling method for digital twinning mechanism model of carrier roller bearing of belt conveyor

Abstract

The invention discloses a modeling method for a digital twinning mechanism model of a carrier roller bearing of a belt conveyor, and the method achieves the estimation and prediction of the full-life-cycle defect size of the carrier roller bearing of the belt conveyor by means of measurable vibration signals and a long-short-term memory network and numerical simulation method. Therefore, a mechanism model of the defect size evolution of the carrier roller bearing of the belt conveyor is established. According to the method, the difficulty in modeling the defect size evolution mechanism model of the carrier roller bearing in the process of building the digital twin platform of the belt conveyor can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment failure prediction, and in particular to a method for modeling a digital twin mechanism model of a belt conveyor roller bearing. Background Art

[0002] In the coal transportation industry, belt conveyors are the most critical transportation equipment, which can continuously transport small pieces of materials. They have the characteristics of large transportation volume, long transportation distance and high transportation efficiency. Rollers are key components on belt conveyors, which support the conveyor belt and the materials it carries and ensure the smooth operation of the conveyor belt. It can reduce the wear of the conveyor belt and effectively extend the service life of the conveyor belt. The healthy state of the rollers is crucial to the stable operation of the belt conveyor. Once the rollers are worn or stuck, it will not only cause the conveyor belt to jump and deviate, causing material accumulation and blockage, but also increase the wear of the conveyor belt, and even cause fires, resulting in serious consequences. Therefore, establishing a digital twin model for roller bearings and predicting faults are of great practical significance for mine safety production.

[0003] Digital twin technology is a technology that maps real equipment to digital space with high fidelity. It can accurately reproduce the state, behavior and operating rules of real objects in digital space, thereby realizing real-time monitoring, state prediction, predictive maintenance and optimization decision-making of physical entities. To realize digital twin technology, it is necessary to build its 3D model and mechanism model, and the key lies in the modeling of the mechanism model. However, the mechanism modeling method of the digital twin of the roller bearing throughout its life cycle faces the following challenges:

[0004] 1) Since the life evolution of roller bearings is time-varying and nonlinear, it is difficult to construct an accurate mechanism model of defect size evolution using traditional modeling methods based on physical laws.

[0005] 2) The roller bearing is installed inside the belt conveyor. The evolution data of its actual defect size during its entire life cycle is difficult to measure. How to use measurable data to estimate its actual defect size is a key issue. Summary of the invention

[0006] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide a method for modeling a mechanism model of a digital twin of a belt conveyor roller bearing, which provides a feasible technical path to realize the mechanism modeling of the defect size evolution of a belt conveyor roller bearing, thereby solving the problem of the driving mechanism for realizing the digital twin of the roller bearing.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] The present invention provides a belt conveyor roller bearing digital twin mechanism modeling method, comprising the following steps:

[0009] Step 1: regularly sample the real vibration signal samples of the roller bearing in the whole life cycle, perform mathematical processing on each collected real vibration signal sample, and calculate the effective value RMS and peak value Pt of each sample;

[0010] Step 2: Design an LSTM time series prediction network, use the effective value RMS of the collected real vibration data for model training, obtain a time series prediction model, and verify the effectiveness of the model;

[0011] Step 3: Collect the vibration signal of the roller bearing in use and calculate its effective value RMS, and input the calculated effective value RMS into the model trained in step 2 to predict the RMS in the future;

[0012] Step 4: Using Hertz theory, a mathematical model of a differential equation group of virtual defect size and simulated vibration signal of the roller bearing is established, and the simulated vibration signal corresponding to different virtual defect sizes Hs is solved by the Runge-Kutta method, and the peak value Ps of the simulated vibration signal is calculated;

[0013] Step 5: Taking the peak value Pt of the actual vibration signal as a reference, traverse the peak values ​​Ps of the simulated vibration signals corresponding to all virtual sizes H. If they are within the allowable error range, it is considered that the real defect size Ht corresponding to the actual vibration signal sample is the virtual defect size Hs, thereby obtaining the mapping relationship between the effective value RMS of the sample and the real defect size Ht, and realizing the estimation of the real defect size.

[0014] Step 6: Use the RMS obtained in step 5 as a feature and the actual defect size Ht as a label to train a fully connected neural network to obtain a model containing the mapping relationship between RMS and Ht. Use the RMS predicted in step 3 as input to the trained model to obtain the predicted defect size Ht for a period of time in the future, thereby establishing a life degradation model for roller bearings.

[0015] Preferably, step 1 comprises the following steps:

[0016] S11. Use the acceleration sensor to sample the real vibration data of the roller bearing throughout its life cycle. The sampling duration of each sample is 1 second. The collected data is stored in a folder. Each file name is the sampling time. A total of 984 sample files are saved until the bearing fails.

[0017] S12. Calculate the effective value RMS and peak value Pt of each sample and store them in a .jason format file. The effective value calculation formula is:

[0018]

[0019] k is the number of discrete signals contained in each sample, y i is the value of a discrete signal.

[0020] Preferably, in step 2, the method for constructing the time series prediction model is specifically as follows:

[0021] S21, prepare a training data set to train a time series prediction model, process the 984 RMS data samples stored in step S12, use the RMS data of the first 20 time units as features, and the RMS data of the next 2 time units as labels, slide backward 2 time units each time, and obtain a total of 482 sets of data;

[0022] S22. The obtained 482 sets of data are sequentially put into the long short-term memory network LSTM to train the model and obtain a time series prediction model.

[0023] Preferably, step 3 specifically includes the following contents:

[0024] S31, using an inspection robot equipped with an acceleration sensor to regularly sample the vibration signal of each belt conveyor roller in use, performing VMD decomposition on the collected vibration signal, selecting a component with the largest kurtosis value, and storing the component;

[0025] S32, extract the latest 20 time unit data samples collected, and calculate the effective value RMS of each sample;

[0026] S33. Put the effective values ​​of the 20 time unit samples into the time series prediction model to perform time series prediction, and obtain the effective value RMS of 2 time units after the current time.

[0027] Preferably, step 4 specifically includes the following contents:

[0028] S41. The collision between the ball and the outer ring is equivalent to the model of the damper and the spring. The Hertz contact theory is used to establish the differential equations between the virtual defect size of the roller bearing and the simulated vibration signal. The expression is:

[0029]

[0030] Where F X ,F Y are the radial forces in the X and Y directions respectively, M is the total mass of the inner ring and the shaft, C is the damping coefficient, K is the stiffness, λ is the effective contact area of ​​the rolling element, δ i is the contact deformation between the ith rolling element and the raceway, θ i is the current angle of the i-th rolling body;

[0031] S42: contact deformation δ between the i-th rolling element and the raceway in step S41 i It is expressed as the following formula:

[0032] δ i =xcosθ i +ysinθ i -cH′

[0033] Where x, y are vibration displacements, c is the radial clearance of the bearing, and H' is the displacement excitation function;

[0034] S43: The effective contact area λ and the contact area δ of the i-th rolling element in step S41 i Related, expressed as:

[0035]

[0036] S44. The displacement excitation function H' in step S42 is expressed by the following formula:

[0037]

[0038] Where H is the value of the virtual defect size, β0 is the starting angle of the virtual defect, and β q is the angle when the rolling element completely falls into the defect, and β is the angle spanned by the virtual defect size;

[0039] S45, setting the virtual defect size Hs from 0.1 mm to 3 mm, with a step length of 0.2 mm for each step, using the Runge-Kutta method to solve the above differential equations, and solving a set of numerical solutions of x and y corresponding to each virtual defect size, each set of solutions is a simulated vibration signal corresponding to each virtual defect size;

[0040] S46. Calculate the peak values ​​Ps of the above-mentioned groups of simulated vibration signals, with each virtual defect size Hs as a key and the corresponding peak value Ps of the simulated vibration signal as a value, and store them in the form of key-value pairs.

[0041] Preferably, step 5 specifically includes the following contents:

[0042] S51, taking the peak value Pt of each real vibration signal sample of the roller bearing in the whole life cycle calculated in step S12 as a reference, traverse the key-value pairs consisting of each virtual defect size Hs and the simulated vibration signal Ps stored in step S46, and find the simulated vibration signal peak value Ps closest to each real peak value Pt, then the virtual size corresponding to Ps is considered to be the real defect size Ht corresponding to the current real peak value Pt, thereby realizing the estimation of the real defect size;

[0043] S52, after matching all the real vibration signals in step S12 with the real defect size Ht using the method of S51, a mapping relationship between the effective value RMS of all real vibration signal samples and the real defect size Ht is obtained.

[0044] Preferably, step 6 specifically includes the following contents:

[0045] S61, taking the effective value RMS of each sample as input and its corresponding real defect size Ht as output, training a fully connected neural network to obtain a model that can reflect the mapping relationship between the effective value RMS and the defect size Ht;

[0046] S62. In step S33, the effective value RMS of the running roller bearing in the next two time units has been obtained. These two effective value RMS are input into the model trained in step S61 to obtain an estimate of the actual defect size in the next two time units, thereby realizing the prediction of the evolution of the roller bearing defects and constructing a mechanism model of the evolution of the roller bearing defect size.

[0047] The beneficial effects of the present invention are:

[0048] 1. The present invention uses a data-driven method to establish a model for the degradation of roller bearing defect size over time, solving the problem that it is difficult to mathematically model the evolution process of belt conveyor roller bearing defect size.

[0049] 2. The present invention proposes a scheme to estimate the real defect size using measurable vibration signals, introduces the concept of virtual defect size, establishes a mathematical model of a differential equation group of virtual defect size and simulated vibration signal of roller bearings, and uses the peak matching method to reasonably estimate the real defect size of roller bearings throughout their life cycle.

[0050] 3. The present invention predicts the faults of belt conveyor roller bearings based on a defect degradation model. Different from the traditional post-detection method, the present invention can effectively improve the operation and maintenance efficiency and reduce the frequency of faults. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 It is a flow chart of a method for modeling a digital twin mechanism model of a belt conveyor roller bearing provided by an embodiment of the present invention;

[0053] Figure 2 is a comparison chart of the actual value and the predicted value of the time series prediction of the bearing RMS provided by an embodiment of the present invention;

[0054] Figure 3 It is a peak diagram of a simulated vibration signal obtained by solving a virtual defect size Hs from 0.1 mm to 3 mm provided by an embodiment of the present invention;

[0055] Figure 4 is a graph showing the change in peak value of a real vibration signal over time provided by an embodiment of the present invention;

[0056] Figure 5 This is a graph showing the variation of the actual defect size over time obtained after peak matching provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] like Figure 1 As shown, this embodiment provides a method for modeling a digital twin mechanism model of a belt conveyor roller bearing, comprising the following steps:

[0059] Step 1: Collect raw data and process the data, including the following:

[0060] S11. Use the acceleration sensor to sample the real vibration data of the roller bearing throughout its life cycle. The sampling duration of each sample is 1s. The collected data is stored in a folder. Each file name is the sampling time. A total of 984 sample files are saved until the bearing fails.

[0061] S12. Calculate the effective value RMS and peak value Pt of each sample and store them in a .jason format file. k is the number of discrete signals contained in each sample, y is i is the value of the discrete signal, and the effective value calculation formula is:

[0062]

[0063] Step 2: Train and verify the time series prediction model, including the following:

[0064] S21. Create a training data set to train a time series prediction model. Process the 984 RMS data samples stored in S12. Use the RMS data of the first 20 time units as features and the RMS data of the next 2 time units as labels. Slide backward 2 time units each time, and get a total of 482 sets of data.

[0065] S22. The obtained 482 sets of data are sequentially put into the long short-term memory network LSTM to train the model and obtain a time series prediction model.

[0066] S33. The trained model is verified on another bearing data set of the same model. The curve between the predicted value and the actual value is as follows: Figure 2 As shown in the figure, the dotted line is the predicted value and the solid line is the actual value. Although there is a gap between the predicted value and the actual value, the overall trend of the RMS curve of the bearing can be reflected.

[0067] Step 3: Predict the RMS value of the roller bearing in use, including the following:

[0068] S31. Use an inspection robot equipped with an acceleration sensor to regularly sample the vibration signal of each belt conveyor roller in use, perform VMD decomposition on the collected vibration signal, select the component with the largest kurtosis value, and store this component.

[0069] S32: extract the latest 20 time unit data samples collected, and calculate the effective value RMS of each sample.

[0070] S33. The effective values ​​of the 20 time unit samples are put into the time series prediction model to perform time series prediction, and the effective value RMS of the next 2 time units from the current time can be obtained.

[0071] Step 4: Establish a mathematical model of the differential equations between the defect size and the vibration signal, and use the Runge-Kutta method to obtain the numerical solution of the vibration signal under different defect sizes, which specifically includes the following contents:

[0072] S41. The collision between the ball and the outer ring is equivalent to the model of the damper and the spring. The Hertz contact theory is used to establish the differential equations between the virtual defect size of the roller bearing and the simulated vibration signal. In the formula, F X ,F Y are the radial forces in the X and Y directions respectively, M is the total mass of the inner ring and the shaft, C is the damping coefficient, K is the stiffness, λ is the effective contact area of ​​the rolling element, δ i is the contact deformation between the ith rolling element and the raceway, θ i is the current angle of the i-th rolling element. The expression is:

[0073]

[0074] S42, contact deformation δ between the i-th rolling element and the raceway in S41 i It can be expressed as the following formula, where x, y are vibration displacements, c is the radial clearance of the bearing, and H' is the displacement excitation function.

[0075] δ i =xcosθ i +ysinθ i -cH′

[0076] S43, the effective contact area λ and the contact area δ of the i-th rolling element in S41 i It can be expressed as:

[0077]

[0078] S44. The displacement excitation function H' in S42 can be expressed by the following formula, where H is the value of the virtual defect size, β0 is the starting angle of the virtual defect, and β q is the angle when the rolling element completely falls into the defect, β is the angle spanned by the virtual defect size, and the expression is as follows:

[0079]

[0080] S45. Set the virtual defect size Hs from 0 to 3 mm, with a step length of 0.2 mm for each step, and use the Runge-Kutta method to solve the above differential equations to obtain a set of numerical solutions of x and y corresponding to each virtual defect size. Each set of solutions is the simulated vibration signal corresponding to each virtual defect size.

[0081] S46, calculate the peak values ​​Ps of the above-mentioned groups of simulated vibration signals, each virtual defect size Hs is a key, and the corresponding peak value Ps of the simulated vibration signal is a value, and store them in the form of key-value pairs. The peak value Ps of the simulated vibration signal changes as follows: Figure 3 shown.

[0082] Step 5: Based on the peak value approach principle, use the virtual defect size to estimate the real defect size, including the following:

[0083] S51, the change of the peak value Pt of the actual vibration signal is as follows Figure 4 As shown, taking the peak value Pt of each real vibration signal sample of the roller bearing in the whole life cycle calculated in step S12 as a reference, traverse the key-value pairs consisting of each virtual defect size Hs and the simulated vibration signal Ps stored in step S46, and find the simulated vibration signal peak value Ps closest to each real peak value Pt, then the virtual size corresponding to Ps is considered to be the real defect size Ht corresponding to the current real peak value Pt, thereby realizing the estimation of the real defect size.

[0084] S52, after all the real vibration signals in step S12 are matched with the real defect size Ht using the method of S51, the mapping relationship between the effective value RMS of all real vibration signal samples and the real defect size Ht is obtained, and the estimated value of the real defect size of each sample is as follows: Figure 5 shown.

[0085] Step 6: Use the previously trained time series prediction model and defect size estimation method to predict the defect size of the roller bearing in use, including the following:

[0086] S61. Take the effective value RMS of each sample as input and its corresponding true defect size Ht as output, train the fully connected neural network, and obtain a model that can reflect the mapping relationship between the effective value RMS and the defect size Ht.

[0087] S62. In step S33, the effective value RMS of the running roller bearing in the next two time units has been obtained. By inputting these two effective value RMS into the model trained in S61, an estimate of the actual defect size in the next two time units can be obtained, thereby realizing the prediction of the evolution of the roller bearing defects and constructing a mechanism model of the evolution of the roller bearing defect size.

[0088] This method uses a time series prediction model and a numerical simulation method to solve the problem that the digital twin mechanism model of the belt conveyor roller bearing cannot be modeled because the actual defect size of the belt conveyor roller bearing is difficult to measure and the degradation law is nonlinear. This provides a feasible technical path for establishing a mechanism model of the evolution of the defect size of the roller bearing.

[0089] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A belt conveyor roller bearing digital twin mechanism modeling method, characterized in that: The following steps are involved: Step 1: regularly sample the real vibration signal samples of the roller bearing in the whole life cycle, perform mathematical processing on each collected real vibration signal sample, and calculate the effective value RMS and peak value Pt of each sample; Step 2: Design an LSTM time series prediction network, use the effective value RMS of the collected real vibration data for model training, obtain a time series prediction model, and verify the effectiveness of the model; Step 3: Collect the vibration signal of the roller bearing in use and calculate its effective value RMS, and input the calculated effective value RMS into the model trained in step 2 to predict the RMS in the future; Step 4: Using Hertz theory, a mathematical model of a differential equation group of virtual defect size and simulated vibration signal of the roller bearing is established, and the simulated vibration signal corresponding to different virtual defect sizes Hs is solved by the Runge-Kutta method, and the peak value Ps of the simulated vibration signal is calculated; Step 5: Taking the peak value Pt of the actual vibration signal as a reference, traverse the peak values ​​Ps of the simulated vibration signals corresponding to all virtual sizes H. If they are within the allowable error range, it is considered that the real defect size Ht corresponding to the actual vibration signal sample is the virtual defect size Hs, thereby obtaining the mapping relationship between the effective value RMS of the sample and the real defect size Ht, and realizing the estimation of the real defect size. Step 6: Use the RMS obtained in step 5 as a feature and the actual defect size Ht as a label to train a fully connected neural network to obtain a model containing the mapping relationship between RMS and Ht. Use the RMS predicted in step 3 as input to the trained model to obtain the predicted defect size Ht for a period of time in the future, thereby establishing a life degradation model for roller bearings.

2. A belt conveyor roller bearing digital twin mechanism modeling method as claimed in claim 1, characterized in that: Step 1 includes the following steps: S11. Use the acceleration sensor to sample the real vibration data of the roller bearing throughout its life cycle. The sampling duration of each sample is 1 second. The collected data is stored in a folder. Each file name is the sampling time. A total of 984 sample files are saved until the bearing fails. S12. Calculate the effective value RMS and peak value Pt of each sample and store them in a .jason format file. The effective value calculation formula is: k is the number of discrete signals contained in each sample, y i is the value of a discrete signal.

3. A belt conveyor roller bearing digital twin mechanism modeling method as claimed in claim 2, characterized in that: In step 2, the construction method of the time series prediction model is as follows: S21, prepare a training data set to train a time series prediction model, process the 984 RMS data samples stored in step S12, use the RMS data of the first 20 time units as features, and the RMS data of the next 2 time units as labels, slide backward 2 time units each time, and obtain a total of 482 sets of data; S22. The obtained 482 sets of data are sequentially put into the long short-term memory network LSTM to train the model and obtain a time series prediction model.

4. The method for modeling a belt conveyor roller bearing digital twin mechanism model according to claim 1, characterized in that: Step 3 specifically includes the following: S31, using an inspection robot equipped with an acceleration sensor to regularly sample the vibration signal of each belt conveyor roller in use, performing VMD decomposition on the collected vibration signal, selecting a component with the largest kurtosis value, and storing the component; S32, extract the latest 20 time unit data samples collected, and calculate the effective value RMS of each sample; S33. Put the effective values ​​of the 20 time unit samples into the time series prediction model to perform time series prediction, and obtain the effective value RMS of 2 time units after the current time.

5. The method for modeling a belt conveyor roller bearing digital twin mechanism model according to claim 1, characterized in that: Step 4 specifically includes the following: S41. The collision between the ball and the outer ring is equivalent to the model of the damper and the spring. The Hertz contact theory is used to establish the differential equations between the virtual defect size of the roller bearing and the simulated vibration signal. The expression is: Where F X ,F Y are the radial forces in the X and Y directions respectively, M is the total mass of the inner ring and the shaft, C is the damping coefficient, K is the stiffness, λ is the effective contact area of ​​the rolling element, δ i is the contact deformation between the ith rolling element and the raceway, θ i is the current angle of the i-th rolling body; S42: contact deformation δ between the i-th rolling element and the raceway in step S41 i It is expressed as the following formula: d i =xcosθ i +ysinθ i -cH′ Where x, y are vibration displacements, c is the radial clearance of the bearing, and H' is the displacement excitation function; S43: The effective contact area λ and the contact area δ of the i-th rolling element in step S41 i Related, expressed as: S44. The displacement excitation function H' in step S42 is expressed by the following formula: Where H is the value of the virtual defect size, β0 is the starting angle of the virtual defect, and β q is the angle when the rolling element completely falls into the defect, and β is the angle spanned by the virtual defect size; S45, setting the virtual defect size Hs from 0.1 mm to 3 mm, with a step length of 0.2 mm for each step, using the Runge-Kutta method to solve the above differential equations, and solving a set of numerical solutions of x and y corresponding to each virtual defect size, each set of solutions is a simulated vibration signal corresponding to each virtual defect size; S46. Calculate the peak values ​​Ps of the above-mentioned groups of simulated vibration signals, with each virtual defect size Hs as a key and the corresponding peak value Ps of the simulated vibration signal as a value, and store them in the form of key-value pairs.

6. A belt conveyor roller bearing digital twin mechanism modeling method as claimed in claim 1, characterized in that: Step 5 specifically includes the following: S51, taking the peak value Pt of each real vibration signal sample of the roller bearing in the whole life cycle calculated in step S12 as a reference, traverse the key-value pairs consisting of each virtual defect size Hs and the simulated vibration signal Ps stored in step S46, and find the simulated vibration signal peak value Ps closest to each real peak value Pt, then the virtual size corresponding to Ps is considered to be the real defect size Ht corresponding to the current real peak value Pt, thereby realizing the estimation of the real defect size; S52, after matching all the real vibration signals in step S12 with the real defect size Ht using the method of S51, a mapping relationship between the effective value RMS of all real vibration signal samples and the real defect size Ht is obtained.

7. The method for modeling a belt conveyor roller bearing digital twin mechanism model according to claim 1, characterized in that: Step 6 specifically includes the following: S61, taking the effective value RMS of each sample as input and its corresponding real defect size Ht as output, training a fully connected neural network to obtain a model that can reflect the mapping relationship between the effective value RMS and the defect size Ht; S62. In step S33, the effective value RMS of the running roller bearing in the next two time units has been obtained. These two effective value RMS are input into the model trained in step S61 to obtain an estimate of the actual defect size in the next two time units, thereby realizing the prediction of the evolution of the roller bearing defects and constructing a mechanism model of the evolution of the roller bearing defect size.

Images