Health monitoring method for planetary gearbox based on modal reconstruction and bagging model

Through variational modal decomposition and Bagging Tree integrated algorithm, the problems of noise interference and nonlinear excitation information in planetary gearbox fault diagnosis are solved, and efficient fault identification and health monitoring are achieved under different working conditions.

CN114861716BActive Publication Date: 2024-11-12SUZHOU VEIZU EQUIPMENT DIAGNOSIS TECHNOLOGY CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210453399.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-27
Publication Date
2024-11-12
Estimated Expiration
2042-04-27

AI Technical Summary

Technical Problem

The prior art is difficult to effectively deal with the nonlinear excitation information caused by complex internal structures and noise interference under harsh service conditions when diagnosing planetary gearbox failures, resulting in insufficient generalization capabilities of diagnostic models.

Method used

A three-phase acceleration sensor is used to collect vibration signals, reduce noise by variational modal decomposition (VMD) and extract eigenmodal components, perform frequency variance reconstruction and energy index extraction, and build a fault diagnosis model with Bagging Tree integrated algorithm.

Benefits of technology

It effectively suppresses noise interference, improves diagnostic efficiency and accuracy, and realizes planetary gearbox health monitoring and fault identification under different service conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114861716B_ABST
    Figure CN114861716B_ABST
Patent Text Reader

Abstract

A planetary gearbox fault diagnosis method based on modal reconstruction and bagging model, firstly, a three-phase acceleration sensor is used to collect the vibration signals of the planetary gearbox in three directions of x, y and z; then, the vibration signal is processed by variational mode decomposition (VMD) to obtain the decomposed intrinsic mode components, so as to weaken the interference of noise on the vibration signal; then, the intrinsic mode components of each order are reconstructed by the frequency variance, and the energy index is extracted; finally, the model is trained by the integrated algorithm of Bagging Tree to obtain an effective model for identifying the state of the planetary gearbox; the present invention can realize the health monitoring of the planetary gearbox under different service conditions.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of planetary gearbox fault diagnosis, and in particular relates to a planetary gearbox fault diagnosis method based on modal reconstruction and bagging model. Background Art

[0002] In recent years, the rapid development of artificial intelligence theory has provided a powerful tool for the intelligent diagnosis of mechanical equipment, and has achieved certain results in the field of planetary gearbox fault diagnosis. Some people have applied EMD to extract bearing fault features and combined SVM's good classification performance for pattern recognition to achieve fault diagnosis (Fu Dapeng, Zhai Yong, Yu Qingmin. Research on rolling bearing fault diagnosis based on EMD and support vector machine [J]. Machine Tool and Hydraulics, 2017, 45(11): 184-187); some people use the original vibration signal as input and automatically extract fault features from it by combining multi-layer nonlinear neural networks to identify fault states (Zhao Lu, Ma Ye. Research on gearbox fault diagnosis based on one-dimensional convolutional neural network [J]. Journal of Testing Technology, 2019, 33(4): 302-306); some people proposed a gearbox fault diagnosis method based on minimum entropy deconvolution and adaptive variational mode decomposition to improve the generalization ability of traditional variational mode decomposition in planetary gearbox fault diagnosis (Zhu Jing, Deng Aidong, Deng Minqiang, et al. Planetary gearbox fault diagnosis method based on MED and adaptive VMD [J]. Journal of Southeast University: Natural Science Edition, 2020, 50(4): 7).

[0003] Although there have been beneficial explorations on rotating parts such as gears and bearings, on the one hand, the complex internal structure of the planetary gearbox causes the excitation information to be coupled with each other, showing obvious nonlinearity, which brings challenges to the interpretation and analysis of dynamic information; on the other hand, the harsh service conditions cause the dynamic information of the planetary gearbox to be seriously disturbed by noise, showing different statistical characteristics, and requiring higher generalization ability of the diagnostic model. Therefore, how to fully explore the damage intrinsic information contained in multiple interference modes under different service conditions is an urgent problem to be solved. Summary of the invention

[0004] In order to overcome the above-mentioned shortcomings of the prior art, the object of the present invention is to provide a planetary gearbox fault diagnosis method based on modal reconstruction and bagging model, which can realize health monitoring of the planetary gearbox under different service conditions.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A planetary gearbox fault diagnosis method based on modal reconstruction and bagging model is proposed. Firstly, the vibration signals of the planetary gearbox in the x, y and z directions are collected by a three-phase accelerometer. Then, the vibration signals are processed by variational mode decomposition (VMD) to obtain the decomposed intrinsic mode components to weaken the interference of noise on the vibration signals. Then, the eigenmode components of each order are reconstructed by their frequency variance, and energy indicators are extracted. Finally, the model is trained by the integrated algorithm of Bagging Tree to obtain an effective model for identifying the state of the planetary gearbox.

[0007] A planetary gearbox fault diagnosis method based on modal reconstruction and bagging model comprises the following steps:

[0008] Step 1: using a three-phase acceleration sensor arranged on a planetary gearbox to collect vibration signals {x, y, z} under different health states;

[0009] Step 2: Perform variational mode decomposition (VMD) on the collected vibration signal to obtain the intrinsic mode components s of each channel k (t);

[0010] s k (t) = a k (t)exp(iφ k (t)) (1)

[0011] where a k (t) is the non-negative signal envelope, φ k (t) is the non-decreasing phase, i is the imaginary unit;

[0012] Step 3: Calculate the variance VF of the intrinsic mode component frequency in each channel according to formula (2), and perform weighted reconstruction on the first five order components based on the size of the frequency variance to obtain the reconstructed signal;

[0013]

[0014] in is the centroid frequency of the signal, f is the frequency, and p is the amplitude corresponding to f;

[0015] Step 4: Calculate the energy E of each order of reconstructed signal according to formula (3), and obtain the normalized feature of each channel by dividing the energy of each order by the energy sum;

[0016]

[0017] Where x i Reconstruct the signal amplitude, n is the signal length;

[0018] Step 5: construct a structured feature set based on the normalized feature values ​​of each channel, use the Bagging Tree integration algorithm to extract a subset with replacement from the structured feature set samples, perform decision tree classification learning and training, and synthesize multiple decision results to obtain a planetary gearbox fault diagnosis model;

[0019] Step six, using the diagnostic model obtained in step five to test the vibration signal collected on the planetary gearbox, so as to identify the health status of the planetary gearbox.

[0020] The specific method of step 2 is as follows:

[0021] The collected vibration signal is subjected to variational mode decomposition (VMD), as shown in formula (4):

[0022]

[0023] Where x(t) is the decomposed vibration signal, K is the number of decomposition levels, {s k} is the kth order eigenmode component after decomposition, {ω k} is the center frequency of the k-th order eigenmode component after decomposition, i is the imaginary unit, δ(t) is the Dirac function, and * is the convolution operator.

[0024] The specific method of step three is as follows:

[0025] The eigenmode components are Fourier transformed and mapped to the frequency domain space, as shown in formula (5),

[0026]

[0027] where s k (t) is the intrinsic mode component, F(n) is the corresponding frequency domain signal, and i is the imaginary unit. The frequency domain signal is matched with the actual frequency using the sampling frequency Fs, as shown in formula (6), and the frequency sequence f of the signal and the corresponding frequency amplitude p are obtained;

[0028]

[0029] The frequency variance VF of each intrinsic mode component is calculated according to formula (2), and the first five order components are reconstructed based on the magnitude of the frequency variance.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] a) The present invention is driven by data information and realizes noise reduction through variational mode decomposition (VMD). At the same time, it proposes weighted reconstruction of each order eigenmode by component frequency variance to enhance fault characteristics. The results show that the present invention not only effectively suppresses random shocks and strong noise interference, but also improves diagnostic efficiency and accuracy.

[0032] b) Different from the traditional method of diagnosing planetary gearboxes based only on single-channel data under stable working conditions, the present invention uses the energy of each order reconstructed signal to construct a structured feature set, and fuses the signal features of each channel through the Bagging Tree integration algorithm, thereby realizing the health monitoring of planetary gearboxes under complex service conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Schematic diagram of a planetary gear fault diagnosis test bench according to an embodiment.

[0034] Figure 2 Schematic diagram of the internal structure of the planetary gearbox in the embodiment.

[0035] Figure 3 Schematic diagram of the planetary gear and its surface processing to produce pitting and tooth surface wear in the embodiment.

[0036] Figure 4 Flow chart of the method of the present invention.

[0037] Figure 5 The vibration signals of the planetary gearbox at different speeds are collected for the embodiment.

[0038] Figure 6 5th order components and remainders of x, y, and z channels in normal conditions of the embodiment.

[0039] Figure 7 Schematic diagram of the fifth-order components and remainders of x, y, and z channels under tooth surface pitting fault.

[0040] Figure 8 5th order components and remainders of x, y, and z channels under tooth surface wear fault of the embodiment.

[0041] Fig. 9 The confusion matrix diagram of the test results of the training set with 10% samples extracted for the embodiment is shown in FIG.

[0042] Fig.10 The following are test results of the method of the embodiment at different rotation speeds.

[0043] Fig.11 The figure is the result of testing the variable speed signal in the embodiment. DETAILED DESCRIPTION

[0044] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0045] The following will take the planetary gear fault diagnosis test bench as an example. The test bench consists of a drive motor, a coupling, a planetary gear box and a magnetic powder brake. The vibration sensor is installed on the planetary gear box. Figure 1The planetary gearbox consists of a sun gear, an inner ring gear and three evenly distributed planetary gears, and the planet carrier is connected to the output shaft, as shown in Figure 2 In order to simulate the early failure of the gearbox, pitting and tooth surface wear are machined on the surface of the planetary gear, as shown in Figure 3 The test bench is driven by a drive motor, which transmits the torque from the input shaft along the planetary gearbox to the magnetic powder brake, which completes the load.

[0046] The specific parameters are as follows: 1) Rated power of driving motor: 1.2kW; 2) Transmission ratio of planetary gearbox: 5.1:1, number of teeth of inner ring gear: 82, module: 1, number of teeth of planetary gear: 31, module: 1, number of teeth of sun gear: 20, module: 1; 3) Torque of magnetic powder brake at rated power: 0.06N*m.

[0047] like Figure 4 As shown, a planetary gearbox health monitoring method based on modal reconstruction and bagging model includes the following steps:

[0048] Step 1: Use the NI2 board to collect the vibration signals {x, y, z} of the planetary gearbox at different speeds (including 1200rpm, 1800rpm, and 2400rpm) through a three-phase acceleration sensor. The sampling frequency is 5000Hz, such as Figure 5 As shown;

[0049] Step 2: intercept the collected vibration signal for 2 seconds, and mark the vibration signal according to the state of the planetary gear (normal, pitting, tooth surface wear);

[0050] like Figure 6-Figure 8 As shown, the vibration signal intercepted in each channel is subjected to variational mode decomposition to obtain the eigenmode component s of each channel k (t), the preprocessing of vibration signal is completed;

[0051] s k (t) = a k (t)exp(iφ k (t)) (1)

[0052] where a k (t) is the non-negative signal envelope, φ k (t) is the non-decreasing phase, i is the imaginary unit;

[0053] In this embodiment, the collected vibration signal is subjected to variational mode decomposition (VMD), as shown in formula (4):

[0054]

[0055] Where x(t) is the decomposed vibration signal, K is the number of decomposition levels, {s k} is the kth order eigenmode component after decomposition, {ω k} is the center frequency of the kth order eigenmode component after decomposition, i is the imaginary unit, δ(t) is the Dirac function, and * is the convolution operator;

[0056] Step 3: Calculate the variance VF of the intrinsic mode component frequency in each channel according to formula (2). Based on the size of the frequency variance, perform weighted reconstruction on the first five order components to obtain the reconstructed signal. The reconstructed energy index is used as the eigenvector of the channel.

[0057]

[0058] in is the centroid frequency of the signal, f is the frequency, and p is the amplitude corresponding to f;

[0059] In this embodiment, the eigenmode component is Fourier transformed and mapped to the frequency domain space, as shown in formula (5),

[0060]

[0061] where s k (t) is the intrinsic mode component, F(n) is the frequency domain signal, and i is the imaginary unit. The frequency domain signal is matched with the actual frequency using the sampling frequency Fs, as shown in formula (6), and the frequency sequence f of the signal and the corresponding frequency amplitude p are obtained;

[0062]

[0063] The frequency variance VF of each intrinsic mode component is calculated according to formula (2), and based on the magnitude of the frequency variance, the first five order components are reconstructed to obtain the reconstructed signal;

[0064] Step 4: Calculate the energy E of each order of reconstructed signal according to formula (3), and obtain the normalized feature of each channel by dividing the energy of each order by the energy sum;

[0065]

[0066] Where x i Reconstruct the signal amplitude, n is the signal length;

[0067] Step 5: construct a structured feature set based on the normalized feature values ​​of each channel, use the Bagging Tree integration algorithm to extract a subset with replacement from the structured feature set samples, perform decision tree classification learning and training, and synthesize multiple decision results to obtain a planetary gearbox fault diagnosis model;

[0068] In this embodiment, 10% of the data in each state is extracted as a training set, and the remaining 90% is used as a test set. The Bagging Tree integrated algorithm is used for training, and each method is distinguished according to the naming method of "decomposing the intrinsic mode component method-the indicator based on which the signal is reconstructed-the characteristic indicator". For example, "VMD-VF-E" means "using variational mode decomposition (VMD)-reconstructing the signal based on frequency variance (VF)-extracting energy (E) normalized features". The test results are as follows: Fig. 9 As shown, it can be seen that the embodiment method accurately distinguishes the normal state and the fault state of the test data, and the overall accuracy rate also reaches 98.37%, which well realizes the fault diagnosis of the planetary gearbox;

[0069] Step six, using the diagnostic model obtained in step five to test the vibration signal collected on the planetary gearbox, so as to identify the health status of the planetary gearbox.

[0070] To verify the generalization ability of this embodiment under different working conditions, tests were conducted under the working conditions of 1200 rpm, 1800 rpm, 2400 rpm and variable speed, and 10% of the samples were randomly selected as the training set. The results are shown in Figure 2. Fig.10 As shown in Figure 3, the results show that the method still maintains a high accuracy under different working conditions.

[0071] In order to verify the generalization ability of this embodiment under complex service conditions, the signal under mixed conditions was tested. The results are as follows: Fig.11 As shown, it can be seen from the figure that the accuracy of the embodiment method is high when there are fewer training samples, and as the number of training set samples increases, the accuracy is stable at more than 98%. Therefore, this embodiment can effectively perform early fault diagnosis on planetary gearboxes under different speed working conditions.

Claims

1. A planetary gearbox fault diagnosis method based on modal reconstruction and bagging model, Features: Firstly, a three-phase acceleration sensor is used to collect the vibration signals of the planetary gearbox in the x, y and z directions. Then, the vibration signals are processed by variational mode decomposition (VMD) to obtain the decomposed intrinsic mode components to weaken the interference of noise on the vibration signals. Then, the intrinsic mode components of each order are reconstructed by their frequency variance and energy index is extracted. Finally, the model is trained by the integrated algorithm of Bagging Tree to obtain an effective model for identifying the state of the planetary gearbox. The planetary gearbox fault diagnosis method based on modal reconstruction and bagging model comprises the following steps: Step 1: using a three-phase acceleration sensor arranged on a planetary gearbox to collect vibration signals {x, y, z} under different health states; Step 2: Perform variational mode decomposition (VMD) on the collected vibration signal to obtain the intrinsic mode components s of each channel. k (t); s k (t)=a k (t)exp(iφ k (t)) (1) where a k (t) is the non-negative signal envelope, φ k (t) is the non-decreasing phase, i is the imaginary unit; Step 3: Calculate the variance VF of the intrinsic mode component frequency in each channel according to formula (2), and perform weighted reconstruction on the first five order components based on the size of the frequency variance to obtain the reconstructed signal; in is the centroid frequency of the signal, f is the frequency, and p is the amplitude corresponding to f; Step 4: Calculate the energy E of each order of reconstructed signal according to formula (3), and obtain the normalized feature of each channel by dividing the energy of each order by the energy sum; Where x i Reconstruct the signal amplitude, n is the signal length; Step 5: construct a structured feature set based on the normalized feature values ​​of each channel, use the Bagging Tree integration algorithm to extract a subset with replacement from the structured feature set samples, perform decision tree classification learning and training, and synthesize multiple decision results to obtain a planetary gearbox fault diagnosis model; Step 6: Using the diagnostic model obtained in step 5, the vibration signal collected from the planetary gearbox is tested to identify the health status of the planetary gearbox; The specific method of step 2 is as follows: The collected vibration signal is subjected to variational mode decomposition VMD, as shown in formula (4): Where x(t) is the decomposed vibration signal, K is the number of decomposition levels, {s k } is the kth order eigenmode component after decomposition, {ω k } is the center frequency of the kth order eigenmode component after decomposition, i is the imaginary unit, δ(t) is the Dirac function, and * is the convolution operator; The specific method of step three is as follows: The eigenmode components are Fourier transformed and mapped to the frequency domain space, as shown in formula (5), where s k (t) is the intrinsic mode component, F(n) is the corresponding frequency domain signal, and i is the imaginary unit. The frequency domain signal is matched with the actual frequency using the sampling frequency Fs, as shown in formula (6), and the frequency sequence f of the signal and the corresponding frequency amplitude p are obtained; The frequency variance VF of each intrinsic mode component is calculated according to formula (2), and the first five order components are reconstructed based on the magnitude of the frequency variance.