Bearing fault diagnosis method based on simulation sound vibration signal driving

Abstract

The invention discloses a bearing fault diagnosis method based on simulation sound vibration signal driving, and belongs to the technical field of mechanical equipment fault diagnosis. Bearing fault sample data are obtained in a simulation mode, fault sample experiments on large bearings are greatly reduced, and the bearing fault sample obtaining cost is greatly reduced; the signal-to-noise ratio of the signal is greatly improved under the condition that the characteristic frequency of the collected actual signal is reserved, the noise interference of the industrial environment is reduced, and the diagnosis accuracy is improved to a certain extent; a bearing sound signal and vibration signal comprehensive diagnosis method is adopted, diagnosis result abnormity caused by fluctuation of a single signal is avoided, the method has the advantages of sound signal fault diagnosis and vibration signal fault diagnosis, and the diagnosis accuracy is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical equipment fault diagnosis, and in particular to a bearing fault diagnosis method based on simulated acoustic vibration signal driving. Background Art

[0002] Bearings are core components of rotating machinery, and their failures can cause equipment downtime or even safety accidents. When diagnosing faults on large bearings, the cost of fault testing is high due to the high cost of bearings. Taking the QJ338 bearing of a mine main ventilation fan as an example, the cost of a single large bearing fault test exceeds 30,000 yuan, and fault samples are difficult to obtain.

[0003] Since vibration and acoustic signals in industrial sites are easily contaminated by environmental noise, traditional noise reduction methods such as wavelet threshold noise reduction can only improve the signal-to-noise ratio of bearing vibration signals by 4.2 to 5.8 dB, and the characteristic frequency retention rate is less than 85%.

[0004] The deep learning model trained based on experimental data has poor adaptability to simulation signals. The accuracy of the ResNet model trained based on CWRU bearing data dropped from 98.6% to 72.3% in industrial field data testing.

[0005] A single vibration or acoustic sensor is easily affected by changes in sensor installation position and operating conditions, with a misjudgment rate exceeding 12%.

[0006] The above problems need to be solved urgently. To this end, the present invention proposes a bearing fault diagnosis method based on simulated acoustic vibration signal drive. Summary of the Invention

[0007] The technical problem to be solved by the present invention is: to generate bearing acoustic vibration signals through high-precision simulation technology, solve the problem of insufficient fault samples, and at the same time greatly reduce the misjudgment rate, and provide a bearing fault diagnosis method driven by simulated acoustic vibration signals.

[0008] The present invention solves the above technical problems through the following technical solutions, which include the following steps:

[0009] S1: Construct a data set of bearing simulation vibration signals and acoustic signals, and perform data enhancement processing;

[0010] S2: Collect the actual vibration and acoustic signals of the experimental bearing and the target bearing to be diagnosed, pre-process the actual vibration and acoustic signals, and then perform denoising on them using the ICEEMDAN-PCA joint method.

[0011] S3: Use the simulated vibration signal and the simulated acoustic signal after data enhancement to train the vibration CNN diagnosis model and the acoustic CNN diagnosis model respectively;

[0012] S4: Transfer learning of the vibration CNN diagnostic model and the acoustic CNN diagnostic model using the actual vibration and acoustic signals of the experimental bearing after noise reduction;

[0013] S5: The actual vibration signal and acoustic signal of the target bearing to be diagnosed after noise reduction are input into the vibration CNN diagnosis model and acoustic CNN diagnosis model after transfer learning respectively to obtain the vibration diagnosis results and acoustic diagnosis results. The vibration diagnosis results and acoustic diagnosis results are then fused at the DS decision level to output the final fault type.

[0014] Furthermore, in step S1, the specific process of constructing the bearing simulation vibration signal data set is as follows:

[0015] S101: Establishing a three-dimensional model of a normal bearing and a faulty bearing with different fault types and fault sizes, wherein the fault types include inner ring fault, outer ring fault, and rolling element fault, and the fault sizes include 0.5 mm fault, 1 mm fault, and 2 mm fault;

[0016] S102: using transient dynamics simulation to obtain original simulated vibration signals of each simulated bearing model, wherein the simulated bearing speeds are set to 400 rpm, 600 rpm, and 800 rpm, respectively, and quantitatively comparing the peak frequency of the signal envelope spectrum with the theoretical fault characteristic frequency;

[0017] S103: performing time shift, amplitude scaling, and noise injection processing on the original simulated vibration signal to obtain hybrid enhanced data, and mixing the enhanced data with the original data.

[0018] Furthermore, in step S1, the specific process of constructing the bearing simulation acoustic signal dataset is as follows:

[0019] S111: Construct a bearing seat acoustic-solid coupling model, set acoustic field boundary conditions, and place a probe in the acoustic field to obtain simulated acoustic signals.

[0020] S112: After performing time-frequency conversion on the original simulated vibration signal of the bearing in step S102, the signal is input as an excitation into the bearing seat acoustic-solid coupling model to calculate the three-dimensional sound field distribution and obtain the corresponding original simulated acoustic signal;

[0021] S113: Verify the consistency between the peak frequency of the envelope spectrum of the original simulated acoustic signal and the theoretical fault characteristic frequency;

[0022] S114: performing time shift, amplitude scaling, and noise injection processing on the original simulated acoustic signal to obtain mixed enhanced data, and mixing the enhanced data with the original data.

[0023] Furthermore, in step S2, the actual vibration signal and acoustic signal collection process of the experimental bearing and the target bearing to be diagnosed is as follows: build a bearing test bench, install acoustic sensors and vibration sensors, select normal bearings and bearings with inner ring fault, outer ring fault, and rolling element fault with a fault size of 1mm as experimental bearings, set the sampling frequency to 32kHz, and collect the actual vibration signal and acoustic signal of the experimental bearing at a speed of 600r / min, select multiple groups of faulty bearings to be diagnosed that include inner ring fault, outer ring fault, rolling element fault and known fault types, set the sampling frequency to 32kHz, and collect the actual vibration signal and acoustic signal of the bearing to be diagnosed at different speeds.

[0024] Furthermore, in step S2, the actual vibration signal and the acoustic signal are respectively averaged and zero-averaged to eliminate sensor bias errors, and the signals are band-pass filtered to remove high-frequency noise and low-frequency components unrelated to bearing faults.

[0025] Furthermore, in step S2, the specific steps of ICEEMDAN-PCA joint denoising are as follows:

[0026] S21: Using ICEEMDAN to decompose the actual collected signal, adding Gaussian white noise and calculating the residual component, from which the IMF components of each order are extracted. The actual collected signal includes the actual vibration signal and the acoustic signal;

[0027] S22: Apply the principal component analysis method to sort the IMF components extracted by ICEEMDAN decomposition, and select the IMF components whose cumulative variance contribution rate is greater than the set value as the main characteristic components, and reconstruct the signal using the main characteristic components.

[0028] Furthermore, in step S3, both the vibration CNN diagnostic model and the acoustic CNN diagnostic model include:

[0029] Input layer: receives input vibration or acoustic signals;

[0030] Convolutional layer: extracts local signal features through multiple convolution kernels. The first convolution layer contains 32 3×3 convolution kernels with a step size of 1; the second convolution layer contains 64 3×3 convolution kernels with a step size of 1; the third convolution layer contains 128 3×3 convolution kernels with a step size of 1;

[0031] Batch normalization layer: normalizes the output of the convolutional layer;

[0032] Activation layer: Use ReLU function to perform nonlinear transformation on the normalized results;

[0033] Max pooling layer: The window is 3×3 and the stride is 2, which reduces the signal dimension and retains the corresponding features;

[0034] Fully connected layer: integrates the features output by the pooling layer;

[0035] Random dropout layer: randomly drops neurons according to a preset ratio after the fully connected layer to suppress overfitting;

[0036] Classification layer: outputs the probability distribution of each fault type;

[0037] Output layer: Determine the fault type label based on the probability distribution.

[0038] Furthermore, in step S4, the specific processing process is as follows:

[0039] S41: Freeze the convolutional layer, batch normalization layer, and pooling layer parameters of the vibration CNN diagnostic model and the acoustic CNN diagnostic model;

[0040] S42: Unfreeze and retrain the fully connected layer and classification layer, inputting the time-frequency feature map of the actual signal of the experimental bearing;

[0041] S43: A dynamic learning strategy is adopted. The initial learning rate is set to 1 / 10 of the training phase in step S3. The multi-task joint loss of the validation set is used as the decay basis. When the validation loss does not decrease for three consecutive epochs, the learning rate decays to 1 / 5 of the original value.

[0042] S44: Set an early stopping mechanism to terminate training when the verification accuracy fluctuates less than the set value for 10 consecutive epochs, thereby obtaining the vibration CNN diagnosis model and acoustic CNN diagnosis model after transfer learning.

[0043] Furthermore, in step S5, the specific process of DS decision-level fusion is as follows:

[0044] S51: Define the fault type identification framework. Assume that the space Θ includes four types of faults: normal, inner race fault, outer race fault, and rolling element fault, where N is normal, I is inner race fault, O is outer race fault, and R is rolling element fault.

[0045] For each fault type θ∈Θ, the trust assignment function of the vibration CNN diagnosis model is defined as:

[0046]

[0047] The trust assignment function of the acoustic CNN diagnosis model is defined as:

[0048]

[0049] in, are the unnormalized output values ​​of the vibration CNN diagnosis model and the acoustic CNN diagnosis model for fault type θ, ∑k∈Θ represents the summation of all fault types k in the spatial set Θ = {N, I, O, R};

[0050] S52: Set is the fusion hypothesis to be determined, They are the diagnosis result subsets of the vibration CNN diagnosis model and the acoustic CNN diagnosis model respectively. The diagnosis result subsets include the corresponding diagnosis results. The confidence level of the diagnosis result subset is defined as:

[0051]

[0052] The confidence levels of the two types of diagnosis results are fused through the DS combination rule, and the fault type with the highest fusion confidence level is determined as the final diagnosis result.

[0053] Furthermore, in step S52, the DS combination rule satisfies:

[0054]

[0055] w a =1-w v

[0056]

[0057] Among them, m fusion (A) is the fusion trust, w v is the vibration signal weight coefficient, w a is the acoustic signal weight coefficient, SNR is the signal-to-noise ratio, SNR v is the vibration signal-to-noise ratio, SNR a is the acoustic signal-to-noise ratio, P s is the target signal power, P n is the background noise power.

[0058] Compared with the prior art, the present invention has the following advantages:

[0059] 1. The simulation data has high fidelity, and the envelope spectrum characteristic frequency matching degree is >95%. The bearing fault sample data is obtained by simulation, which reduces the fault sample experiments of large bearings and greatly reduces the cost of obtaining bearing fault samples.

[0060] 2. While retaining the actual characteristic frequency of the collected signal, the signal-to-noise ratio is greatly improved, the noise interference of the industrial environment is reduced, and the diagnostic accuracy is improved to a certain extent.

[0061] 3. The comprehensive diagnosis method of bearing sound signal and vibration signal is adopted to avoid abnormal diagnosis results due to fluctuation of a single signal. It combines the advantages of sound signal fault diagnosis and vibration signal fault diagnosis, greatly improving the diagnosis accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 1 is a flow chart of a bearing fault diagnosis method based on simulated acoustic vibration signal driving in an embodiment of the present invention;

[0063] Figure 2 is a vibration signal simulation model in an embodiment of the present invention;

[0064] Figure 3 : is an envelope spectrum of a simulated vibration signal of an outer race fault at a speed of 600 r / min in an embodiment of the present invention;

[0065] Figure 4 is the acoustic signal simulation model in the embodiment of the present invention;

[0066] Figure 5 is the low-frequency portion of the outer race fault simulation acoustic signal at a speed of 600 r / min in the embodiment of the present invention;

[0067] Figure 6 This is a physical picture of a bearing test bench for collecting actual bearing signals in an embodiment of the present invention, where 1 is a vibration sensor, 2 is an acoustic sensor, 3 is a speed regulator, and 4 is an acquisition card;

[0068] Figure 7 : is the IMF component diagram of the actual outer race fault signal after ICEEMDAN decomposition in an embodiment of the present invention;

[0069] Figure 8 : This is a comparison diagram of the actual outer race fault signal before and after ICEEMDAN-PCA noise reduction in an embodiment of the present invention;

[0070] Figure 9 Schematic diagram of the structure of the CNN diagnostic model in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0072] like Figure 1As shown, this embodiment provides a technical solution: a bearing fault diagnosis method based on simulated acoustic vibration signal drive, using ICEEMDAN-PCA joint noise reduction technology to improve the signal-to-noise ratio and retain the fault characteristic frequency; designing a dynamic learning rate migration strategy to achieve rapid adaptation of the simulation model to the industrial scenario, the dynamic learning rate strategy increases the convergence speed of the model in mixed data training by 2.3 times (the number of iterations is reduced from 120 to 52), and the training time is shortened from 4.2 hours to 2.7 hours, a reduction of 37% in training time; integrating vibration and acoustic dual-modal diagnosis results, and reducing the misjudgment rate based on the DS evidence theory weighted by the signal-to-noise ratio; specifically comprising the following steps:

[0073] Step 1: Taking the 6012 bearing as an example, build a bearing transient dynamics model to obtain the bearing simulation vibration signal; input the obtained simulation vibration signal into the bearing seat acoustic-solid coupling model to obtain the bearing simulation acoustic signal.

[0074] Step 1.1: Build a bearing transient dynamics model to obtain the bearing simulation vibration signal, including:

[0075] Build three-dimensional models of normal bearings, inner ring faults, outer ring faults, and rolling element faults of 6012 bearings. The fault sizes include 0.5mm fault, 1mm fault, and 2mm fault.

[0076] like Figure 2 As shown in the figure, an ANSYS transient dynamics model was constructed for a 3D bearing model with different fault types and sizes. The mesh was created, the sampling frequency was set to 32 kHz, and the speed gradients were set to 400, 600, and 800 rpm, covering the typical operating range of the bearing. The simulation lasted 0.5 seconds, and simulated vibration signals were obtained under different fault conditions.

[0077] like Figure 3 As shown in the figure, the envelope spectrum of the simulated vibration signal for the outer ring fault at a speed of 600r / min is shown. The characteristic frequency of the outer ring fault is calculated according to the characteristic frequency calculation formula of the outer ring fault:

[0078]

[0079] Where n is the number of rolling elements, d is the diameter of the rolling element, and D m is the pitch diameter, α is the contact angle, f r is the speed frequency.

[0080] Substitute the 6012 bearing parameters n = 14, d = 10mm, D m =77.5mm,α=0°,f r=10Hz, the characteristic frequency of the 6012 bearing outer ring fault is calculated to be about 60.97Hz. The characteristic frequency of the bearing outer ring fault and its multiples can be clearly found on the envelope spectrum of the outer ring fault, verifying the correctness of the simulation signal.

[0081] Step 1.2: Input the obtained simulated vibration signal into the bearing seat acoustic-solid coupling model to obtain the bearing simulated acoustic signal, specifically including:

[0082] like Figure 4 As shown in the figure, a COMSOL bearing seat acoustic-solid coupling model is built, and a perfect matching layer is set outside the sound field as an ideal sound wave absorber to isolate the sound wave reflection interference.

[0083] Since the excitation of the COMSOL acoustic-solid coupling model is a frequency domain signal, the simulated vibration signal is subjected to FFT transformation.

[0084] The simulated vibration acceleration signal after FFT transformation is multiplied by the equivalent mass of the bearing to simulate the excitation force. The force is applied to the bearing seat for simulation, and a probe is set in the sound field to obtain the simulated acoustic signal.

[0085] like Figure 5 As shown in the figure, the outer race fault simulation acoustic signal obtained at a speed of 600 r / min can also clearly find the characteristic frequency of the outer race fault and its multiples, verifying the correctness of the simulation signal.

[0086] Step 2: Collect the actual vibration and acoustic signals of the experimental bearing and the target bearing to be diagnosed, including:

[0087] like Figure 6 As shown, a bearing test bench was built, and acoustic sensors and vibration sensors were installed. Normal bearings and bearings with inner ring fault, outer ring fault, and rolling element fault with a fault size of 1mm were selected as experimental bearings. The sampling frequency was set to 32kHz, and the actual vibration signals and acoustic signals of the experimental bearings were collected at a speed of 600r / min. Multiple groups of faulty bearings to be diagnosed, including inner ring fault, outer ring fault, rolling element fault and with known fault types, were selected. The sampling frequency was set to 32kHz, and the actual vibration signals and acoustic signals of the bearings to be diagnosed were collected at different speeds.

[0088] Step 3: Preprocess the collected signals by calculating their means and performing zero-mean processing to eliminate sensor bias errors. Bandpass filter the signals to remove high-frequency noise and low-frequency components unrelated to bearing failures. The preprocessed actual signals undergo a combined ICEEMDAN-PCA noise reduction process, specifically including:

[0089] like Figure 7As shown in FIG, the actual signal after preprocessing is decomposed by ICEEMDAN, Gaussian white noise is added, the residual component is calculated, and the intrinsic mode function (IMF component) is extracted.

[0090] Table 1 Contribution rate of each IMF signal to the variance of the entire data matrix

[0091]

[0092]

[0093] Principal component analysis (PCA) was used to select IMF components with cumulative variance contributions exceeding 95% for signal reconstruction. When the number of principal components reached three, the cumulative variance contribution reached 96.41%, indicating that the first three principal components were highly representative of the decomposed sequence and that the data information was intact. Therefore, IMF1, IMF3, and IMF2 were selected to reconstruct the signal, ensuring that the signal was denoised while preserving as much complete data information as possible.

[0094] like Figure 8 As shown in the figure, a comparison diagram of the signal before and after noise reduction of the 6012 bearing outer ring fault signal is shown. The main features of the signal are retained while irrelevant noise is removed to achieve signal noise reduction.

[0095] Step 4: Build and train the vibration CNN diagnostic model and acoustic CNN diagnostic model for fault diagnosis. The specific contents include:

[0096] like Figure 9 As shown in the figure, a CNN diagnostic model with a suitable structure is built.

[0097] Data enhancement is performed on the simulated vibration signal and the simulated acoustic signal. The original simulated vibration signal is subjected to a time shift of ±100ms, an amplitude scaling of 0.8 to 1.2, and a noise injection of 6% to 18%, and the enhanced data is mixed with the original data.

[0098] The vibration CNN diagnostic model is trained with the simulated vibration signal after data enhancement, and the acoustic CNN diagnostic model is trained with the simulated acoustic signal after data enhancement.

[0099] The diagnostic model was trained using actual vibration and acoustic signals from experimental bearings. The signals were segmented into 1024 points / sample using a sliding window, with a batch size of 64. After training, the model's diagnostic accuracy for target bearing data increased from 92.4% to 98.2%. The details include:

[0100] Freeze the convolutional layer, batch normalization layer, and pooling layer parameters of the vibration CNN diagnosis model and the acoustic CNN diagnosis model.

[0101] Unfreeze and retrain the fully connected layer and classification layer, and input the time-frequency feature map of the actual signal of the experimental bearing.

[0102] A dynamic learning strategy is adopted, with the initial learning rate set to 1 / 10 of the pre-training stage with augmented data. The multi-task joint loss of the validation set is used as the decay basis. When the validation loss does not decrease for three consecutive epochs, the learning rate decays to 1 / 5 of the original value. The dynamic learning strategy increases the convergence speed of the model on industrial field data by 2.3 times and reduces the training time by 37% compared with the fixed learning rate strategy.

[0103] An early stopping mechanism is set to terminate training when the verification accuracy fluctuates less than ±0.3% for 10 consecutive epochs.

[0104] Step 5: Perform fault diagnosis on the noise-reduced target bearing signal. The specific contents include:

[0105] The denoised vibration signal of the target bearing to be diagnosed is input into the trained vibration CNN model for fault diagnosis.

[0106] The denoised acoustic signal of the target bearing to be diagnosed is input into the trained acoustic CNN model for fault diagnosis.

[0107] The diagnosis results of the vibration CNN model and the acoustic CNN model are fused at the DS decision level to obtain the final diagnosis result. The specific contents of the DS decision level fusion include:

[0108] S51: Define the fault type identification framework. Assume that the space Θ includes four types of faults: normal, inner race fault, outer race fault, and rolling element fault, where N is normal, I is inner race fault, O is outer race fault, and R is rolling element fault.

[0109] For each fault type θ∈Θ, the trust assignment function of the vibration CNN diagnosis model is defined as:

[0110]

[0111] The trust assignment function of the acoustic CNN diagnosis model is defined as:

[0112]

[0113] in, are the unnormalized output values ​​of the vibration CNN diagnosis model and the acoustic CNN diagnosis model for fault type θ, ∑ k∈Θ The sum of all fault types k in the space set Θ = {N, I, O, R} is shown;

[0114] It should be noted that in the above formula, θ and k are both elements from the fault type set Θ = {N, I, O, R}, but their meanings are slightly different. θ represents the target fault type for which the trustworthiness is currently being calculated. k represents the intermediate variable used to traverse all fault types during the normalization process. To facilitate understanding, the following example is provided:

[0115] Assume that the unnormalized output values ​​of the vibration CNN diagnosis model for the four types of faults are: At this time, if the trust distribution value m corresponding to the outer ring fault O is required v (O), then let θ = O, that is, we require the trust of class O, and the sum variable k traverses N, I, O, and R in turn. Substituting into the formula, we get:

[0116]

[0117] Thus, θ=O indicates that the trustworthiness of the outer ring fault is currently being sought, and k∈Θ indicates that all fault types are summed up and normalized in the denominator.

[0118] S52: Set is the fusion hypothesis to be determined, They are the diagnosis result subsets of the vibration CNN diagnosis model and the acoustic CNN diagnosis model respectively. The diagnosis result subsets include the corresponding diagnosis results. The confidence level of the diagnosis result subset is defined as:

[0119]

[0120] The confidence levels of the two types of diagnosis results are fused through the DS combination rule, and the fault type with the highest fusion confidence level is determined as the final diagnosis result.

[0121] Among them, the DS combination rule satisfies:

[0122]

[0123] w a =1-w v

[0124]

[0125] Among them, m fusion (A) is the fusion trust, w v is the vibration signal weight coefficient, w a is the acoustic signal weight coefficient, SNR is the signal-to-noise ratio, SNR v is the vibration signal-to-noise ratio, SNR a is the acoustic signal-to-noise ratio, P s is the target signal power, P n is the background noise power.

[0126] Taking the outer ring fault experiment bearing signal fault diagnosis as an example, the vibration signal noise ratio SNR v =18dB, acoustic signal-to-noise ratio SNR a =12dB, weight calculation is w v =0.60, w a =0.40, and the comprehensive confidence of the outer ring fault after fusion is 0.88. The fault diagnosis accuracy of the target bearing data is 92.5% for the single vibration signal and 85.3% for the acoustic signal. The comprehensive accuracy after fusion reaches 96.8%, and the misjudgment rate is reduced from 7.5% for the single vibration signal to 3.2%.

[0127] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A bearing fault diagnosis method based on simulated acoustic vibration signal drive, characterized in that: The following steps are involved: S1: Construct a data set of bearing simulation vibration signals and acoustic signals, and perform data enhancement processing; S2: Collect the actual vibration and acoustic signals of the experimental bearing and the target bearing to be diagnosed, pre-process the actual vibration and acoustic signals, and then perform denoising on them using the ICEEMDAN-PCA joint method. S3: Use the simulated vibration signal and the simulated acoustic signal after data enhancement to train the vibration CNN diagnosis model and the acoustic CNN diagnosis model respectively; S4: Transfer learning of the vibration CNN diagnostic model and the acoustic CNN diagnostic model using the actual vibration and acoustic signals of the experimental bearing after noise reduction; S5: The actual vibration signal and acoustic signal of the target bearing to be diagnosed after noise reduction are input into the vibration CNN diagnosis model and acoustic CNN diagnosis model after transfer learning respectively to obtain the vibration diagnosis results and acoustic diagnosis results. The vibration diagnosis results and acoustic diagnosis results are then fused at the DS decision level to output the final fault type.

2. A bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 1, characterized in that: In step S1, the specific process of constructing the bearing simulation vibration signal data set is as follows: S101: Establishing a three-dimensional model of a normal bearing and a faulty bearing with different fault types and fault sizes, wherein the fault types include inner ring fault, outer ring fault, and rolling element fault, and the fault sizes include 0.5 mm fault, 1 mm fault, and 2 mm fault; S102: using transient dynamics simulation to obtain original simulated vibration signals of each simulated bearing model, wherein the simulated bearing speeds are set to 400 rpm, 600 rpm, and 800 rpm, respectively, and quantitatively comparing the peak frequency of the signal envelope spectrum with the theoretical fault characteristic frequency; S103: performing time shift, amplitude scaling, and noise injection processing on the original simulated vibration signal to obtain hybrid enhanced data, and mixing the enhanced data with the original data.

3. The bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 2 is characterized in that: In step S1, the specific process of constructing the bearing simulation acoustic signal data set is as follows: S111: Construct a bearing seat acoustic-solid coupling model, set acoustic field boundary conditions, and place a probe in the acoustic field to obtain simulated acoustic signals. S112: After performing time-frequency conversion on the original simulated vibration signal of the bearing in step S102, the signal is input as an excitation into the bearing seat acoustic-solid coupling model to calculate the three-dimensional sound field distribution and obtain the corresponding original simulated acoustic signal; S113: Verify the consistency between the peak frequency of the envelope spectrum of the original simulated acoustic signal and the theoretical fault characteristic frequency; S114: performing time shift, amplitude scaling, and noise injection processing on the original simulated acoustic signal to obtain mixed enhanced data, and mixing the enhanced data with the original data.

4. The bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 1 is characterized in that: In step S2, the actual vibration signal and acoustic signal collection process of the experimental bearing and the target bearing to be diagnosed is as follows: build a bearing test bench, install acoustic sensors and vibration sensors, select normal bearings and bearings with inner ring fault, outer ring fault, and rolling element fault with a fault size of 1mm as experimental bearings, set the sampling frequency to 32kHz, and collect the actual vibration signal and acoustic signal of the experimental bearing at a speed of 600r / min, select multiple groups of faulty bearings to be diagnosed that include inner ring fault, outer ring fault, rolling element fault and known fault types, set the sampling frequency to 32kHz, and collect the actual vibration signal and acoustic signal of the bearing to be diagnosed at different speeds.

5. The bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 1 is characterized in that: In step S2, the actual vibration signal and the acoustic signal are respectively averaged and zero-averaged to eliminate sensor bias errors, and the signals are band-pass filtered to remove high-frequency noise and low-frequency components unrelated to bearing faults.

6. The bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 1 is characterized in that: In step S2, the specific steps of ICEEMDAN-PCA joint denoising are as follows: S21: Using ICEEMDAN to decompose the actual collected signal, adding Gaussian white noise and calculating the residual component, from which the IMF components of each order are extracted. The actual collected signal includes the actual vibration signal and the acoustic signal; S22: Apply the principal component analysis method to sort the IMF components extracted by ICEEMDAN decomposition, and select the IMF components whose cumulative variance contribution rate is greater than the set value as the main characteristic components, and reconstruct the signal using the main characteristic components.

7. The bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 6, characterized in that: In step S3, both the vibration CNN diagnostic model and the acoustic CNN diagnostic model include: Input layer: receives input vibration or acoustic signals; Convolutional layer: extracts local signal features through multiple convolution kernels. The first convolution layer contains 32 3×3 convolution kernels with a step size of 1; the second convolution layer contains 64 3×3 convolution kernels with a step size of 1; the third convolution layer contains 128 3×3 convolution kernels with a step size of 1; Batch normalization layer: normalizes the output of the convolutional layer; Activation layer: Use ReLU function to perform nonlinear transformation on the normalized results; Max pooling layer: The window is 3×3 and the stride is 2, which reduces the signal dimension and retains the corresponding features; Fully connected layer: integrates the features output by the pooling layer; Random dropout layer: randomly drops neurons according to a preset ratio after the fully connected layer to suppress overfitting; Classification layer: outputs the probability distribution of each fault type; Output layer: Determine the fault type label based on the probability distribution.

8. The bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 7 is characterized in that: In step S4, the specific processing process is as follows: S41: Freeze the convolutional layer, batch normalization layer, and pooling layer parameters of the vibration CNN diagnostic model and the acoustic CNN diagnostic model; S42: Unfreeze and retrain the fully connected layer and classification layer, inputting the time-frequency feature map of the actual signal of the experimental bearing; S43: A dynamic learning strategy is adopted. The initial learning rate is set to 1 / 10 of the training phase in step S3. The multi-task joint loss of the validation set is used as the decay basis. When the validation loss does not decrease for three consecutive epochs, the learning rate decays to 1 / 5 of the original value. S44: Set an early stopping mechanism to terminate training when the verification accuracy fluctuates less than the set value for 10 consecutive epochs, thereby obtaining the vibration CNN diagnosis model and acoustic CNN diagnosis model after transfer learning.

9. The bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 8, characterized in that: In step S5, the specific process of DS decision-level fusion is as follows: S51: Define the fault type identification framework. Assume that the space Θ includes four types of faults: normal, inner race fault, outer race fault, and rolling element fault, where N is normal, I is inner race fault, O is outer race fault, and R is rolling element fault. For each fault type θ∈Θ, the trust assignment function of the vibration CNN diagnosis model is defined as: The trust assignment function of the acoustic CNN diagnosis model is defined as: in, are the unnormalized output values ​​of the vibration CNN diagnosis model and the acoustic CNN diagnosis model for fault type θ, ∑ k∈Θ represents the summation of all fault types k in the spatial set Θ = {N, I, O, R}; S52: Set is the fusion hypothesis to be determined, They are the diagnosis result subsets of the vibration CNN diagnosis model and the acoustic CNN diagnosis model respectively. The diagnosis result subsets include the corresponding diagnosis results. The confidence level of the diagnosis result subset is defined as: The confidence levels of the two types of diagnosis results are fused through the DS combination rule, and the fault type with the highest fusion confidence level is determined as the final diagnosis result.

10. The bearing fault diagnosis method based on simulated acoustic vibration signal drive according to claim 9, characterized in that: In step S52, the DS combination rule satisfies: Among them, m fusion (A) is the fusion trust, w v is the vibration signal weight coefficient, w a is the acoustic signal weight coefficient, SNR is the signal-to-noise ratio, SNR v is the vibration signal-to-noise ratio, SNR a is the acoustic signal-to-noise ratio, P s is the target signal power, P n is the background noise power.

Images