Bearing fault determination method and device, equipment, storage medium and product

Abstract

The invention provides a bearing fault determination method and device, equipment, a storage medium and a product, and the method comprises the steps: receiving vibration signals collected by a plurality of acceleration sensors; the vibration signal is a vibration amplitude signal related to the bearing; obtaining a target fusion signal based on the plurality of vibration signals; and determining that the bearing is in a fault state based on the target fusion signal. According to the scheme, a plurality of vibration signals can be received so as to realize fusion of data-level vibration signals, and determination of bearing fault accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of computers, and in particular to a bearing fault determination method, device, equipment, storage medium and product. Background Art

[0002] In industrial manufacturing scenarios, bearings are core components of rotating machinery, and their health status directly determines the safety and reliability of the entire rotating machinery. However, the working environment of rotating machinery is relatively complex and harsh, and its components are prone to failure. In order to avoid losses and even safety accidents caused by bearing failures, it is necessary to carry out fault diagnosis of bearings.

[0003] At present, acceleration sensors are commonly used to monitor the condition of bearings. Specifically, the health condition of the bearing is determined by obtaining the vibration signal collected by the acceleration sensor.

[0004] However, due to the limitations of the coverage range and accuracy of the acceleration sensor, a single acceleration sensor cannot fully reflect the health status of the bearing, resulting in low fault diagnosis accuracy. Summary of the invention

[0005] The present application provides a bearing fault determination method, device, equipment, storage medium and product to solve the problem of low fault diagnosis accuracy.

[0006] In a first aspect, the present application provides a bearing fault determination method, the method comprising:

[0007] Receiving vibration signals collected by multiple acceleration sensors; the vibration signals are vibration amplitude signals related to the bearing;

[0008] Obtaining a target fusion signal based on the plurality of vibration signals;

[0009] The bearing is determined to be in a fault state based on the target fusion signal.

[0010] In one embodiment, obtaining a target fusion signal based on the plurality of vibration signals includes:

[0011] Obtaining initial weights of a plurality of the vibration signals;

[0012] Inputting the plurality of initial weights into a fusion signal algorithm to obtain an initial fusion signal;

[0013] The target fusion signal is obtained based on the initial fusion signal.

[0014] In one embodiment, obtaining the target fusion signal based on the initial fusion signal includes:

[0015] Calculating a corresponding kurtosis based on the initial fusion signal;

[0016] Inputting the initial fusion signal into a weight algorithm to obtain initial weights of the fusion signal and multiple vibration signals;

[0017] A target fusion signal is obtained based on the kurtosis and the initial weight.

[0018] In one embodiment, the step of inputting the initial fusion signal into a weight algorithm to obtain initial weights of the fusion signal and the plurality of vibration signals includes:

[0019] Inputting the initial fusion signal, multiple vibration signals and the average value of the vibration signals into a correlation coefficient algorithm to obtain a correlation coefficient;

[0020] The correlation coefficient is input into a weight algorithm to obtain the initial weight.

[0021] In one embodiment, obtaining a target fusion signal based on the kurtosis and the initial weight includes:

[0022] Inputting the kurtosis and the initial weight into a correlation kurtosis algorithm to obtain an initial correlation kurtosis value of the initial fusion signal;

[0023] The target fusion signal is obtained by using the initial correlation kurtosis value.

[0024] In one embodiment, the step of obtaining a target fusion signal by using the initial correlation kurtosis value includes:

[0025] Taking the reciprocal of the initial correlation kurtosis value as the initial loss value;

[0026] Inputting the initial loss value into a weight optimization algorithm to obtain optimization weights corresponding to a plurality of vibration signals;

[0027] Within the number of iterations, the optimization weight corresponding to each iteration is continuously calculated based on the optimization weight and related algorithms; the related algorithms include fusion signal algorithm, weight algorithm, correlation coefficient algorithm, correlation kurtosis algorithm and weight optimization algorithm;

[0028] Determine the target weight from the optimization weights corresponding to all iterations;

[0029] The target weight is input into the fusion algorithm to obtain a target fusion signal.

[0030] In one embodiment, determining that the bearing is in a fault state based on the target fusion signal includes:

[0031] Performing continuous wavelet transform on the target fusion signal to obtain a wavelet time-frequency diagram;

[0032] Using a one-dimensional neural network to extract features from the target fusion signal to obtain a first feature;

[0033] Using a two-dimensional neural network to extract features from the wavelet time-frequency graph to obtain a second feature;

[0034] Fusing the first feature and the second feature to obtain a target fused feature;

[0035] The bearing is determined to be in a fault state based on the target fusion feature.

[0036] In one embodiment, determining that a bearing is in a fault state based on the target fusion feature includes:

[0037] Inputting the target fusion feature into a classification model that has been pre-trained to convergence to obtain an output result of at least one bearing;

[0038] In response to determining that the output result is a fault category, determining that the corresponding bearing is in a fault state;

[0039] The method further comprises:

[0040] In response to determining that the output result is of the normal category, the corresponding bearing is determined to be in a normal state.

[0041] In a second aspect, the present application provides a bearing fault determination device, the device comprising:

[0042] A receiving module, used to receive vibration signals collected by multiple acceleration sensors; the vibration signals are vibration amplitude signals related to the bearing;

[0043] An acquisition module, used for acquiring a target fusion signal based on the plurality of vibration signals;

[0044] A determination module is used to determine whether the bearing is in a fault state based on the target fusion signal.

[0045] In a third aspect, the present application provides an electronic device, comprising: a processor, and a memory and a transceiver communicatively connected to the processor;

[0046] The memory stores computer-executable instructions; the transceiver is used to send and receive data;

[0047] The processor executes the computer-executable instructions stored in the memory to implement the method as described in the first aspect or any one of the above-mentioned methods.

[0048] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used for the method described in any one of the above-mentioned first aspects.

[0049] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method as described in any one of the above-mentioned first aspects.

[0050] The present application provides a bearing fault determination method, device, equipment, storage medium and product. In the present application, the fault determination device receives vibration signals collected by multiple acceleration sensors, then obtains a target fusion signal based on the multiple vibration signals, and further determines that the bearing is in a fault state based on the target fusion signal. In the present application, multiple vibration signals are relied upon to determine the fault state, so the target fusion signal in the present application is considered from multiple aspects, not a single aspect, so the target fusion signal can better represent the true health state of the bearing, so that the present application can improve the accuracy of the diagnosis of the health state of the bearing. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0052] Figure 1 An application scenario diagram of a bearing fault determination method provided in this application;

[0053] Figure 2 A schematic diagram of a bearing fault determination method provided in Example 1;

[0054] Figure 3 A schematic flow chart of a bearing fault determination method provided in Embodiment 2;

[0055] Figure 4 A schematic flow chart of a bearing fault determination method provided in Embodiment 3;

[0056] Figure 5 A flow chart of a bearing fault determination method provided in Example 6;

[0057] Figure 6 A flow chart of a bearing fault determination method provided in Embodiment 7;

[0058] Figure 7 A bearing fault determination flow chart of a proposed fusion method provided in Example 9;

[0059] Figure 8 A schematic diagram of a data-level fusion signal algorithm provided in Example 9;

[0060] Fig. 9 A schematic diagram of average accuracy results of multiple fusion methods provided in Example 9;

[0061] Fig.10A structural schematic diagram of a bearing fault determination device provided in Embodiment 10;

[0062] Fig.11 A schematic diagram of the structure of an electronic device provided in Example 11.

[0063] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0064] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0065] In the prior art, currently, acceleration sensors are commonly used to monitor the state of bearings. Specifically, the health state of the bearing is determined by obtaining vibration signals collected by the acceleration sensors.

[0066] However, in the prior art, the vibration signal collected by an acceleration sensor is used for monitoring status diagnosis. However, due to the limitations of the coverage, quantity and accuracy of the acceleration sensor, a single acceleration sensor cannot fully reflect the health status of the bearing, resulting in low fault diagnosis accuracy.

[0067] In order to solve the defects of the prior art, the inventor of this scheme has designed a new scheme after creative research. This scheme provides a bearing fault determination method. In order to solve the problem of low diagnostic accuracy of the bearing health status in the prior art, the bearing fault determination device in this scheme can receive vibration signals collected by multiple acceleration sensors, so that this scheme can obtain multiple vibration signals, and then fuse the multiple vibration signals to obtain the target fusion signal, and then determine that the bearing is in a fault state based on the target fusion signal. In this scheme, multiple vibration signals can more comprehensively reflect the current state of the bearing, so this scheme considers more aspects, and multiple vibration signals come from multiple acceleration sensors, thereby improving the coverage of the acceleration sensor, and can also avoid the problem of inaccurate vibration signals due to the accuracy of the sensor itself, because the number of acceleration sensors in this scheme is multiple, and in addition, multiple vibration signals are fused in this scheme, so the fusion signal is considered, so this scheme can improve the accuracy.

[0068] The following introduces the application scenarios of a bearing fault determination method, device, equipment, storage medium and product provided by the present application.

[0069] Figure 1 This is an application scenario diagram of a bearing fault determination method provided in this application. Figure 1 As shown, the application scenario diagram includes an electronic device 101 and a plurality of acceleration sensors 102 .

[0070] The electronic device 101 may be any device capable of calculating and receiving vibration signals, and is not limited here.

[0071] The plurality of acceleration sensors 102 include three acceleration sensors.

[0072] Specifically, the multiple acceleration sensors 102 collect vibration signals and send the multiple vibration signals to the electronic device 101. Then the electronic device 101 receives the multiple vibration signals. Further, the electronic device 101 obtains a target fusion signal based on the multiple vibration signals.

[0073] Further, the electronic device 101 determines that the bearing is in a fault state based on the target fusion signal.

[0074] It should be noted that the multiple acceleration sensors 102 respectively collect vibration signals of the bearings in their respective areas.

[0075] It should be noted that one acceleration sensor 102 can collect vibration signals of at least one bearing.

[0076] It should be noted that the bearing fault determination method, device, equipment, storage medium and product provided by the present application can be applied to the computer field or to any field other than computers, and the application field of the present application is not limited.

[0077] The present application provides a bearing fault determination method, which aims to solve the above technical problems in the prior art.

[0078] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0079] Embodiment 1

[0080] The execution body of Embodiments 1 to 9 of the present application is a bearing fault determination device, and the bearing fault determination device (referred to as the determination device) is located in the electronic device.

[0081] Figure 2The following is a flow chart of a bearing fault determination method provided in Example 1. Figure 2 As shown, specifically including:

[0082] S201, receiving vibration signals collected by multiple acceleration sensors; the vibration signals are vibration amplitude signals related to the bearing.

[0083] The acceleration sensor is a sensor capable of measuring acceleration. It should be noted that the acceleration sensor can collect the vibration amplitude related to the bearing, that is, the vibration signal.

[0084] It should be noted that before the acceleration sensor collects the vibration signal, the length of the collected vibration signal can be preset.

[0085] S202, obtaining a target fusion signal based on multiple vibration signals.

[0086] The target fusion signal refers to the final or optimal fusion signal after fusing multiple vibration signals.

[0087] In one embodiment, a bearing fault determination device (hereinafter referred to as the determination device) fuses multiple vibration signals based on the weights of each acceleration sensor to obtain a target fusion signal.

[0088] In one embodiment, the determination device may perform feature extraction on multiple vibration signals to obtain a target fusion signal.

[0089] S203: Determine that the bearing is in a fault state based on the target fusion signal.

[0090] The fault state refers to that the bearing is faulty.

[0091] It should be noted that the determination device can also determine that the bearing is in a normal state based on the target fusion signal.

[0092] It should be noted that in the present application, the state of at least one bearing can be determined based on the target fusion signal.

[0093] This embodiment provides a bearing fault determination method. In this embodiment, the determination device can receive vibration signals collected by multiple acceleration sensors, and obtain a target fusion signal based on the multiple vibration signals, so as to determine that the bearing is in a fault state. In this scheme, the fault state is determined based on multiple vibration signals, so it is considered comprehensively from multiple aspects and does not rely on a single vibration signal. In addition, there are multiple acceleration sensors in this scheme, which have a wider coverage range and the obtained vibration signals are more comprehensive, so that this scheme can determine the fault state more accurately.

[0094] Embodiment 2

[0095] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to obtain a target fusion signal based on multiple vibration signals.

[0096] Figure 3 A flow chart of a bearing fault determination method provided in Example 2. Figure 3 As shown, specifically including:

[0097] S301, obtaining initial weights of multiple vibration signals.

[0098] The initial weight is randomly generated for the corresponding acceleration sensor, or may be preset.

[0099] In one embodiment, the determining device obtains initial weights of the plurality of vibration signals from a storage center.

[0100] S302: Input a plurality of initial weights into a fusion signal algorithm to obtain an initial fusion signal.

[0101] Among them, the fusion signal algorithm is an algorithm that calculates a fusion signal based on multiple vibration signals.

[0102] In one approach, the fusion signal algorithm is shown in equation (1):

[0103]

[0104] Among them, S i represents the vibration signal collected by the i-th acceleration sensor, M represents the number of acceleration sensors, and W i represents the weight corresponding to the vibration signal collected by the i-th acceleration sensor, S f represents the fusion signal after the data set is fused. i After iteration, the weight corresponding to each iteration is obtained.

[0105] It should be noted that the determining device inputs the initial weight into the fusion signal algorithm formula (1) to obtain the initial fusion signal.

[0106] The initial fusion signal refers to a fusion signal calculated based on the initial weights.

[0107] For example, it is assumed that there are three vibration signals, namely, a first vibration signal S1, a second vibration signal S2 and a third vibration signal S3. M is 3. i m Represents the optimization weight of the mth iteration of the i-th vibration signal. W1 0 Represents the initial weight corresponding to the first vibration signal, W2 0 Represents the initial weight corresponding to the second vibration signal, W3 0represents the initial weight corresponding to the third vibration signal. In this embodiment, W i 0 Represents the initial weight corresponding to the i-th vibration signal.

[0108] Furthermore, M=3, W 0 , S1, S2 and S3 are input into formula (1) to obtain S f1 , where S f1 is the initial fusion signal. 0 are the initial weights corresponding to the three vibration signals (S1, S2 and S3).

[0109] S303: Obtain a target fusion signal based on the initial fusion signal.

[0110] In one embodiment, the determining device performs iterative calculation based on the initial fusion signal to obtain the target fusion signal.

[0111] This embodiment provides a bearing fault determination method. In this embodiment, a determination device obtains an initial weight, thereby calculating an initial fusion signal based on a fusion signal algorithm, thereby obtaining a target fusion signal. This solution can improve the accuracy.

[0112] Embodiment 3

[0113] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to obtain a target fusion signal based on an initial fusion signal.

[0114] Figure 4 The following is a flow chart of a bearing fault determination method provided in Example 3. Figure 4 As shown, specifically including:

[0115] S401, calculating the corresponding kurtosis based on the initial fusion signal.

[0116] In this step, the kurtosis algorithm is used to calculate the kurtosis corresponding to the fusion signal, as shown in formula (2):

[0117]

[0118] Among them, S f (n) represents the nth point signal in the fused signal, N represents the length of the fused signal, represents the average value of the fused signal, and n represents the point signal in the fused signal.

[0119] According to the exemplary embodiment of step S302 above, an initial fusion signal is calculated based on three vibration signals, assuming that S f1 , based on S f1 and S f1The kurtosis of the initial fusion signal at this time is calculated by the average value, which can be determined as the initial kurtosis. f1 is the initial fusion signal.

[0120] S402: Input the initial fusion signal into a weight algorithm to obtain initial weights of the fusion signal and multiple vibration signals.

[0121] Among them, the weight algorithm introduces the correlation function between the fusion signal and multiple vibration signals. The weight algorithm is shown in formula (3):

[0122]

[0123] Where M represents the number of acceleration sensors, S i represents the i-th vibration signal, S f represents the fused signal after data-level fusion, and C(.) represents the i-th vibration signal S i With the fusion signal S f The correlation coefficient between them is λ, and λ represents the weight of the fusion signal and multiple vibration signals.

[0124] S403, obtaining a target fusion signal based on the kurtosis and the initial weight.

[0125] This embodiment provides a bearing fault determination method. In this embodiment, the initial fusion signal is input into a weight algorithm to obtain a weight, so the target fusion signal obtained based on kurtosis and weight in this embodiment will be more accurate.

[0126] Embodiment 4

[0127] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way of inputting the initial fusion signal into the weight algorithm to obtain the initial weights of the fusion signal and multiple vibration signals.

[0128] This embodiment includes: inputting the initial fusion signal, multiple vibration signals and the average value of the vibration signals into a correlation coefficient algorithm to obtain a correlation coefficient, and inputting the correlation coefficient into a weight algorithm to obtain an initial weight.

[0129] Among them, the correlation coefficient refers to the i-th vibration signal S i With the fusion signal S f The correlation coefficient between .

[0130] In this step, the correlation coefficient algorithm is shown in formula (4):

[0131]

[0132] Among them, S i represents the i-th vibration signal, S i(n) represents the nth point signal in the ith vibration signal, S f represents the fused signal after data-level fusion, S f (n) represents the nth point signal in the fusion signal, N represents the length of the vibration signal, represents the average value of the i-th vibration signal, represents the average value of fused signals.

[0133] Among them, length refers to the number of point signals.

[0134] in, It can be calculated based on the length of the vibration signal.

[0135] in, It can also be calculated based on the length of the fused signal.

[0136] According to the above exemplary method, continue to calculate S f1 The average value is the average value of the initial fusion signal. N, S1(n), S2(n), S3(n), the average value of the first vibration signal, the average value of the second vibration signal, the average value of the third vibration signal, and the nth point signal of the initial fusion signal are input into the correlation coefficient algorithm to obtain the correlation coefficients corresponding to each vibration signal, which are recorded as C1, C2 and C3 respectively.

[0137] Furthermore, the determination device inputs C1, C2 and C3 into a weight algorithm to calculate an initial weight λ1.

[0138] This embodiment provides a bearing fault determination method. In this embodiment, the correlation coefficient is calculated with the help of a correlation coefficient algorithm so that the obtained weights are more comprehensive.

[0139] Embodiment 5

[0140] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to obtain a target fusion signal based on kurtosis and initial weights.

[0141] This embodiment includes: inputting the kurtosis and the initial weight into the correlation kurtosis algorithm to obtain the initial correlation kurtosis value of the initial fusion signal; and using the initial correlation kurtosis value to obtain the target fusion signal.

[0142] Among them, the relevant kurtosis algorithm is shown in formula (5):

[0143] (5)CK=λ·K

[0144] Where CK represents the relevant kurtosis value.

[0145] Exemplarily, λ1 is input into the correlation kurtosis algorithm to calculate the initial correlation kurtosis value CK1.

[0146] Further, the determination device uses the initial correlation kurtosis value CK1 to obtain the target fusion signal.

[0147] It should be noted that the present application takes into account the kurtosis and related kurtosis values ​​so that the target fusion signal is fused from multiple angles and aspects, so the target fusion signal has high accuracy.

[0148] This embodiment provides a bearing fault determination method. In this embodiment, an initial correlation kurtosis value is calculated, and then the initial correlation kurtosis value is used to accurately determine a target fusion signal.

[0149] Embodiment 6

[0150] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to obtain a target fusion signal by using an initial correlation kurtosis value.

[0151] Figure 5 A flow chart of a bearing fault determination method provided in Example 6. Figure 5 As shown, specifically including:

[0152] S501, taking the inverse of the initial correlation kurtosis value as the initial loss value.

[0153] S502, inputting the initial loss value into a weight optimization algorithm to obtain optimization weights corresponding to a plurality of vibration signals.

[0154] The weight optimization algorithm may be a stochastic gradient descent optimization algorithm, or others, which are not limited here.

[0155] It should be noted that the optimization weight includes weights corresponding to multiple vibration signals after being iterated in sequence.

[0156] S503, within the number of iterations, continue to calculate the optimization weight corresponding to each iteration based on the optimization weight and related algorithms.

[0157] Among them, the relevant algorithms include fusion signal algorithm, weight algorithm, correlation coefficient algorithm, correlation kurtosis algorithm and weight optimization algorithm.

[0158] It should be noted that when the determination device calculates the optimization weight W corresponding to the initial loss value in S502, i 1 , where W i 1 W represents the optimization weight of the first iteration of the i-th vibration signal. i m Represents the optimization weight of the mth iteration of the ith vibration signal.

[0159] Furthermore, after each iteration, the optimization weight corresponding to the i-th vibration signal is obtained, and then the optimization weight corresponding to the i-th vibration signal is input into formula (1), and the next iteration is entered. The corresponding pending fusion signal is obtained from formula (1), and the pending fusion signal is input into formula (2) to obtain the kurtosis corresponding to the pending fusion signal. Then, the pending fusion signal is input into formulas (3) and (4) to obtain the optimization weight, and then the optimization weight and the corresponding kurtosis are input into formula (5) to obtain the optimization related kurtosis value, and then the inverse of the optimization related kurtosis value is used as the optimization loss value, and the optimization loss value is input into the weight optimization algorithm to obtain the optimization weight corresponding to this iteration, and then the above steps are continued to be performed to obtain the optimization weight corresponding to each iteration.

[0160] It should be noted that in each iteration, through the new optimization weight, the corresponding undetermined fusion signal, the corresponding kurtosis of the undetermined fusion signal, the corresponding optimization weight and the corresponding related kurtosis value can be obtained.

[0161] Exemplarily, assuming that the number of iterations is 50, after 50 iterations, 50 pending fusion signals and the corresponding kurtosis are calculated, 50 related kurtosis values ​​(including the initial related kurtosis value and the subsequent optimized related kurtosis value) can be calculated, the corresponding loss value (including the initial loss value and the subsequent optimized loss value), and the corresponding optimization weight.

[0162] It should be noted that each iteration obtains the corresponding optimization weight, which includes the optimization weights corresponding to all vibration signals. According to the above example, the optimization weights after the last iteration include W1 50 , W2 50 and W3 50 , that is, the optimization weight of the 50th iteration corresponding to each of the three vibration signals.

[0163] S504: Determine a target weight from the optimization weights corresponding to all iterations.

[0164] Furthermore, the determination device obtains the loss value of the optimization weight corresponding to all iterations, and then obtains all loss values ​​in the iteration (including the initial loss value and the subsequent optimization loss values), determines the smallest optimization loss value as the target loss value, and determines the optimization weight corresponding to the target loss value as the target weight.

[0165] For example, it is assumed that in the above 50 iterations, the optimization loss value of the 30th iteration is the smallest among all the loss values, so that the optimization loss value of the 30th iteration is determined to be the target loss value. Further, the optimization weight calculated by the target loss value, that is, W i 30 Determined as the target weight.

[0166] S505, inputting the target weight into the fusion algorithm to obtain a target fusion signal.

[0167] Furthermore, the above W i 30 Input into formula (1) to obtain the target fusion signal.

[0168] It should be noted that the target fusion signal is the pending fusion signal corresponding to the 30th iteration.

[0169] This embodiment provides a bearing fault determination method. In this embodiment, the optimization weight corresponding to each iteration can be calculated within the number of iterations, and the target weight can be determined from all the optimization weights to obtain the target fusion signal. In this embodiment, data-level iterative optimization can be achieved, so that the determined target fusion signal is more accurate and achieves a better fusion effect.

[0170] Embodiment 7

[0171] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to determine that a bearing is in a fault state based on a target fusion signal.

[0172] Figure 6 A flow chart of a bearing fault determination method provided in Example 7. Figure 6 As shown, specifically including:

[0173] S601, performing continuous wavelet transform on the target fusion signal to obtain a wavelet time-frequency diagram.

[0174] Among them, continuous wavelet transform can be used to decompose the wavelet signal on the target fusion signal. Wavelets are small and highly localized oscillations in time.

[0175] Among them, wavelet time-frequency diagram is a signal analysis tool, which can reveal the characteristics of the signal in the time domain and frequency domain.

[0176] In one approach, the continuous wavelet transform employs the Morlet wavelet.

[0177] S602, using a one-dimensional neural network to extract features of the target fusion signal to obtain a first feature.

[0178] The one-dimensional neural network includes 3 convolutional layers, 3 maximum pooling layers and 1 global average pooling layer, and each convolutional layer is followed by a batch normalization layer (i.e., BN layer) and an activation function layer (i.e., ReLU layer). The convolution kernel size and number of convolution kernels of the 3 convolutional layers can be (3x1, 16), (3x1, 32), (3x1, 64), respectively, while the convolution kernel size and stride of the 3 maximum pooling layers can be (2x1, 2x1), respectively.

[0179] Specifically, the determination device may use a one-dimensional neural network to perform feature extraction on a one-dimensional target fusion signal to obtain a first feature.

[0180] Among them, the first feature is output by the global average pooling layer in the one-dimensional neural network.

[0181] Among them, the first feature is the extracted one-dimensional feature.

[0182] S603, using a two-dimensional neural network to extract features from the wavelet time-frequency graph to obtain a second feature.

[0183] The two-dimensional neural network includes 3 convolutional layers, 3 maximum pooling layers and 1 global average pooling layer, and each convolutional layer is followed by a batch normalization layer (i.e., BN layer) and an activation function layer (i.e., ReLU layer). The convolution kernel sizes and numbers of the 3 convolutional layers are (3x3, 16), (3x3, 32), and (3x3, 64), respectively, while the convolution kernel sizes and strides of the 3 maximum pooling layers are (2x2, 2x2).

[0184] Specifically, the determining device may use a two-dimensional neural network to perform feature extraction on a two-dimensional wavelet time-frequency graph to obtain a first feature.

[0185] Among them, the second feature is output by the global average pooling layer in the two-dimensional neural network.

[0186] Among them, the second feature is the extracted two-dimensional feature.

[0187] S604: Fuse the first feature and the second feature to obtain a target fused feature.

[0188] It should be noted that in the present application, the first feature is a one-dimensional feature and the second feature is a two-dimensional feature. In this step, the first feature and the second feature are fused so that the target fusion feature includes both one-dimensional features and two-dimensional features, thereby making the target fusion feature more comprehensive and therefore more accurate.

[0189] S605: Determine that the bearing is in a fault state based on the target fusion feature.

[0190] It should be noted that the target fusion signal in the present application is comprehensively considered from the data level and the feature level, so it has high accuracy.

[0191] The fault state refers to the bearing being in a fault state.

[0192] In one embodiment, the determination device can classify the target fusion features and determine that the bearing is in a fault state.

[0193] The present embodiment provides a bearing fault determination method. In the present embodiment, when the determination device determines that the fault state is based on the target fusion signal, the determination device first performs continuous wavelet transform on the target fusion signal to obtain a wavelet time-frequency diagram, and further performs feature extraction on the target fusion signal and the wavelet time-frequency diagram to obtain a first feature and a second feature, respectively, wherein the two features are obtained through a one-dimensional neural network and a two-dimensional neural network, respectively, so that the target fusion feature can include a one-dimensional feature and a two-dimensional feature, so the target fusion feature is more comprehensive.

[0194] Embodiment 8

[0195] This embodiment is a further refinement of any of the above embodiments. This embodiment is an optional way to determine that a bearing is in a fault state based on target fusion features.

[0196] This embodiment includes: inputting the target fusion feature into a classification model that has been pre-trained to convergence, obtaining an output result of at least one bearing, and in response to determining that the output result is a fault category, determining that the corresponding bearing is in a fault state.

[0197] There is no restriction on the classification model.

[0198] The output result may correspond to a variety of bearing fault categories, and the number of bearing fault categories may be preset and is not limited here.

[0199] Among them, the bearing fault category may include 5 categories, namely normal state, rolling element fault category, inner ring fault category, outer ring fault category and conforming fault category. Alternatively, the bearing fault category may include 2 categories, namely fault category and normal category.

[0200] Exemplarily, assume that the output results include the output results of a bearing, which are (0.8, 0.2), where 0.8 represents the probability of category one, and 0.2 represents the probability of category two. Assuming that category one is a fault category and category two is a normal category, it is determined that the probability of category one is greater than that of category two, thereby determining that the output result of the bearing is a fault category, and then determining that it is in a fault state.

[0201] This embodiment also includes:

[0202] In response to determining that the output result is of the normal category, the corresponding bearing is determined to be in a normal state.

[0203] Exemplarily, assume that the output results include the output results of a bearing, which are (0.1, 0.9), where 0.1 represents the probability of category one, and 0.9 represents the probability of category two. Assuming that category one is a fault category and category two is a normal category, it is determined that the probability of category two is greater than that of category one, thereby determining that the output result of the bearing is a normal category, and then determining that it is in a normal state.

[0204] It should be noted that, in this embodiment, it is also possible to diagnose the output results of multiple bearings, and then determine the health status of the multiple bearings.

[0205] It can be seen that in this embodiment, the health status diagnosis of the bearing can be achieved at an order of magnitude.

[0206] The present embodiment provides a bearing fault determination device. In the present embodiment, the determination device inputs the target fusion feature into a classification model that has been pre-trained to convergence, thereby obtaining an output result of at least one bearing. If the output result is a fault category, the corresponding bearing is determined to be in a fault state. If the output result is a normal category, the corresponding bearing is determined to be in a normal state. In the present embodiment, orders of magnitude of bearing health status diagnosis can be achieved. At the same time, the target fusion feature in the present embodiment is a state fused from multiple aspects, so that the final output result is more accurate.

[0207] Embodiment 9

[0208] Figure 7 A bearing fault determination flow chart of the proposed fusion method provided in Example 9.

[0209] It should be noted that in this application, the target weight (i.e., the optimal weight) can be obtained by iterating the vibration signal of the data set for multiple times, and the target fusion signal is obtained based on the target weight fusion, so it is at the data level. In addition, in this application, feature extraction is performed on the target fusion signal and the wavelet time-frequency diagram, and feature-level fusion is achieved to obtain the target fusion feature.

[0210] Exemplarily, the vibration signals corresponding to the three channels of x, y and z collected by the three-axis acceleration sensor arranged on the planetary gearbox are received as input, and the three vibration signals are fused using a data-level fusion method based on the relevant kurtosis rule. After a series of iterative calculations, the target fusion signal is obtained. Then, the target fusion signal is subjected to a continuous wavelet transform to obtain its corresponding wavelet time-frequency diagram. After that, a one-dimensional convolutional neural network and a two-dimensional convolutional neural network are used to extract features from the target fusion signal and the wavelet time-frequency diagram, respectively, and the first feature and the second feature are distributed. Then, the target fusion feature is obtained by splicing, and finally, the target fusion feature is input into a classifier pre-trained to convergence, and the health status of at least one bearing is classified to complete the bearing fault diagnosis.

[0211] like Figure 7 As shown, S1, S2 to S i A total of i vibration signals are input into the fusion signal algorithm, and a data-level target fusion signal is obtained after iteration. Then, the one-dimensional signal in the target fusion signal is extracted, and the first feature is output through the global average pooling layer A in the one-dimensional neural network. At the same time, the target fusion signal is transformed to obtain a two-dimensional wavelet time-frequency diagram, and a two-dimensional signal is extracted from the wavelet time-frequency diagram, and the second feature is output through the global average pooling layer B in the two-dimensional neural network. The first feature and the second feature are spliced ​​to obtain the target fusion feature, which is a feature-level fusion feature. Then, the target fusion feature is input into the classification model to obtain the output result, which includes five bearing fault categories.

[0212] Figure 8 A schematic diagram of a data-level signal fusion algorithm provided in Example 9.

[0213] like Figure 8 As shown, S1, S2 to S i A total of i vibration signals are input into the fusion signal algorithm, and based on the corresponding weights (W1, W2 to W i ) The fusion signal is calculated through weighted fusion, and then the weight of the vibration signal is updated based on the relevant algorithm in back propagation to obtain the optimized weight. The fusion signal is recalculated based on the optimized weight through weighted fusion, and finally an optimal weight (i.e., target weight) is obtained, and the target fusion signal is calculated based on the target weight.

[0214] Fig. 9 This is a schematic diagram of the average accuracy results of multiple fusion methods provided in Example 9. Fig. 9 As shown, specifically including:

[0215] In order to further illustrate the advantages of the present application in bearing fault determination, the method mentioned in the present application is compared with other comparative fusion methods, and the other comparative fusion methods include a first fusion method, a second fusion method, a third fusion method, a fourth fusion method and a fifth fusion method.

[0216] Among them, the first fusion method can be a convolutional neural network with a first layer of wide convolution kernel (abbreviated as WDCNN).

[0217] Among them, the second fusion method can be a multi-scale convolutional neural network method (MSCNN).

[0218] Among them, the third fusion method can be a feature fusion stacked autoencoder method (FFSAE).

[0219] Among them, the fourth fusion method can be an information fusion method (2MNet) based on a multi-sensor multi-scale model.

[0220] Among them, the fifth fusion method can be a data-level fusion based on an adaptive weighted fusion layer and a feature-level fusion method based on splicing (MLVAF-CNN).

[0221] Furthermore, in this application, the proposed method and the above five comparative fusion methods are used to process the three collected vibration signals. Each method is tested 5 times in total. The average accuracy results of each method are as follows: Figure 8 shown.

[0222] Among them, the average result of the first fusion method is 96.5%, the average result of the second fusion method is 96.2%, the average result of the third fusion method is 95.6%, the average result of the fourth fusion method is 97.4%, the average result of the fifth fusion method is 97.8%, and the average result of the method proposed in this application is 99.00%.

[0223] It can be seen that the method proposed in this application achieved the highest average result of 99.00%, indicating that the method of this application can make full use of multiple vibration signals for fusion, give play to the advantages of data-level and feature-level fusion, and accurately complete the bearing fault determination under complex working conditions.

[0224] Embodiment 10

[0225] The following is an embodiment of the device of the present application. Fig.10 This is a structural diagram of a bearing fault determination device provided in Example 10. The bearing fault determination device 100 is located in an electronic device, and the bearing fault determination device 100 includes the following modules:

[0226] The receiving module 1001 is used to receive vibration signals collected by multiple acceleration sensors; the vibration signals are vibration amplitude signals related to the bearings;

[0227] An acquisition module 1002 is used to acquire a target fusion signal based on multiple vibration signals;

[0228] The determination module 1003 is used to determine whether the bearing is in a fault state based on the target fusion signal.

[0229] In one embodiment, when obtaining the target fusion signal based on the multiple vibration signals, the obtaining module 1002 is specifically configured to:

[0230] Obtaining initial weights of multiple vibration signals;

[0231] Inputting a plurality of initial weights into a fusion signal algorithm to obtain an initial fusion signal;

[0232] A target fusion signal is obtained based on the initial fusion signal.

[0233] In one embodiment, when obtaining the target fusion signal based on the initial fusion signal, the obtaining module 1002 is specifically configured to:

[0234] Calculate the corresponding kurtosis based on the initial fusion signal;

[0235] Inputting the initial fusion signal into the weight algorithm to obtain the initial weights of the fusion signal and multiple vibration signals;

[0236] The target fusion signal is obtained based on the kurtosis and the initial weights.

[0237] In one embodiment, when the obtaining module 1002 inputs the initial fusion signal into the weight algorithm to obtain the initial weights of the fusion signal and the multiple vibration signals, it is specifically used to:

[0238] Inputting the initial fusion signal, the multiple vibration signals and the average value of the vibration signals into a correlation coefficient algorithm to obtain a correlation coefficient;

[0239] The correlation coefficient is input into the weight algorithm to obtain the initial weight.

[0240] In one embodiment, when obtaining the target fusion signal based on the kurtosis and the initial weight, the obtaining module 1002 is specifically used to:

[0241] Inputting the kurtosis and the initial weight into the correlation kurtosis algorithm to obtain the initial correlation kurtosis value of the initial fusion signal;

[0242] The target fusion signal is obtained using the initial correlation kurtosis value.

[0243] In one embodiment, when the obtaining module 1002 uses the initial correlation kurtosis value to obtain the target fusion signal, it is specifically used to:

[0244] The inverse of the initial correlation kurtosis value is used as the initial loss value;

[0245] The initial loss value is input into the weight optimization algorithm to obtain the optimization weights corresponding to the multiple vibration signals;

[0246] Within the number of iterations, the optimization weight corresponding to each iteration is continuously calculated based on the optimization weight and related algorithms; the related algorithms include fusion signal algorithm, weight algorithm, correlation coefficient algorithm, correlation kurtosis algorithm and weight optimization algorithm;

[0247] Determine the target weight from the optimization weights corresponding to all iterations;

[0248] The target weight is input into the fusion algorithm to obtain the target fusion signal.

[0249] In one embodiment, when determining that the bearing is in a fault state based on the target fusion signal, the determination module 1003 is specifically configured to:

[0250] Perform continuous wavelet transform on the target fusion signal to obtain the wavelet time-frequency diagram;

[0251] A one-dimensional neural network is used to extract features of the target fusion signal to obtain the first feature;

[0252] A two-dimensional neural network is used to extract features from the wavelet time-frequency graph to obtain the second feature;

[0253] Fusing the first feature and the second feature to obtain a target fused feature;

[0254] The bearing is determined to be in a faulty state based on the target fusion features.

[0255] In one embodiment, when determining that the bearing is in a fault state based on the target fusion feature, the determination module 1003 is specifically configured to:

[0256] Inputting the target fusion features into a classification model that has been pre-trained to convergence to obtain an output result of at least one bearing;

[0257] In response to determining that the output result is a fault category, determining that the corresponding bearing is in a fault state;

[0258] The determination module 1003 is further used to: in response to determining that the output result is a normal category, determine that the corresponding bearing is in a normal state.

[0259] Embodiment 11

[0260] Fig.11 This is a schematic diagram of the structure of an electronic device provided in Example 11. Fig.11As shown, the electronic device 1100 may include: a processor 1101, and a memory 1102 and a transceiver 1103 that are communicatively connected to the processor 1101. The memory 1102 stores computer-executable instructions; the transceiver 1103 is used to send and receive data; the processor 1101 executes the computer-executable instructions stored in the memory 1102 to implement any one of the method embodiments in the above-mentioned embodiments 1 to 9, and the specific implementation methods and technical effects are similar, which will not be repeated here.

[0261] In this embodiment, the transceiver 1103, the memory 1102 and the processor 1101 are connected via a bus. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0262] Embodiment 12

[0263] The present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement any one of the method embodiments of Embodiments 1 to 9 above. The specific implementation methods and technical effects are similar and will not be repeated here.

[0264] Embodiment 13

[0265] The present application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements any one of the method embodiments of Embodiments 1 to 9 above. The specific implementation methods and technical effects are similar and will not be repeated here.

[0266] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0267] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0268] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0269] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable interrupt processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0270] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0271] In addition, although each operation is described in a specific order, this should be understood as requiring such operation to be performed in the specific order shown or in a sequential order, or requiring that all illustrated operations should be performed to obtain desired results. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present invention. Some features described in the context of a separate embodiment can also be implemented in a single implementation in combination. On the contrary, the various features described in the context of a single implementation can also be implemented in multiple implementations individually or in any suitable sub-combination mode.

[0272] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0273] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for determining a bearing fault, characterized in that: The method comprises: Receiving vibration signals collected by multiple acceleration sensors; the vibration signals are vibration amplitude signals related to the bearing; A target fusion signal is obtained based on the multiple vibration signals; the target fusion signal is obtained through multiple iterations of fusion; The bearing is determined to be in a fault state based on the target fusion signal.

2. The method according to claim 1, characterized in that The obtaining of a target fusion signal based on the plurality of vibration signals comprises: Obtaining initial weights of a plurality of the vibration signals; Inputting the plurality of initial weights into a fusion signal algorithm to obtain an initial fusion signal; The target fusion signal is obtained based on the initial fusion signal.

3. The method according to claim 2, characterized in that The obtaining the target fusion signal based on the initial fusion signal comprises: Calculating a corresponding kurtosis based on the initial fusion signal; Inputting the initial fusion signal into a weight algorithm to obtain initial weights of the fusion signal and multiple vibration signals; A target fusion signal is obtained based on the kurtosis and the initial weight.

4. The method according to claim 3, characterized in that The step of inputting the initial fusion signal into a weight algorithm to obtain initial weights of the fusion signal and the plurality of vibration signals includes: Inputting the initial fusion signal, multiple vibration signals and the average value of the vibration signals into a correlation coefficient algorithm to obtain a correlation coefficient; The correlation coefficient is input into a weight algorithm to obtain the initial weight.

5. The method according to claim 4, characterized in that The obtaining of a target fusion signal based on the kurtosis and the initial weight includes: Inputting the kurtosis and the initial weight into a correlation kurtosis algorithm to obtain an initial correlation kurtosis value of the initial fusion signal; The target fusion signal is obtained by using the initial correlation kurtosis value.

6. The method according to claim 5, characterized in that The step of obtaining a target fusion signal by using the initial correlation kurtosis value comprises: Taking the reciprocal of the initial correlation kurtosis value as the initial loss value; Inputting the initial loss value into a weight optimization algorithm to obtain optimization weights corresponding to a plurality of vibration signals; Within the number of iterations, the optimization weight corresponding to each iteration is continuously calculated based on the optimization weight and related algorithms; the related algorithms include fusion signal algorithm, weight algorithm, correlation coefficient algorithm, correlation kurtosis algorithm and weight optimization algorithm; Determine the target weight from the optimization weights corresponding to all iterations; The target weight is input into the fusion algorithm to obtain a target fusion signal.

7. The method according to claim 1, characterized in that The determining that the bearing is in a fault state based on the target fusion signal includes: Performing continuous wavelet transform on the target fusion signal to obtain a wavelet time-frequency diagram; Using a one-dimensional neural network to extract features from the target fusion signal to obtain a first feature; Using a two-dimensional neural network to extract features from the wavelet time-frequency graph to obtain a second feature; Fusing the first feature and the second feature to obtain a target fused feature; The bearing is determined to be in a fault state based on the target fusion feature.

8. The method according to claim 7, characterized in that The determining that the bearing is in a fault state based on the target fusion feature includes: Inputting the target fusion feature into a classification model that has been pre-trained to convergence to obtain an output result of at least one bearing; In response to determining that the output result is a fault category, determining that the corresponding bearing is in a fault state; The method further comprises: In response to determining that the output result is of the normal category, the corresponding bearing is determined to be in a normal state.

9. A bearing fault determination device, characterized in that: The device comprises: A receiving module, used to receive vibration signals collected by multiple acceleration sensors; the vibration signals are vibration amplitude signals related to the bearing; An acquisition module, used for acquiring a target fusion signal based on the plurality of vibration signals; A determination module is used to determine whether the bearing is in a fault state based on the target fusion signal.

10. An electronic device comprising: A processor, and a memory and a transceiver communicatively connected to the processor; The memory stores computer-executable instructions; the transceiver is used to send and receive data; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 8 when executed by a processor.

12. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images