Bearing fault intelligent diagnosis model, diagnosis method and system

Abstract

The present invention relates to the field of intelligent diagnosis technology, and provides a bearing fault intelligent diagnosis model, method and system, comprising a first sub-model, a second sub-model, a third sub-model and a fourth sub-model, wherein the first sub-model and the third sub-model are respectively connected to the second sub-model; the first sub-model comprises a two-dimensional convolution module, a BN batch normalization module, a Swish activation function module and an SENet module connected in series; the second sub-model comprises a plurality of modules connected in a stack; the third sub-model comprises a plurality of modules connected in a stack; the fourth sub-model comprises a convolution layer, a global average pooling and a fully connected layer connected in series; by constructing a neural network structure with different connection modes, the sample features of bearing vibration data are extracted, thereby improving the problems of insufficient concentration and unreasonable feature extraction, and then being able to accurately identify the characteristics of the bearing fault, the severity of the bearing fault, and whether corresponding maintenance measures are adopted.

Description

Technical Field

[0001] The present invention relates to the field of intelligent diagnosis technology, and in particular to an intelligent diagnosis model, a diagnosis method and a system for bearing faults. Background Art

[0002] Intelligent Fault Diagnosis (IFD) integrates expert knowledge with artificial intelligence technology during the diagnostic process. It consists of hardware that simulates brain functions, its necessary external equipment, physical components, and the software that supports this hardware. Rolling bearings, as key components of rotating machinery, are widely used in industrial production equipment such as aerospace, rail transportation, petrochemicals, and wind power generation. Bearing fault diagnosis is crucial for preventing potential accidents, ensuring safe equipment operation, and reducing economic losses.

[0003] Rolling bearing fault diagnosis mainly includes two categories: qualitative diagnosis and quantitative diagnosis. Qualitative diagnosis refers to determining whether there is a bearing fault and the specific location of the fault (such as inner ring fault, outer ring fault, etc.); quantitative diagnosis refers to determining the size of the bearing fault and the severity of the fault. CN114894482A discloses a rolling bearing fault diagnosis method based on NMF data enhancement and CNN, including step 1, signal preprocessing, selecting a suitable sampling rate according to the bearing speed, and dividing the original vibration signal into sample signals of equal size; step 2, data enhancement, using Gaussian, Kaiser, rectangular window functions, and overlapping of different proportions to process the sample signal, generate new signal samples, select a fixed-length window function, and obtain the two-dimensional time-frequency spectrum corresponding to each signal sample through STFT; step 3, converting the time-frequency diagram generated for the original signal sample into matrix data, and using NMF to extract it The fault characteristics are calculated, the closed-form solution after iteration is calculated, and a time-frequency diagram with its characteristics is generated to achieve data expansion; Step 4, a lightweight CNN model for fault diagnosis is constructed. By setting appropriate hyperparameters, the convolution speed is improved with as few layers as possible. The model extracts features from the expanded training samples through convolution, batch normalization, and pooling operations to classify the fault types; Step 5, multiple data sets and reference groups are set to refer to and evaluate the classification effect of the diagnosis model. The fault diagnosis performance under different loads is compared in terms of accuracy, precision, recall rate and f1 value, the cross-domain diagnosis performance of the model is tested, and the advantages and disadvantages of this method are compared with similar methods.

[0004] Existing bearing fault diagnosis methods are based on determining whether a fault exists and the location of the bearing fault. It is difficult to accurately identify the characteristics of the bearing fault, the severity of the bearing fault, and whether to adopt corresponding maintenance measures. Summary of the Invention

[0005] After long-term practice, it was found that the existing bearing fault diagnosis methods are all based on the determination of whether there is a fault and the location of the bearing fault. However, in the process of simply using the EfficientNetV2 intelligent diagnosis model to extract bearing sample features, there are problems such as insufficient focus and unreasonable feature extraction, which makes it difficult to accurately identify bearing fault characteristics, the severity of bearing faults, and whether to adopt corresponding maintenance measures.

[0006] In view of this, the present invention aims to propose an intelligent diagnosis model for bearing faults, which includes a first sub-model, a second sub-model, a third sub-model, and a fourth sub-model, and the first sub-model and the third sub-model are respectively connected to the second sub-model; the first sub-model includes a two-dimensional convolution module with a convolution kernel size of 3×3 and a step size of 2, a BN batch normalization module, a Swish activation function module, and a SENet module connected in series; the second sub-model includes 10 Fused-MBConv modules forming a stacked connection, and the Fused-MBConv module includes 1 dimensionality increase convolution layer and 1 dimensionality reduction convolution layer; the third sub-model includes 30 MBConv modules forming a stacked connection; the fourth sub-model includes 1 convolution layer, 1 global average pooling, and 1 fully connected layer connected in series.

[0007] Preferably, in the second sub-model, the size of the convolution kernel in the dimensionality-increasing convolution layer is 3×3, and the stride is 2; the size of the convolution kernel in the dimensionality-reducing convolution layer is 1×1, and the stride is 1.

[0008] Preferably, in the third sub-model, the MBConv module includes 1 dimensionality-increasing convolution layer, 1 depth-separable convolution, 1 SENet module, and 1 dimensionality-reducing convolution layer; the dimensionality-increasing convolution layer and the dimensionality-reducing convolution layer use two-dimensional convolution with a convolution kernel size of 1×1 and a stride of 1; the depth-separable convolution uses a two-dimensional convolution with a convolution kernel size of 3×3 and a stride of 1; SENet uses the Swish activation function.

[0009] Preferably, in the fourth sub-model, the convolution layer includes 1 two-dimensional convolution with a convolution kernel size of 1×1 and a stride of 1, 1 BN batch normalization, and 1 Swish activation function.

[0010] Preferably, the bearing fault intelligent diagnosis model is obtained through supervised learning training of a neural network model.

[0011] The present invention also discloses a bearing intelligent fault diagnosis method, which includes:

[0012] Step S1, collecting raw vibration data of the rolling bearing in normal, inner ring mild fault, inner ring moderate fault, inner ring severe fault, outer ring mild fault, outer ring moderate fault, and outer ring severe fault states through multiple different types of sensors;

[0013] Step S2: Initialize the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples;

[0014] Step S3: According to the initially set number of sampling points SL of a single sample, the first sample is intercepted at the initial position of the truncated vibration data, a sampling offset SO is determined to be moved from the initial position, and the next sample is intercepted to generate training samples and test samples;

[0015] Step S4, normalize the training samples and test samples to convert them into data with a mean of 0 and a variance of 1;

[0016] Step S5: extracting valid information from the training samples and the test samples, converting the standardized training samples and the test samples into grayscale images with a resolution of 64×64, respectively, to obtain the training set and the test set;

[0017] Step S6, initializing the parameters of the above-mentioned bearing fault intelligent diagnosis model; inputting the training set into the bearing fault intelligent diagnosis model for training, and optimizing the weights and biases of the parameters in the model through the loss function;

[0018] Step S7, determining whether the number of iterations reaches the initially set value. If so, the training of the bearing fault intelligent diagnosis model is completed and the optimal model is saved; if not, proceeding to step S6;

[0019] In step S8, the test set is input into the optimal model, and the optimal model outputs the predicted label for each sample in the test set; the predicted label is compared with the label in the training set to determine the fault type and severity of the current bearing.

[0020] The present invention also provides a system for executing the above-mentioned bearing intelligent fault diagnosis method, the system comprising:

[0021] An acquisition unit includes multiple sensors of different types for collecting raw vibration data of the rolling bearing in normal, inner ring mild fault, inner ring moderate fault, inner ring severe fault, outer ring mild fault, outer ring moderate fault, and outer ring severe fault states;

[0022] Initialization unit, used to initialize the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples;

[0023] The sample data acquisition unit is used to intercept the first sample at the initial position of the truncated vibration data according to the initially set number of sampling points SL of a single sample, determine a sampling offset SO from the initial position, and intercept the next sample to generate training samples and test samples;

[0024] The normalization processing unit is used to normalize the training samples and test samples and convert them into data with a mean of 0 and a variance of 1;

[0025] Grayscale processing unit, used to extract effective information from training samples and test samples, convert the standardized training samples and test samples into grayscale images with a resolution of 64×64, and obtain training sets and test sets respectively;

[0026] A parameter setting unit is used to initialize the parameters of the above-mentioned bearing fault intelligent diagnosis model; the training set is input into the bearing fault intelligent diagnosis model for training, and the weights and biases of the parameters in the model are optimized through the loss function;

[0027] A judgment unit is used to judge whether the number of iterations reaches the initially set value. If so, the training of the bearing fault intelligent diagnosis model is completed and the optimal model is saved; if not, step S6 is continued;

[0028] The output unit is used to input the test set into the optimal model and output the predicted label of the optimal model for each sample in the test set; by comparing the predicted label with the label in the training set, the fault type and severity of the current bearing can be determined.

[0029] Preferably, the acquisition unit includes multiple sensors of different types, including at least 2 piezoelectric acceleration sensors and 1 vibration acceleration sensor. The piezoelectric acceleration sensors and vibration acceleration sensors are fixedly mounted on the measured bearing base. A bearing is provided in the measured bearing base, and the bearing is fixedly connected to the motor output shaft.

[0030] The present invention discloses an electronic device, comprising a memory and a processor: the memory is used to store a computer program; the processor is used to implement the above-mentioned bearing intelligent fault diagnosis method when executing the computer program.

[0031] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the method provided by the present invention when the program is executed by a processor.

[0032] Compared with the prior art, the bearing fault intelligent diagnosis model provided by the present invention includes a first sub-model, a second sub-model, a third sub-model, and a fourth sub-model, and the first sub-model and the third sub-model are respectively connected to the second sub-model; the first sub-model includes a two-dimensional convolution module with a convolution kernel size of 3×3 and a step size of 2, a BN batch normalization module, a Swish activation function module, and a SENet module connected in series; the second sub-model includes 10 Fused-MBConv modules forming a stacked connection, and the Fused-MBConv module includes 1 dimensionality increase convolution layer and 1 dimensionality reduction convolution layer; the third sub-model includes 30 MBConv modules forming a stacked connection; the fourth sub-model includes 1 convolution layer, 1 global average pooling and 1 fully connected layer connected in series; forming a network structure that is more powerful than the traditional EfficientNetV2 neural network structure, which can better diagnose the current bearing fault type and fault severity. The present invention also provides a bearing intelligent fault diagnosis method and system, which uses multiple different types of sensors to collect original vibration data of rolling bearings in normal, inner ring mild fault, inner ring moderate fault, inner ring severe fault, outer ring mild fault, outer ring moderate fault, and outer ring severe fault states; initializes the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples; according to the initially set number of sampling points SL of a single sample, intercepts the first sample at the initial position of the truncated vibration data, determines to move a sampling offset SO from the initial position, and intercepts the next sample to generate training samples and test samples; normalizes the training samples and test samples and converts them into data with a mean of 0 and a variance of 1; extracts training samples and test samples; The effective information in the training and test samples is converted into grayscale images with a resolution of 64×64 after standardization to obtain the training set and test set respectively; the parameters in the above-mentioned bearing fault intelligent diagnosis model are initialized; the training set is input into the bearing fault intelligent diagnosis model for training, and the weights and biases of the parameters in the model are optimized through the loss function; it is determined whether the number of iterations reaches the initial set value. If so, the training of the bearing fault intelligent diagnosis model is completed and the optimal model is saved; if not, the execution continues; the test set is input into the optimal model, and the optimal model predicts the label for each sample in the test set. The predicted label is compared with the label in the training set to determine the current bearing fault type and fault severity. By constructing a bearing fault intelligent diagnosis model with a neural network structure of different connection methods, the characteristics of the bearing vibration data samples are extracted, thereby improving the problems of insufficient attention and unreasonable feature extraction, and then accurately identifying the bearing fault characteristics, the severity of the bearing fault, and whether the corresponding maintenance measures should be adopted.

[0033] Other features and advantages of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings, which constitute part of the present invention, are intended to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are intended to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0035] Figure 1 This is a structural diagram of EfficientNetV2 in the bearing intelligent fault diagnosis model of the present invention;

[0036] Figure 2 This is the SENet structure diagram of the bearing intelligent fault diagnosis model of the present invention;

[0037] Figure 3 This is the Fused-MBConv structure diagram of the bearing intelligent fault diagnosis model of the present invention;

[0038] Figure 4 This is the MBConv structure diagram of the bearing intelligent fault diagnosis model of the present invention;

[0039] Figure 5 This is a flow chart of the bearing intelligent fault diagnosis method of the present invention;

[0040] Figure 6 This is a schematic diagram of sliding interception of bearing samples for the bearing intelligent fault diagnosis model of the present invention. DETAILED DESCRIPTION

[0041] The following describes the specific embodiments of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0042] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0043] It should be noted that the terms "first," "second," "third," "fourth," etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged where appropriate to facilitate the embodiments of the present invention described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or are inherent to these processes, methods, products, or apparatuses.

[0044] In order to solve the problem pointed out in the background technology part that the existing bearing fault diagnosis methods are based on the judgment of whether there is a fault and the location of the bearing fault, but in the process of extracting bearing sample features by simply using the EfficientNetV2 intelligent diagnosis model, there are problems such as insufficient concentration and unreasonable feature extraction, which makes it difficult to accurately identify the bearing fault characteristics, the severity of the bearing fault, and whether to adopt corresponding maintenance measures. The present invention provides a bearing fault intelligent diagnosis model, such as Figures 1-6 As shown, the bearing fault intelligent diagnosis model includes:

[0045] The bearing fault intelligent diagnosis model includes a first sub-model, a second sub-model, a third sub-model, and a fourth sub-model, and the first sub-model and the third sub-model are respectively connected to the second sub-model; the first sub-model includes a two-dimensional convolution module with a convolution kernel size of 3×3 and a step size of 2, a BN batch normalization module, a Swish activation function module, and a SENet module connected in series; the second sub-model includes 10 Fused-MBConv modules forming a stacked connection, and the Fused-MBConv module includes 1 dimensionality increase convolution layer and 1 dimensionality reduction convolution layer; the third sub-model includes 30 MBConv modules forming a stacked connection; the fourth sub-model includes 1 convolution layer, 1 global average pooling, and 1 fully connected layer connected in series.

[0046] The bearing fault intelligent diagnosis model provided by the present invention includes a first sub-model, a second sub-model, a third sub-model, and a fourth sub-model, wherein the first sub-model and the third sub-model are respectively connected to the second sub-model; the first sub-model includes a two-dimensional convolution module with a convolution kernel size of 3×3 and a step size of 2, a BN batch normalization module, a Swish activation function module, and a SENet module connected in series; the second sub-model includes 10 Fused-MBConv modules forming a stacked connection, and the Fused-MBConv module includes 1 dimensionality increase convolution layer and 1 dimensionality reduction convolution layer, such as Figure 3As shown in the figure, the third sub-model includes 30 MBConv modules in a stacked connection. The fourth sub-model consists of one convolutional layer, one global average pooling layer, and one fully connected layer connected in series. This results in a more robust network structure than the traditional EfficientNetV2 neural network, enabling better diagnosis of the current bearing fault type and severity. By constructing an intelligent bearing fault diagnosis model with neural network structures using different connection methods, the characteristics of bearing vibration data samples are extracted, thereby improving problems such as insufficient attention and inappropriate feature extraction. This allows accurate identification of bearing fault characteristics, the severity of the bearing fault, and the need for appropriate maintenance measures.

[0047] EfficientNetV2 uses a non-uniform scaling strategy and an improved progressive learning method. It is a deep learning model with fewer parameters and higher accuracy. Figure 1 shown.

[0048] In order to better preprocess the vibration data and reduce the computational complexity of the model, in the preferred embodiment of the present invention, in the second sub-model, the size of the convolution kernel in the dimensionality-increasing convolution layer is 3×3 and the stride is 2; the size of the convolution kernel in the dimensionality-reducing convolution layer is 1×1 and the stride is 1.

[0049] In order to extract the sample features of bearing vibration data, thereby improving the lack of concentration and unreasonable feature extraction, in the preferred case of the present invention, in the third sub-model, the MBConv module includes 1 dimensionality-raising convolution layer, 1 depth-separable convolution, 1 SENet module, and 1 dimensionality-reducing convolution layer; the dimensionality-raising convolution layer and the dimensionality-reducing convolution layer use two-dimensional convolution with a convolution kernel size of 1×1 and a stride of 1; the depth-separable convolution uses a two-dimensional convolution with a convolution kernel size of 3×3 and a stride of 1; the SENet module uses the Swish activation function, such as Figure 4 To better handle the inattention problem in vibration data, the SENet model preferably replaces the ReLU activation function with the Swish activation function.

[0050] In order to reduce the number of invalid features of a single sample, reduce the amount of single sample data, and highlight the characteristic value of the bearing vibration data sample, in the preferred case of the present invention, in the fourth sub-model, the convolution layer includes 1 two-dimensional convolution with a convolution kernel size of 1×1 and a stride of 1, 1 BN batch normalization, and 1 Swish activation function.

[0051] The SENet module is a channel attention module that enables the model to autonomously learn the importance of features on different channels by mining the relationship between feature channels, thereby improving the performance of the model. Figure 2As shown in the SENet module, the input feature u of each channel is first globally averaged pooled, and the input feature u of each channel is compressed into a one-dimensional real number z∈ c The cth element in z can be calculated as follows.

[0052]

[0053] Where c represents the number of channels; H represents the height of the input image; and W represents the width of the input image. Secondly, by building two fully connected layers and an activation function, we globally capture the inter-channel scale dependencies. The weight vector s, consisting of the weights for each channel, can be calculated using the following formula.

[0054] s=F ex (z,W)=σ(W2δ(W1z))

[0055] Where: W1 and W2 both represent fully connected layers; σ represents the Sigmoid function, and δ represents the ReLU activation function, which is a real number domain.

[0056] The obtained weight vector s and the original feature u c Perform mapping point multiplication to obtain weighted feature map As shown in the following formula.

[0057]

[0058] In order to enable the intelligent diagnosis model for bearing faults to accurately predict, identify the characteristics of bearing faults, the severity of bearing faults, and whether corresponding maintenance measures should be adopted, in the preferred embodiment of the present invention, the intelligent diagnosis model for bearing faults is obtained through supervised learning training of a neural network model.

[0059] The present invention also discloses a bearing intelligent fault diagnosis method, which includes:

[0060] Step S1, collecting raw vibration data of the rolling bearing in normal, inner ring mild fault, inner ring moderate fault, inner ring severe fault, outer ring mild fault, outer ring moderate fault, and outer ring severe fault states through multiple different types of sensors;

[0061] Step S2: Initialize the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples;

[0062] Step S3: According to the initially set number of sampling points SL of a single sample, the first sample is intercepted at the initial position of the truncated vibration data, a sampling offset SO is determined to be moved from the initial position, and the next sample is intercepted to generate training samples and test samples;

[0063] Step S4, normalize the training samples and test samples to convert them into data with a mean of 0 and a variance of 1;

[0064] Step S5: extracting valid information from the training samples and the test samples, converting the standardized training samples and the test samples into grayscale images with a resolution of 64×64, respectively, to obtain the training set and the test set;

[0065] Step S6, initializing the parameters of the above-mentioned bearing fault intelligent diagnosis model; inputting the training set into the bearing fault intelligent diagnosis model for training, and optimizing the weights and biases of the parameters in the model through the loss function;

[0066] Step S7, determining whether the number of iterations reaches the initially set value. If so, the training of the bearing fault intelligent diagnosis model is completed and the optimal model is saved; if not, proceeding to step S6;

[0067] In step S8, the test set is input into the optimal model, and the optimal model outputs the predicted label for each sample in the test set; the predicted label is compared with the label in the training set to determine the fault type and severity of the current bearing.

[0068] The present invention also provides a bearing intelligent fault diagnosis method, such as Figure 5As shown, the original vibration data of the rolling bearing in normal, inner ring light fault, inner ring moderate fault, inner ring severe fault, outer ring light fault, outer ring moderate fault, and outer ring severe fault states are collected by multiple different types of sensors; the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples are initialized; according to the initially set number of sampling points SL of a single sample, the first sample is intercepted at the initial position of the truncated vibration data, a sampling offset SO is determined to be moved from the initial position, and the next sample is intercepted to generate training samples and test samples; the training samples and test samples are standardized and converted into data with a mean of 0 and a variance of 1; the training samples and test samples are extracted Based on the effective information of the training sample and the test sample, the standardized training sample and the test sample are converted into grayscale images with a resolution of 64×64 to obtain the training set and the test set respectively; the parameters in the above-mentioned bearing fault intelligent diagnosis model are initialized; the training set is input into the bearing fault intelligent diagnosis model for training, and the weights and biases of the parameters in the model are optimized through the loss function; it is judged whether the number of iterations reaches the initial set value. If so, the training of the bearing fault intelligent diagnosis model is completed and the optimal model is saved; if not, the execution continues; the test set is input into the optimal model, and the predicted label of each sample in the test set is output by the optimal model; the predicted label is compared with the label in the training set to determine the current bearing fault type and the severity of the fault. By constructing a bearing fault intelligent diagnosis model with a neural network structure of different connection methods, the characteristics of the bearing vibration data samples are extracted, thereby improving the problems of insufficient attention and unreasonable feature extraction, and then accurately identifying the bearing fault characteristics, the severity of the bearing fault, and whether the corresponding maintenance measures are adopted.

[0069] The original signal data of the bearing vibration is obtained, and the original signal data is truncated into training set data and test set data according to the ratio value; wherein the truncation position X is,

[0070] X=TrR×(TL-SL+SO)+TeR×(SL-SO)

[0071] TrR is the ratio of the training set, TeR is the ratio of the test set, TL is the total number of sampling points of the bearing vibration raw signal data, SL is the number of sampling points of a single sample, and SO is the sampling offset;

[0072] like Figure 6 As shown. According to the initial set number of sampling points SL of a single sample, the first sample is intercepted at the initial position of the truncated vibration data, a sampling offset SO is determined to be moved from the initial position, and the next sample is intercepted to generate the training set and test set; wherein, the number of samples TrN in the training set is,

[0073]

[0074] The number of samples TeN in the test set is,

[0075]

[0076] The present invention also discloses a system for executing the above-mentioned bearing intelligent fault diagnosis method, the system comprising:

[0077] An acquisition unit includes multiple sensors of different types for collecting raw vibration data of the rolling bearing in normal, inner ring mild fault, inner ring moderate fault, inner ring severe fault, outer ring mild fault, outer ring moderate fault, and outer ring severe fault states;

[0078] Initialization unit, used to initialize the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples;

[0079] The sample data acquisition unit is used to intercept the first sample at the initial position of the truncated vibration data according to the initially set number of sampling points SL of a single sample, determine a sampling offset SO from the initial position, and intercept the next sample to generate training samples and test samples;

[0080] The normalization processing unit is used to normalize the training samples and test samples and convert them into data with a mean of 0 and a variance of 1;

[0081] Grayscale processing unit, used to extract effective information from training samples and test samples, convert the standardized training samples and test samples into grayscale images with a resolution of 64×64, and obtain training sets and test sets respectively;

[0082] A parameter setting unit is used to initialize the parameters of the above-mentioned bearing fault intelligent diagnosis model; the training set is input into the bearing fault intelligent diagnosis model for training, and the weights and biases of the parameters in the model are optimized through the loss function;

[0083] A judgment unit is used to judge whether the number of iterations reaches the initially set value. If so, the training of the bearing fault intelligent diagnosis model is completed and the optimal model is saved; if not, step S6 is continued;

[0084] The output unit is used to input the test set into the optimal model and output the predicted label of the optimal model for each sample in the test set; by comparing the predicted label with the label in the training set, the fault type and severity of the current bearing can be determined.

[0085] The present invention also provides an intelligent bearing fault diagnosis system, which uses multiple different types of sensors in an acquisition unit to collect original vibration data of a rolling bearing in normal, inner ring mild fault, inner ring moderate fault, inner ring severe fault, outer ring mild fault, outer ring moderate fault, and outer ring severe fault states; the initialization unit initializes the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples; the sample data acquisition unit intercepts the first sample at the initial position of the truncated vibration data according to the initially set number of sampling points SL of the single sample, determines to move a sampling offset SO from the initial position, and intercepts the next sample to generate training samples and test samples; the standardization processing unit standardizes the training samples and test samples and converts them into data with a mean of 0 and a variance of 1; The processing unit extracts valid information from the training and test samples, converting the standardized training and test samples into grayscale images with a resolution of 64×64, respectively, to obtain the training and test sets. The parameter setting unit initializes the parameters of the aforementioned intelligent bearing fault diagnosis model. The training set is input into the intelligent bearing fault diagnosis model for training, optimizing the weights and biases of the model parameters using a loss function. The judgment unit determines whether the number of iterations has reached the initial set value. If so, the training of the intelligent bearing fault diagnosis model is complete and the optimal model is saved. If not, execution continues. The test set is input into the optimal model and the optimal model's predicted label for each sample in the test set is output. The output unit compares the predicted label with the label in the training set to determine the current bearing fault type and fault severity. By constructing an intelligent bearing fault diagnosis model with a neural network structure using different connection methods, the characteristics of bearing vibration data samples are extracted, thereby improving problems such as insufficient attention and inappropriate feature extraction. The model can then accurately identify bearing fault characteristics, the severity of bearing faults, and whether corresponding maintenance measures should be adopted.

[0086] In order to obtain bearing vibration data truthfully, accurately and in real time, in a more preferred embodiment of the present invention, the acquisition unit includes multiple sensors of different types, and the sensors include at least 2 piezoelectric acceleration sensors and 1 vibration acceleration sensor. The piezoelectric acceleration sensor and the vibration acceleration sensor are fixedly mounted on the measured bearing base, and a bearing is provided in the measured bearing base, and the bearing is fixedly connected to the motor output shaft.

[0087] The present invention also discloses an electronic device, comprising a memory and a processor: the memory is used to store a computer program; the processor is used to implement the above-mentioned bearing intelligent fault diagnosis method when executing the computer program.

[0088] Furthermore, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the method provided by the present invention when the program is executed by a processor.

[0089] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations. However, those skilled in the art should be aware that the present invention is not limited by the order of the actions described, because according to the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0090] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0091] In the several embodiments provided herein, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units described is merely a logical functional division. In actual implementation, there may be other division methods, such as multiple units or components being combined or integrated into another system, or some features being ignored or not implemented. In addition, the coupling or direct coupling or communication connection shown or discussed may be through some interface, and the indirect coupling or communication connection of devices or units may be electrical or other forms.

[0092] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0093] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0094] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, mobile terminal, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0095] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. An intelligent diagnosis model for bearing faults, characterized in that: The bearing fault intelligent diagnosis model includes a first sub-model, a second sub-model, a third sub-model, and a fourth sub-model, wherein the first sub-model and the third sub-model are respectively connected to the second sub-model; the first sub-model includes a two-dimensional convolution module with a convolution kernel size of 3×3 and a stride of 2, a BN batch normalization module, a Swish activation function module, and an SENet module connected in series; the second sub-model includes 10 Fused-MBConv modules forming a stacked connection, wherein the Fused-MBConv module includes 1 dimensionality increase convolution layer and 1 dimensionality reduction convolution layer; the third sub-model includes 30 MBConv modules forming a stacked connection; the fourth sub-model includes 1 convolution layer, 1 global average pooling, and 1 fully connected layer connected in series; In the second sub-model, the size of the convolution kernel in the dimension-increasing convolution layer is 3×3 and the stride is 2; the size of the convolution kernel in the dimension-reducing convolution layer is 1×1 and the stride is 1; In the third sub-model, the MBConv module includes one dimensionality-increasing convolution layer, one depthwise separable convolution, one SENet module, and one dimensionality-reducing convolution layer; the dimensionality-increasing convolution layer and the dimensionality-reducing convolution layer use two-dimensional convolution with a kernel size of 1×1 and a stride of 1; the depthwise separable convolution uses a two-dimensional convolution with a kernel size of 3×3 and a stride of 1; SENet uses the Swish activation function; In the fourth sub-model, the convolution layer includes a 2D convolution with a kernel size of 1×1 and a stride of 1, a batch normalization (BN), and a Swish activation function. The bearing fault intelligent diagnosis model is obtained through supervised learning training of a neural network model.

2. A bearing intelligent fault diagnosis method, characterized in that: The bearing intelligent fault diagnosis method includes: Step S1, collecting raw vibration data of the rolling bearing in normal, inner ring mild fault, inner ring moderate fault, inner ring severe fault, outer ring mild fault, outer ring moderate fault, and outer ring severe fault states through multiple different types of sensors; Step S2: Initialize the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples; Step S3: According to the initially set number of sampling points SL of a single sample, the first sample is intercepted at the initial position of the truncated vibration data, a sampling offset SO is determined to be moved from the initial position, and the next sample is intercepted to generate training samples and test samples; Step S4, normalize the training samples and test samples to convert them into data with a mean of 0 and a variance of 1; Step S5: extracting valid information from the training samples and the test samples, converting the standardized training samples and the test samples into grayscale images with a resolution of 64×64, respectively, to obtain the training set and the test set; Step S6, initializing the parameters of the bearing fault intelligent diagnosis model according to claim 1; inputting the training set into the bearing fault intelligent diagnosis model for training, and optimizing the weights and biases of the parameters in the model through the loss function; Step S7, determining whether the number of iterations reaches the initially set value. If so, the training of the bearing fault intelligent diagnosis model is completed and the optimal model is saved; if not, proceeding to step S6; In step S8, the test set is input into the optimal model, and the optimal model outputs the predicted label for each sample in the test set; the predicted label is compared with the label in the training set to determine the fault type and severity of the current bearing.

3. A system for executing the bearing intelligent fault diagnosis method as claimed in claim 2, characterized in that: The system comprises, An acquisition unit includes multiple sensors of different types for collecting raw vibration data of the rolling bearing in normal, inner ring mild fault, inner ring moderate fault, inner ring severe fault, outer ring mild fault, outer ring moderate fault, and outer ring severe fault states; Initialization unit, used to initialize the number of sampling points of a single sample, the size of the sampling offset, and the ratio of training samples to test samples; The sample data acquisition unit is used to intercept the first sample at the initial position of the truncated vibration data according to the initially set number of sampling points SL of a single sample, determine a sampling offset SO from the initial position, and intercept the next sample to generate training samples and test samples; The normalization processing unit is used to normalize the training samples and test samples and convert them into data with a mean of 0 and a variance of 1; Grayscale processing unit, used to extract effective information from training samples and test samples, convert the standardized training samples and test samples into grayscale images with a resolution of 64×64, and obtain training sets and test sets respectively; A parameter setting unit for initializing the parameters of the bearing fault intelligent diagnosis model according to claim 1; inputting the training set into the bearing fault intelligent diagnosis model for training, and optimizing the weights and biases of the parameters in the model through a loss function; A judgment unit is used to judge whether the number of iterations reaches the initially set value. If so, the training of the bearing fault intelligent diagnosis model is completed and the optimal model is saved; if not, step S6 is continued; The output unit is used to input the test set into the optimal model and output the predicted label of the optimal model for each sample in the test set; by comparing the predicted label with the label in the training set, the fault type and severity of the current bearing can be determined.

4. The system according to claim 3, characterized in that The acquisition unit includes multiple sensors of different types, including at least two piezoelectric acceleration sensors and one vibration acceleration sensor. The piezoelectric acceleration sensors and the vibration acceleration sensor are fixedly installed on the measured bearing base. A bearing is provided in the measured bearing base, and the bearing is fixedly connected to the motor output shaft.

5. An electronic device, characterized in that: It comprises a memory and a processor: the memory is used to store a computer program; the processor is used to implement the bearing intelligent fault diagnosis method according to claim 2 when executing the computer program.

6. A machine-readable storage medium having instructions stored thereon, the instructions being used to enable a machine to execute the bearing intelligent fault diagnosis method according to claim 2 of the present application.

Images