Equipment part fault diagnosis method, device and equipment based on mixed noise suppression of sensing data

Abstract

The application discloses an equipment part fault diagnosis method and device based on mixed noise suppression of sensing data and equipment, and relates to the field of fault diagnosis.The diagnosis method comprises the following steps: acquiring vibration signals of equipment parts under different fault types; performing data preprocessing on the vibration signals to obtain samples, taking the fault types corresponding to the vibration signals as real labels, and constructing a training set and a verification set; using the training set and the verification set, using a transferable Adaboost algorithm and an adaptive cost-sensitive learning method to train a plurality of deep learning models, and obtaining a plurality of trained deep learning models; and using the plurality of trained deep learning models to perform fault diagnosis on the to-be-tested parts of the equipment.The application improves the accuracy of equipment part fault diagnosis in a mixed noise environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fault diagnosis, in particular to an equipment part fault diagnosis method and device based on mixed noise suppression of sensing data and equipment. BACKGROUND

[0002] Bearing, gear, motor and other complex equipment key parts are widely used in many fields of modern industry. However, many parts work in harsh conditions, which inevitably leads to their failure, which may cause serious accidents and huge losses. Therefore, it is of great significance to accurately and timely diagnose the fault of complex equipment key parts and intervene in time when early fault signs are found to prevent major accidents and ensure personal safety.

[0003] In academia and industry, deep learning-based fault diagnosis methods have received increasing attention, including convolutional neural networks (CNN), recurrent neural networks (RNN), autoencoders (AE), deep belief networks (DBN), etc. However, due to the harsh working environment of complex equipment, the data collected from many complex equipment key parts often contains a large amount of noise, which reduces the quality of the data. The interference of noise makes it easy to be submerged in valuable fault features, and the feature extraction capability of deep learning often decreases, which may not be able to learn fault-related features well, resulting in a decrease in fault diagnosis accuracy.

[0004] To solve the problem of noisy data, the most representative method in existing methods is the deep residual shrinkage network and its various improved models. However, due to the influence of random noise and other factors, the noise distribution in actual industrial data is uneven, resulting in different signal-to-noise ratios of each sample. Most existing methods only implement fault diagnosis under multiple fixed signal-to-noise ratios, and for fault diagnosis under mixed noise environment, samples with stronger noise are more difficult to train, and due to the different training difficulties of each sample, these methods are difficult to obtain high diagnostic accuracy. SUMMARY

[0005] The purpose of the present application is to provide an equipment part fault diagnosis method, device and equipment based on mixed noise suppression of sensing data, which can improve the accuracy of equipment part fault diagnosis under mixed noise environment.

[0006] To achieve the above purpose, the present application provides the following scheme:

[0007] In a first aspect, the present application provides an equipment part fault diagnosis method based on mixed noise suppression of sensing data, which comprises:

[0008] Obtaining the vibration signals of the equipment part under different fault types;

[0009] Data preprocessing is performed on the vibration signal to obtain samples, and the fault type corresponding to the vibration signal is taken as a true label to construct a training set and a verification set;

[0010] A plurality of deep learning models are trained using a migratory Adaboost algorithm and an adaptive cost-sensitive learning method based on the training set and the verification set, and a plurality of trained deep learning models are obtained;

[0011] The plurality of trained deep learning models are used to perform fault diagnosis on the parts to be tested of the equipment.

[0012] Optionally, the plurality of deep learning models are trained using a migratory Adaboost algorithm and an adaptive cost-sensitive learning method based on the training set and the verification set, and the plurality of trained deep learning models are obtained, and the method specifically comprises the following steps:

[0013] M same deep learning models are constructed;

[0014] The weight of each sample in the training set is initialized;

[0015] The value of m is set to 1;

[0016] The mth deep learning model is trained based on the training set and using a target loss function, and d trained mth deep learning models are obtained; when m is equal to 1, the loss function is a cross-entropy loss function, and when m is greater than 1, the loss function is an adaptive weighted cross-entropy loss function, and the number of iterations in the training is less than that when m is equal to 1;

[0017] The d trained mth deep learning models are verified based on the verification set, and a trained mth deep learning model with the highest verification accuracy is obtained as the mth trained deep learning model; the weight of each sample in the training set is updated according to the training result when the mth deep learning model has the highest verification accuracy;

[0018] It is determined whether the value of m is less than M, and a determination result is obtained;

[0019] If the determination result indicates yes, the parameters of the mth trained deep learning model are migrated to the m+1th deep learning model; the value of m is increased by 1, and the step of training the mth deep learning model based on the training set and using the target loss function is returned to;

[0020] If the determination result indicates no, the M trained deep learning models are output.

[0021] Optionally, the formula for initializing the weight of each sample in the training set is:

[0022]

[0023] wherein, is the initialized weight of the i-th sample, and n is the number of samples in the training set.

[0024] Optionally, the adaptive weighted cross-entropy loss function is:

[0025]

[0026] wherein, l (m) is the adaptive weighted cross-entropy loss for training the m-th deep learning model, g is the batch size set during training, Sigmoid() is the Sigmoid function, and are the weight of the i-th sample and the average of all sample weights for training the m-th deep learning model, respectively, y i is the true label of the i-th sample, p (m )(x i ) is the result obtained after the i-th sample passes through the m-th deep learning model and the Softmax operation during training.

[0027] Optionally, the formula for updating the weight of each sample in the training set according to the training result when the m-th deep learning model has the highest validation accuracy is:

[0028]

[0029] δ i = (a1, a2, ···, a b , ···, a K ), 1≤b≤K;

[0030]

[0031] wherein, is the weight of the i-th sample when training the m+1-th deep learning model, and are the weights of the i-th and j-th samples when training the m-th deep learning model, respectively, β is a stability coefficient, K is the number of fault types, δ i and δ j are the class vectors obtained from the true labels of the i-th sample and the j-th sample, respectively, a1, a2, a b , and a K are the 1st, 2nd, bth, and Kth elements in δ i , respectively, p (m) (xi ) and p (m) (x j ) are the results of the i-th sample and the j-th sample after the m-th deep learning model and the Softmax operation during training, and n is the number of samples in the training set.

[0032] Optionally, the formula for fault diagnosis of the to-be-tested part of the equipment using multiple trained deep learning models is:

[0033]

[0034] wherein C(x) is the fault diagnosis result corresponding to the sample x of the to-be-tested part, and are the probabilities of the sample x of the to-be-tested part belonging to the k-th class and the k'-th class after the m-th trained deep learning model and the Softmax operation, M is the number of trained deep learning models, and K is the number of fault types.

[0035] Optionally, the deep learning model is a deep residual shrinkage network with a selective kernel convolution module and a global parameterized ReLU.

[0036] Optionally, the selective kernel convolution module comprises a first convolution layer, a second convolution layer, a first batch normalization layer, a second batch normalization layer, a first activation function module, a second activation function module, a first element-wise addition module, a global average pooling layer, a first full connection layer, a third batch normalization layer, a third activation function module, a second full connection layer, a third full connection layer, a Softmax module, a first channel-wise multiplication module, a second channel-wise multiplication module, and a second element-wise addition module.

[0037] The first convolution layer is connected with the first batch normalization layer, and the first batch normalization layer is connected with the first activation function module.

[0038] The second convolution layer is connected with the second batch normalization layer, and the second batch normalization layer is connected with the second activation function module.

[0039] The first activation function module is connected with the first input end of the first element-wise addition module and the first input end of the first channel-wise multiplication module, respectively; and the second activation function module is connected with the second input end of the first element-wise addition module and the first input end of the second channel-wise multiplication module, respectively.

[0040] The output end of the first element-by-element addition module is connected with the global average pooling layer, the global average pooling layer is connected with the first full connection layer, the first full connection layer is connected with the third batch normalization layer, the third batch normalization layer is connected with the third activation function module, and the third activation function module is connected with the second full connection layer and the third full connection layer respectively;

[0041] The second full connection layer is connected with the first input end of the Softmax module, and the first output end of the Softmax module is connected with the second input end of the first channel-by-channel multiplication module.

[0042] The third full connection layer is connected with the second input end of the Softmax module, and the second output end of the Softmax module is connected with the second input end of the second channel-by-channel multiplication module.

[0043] The output end of the first channel-by-channel multiplication module is connected with the first input end of the second element-by-element addition module, and the output end of the second channel-by-channel multiplication module is connected with the second input end of the second element-by-element addition module.

[0044] In a second aspect, the application provides an equipment part fault diagnosis device based on mixed noise suppression of sensing data, the diagnosis device applying the diagnosis method described above, and the diagnosis device comprising:

[0045] A signal acquisition module is configured to acquire vibration signals of equipment parts under different fault types.

[0046] A data preprocessing module is configured to perform data preprocessing on the vibration signals to obtain samples, and the fault types corresponding to the vibration signals are taken as real labels to construct a training set and a verification set.

[0047] A training module is configured to train a plurality of deep learning models using a transferable Adaboost algorithm and an adaptive cost-sensitive learning method by using the training set and the verification set, and obtain a plurality of trained deep learning models.

[0048] A diagnosis module is configured to perform fault diagnosis on a to-be-tested part of equipment by using the plurality of trained deep learning models.

[0049] In a third aspect, a computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the equipment part fault diagnosis method based on mixed noise suppression of sensing data.

[0050] According to the embodiments of the application, the following technical effects are achieved:

[0051] The application provides an equipment part fault diagnosis method based on mixed noise suppression of sensing data, a device and equipment.

[0052] Because the training difficulty of each sample in the mixed noise data is different, the application increases the weight of the difficult-to-train sample and reduces the weight of the easy-to-train sample through the principle of the transferable Adaboost, updates the weight, and proposes an adaptive weighted cross-entropy loss function to realize cost-sensitive learning. The sample weight obtained through the transferable Adaboost and the adaptive weighted cross-entropy loss function enable the deep learning model to pay more attention to the strong noise samples. In addition, the transferable Adaboost improves the diagnosis performance in the mixed noise environment, and the number of iterations for training other deep learning models is less than the number of iterations for training the first deep learning model, thereby reducing the model training time while ensuring good diagnosis effect.

[0053] The application proposes a deep residual shrinkage network with a selective kernel convolution module and a global parameterized ReLU, introduces the selective kernel convolution module, and sets different convolution kernel sizes in two branches. Because the large convolution kernel is suitable for processing strong noise samples and the small convolution kernel is suitable for processing weak noise samples, the selective kernel convolution module can effectively combine the respective advantages of the large and small convolution kernels, thereby improving the fault diagnosis effect in the mixed noise environment. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings in the following description only constitute some embodiments of the application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0055] Figure 1 An application environment diagram of an equipment part fault diagnosis method based on mixed noise suppression of sensing data in an embodiment of the application;

[0056] Figure 2 A flowchart of an equipment part fault diagnosis method based on mixed noise suppression of sensing data provided by an embodiment of the application;

[0057] Figure 3A schematic diagram of a method for equipment part fault diagnosis based on mixed noise suppression of sensing data is provided for an embodiment of the present application.

[0058] Figure 4 A structural diagram of a deep residual shrink network with a selective kernel convolution module and a global parameterized ReLU is provided for an embodiment of the present application.

[0059] Figure 5 A structural diagram of a residual block is provided for an embodiment of the present application.

[0060] Figure 6 A structural diagram of a selective kernel convolution module is provided for an embodiment of the present application.

[0061] Figure 7 A structural diagram of a global parameterized ReLU activation function module is provided for an embodiment of the present application.

[0062] Figure 8 A structural schematic diagram of a computer device is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0064] The above purposes, features and advantages of the present application can be more obvious and easy to understand. The present application will be described in further detail below with reference to the drawings and specific embodiments.

[0065] The method for equipment part fault diagnosis based on mixed noise suppression of sensing data provided by the embodiments of the present application can be applied to, for example, Figure 1The application environment shown. Among them, the terminal 101 communicates with the server 102 through the network. The data storage system can store the data required by the server 102 to process. The data storage system can be set up separately, or integrated on the server 102, or placed on the cloud or other servers. The terminal 101 can send the vibration signals of the equipment parts under different fault types to the server 102, and the server 102 receives the vibration signals of the equipment parts under different fault types; the server 102 pre-processes the vibration signals to obtain samples, and the fault type corresponding to the vibration signal is taken as the true label, and a training set and a verification set are constructed; using the training set and the verification set, a plurality of deep learning models are trained using a transferable Adaboost algorithm and an adaptive cost-sensitive learning method, and a plurality of trained deep learning models are obtained; using the plurality of trained deep learning models, the fault diagnosis of the equipment to be tested is performed. The server 102 can feed back the obtained diagnosis result to the terminal 101. In addition, in some embodiments, the equipment part fault diagnosis method based on mixed noise suppression of sensing data can also be implemented by the server 102 or the terminal 101 alone, such as the terminal 101 can directly process the vibration signals of the equipment parts under different fault types, or the server 102 can obtain the vibration signals of the equipment parts under different fault types from the data storage system and process the vibration signals of the equipment parts under different fault types.

[0066] Among them, the terminal 101 can be, but not limited to, various desktop computers, notebook computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices, and the Internet of Things devices can be smart speakers, smart televisions, smart air conditioners, smart vehicle devices, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The server 102 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0067] In an exemplary embodiment, as Figure 2 shown, an equipment part fault diagnosis method based on mixed noise suppression of sensing data is provided, which is executed by a computer device, specifically by a terminal or a server computer device alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 the server 102 in the application environment shown. The following steps 201 to 204 are included. Among them:

[0068] Step 201, obtaining vibration signals of equipment parts under different fault types.

[0069] Step 202, pre-processing the vibration signals to obtain samples, and taking the fault type corresponding to the vibration signal as the true label, and constructing a training set and a verification set.

[0070] Step 203: Using the training set and the validation set, train multiple deep learning models using the transferable Adaboost algorithm and the adaptive cost-sensitive learning method to obtain multiple trained deep learning models.

[0071] Step 204: Use multiple trained deep learning models to perform fault diagnosis on the parts to be tested in the equipment.

[0072] By implementing steps 201 to 204 above, the accuracy of fault diagnosis of equipment parts in mixed noise environment can be improved. The above technical solution of the present invention can be applied to fault diagnosis of key parts of complex equipment.

[0073] In another exemplary embodiment of the present invention, in order to achieve data preprocessing, such as Figure 3 As shown, steps 201 and 202 above are replaced by the following steps S1-S2:

[0074] S1: Acquire the initial vibration signal of the critical component of complex equipment during operation through sensors. The vibration signal contains mixed noise.

[0075] This embodiment uses the Case Western Reserve University (CWRU) rolling bearing dataset to experimentally verify the equipment component fault diagnosis method based on sensor data hybrid noise suppression proposed in this invention.

[0076] The test bench for acquiring the CWRU dataset includes a motor, a torque sensor / encoder, a power meter, and an electronic controller. Single-point faults are introduced into the test bearings using electrical discharge machining (EDM). Vibration signals from the bearings are acquired using an accelerometer, and speed and horsepower data are collected using the torque sensor / encoder. The CWRU dataset includes normal baseline data, drive-end bearing fault data at a 12kHz sampling frequency, drive-end bearing fault data at a 48kHz sampling frequency, and fan-end bearing fault data at a 12kHz sampling frequency. The CWRU dataset includes four operating conditions.

[0077] In this embodiment, normal baseline data and drive-end bearing fault data at a sampling frequency of 48kHz were selected for the experiment. The selected operating conditions were 2 horsepower and 1750 r / min. The data used are described in Table 1.

[0078] Table 1 describes the selected data.

[0079] Description Label Normal 0 Inner race failure with fault diameter of 7 mils 1 Inner race failure with fault diameter of 21 mils 2 Rolling element failure with fault diameter of 7 mils 3 Rolling element failure with fault diameter of 21 mils 4 Outer race failure with fault diameter of 7 mils 5 Outer race failure with fault diameter of 21 mils 6

[0080] S2: After preprocessing the initial vibration signal, samples of the training set and validation set are obtained. The training set and validation set are composed of the samples of the training set and validation set and the corresponding fault types.

[0081] In this embodiment, the sliding window length is set to 1024, the sliding step is set to 1024; part of the vibration signal is selected for sample division, the number of each type of sample in the training set is 120, and the number of each type of sample in the verification set is 90; because the noise intensity of the vibration signal in the CWRU data set is small and unknown, Gaussian white noise is added to the training set sample and the verification set sample to simulate mixed noise data, and 10 signal-to-noise ratios from -6dB to 3dB are set, and the sample number corresponding to each signal-to-noise ratio accounts for 10% of the total sample number.

[0082] In another exemplary embodiment of the present application, in order to realize the construction and training of the deep learning model, as shown in the above step 203, the following steps S31-S39 are used instead: Figure 3

[0083] S31: Construct M (M>1) deep learning models which are the same and have strong anti-noise performance.

[0084] S32: Initialize the parameters of the first deep learning model; initialize the training set sample weight by the following formula:

[0085]

[0086] wherein, is the weight of the i-th sample after initialization, and n is the number of samples in the training set.

[0087] S33: Set a large number of iteration times, train the first deep learning model using the cross-entropy loss function, and select the model with the highest verification accuracy for subsequent model parameter migration and model testing.

[0088] S34: When the verification accuracy is the highest, update the sample weight by the following formula for training the next deep learning model.

[0089]

[0090] δ i =(a1,a2,···,a b ,···,a K ),1≤b≤K.

[0091]

[0092] wherein, is the weight of the i-th sample for training the second deep learning model, β is a stability coefficient, K is the number of fault types, and δ i is the category vector obtained from the true label of the i-th sample, a b is the b-th element in δ i , and p​(1) (x i ) is the result of the first deep learning model and the Softmax operation on the ith sample during training.

[0093] S35: Migrate model parameters to the next deep learning model.

[0094] S36: Update the adaptive weighted cross-entropy loss function according to the current sample weight; the adaptive weighted cross-entropy loss function formula is as follows:

[0095]

[0096] wherein, l (m) is the adaptive weighted cross-entropy loss for training the mth deep learning model, g is the batch size set during training, Sigmoid() is the Sigmoid function, and are the weight of the ith sample and the average of all sample weights for training the mth deep learning model, y i is the true label of the ith sample, p (m) (x i ) is the result of the mth deep learning model and the Softmax operation on the ith sample during training.

[0097] S37: Set a smaller number of iterations, train the current deep learning model using the adaptive weighted cross-entropy loss function, and select the model with the highest validation accuracy for subsequent model parameter migration and model testing.

[0098] S38: When the validation accuracy is the highest, update the sample weight by the following formula for training the next deep learning model.

[0099]

[0100] wherein, is the weight of the ith sample for training the m+1th deep learning model, at this time, m is greater than 1, and are the weights of the ith and jth samples for training the mth deep learning model, β is a stability coefficient, K is the number of fault types, δ i and δ j are the class vectors obtained from the true labels of the ith sample and the jth sample, p (m) (x i ) and p (m) (x j) are the results of the ith sample and the jth sample after the mth deep learning model and the Softmax operation during training, and n is the number of samples in the training set.

[0101] S39: repeating S35-S38 until the M deep learning models are all trained.

[0102] The deep learning model with strong noise immunity in S31 is specifically a deep residual shrinkage network with a global parameterized ReLU (GPReLU), in which the first convolutional layer of the model is replaced with a selective kernel convolution module; the deep residual shrinkage network with the selective kernel convolution module and the GPReLU is as shown in FIG. 2, wherein the residual block is as shown in FIG. 3. Figure 4 Figure 5

[0103] Figure 6 In the formula, C represents the number of feature channels, W1 represents the length of a sample along the time dimension, W2 represents the length of a feature along the time dimension, and M represents the number of neurons in a fully connected (FC) layer. As shown in FIG. 4, the selective kernel convolution module includes: Figure 6

[0104] a first convolutional layer, a second convolutional layer, a first batch normalization (BN) layer, a second BN layer, a first activation function module, a second activation function module, a first element-wise addition module, a global average pooling (GAP) layer, a first FC layer, a third BN layer, a third activation function module, a second FC layer, a third FC layer, a Softmax module, a first channel-wise multiplication module, a second channel-wise multiplication module, and a second element-wise addition module;

[0105] The first convolutional layer is connected with the first BN layer, and the first BN layer is connected with the first activation function module.

[0106] The second convolutional layer is connected with the second BN layer, and the second BN layer is connected with the second activation function module.

[0107] The first activation function module is connected with a first input end of the first element-wise addition module and a first input end of the first channel-wise multiplication module respectively, and the second activation function module is connected with a second input end of the first element-wise addition module and a first input end of the second channel-wise multiplication module respectively.

[0108] An output end of the first element-wise addition module is connected with the GAP layer, the GAP layer is connected with the first FC layer, the first FC layer is connected with the third BN layer, the third BN layer is connected with the third activation function module, and the third activation function module is connected with the second FC layer and the third FC layer respectively.

[0109] ​​​The second FC layer is connected with a first input end of the Softmax module, a first output end of the Softmax module is connected with a second input end of the first channel-by-channel multiplication module;

[0110] The third FC layer is connected with a second input end of the Softmax module, and a second output end of the Softmax module is connected with a second input end of the second channel-by-channel multiplication module.

[0111] The output end of the first channel-by-channel multiplication module is connected with a first input end of the second element-by-element addition module, and the output end of the second channel-by-channel multiplication module is connected with a second input end of the second element-by-element addition module.

[0112] The activation function module is a GPReLU activation function module, as shown in Figure 7 .

[0113] In a specific implementation, the structure parameters of the deep residual shrinkage network with the selective kernel convolution module and the GPReLU are shown in Table 2; the convolution kernel sizes of the first convolution layer and the second convolution layer in the selective kernel convolution module are 7 and 31 respectively; and the neuron numbers of the four FC layers in the activation function module are equal to the channel numbers of the module input features.

[0114] Table 2 is the structure parameters of the deep residual shrinkage network with the selective kernel convolution module and the GPReLU

[0115] Network layer name Number of units Output size Selective kernel convolution module 1 4×512×1 Residual block (4, 3, / 2) 1 4×256×1 Residual block (4, 3) 3 4×256×1 Residual block (8, 3, / 2) 1 8×128×1 Residual block (8, 3) 3 8×128×1 Residual block (16, 3, / 2) 1 16×64×1 Residual block (16, 3) 3 16×64×1 BN, GPReLU, GAP 1 16 FC 1 7

[0116] In this embodiment, the batch size is 32, the optimizer is Adam, the learning rate is 0.002, the learning rate adjustment strategy uses CosineAnnealingLR, and the stability coefficient is 0.1; when training the first deep learning model, the loss function uses the cross-entropy loss function, and the iteration number is 100; when training other deep learning models, the loss function uses the adaptive weighted cross-entropy loss function, and the iteration number is 20.

[0117] In another exemplary embodiment of the present application, in order to realize fault diagnosis, as shown in Figure 3 , the above step 204 is replaced by the following step S4:

[0118] S4: After pre-processing the vibration signal of the part to be tested, the test sample is obtained and input into the trained multiple deep learning models, and the fault type of the current part is obtained through the model test based on the transferable Adaboost.

[0119] The model test based on the transferable Adaboost is specifically that the fault type of the current part is obtained through the following formula:

[0120]

[0121] wherein C(x) is a fault diagnosis result corresponding to a sample x of the part to be tested, and are probabilities of the sample x of the part to be tested belonging to the kth and k'th classes after the sample x of the part to be tested is subjected to the mth trained deep learning model and the Softmax operation, M is the number of the trained deep learning models, and K is the number of fault types.

[0122] In the specific implementation, the sliding window length is set to 1024, and the sliding step length is set to 1024; part of the vibration signals are selected to divide the samples, the number of samples of each class in the test set is 90, and the samples do not overlap with the training set samples and the validation set samples; because the noise intensity of the vibration signals in the CWRU data set is small and unknown, Gaussian white noise is added to the training set samples and the validation set samples to simulate mixed noise data, and 10 kinds of signal-to-noise ratios from -6 dB to 3 dB are set, and the number of samples corresponding to each signal-to-noise ratio accounts for 10% of the total number of samples.

[0123] The experiment sets four different numbers of deep learning models, in order to reduce the influence of random factors, 10 experiments are performed under each number of deep learning models, and the experimental results are shown in Table 3. It can be seen that the accuracy obtained by the experiment is high, which verifies the effectiveness of the method proposed in the application.

[0124] Table 3 is the fault diagnosis test accuracy (%) under different numbers of deep learning models

[0125]

[0126]

[0127] The above embodiment is realized as an optimized result of the application on the CWRU bearing data set, but the specific implementation of the application is not limited to the above embodiment. The above embodiment is only used to help understand the method of the application and its core idea; meanwhile, for those skilled in the art, according to the idea of the application, the specific implementation and application range will be changed. In view of the above, the content of the specification should not be understood as a limitation of the application.

[0128] Based on the same inventive concept, the embodiment of the application also provides an equipment part fault diagnosis device for implementing the equipment part fault diagnosis method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more equipment part fault diagnosis device embodiments based on mixed noise suppression of sensing data provided below can be referred to the limitations of the equipment part fault diagnosis method based on mixed noise suppression of sensing data in the above, which will not be repeated here.

[0129] In one example embodiment, a device for equipment part fault diagnosis based on mixed noise suppression of sensing data is provided, comprising:

[0130] A signal acquisition module is configured to acquire vibration signals of the equipment part under different fault types.

[0131] A data preprocessing module is configured to perform data preprocessing on the vibration signals to obtain samples, the fault type corresponding to the vibration signals being taken as a true label, and a training set and a verification set are constructed.

[0132] A training module is configured to train a plurality of deep learning models using a transferable Adaboost algorithm and an adaptive cost-sensitive learning method by using the training set and the verification set, and obtain a plurality of trained deep learning models.

[0133] A diagnosis module is configured to perform fault diagnosis on a to-be-tested part of the equipment by using the plurality of trained deep learning models.

[0134] In one example embodiment, a computer device can be a server or a terminal, and an internal structure diagram thereof can be as shown in Figure 8 The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is configured to store data determined by complementarity. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through network connection. The computer program is executed by the processor to implement a method for equipment part fault diagnosis based on mixed noise suppression of sensing data.

[0135] Those skilled in the art can understand that Figure 8 the structure shown in the above

[0136] In an example embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the above-mentioned equipment part fault diagnosis method based on mixed noise suppression of sensing data.

[0137] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, the processes of the above-mentioned embodiments can be included. Any reference to memory, database or other medium in each embodiment of the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0138] The database involved in each embodiment of the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in each embodiment of the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, etc., without being limited thereto.

[0139] Each technical feature of the above embodiments can be combined arbitrarily. To make the description concise, all possible combinations of each technical feature in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0140] The principles and implementation manners of the present application are described by using specific examples in the present application, and the above examples are only used to help understand the method of the present application and its core idea; meanwhile, for the general technical personnel in the art, the specific implementation manners and application ranges will be changed according to the idea of the present application. In conclusion, the content of the present specification should not be understood as the limitation of the present application.

Claims

1. A method for fault diagnosis of equipment parts based on mixed noise suppression of sensor data, characterized in that, The diagnostic method includes: Acquire vibration signals of equipment parts under different fault types; Vibration signals are preprocessed to obtain samples. The fault type corresponding to the vibration signal is used as the true label to construct training and validation sets. Using the training set and the validation set, multiple deep learning models are trained using the transferable Adaboost algorithm and the adaptive cost-sensitive learning method to obtain multiple well-trained deep learning models. Using multiple trained deep learning models, fault diagnosis is performed on the parts under test of the equipment. The formula is as follows: Where C(x) is the fault diagnosis result corresponding to sample x of the part under test. and Let M and K be the probabilities of a sample x of the part under test belonging to class k and class k' respectively after passing through the m-th trained deep learning model and performing a Softmax operation. M is the number of trained deep learning models, and K is the number of fault types.

2. The equipment component fault diagnosis method based on sensor data hybrid noise suppression according to claim 1, characterized in that, Using the training set and the validation set, multiple deep learning models are trained using the transferable Adaboost algorithm and the adaptive cost-sensitive learning method to obtain multiple trained deep learning models, specifically including: Construct M identical deep learning models; Initialize the weights for each sample in the training set; Let the value of m be 1; The m-th deep learning model is trained using the training set and the target loss function to obtain d trained m-th deep learning models. The d trained m-th deep learning models are obtained by performing d iterations on the m-th deep learning model. When m equals 1, the loss function is the cross-entropy loss function. When m is greater than 1, the loss function is the adaptive weighted cross-entropy loss function. In this case, the number of training iterations is less than the number of iterations when m equals 1. The m-th deep learning model after training is validated using the validation set. The m-th deep learning model with the highest validation accuracy is obtained and used as the m-th well-trained deep learning model. The weight of each sample in the training set is updated based on the training result when the m-th deep learning model has the highest validation accuracy. Determine whether the value of m is less than M, and obtain the result. If the judgment result indicates yes, then transfer the parameters of the m-th trained deep learning model to the (m+1)-th deep learning model; increment the value of m by 1, and return to the step of "training the m-th deep learning model using the training set and the target loss function to obtain d trained m-th deep learning models". If the judgment result indicates no, then M trained deep learning models are output.

3. The equipment component fault diagnosis method based on sensor data hybrid noise suppression according to claim 2, characterized in that, The formula for initializing the weights of each sample in the training set is: in, Here, n represents the weight of the i-th sample after initialization, and n is the number of samples in the training set.

4. The equipment component fault diagnosis method based on sensor data hybrid noise suppression according to claim 2, characterized in that, The adaptive weighted cross-entropy loss function is: Among them, l (m) Here, g is the adaptive weighted cross-entropy loss used to train the m-th deep learning model, g is the batch size set during training, and sigmoid() is the sigmoid function. and Let y be the weight of the i-th sample used to train the m-th deep learning model and the average weight of all samples, respectively. i p is the true label of the i-th sample. (m) (x i ) represents the result obtained by training the i-th sample through the m-th deep learning model and the Softmax operation.

5. The equipment component fault diagnosis method based on sensor data hybrid noise suppression according to claim 2, characterized in that, The formula for updating the weights of each sample in the training set based on the training result that achieves the highest accuracy with the m-th deep learning model is: δ i =(a1,a2,···,a b ,···,a K ),1≤b≤K; in, The weights of the i-th sample when training the (m+1)-th deep learning model. and Here, β represents the weights of the i-th and j-th samples when training the m-th deep learning model, K represents the number of fault types, and δ represents the weights of the samples. i and δ j Let a1, a2, and a be the class vectors obtained from the true labels of the i-th and j-th samples, respectively. b and a K For δ i The first, second, b-th, and k-th elements in the array, p (m) (x i ) and p (m) (x j ) are the results obtained by the i-th sample and the j-th sample after training through the m-th deep learning model and the Softmax operation, respectively, where n is the number of samples in the training set.

6. The equipment component fault diagnosis method based on sensor data hybrid noise suppression according to claim 1, characterized in that, The deep learning model is a deep residual shrinking network with selective kernel convolution modules and globally parameterized ReLU.

7. The equipment component fault diagnosis method based on sensor data hybrid noise suppression according to claim 6, characterized in that, The selective kernel convolution module includes a first convolutional layer, a second convolutional layer, a first batch normalization layer, a second batch normalization layer, a first activation function module, a second activation function module, a first element-wise addition module, a global average pooling layer, a first fully connected layer, a third batch normalization layer, a third activation function module, a second fully connected layer, a third fully connected layer, a Softmax module, a first channel-wise multiplication module, a second channel-wise multiplication module, and a second element-wise addition module. The first convolutional layer is connected to the first batch of normalized layers, and the first batch of normalized layers is connected to the first activation function module; The second convolutional layer is connected to the second batch normalization layer, and the second batch normalization layer is connected to the second activation function module; The first activation function module is connected to the first input terminal of the first element-wise addition module and the first input terminal of the first channel-wise multiplication module, respectively; the second activation function module is connected to the second input terminal of the first element-wise addition module and the first input terminal of the second channel-wise multiplication module, respectively. The output of the first element-wise addition module is connected to the global average pooling layer, the global average pooling layer is connected to the first fully connected layer, the first fully connected layer is connected to the third batch normalization layer, the third batch normalization layer is connected to the third activation function module, and the third activation function module is connected to the second fully connected layer and the third fully connected layer respectively. The second fully connected layer is connected to the first input of the Softmax module, and the first output of the Softmax module is connected to the second input of the first channel-by-channel multiplication module. The third fully connected layer is connected to the second input of the Softmax module, and the second output of the Softmax module is connected to the second input of the second channel-by-channel multiplication module. The output of the first channel-by-channel multiplication module is connected to the first input of the second element-by-element addition module, and the output of the second channel-by-channel multiplication module is connected to the second input of the second element-by-element addition module.

8. A fault diagnosis device for equipment parts based on mixed noise suppression of sensor data, characterized in that, The diagnostic device applies the diagnostic method according to any one of claims 1-7, and the diagnostic device comprises: The signal acquisition module is used to acquire vibration signals of equipment parts under different fault types. The data preprocessing module is used to preprocess the vibration signals to obtain samples. The fault type corresponding to the vibration signal is used as the real label to construct the training set and the validation set. The training module is used to train multiple deep learning models using the training set and the validation set, employing the transferable Adaboost algorithm and the adaptive cost-sensitive learning method, to obtain multiple trained deep learning models. The diagnostic module is used to diagnose faults in the equipment's test parts using multiple trained deep learning models.

9. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the equipment part fault diagnosis method based on sensor data hybrid noise suppression as described in any one of claims 1-7.

Images