CN113740064B - Rolling bearing fault type diagnosis method, device, equipment and readable storage medium - Google Patents

Abstract

The invention discloses a rolling bearing fault type diagnosis method, device, equipment and a readable storage medium, comprising: acquiring a rolling bearing vibration signal; converting the rolling bearing vibration signal into a frequency spectrum; inputting the frequency spectrum into a pre-built hybrid intelligent diagnosis model, Output the diagnosis result of rolling bearing fault type. The present invention is more suitable for the application environment of rare fault data in the actual industry.

Description

Rolling bearing fault type diagnosis method, device, equipment and readable storage medium

Technical Field

The present invention belongs to the field of fault diagnosis, and in particular relates to a rolling bearing fault type diagnosis method, device, equipment and a readable storage medium.

Background Art

As the core and key components of rotating machinery, the monitoring and detection technology of rolling bearings is crucial for the reliable and smooth operation of the equipment. Deep models are widely used in the field of mechanical fault diagnosis due to their powerful modeling and characterization capabilities. Existing research mainly includes basic models such as deep belief networks (DBN), deep convolutional networks (CNN), stacked autoencoders (SAE) and recurrent neural networks (RNN) and their variants. The input of the network involves different representations of vibration signals in the time domain, frequency domain and time-frequency domain. Through deep models, the deep features of the data are mined to achieve the fault feature extraction and health status identification of bearings. Applications in recent years have shown that deep networks have a long training time due to their large number of parameters. Studies have found that the huge parameters of deep models put forward high requirements on the scale and quality of training data. The training process requires a large amount of labeled data, otherwise it will lead to underfitting and affect the performance of the model. In order to overcome the problem of large amounts of data and high time cost required for deep model training, it is necessary to design special intelligent diagnosis models to improve the rapid learning ability of the models.

Summary of the invention

In view of the problems existing in the prior art, the present invention provides a rolling bearing fault type diagnosis method, device, equipment and readable storage medium, which are more adaptable to the application environment in actual industry where fault data is scarce.

In order to solve the above technical problems, the present invention is implemented by the following technical solutions:

A rolling bearing fault type diagnosis method, comprising:

Obtain rolling bearing vibration signal;

Converting the rolling bearing vibration signal into a frequency spectrum;

The frequency spectrum is input into a pre-built hybrid intelligent diagnosis model, and a rolling bearing fault type diagnosis result is output.

Furthermore, the hybrid intelligent diagnosis model is composed of a random kernel convolutional network and a deep belief network;

The random kernel convolution network includes c layers of one-dimensional feedforward convolution networks, where the value of c is 2 or 3, and each layer of the one-dimensional feedforward convolution network includes a convolution layer and a pooling layer; after the c layers of one-dimensional feedforward convolution networks, the feature outputs corresponding to all channels are averaged to obtain a feature vector of the rolling bearing;

The deep belief network is composed of a z-layer restricted Boltzmann machine, where the value of z is 2, 3 or 4. The deep belief network is used to process the feature vector of the rolling bearing and output a rolling bearing fault type diagnosis result.

Furthermore, assuming that the input of the lth convolutional layer of the random kernel convolutional network is X ^l-1 , specifically:

Where, s is the data length;

The l-th convolution layer K ^l is composed of m convolution kernels, K ^l = [k ₁ , k _i …k _m ], the i-th convolution kernel k _i = [k _i,1 ,k _i,2 …k _i,t ,k _i,n ], n represents the length of the convolution kernel, and t represents the t-th value of the i-th convolution kernel;

具体为：When performing a convolution operation, the step length of each convolution kernel sliding is 1. When the i-th convolution kernel moves j steps, the output is

Specifically:

The value of the convolution kernel k _i,t is randomly generated from {0,1}.

Furthermore, the convolution layer of the random kernel convolution network adopts the Leaky-ReLu function as the activation function, and the Leaky-ReLu function is specifically as follows:

为激活函数的输入，

为激活函数的输出。In the formula,

is the input of the activation function,

is the output of the activation function.

Furthermore, the pooling layer of the random kernel convolutional network uses a non-overlapping maximum pooling operation to process the output of the activation function, as follows:

为第l层卷积层中第i个卷积核的输出。In the formula,

is the output of the i-th convolution kernel in the l-th convolution layer.

Furthermore, after the c-layer one-dimensional feedforward convolutional network, the feature outputs of different channels are averaged to obtain the feature vector of the rolling bearing, which specifically includes:

Generate a single convolution kernel K = [k ₁ , k ₂ …, k _t , … k _n ], perform an average operation on the feature outputs corresponding to all channels without overlap, and average the convolution operation results at the same position of all channels to obtain the feature vector O = [o ₁ , o ₂ , …, o _p , … o _q ] of the rolling bearing, where:

的长度s决定；u和g分别表示该层网络输出的总通道数和第g个通道。In the formula, o _p represents the pth eigenvalue in the eigenvector of the rolling bearing, q represents the dimension of the eigenvector, and q is the output data of the lth layer network.

It is determined by the length s of ; u and g represent the total number of channels and the g-th channel of the network output of this layer respectively.

Furthermore, the deep belief network is used to process the feature vector of the rolling bearing and output a rolling bearing fault type diagnosis result, which is as follows:

The acquired rolling bearing vibration signal is divided into training samples and test samples;

The extracted training sample feature vector is used as the input of the deep belief network, and the parameters of the deep belief network are initialized by greedy pre-training layer by layer;

Calculate the error between the preset labels of the training samples and the actual output of the deep belief network, fine-tune the initialization parameters of the deep belief network, and obtain the trained deep belief network;

The trained deep belief network is used to classify the test samples and output the rolling bearing fault type diagnosis results.

A rolling bearing fault type diagnosis device, comprising:

An acquisition module, used for acquiring a rolling bearing vibration signal;

A conversion module, used for converting the rolling bearing vibration signal into a frequency spectrum;

The result output module is used to input the frequency spectrum into a pre-built hybrid intelligent diagnosis model and output the rolling bearing fault type diagnosis result.

A device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the rolling bearing fault type diagnosis method are implemented.

A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of a rolling bearing fault type diagnosis method are implemented.

Compared with the prior art, the present invention has at least the following beneficial effects: the present invention obtains the vibration signal of the rolling bearing and converts it into a frequency spectrum, and uses a hybrid intelligent fault diagnosis model based on a random kernel convolutional network and a deep belief network to classify and identify the vibration signal of the rolling bearing, obtain a fault diagnosis result, and realize the fault diagnosis of the rolling bearing. The hybrid intelligent diagnosis model in the present invention integrates the convolutional network and the deep belief network, and does not learn all the network layers of the deep model, but only trains and adjusts the parameters of the deep belief network. Under the premise of ensuring the recognition accuracy, the computational complexity of the model can be reduced, thereby improving the training efficiency of the model. The convolutional network itself has the characteristics of local receptive field and weight sharing. Since in the hybrid intelligent fault diagnosis model, the network parameters are generated according to a certain distribution and are not adjusted during the training process, the scale of the training parameters can be reduced, thereby reducing the number of training data requirements, and can better adapt to the application environment where fault data is scarce in actual industry.

In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solutions in the specific implementation modes of the present invention, the drawings required for use in the description of the specific implementation modes will be briefly introduced below. Obviously, the drawings described below are some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

FIG1 is a main flow chart of a rolling bearing fault type diagnosis method according to the present invention;

FIG2 is the overall architecture of the hybrid intelligent fault diagnosis model in the present invention;

FIG3 is a diagram of the architecture of a random kernel convolutional network of the present invention;

FIG4 is a feature extraction result of each layer of the random kernel convolutional network of the present invention;

FIG5 shows the feature results of different types of bearings extracted by the random kernel convolutional network of the present invention.

DETAILED DESCRIPTION

In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

As a specific embodiment of the present invention, as shown in FIG1 , a rolling bearing fault type diagnosis method specifically includes the following steps:

Step 1: Obtain rolling bearing vibration signal.

Step 2: Convert the rolling bearing vibration signal into a frequency spectrum.

Specifically, the present invention uses Fourier transform to process the rolling bearing vibration signal to obtain a frequency spectrum.

Step 3: Input the spectrum into the pre-built hybrid intelligent diagnosis model and output the rolling bearing fault type diagnosis result.

Specifically, the hybrid intelligent diagnosis model consists of a random kernel convolutional network and a deep belief network.

In the hybrid intelligent diagnosis model, the random kernel convolution network is used as a feature extractor, and the random kernel convolution network includes c layers of one-dimensional feedforward convolution networks, where the value of c is 2 or 3, and each layer of one-dimensional feedforward convolution network includes a convolution layer and a pooling layer. In this embodiment, preferably, the random kernel convolution network includes 2 layers of one-dimensional feedforward convolution networks. In this embodiment, as shown in Figure 3, the spectrum result is used as the input of the random kernel convolution network, and after 2 layers of convolution layers and pooling layers, the feature mapping results of each layer of convolution and pooling operations are obtained.

After the c-layer one-dimensional feedforward convolutional network, the feature outputs corresponding to all channels are averaged to obtain the feature vector of the rolling bearing, which includes:

Generate a single convolution kernel K = [k ₁ , k ₂ …, k _t , … k _n ], perform an average operation on the feature outputs corresponding to all channels without overlap, and average the convolution operation results at the same position of all channels to obtain the feature vector O = [o ₁ , o ₂ , …, o _p , … o _q ] of the rolling bearing, where:

的长度s决定；u和g分别表示该层网络输出的总通道数和第g个通道；k应避免取为零向量。In the formula, o _p represents the pth eigenvalue in the eigenvector of the rolling bearing, q represents the dimension of the eigenvector, and q is the output data of the lth layer network.

u and g represent the total number of channels and the g-th channel of the network output of this layer respectively; k should avoid being a zero vector.

In this embodiment, after the second convolutional network layer, the feature outputs corresponding to all channels are averaged to extract the feature vector reflecting the local frequency band in the spectrum, as shown in Figure 4. Each dimension of the feature vector corresponds to the local frequency band of the original spectrum, and its physical meaning is the energy of the local frequency band. Finally, high-dimensional local features of different rolling bearing fault types are obtained as shown in Figure 5.

The construction method of the random kernel convolutional network is as follows:

Assume that the input of the lth convolutional layer of the random kernel convolutional network is X ^l-1 , specifically:

Where, s is the data length;

The l-th convolution layer K ^l is composed of m convolution kernels, K ^l = [k ₁ , k _i …k _m ], the i-th convolution kernel k _i = [k _i,1 ,k _i,2 …k _i,t ,k _i,n ], n represents the length of the convolution kernel, and t represents the t-th value of the i-th convolution kernel;

具体为：When performing a convolution operation, the step length of each convolution kernel sliding is 1. When the i-th convolution kernel moves j steps, the output is

Specifically:

The value of the convolution kernel k _i,t is randomly generated from {0,1} and is frozen once generated and not adjusted during the training phase.

As a preferred implementation, the convolution layer of the random kernel convolution network uses the Leaky-ReLu function as the activation function, and uses the Leaky-ReLu function instead of the traditional S-type function to avoid the phenomenon of gradient disappearance. The Leaky-ReLu function is as follows:

为激活函数的输入，

为激活函数的输出。In the formula,

is the input of the activation function,

is the output of the activation function.

In the present invention, in order to make the obtained two-dimensional random convolution features have a certain invariance to the target translation, the pooling layer of the random kernel convolution network adopts a non-overlapping maximum pooling operation, that is, the maximum value of a specific area of 3*3 or 5*5 is selected as the output of the pooling layer, and the output of the activation function is processed as follows:

为第l层卷积层中第i个卷积核的输出。In the formula,

is the output of the i-th convolution kernel in the l-th convolution layer.

In this embodiment, the convolution kernel parameters of the random kernel convolution network are generated according to a 0-1 distribution, and each layer of the network uses convolution and maximum pooling operations to perform local weighting and dimensionality reduction processing on the input spectrum.

In the hybrid intelligent diagnosis model, the deep belief network is composed of z layers of restricted Boltzmann machines (RBM), and the value of z is 2, 3 or 4. In this embodiment, preferably, the deep belief network is composed of 2 layers of restricted Boltzmann machines. The deep belief network is used to process the feature vector of the rolling bearing and output the rolling bearing fault type diagnosis result. In this embodiment, preferably, the rolling bearing fault type diagnosis result is output in combination with the Soft-max classifier.

As a specific implementation of the present invention, a deep belief network is used to process the feature vector of the rolling bearing and output the rolling bearing fault type diagnosis result, which is as follows:

The acquired rolling bearing vibration signal is divided into training samples and test samples;

The extracted training sample feature vector is used as the input of the deep belief network, and the parameters of the deep belief network are initialized by greedy pre-training layer by layer;

Calculate the error between the preset label of the training sample and the actual output of the deep belief network, fine-tune the initialization parameters of the deep belief network, and obtain a trained deep belief network; in this embodiment, the first RBM is trained with the input until energy balance is achieved; the output obtained by learning the first layer of the deep belief network is used as the input of the second RBM to continue training until the energy balance of the second RBM is achieved; add a Soft-max function after the second RBM, and use the expected output to fine-tune the network parameters to achieve the training of the deep belief network; it should be emphasized that only the parameters of the deep belief network and the Soft-max classifier are trained in the training stage, and the feature extractor parameters are no longer adjusted once generated to process all data including training samples and test samples;

The trained deep belief network is used to classify the test samples and output the rolling bearing fault type diagnosis results.

The present invention provides a rolling bearing fault type diagnosis device, comprising:

An acquisition module, used for acquiring a rolling bearing vibration signal;

A conversion module, used for converting the rolling bearing vibration signal into a frequency spectrum;

The result output module is used to input the spectrum into the pre-built hybrid intelligent diagnosis model and output the rolling bearing fault type diagnosis result.

In one embodiment of the present invention, a computer device is provided, the computer device comprising a processor and a memory, the memory being used to store a computer program, the computer program comprising program instructions, and the processor being used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement corresponding method flows or corresponding functions; the processor described in the embodiment of the present invention can be used for the operation of a rolling bearing fault type diagnosis method.

In one embodiment of the present invention, a rolling bearing fault type diagnosis method can be stored in a computer-readable storage medium if it is implemented in the form of a software functional unit and sold or used as an independent product. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data.

The computer storage medium can be any available medium or data storage device that can be accessed by a computer, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NANDFLASH), solid-state drive (SSD)), etc.

Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

The present application is described with reference to the flowchart and/or block diagram of the method, device (system) and computer program product according to the embodiment of the present application. It should be understood that each process and/or box in the flowchart and/or block diagram, and the combination of the process and/or box in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for realizing the function specified in one process or multiple processes in the flowchart and/or one box or multiple boxes in the block diagram.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (3)

1. A rolling bearing fault type diagnosis method, characterized by comprising: Obtain rolling bearing vibration signal; Convert the rolling bearing vibration signal into a frequency spectrum using Fourier transform; Inputting the frequency spectrum into a pre-built hybrid intelligent diagnosis model to output a rolling bearing fault type diagnosis result; The hybrid intelligent diagnosis model is composed of a random kernel convolutional network and a deep belief network; The random kernel convolution network includes c layers of one-dimensional feedforward convolution networks, where the value of c is 2 or 3, and each layer of the one-dimensional feedforward convolution network includes a convolution layer and a pooling layer; after the c layers of one-dimensional feedforward convolution networks, the feature outputs corresponding to all channels are averaged to obtain a feature vector of the rolling bearing; The deep belief network is composed of a z-layer restricted Boltzmann machine, where the value of z is 2, 3 or 4, and the deep belief network is used to process the feature vector of the rolling bearing and output a rolling bearing fault type diagnosis result; Assume that the input of the lth convolutional layer of the random kernel convolutional network is X ^l-1 , specifically:

Where, s is the data length; The l-th convolution layer K ^l is composed of m convolution kernels, K ^l = [k ₁ , k _i ... k _m ], the i-th convolution kernel k _i = [k _{i, 1} , k _{i, 2} ... k _{i, t} , k _{i, n} ], n represents the length of the convolution kernel, and t represents the t-th value of the i-th convolution kernel; When performing a convolution operation, the step length of each convolution kernel sliding is 1. When the i-th convolution kernel moves j steps, the output is

Specifically:

The value of the convolution kernel k _i,t is randomly generated from {0, 1}; The convolution layer of the random kernel convolution network adopts the Leaky-ReLu function as the activation function. The Leaky-ReLu function is as follows:

In the formula,

is the input of the activation function,

is the output of the activation function; The pooling layer of the random kernel convolutional network uses a non-overlapping maximum pooling operation to process the output of the activation function, as follows:

In the formula,

is the output of the i-th convolution kernel in the l-th convolution layer; After the c-layer one-dimensional feedforward convolutional network, the feature outputs of different channels are averaged to obtain the feature vector of the rolling bearing, which specifically includes: Generate a single convolution kernel K = [k ₁ , k ₂ ..., k _t , ... k _n ], perform an average operation on the feature outputs corresponding to all channels without overlap, and average the convolution operation results at the same position of all channels to obtain the feature vector O = [o ₁ , o ₂ , ..., o _p , ... o _q ] of the rolling bearing, where:

In the formula, o _p represents the pth eigenvalue in the eigenvector of the rolling bearing, q represents the dimension of the eigenvector, and q is the output data of the lth layer network.

u and g represent the total number of channels and the g-th channel of the network output of this layer respectively; The deep belief network is used to process the feature vector of the rolling bearing and output the rolling bearing fault type diagnosis result, which is as follows: The acquired rolling bearing vibration signal is divided into training samples and test samples; The extracted training sample feature vector is used as the input of the deep belief network, and the parameters of the deep belief network are initialized by greedy pre-training layer by layer; Calculate the error between the preset labels of the training samples and the actual output of the deep belief network, fine-tune the initialization parameters of the deep belief network, and obtain the trained deep belief network; The trained deep belief network is used to classify the test samples and output the rolling bearing fault type diagnosis results. 2. A device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a rolling bearing fault type diagnosis method as described in claim 1 when executing the computer program. 3. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of a rolling bearing fault type diagnosis method as claimed in claim 1.

Images