A method, system, device and medium for davit early fault diagnosis

Abstract

The present invention provides a method, system, device and medium for early fault diagnosis of a boat davit, which relates to the field of fault diagnosis. Aiming at the problem that it is difficult to perform early fault diagnosis of a boat davit in a complex electromechanical and hydraulic noise environment, the early operating state of the boat davit is obtained, and after noise reduction and decomposition, the long-term dependency of a long sequence signal is obtained by using a feature extraction network, and a proxy attention mechanism is introduced to improve the Transformer network. While obtaining the global position information of the early weak signal, the calculation amount of the model is reduced, so that the early fault diagnosis of the complex system of the boat davit under the condition of early strong noise interference is realized.

Description

Technical Field

[0001] The present invention relates to the field of fault diagnosis, and in particular to a method, system, equipment and medium for diagnosing early faults of a davit. Background Art

[0002] As a key equipment for ships to perform lifting tasks, the davit plays an important role in ship operations and safe operations at sea. However, since the davit is a complex system integrating mechanical, electrical and hydraulic functions, it has characteristics such as uncertainty, nonlinearity and time-varying, and the working environment contains strong noise interference, which seriously affects the accuracy of early fault diagnosis and increases the difficulty of early reliability maintenance of the equipment.

[0003] A bearing fault diagnosis method based on GA-VMD and adaptive stochastic resonance is disclosed in a Chinese patent (publication number CN115628906A). The optimal variational mode decomposition algorithm is used to reduce the noise generated by the harsh external environment in the original rolling bearing signal, and the bearing vibration signal with preliminary noise reduction by the optimal variational mode decomposition algorithm is obtained. The rolling bearing vibration signal is subjected to random resonance by an improved stochastic resonance system, and the noise contained in the rolling bearing vibration signal is further reduced, and the signal-to-noise ratio of the rolling bearing vibration signal is improved. The characteristic frequency of the early weak fault of the rolling bearing is accurately identified by analysis, and the early fault characteristics of the rolling bearing can be accurately extracted and the fault type can be accurately identified in a harsh environment with strong noise. When diagnosing the fault of the davit, not only the traditional noise environment needs to be considered, but also the complex noise under the joint action of electromechanical and hydraulic. The current fault diagnosis method is difficult to cope with the signal processing and analysis under the complex noise, which makes it difficult to effectively improve the accuracy of the early fault diagnosis of the davit, affecting the difficulty of maintenance. Summary of the invention

[0004] The purpose of the present invention is to provide a method, system, device and medium for early fault diagnosis of a boat davit in view of the defects of the prior art, so as to obtain the early operating status of the boat davit, obtain the long-term dependency of long sequence signals by using a feature extraction network after denoising and decomposition, and introduce a proxy attention mechanism to improve the Transformer network, so as to reduce the amount of model calculation while obtaining the global position information of early weak signals, and realize early fault diagnosis of the complex system of the boat davit under the condition of early strong noise interference.

[0005] The first object of the present invention is to provide a method for diagnosing early faults of a davit, which adopts the following scheme:

[0006] include:

[0007] Get real-time status data of davits;

[0008] The real-time status data of the davit is input into the early fault diagnosis model, and the fault diagnosis results are output;

[0009] Among them, the construction of the early fault diagnosis model includes:

[0010] Obtain the operating data of the early failure of the davit and remove the early noise that obscures the operating data;

[0011] Extract fault features from the operating data after removing early noise, and obtain the correlation between early time-frequency information and fault features;

[0012] The early time-frequency information with fault characteristics is input into the fault diagnosis model for training to obtain a trained early fault diagnosis model.

[0013] Furthermore, the improved variational mode decomposition denoising network is used to decompose the operating data of the davit at the early stage of failure, and the specific properties of the components obtained after constrained decomposition are screened to remove the early noise.

[0014] Furthermore, the operating data after removing the early noise is transformed to obtain the time-frequency information of the non-stationary signal in the davit, and the constructed feature extraction network is used to obtain the temporal long-term dependency of the fault features in the time-frequency information.

[0015] Furthermore, the forget gate in the feature extraction network is used to filter the fault features, and the filtered fault features are input into the fault diagnosis model.

[0016] Furthermore, the time-frequency features corresponding to the time-frequency information are encoded, so that the long sequence features extracted by the feature extraction module carry the position information in the time series signal through encoding.

[0017] Furthermore, in the long sequence features, the encoding vectors corresponding to any two different positions are not equal in each dimension.

[0018] Furthermore, the residual network is used to accelerate the convergence of the training process, and the output of the residual network is converted into a probability distribution to achieve early fault classification of the davit.

[0019] A second object of the present invention is to provide a davit early fault diagnosis system, comprising:

[0020] The acquisition module is configured to: obtain real-time status data of the davit;

[0021] The fault diagnosis module is configured to: input the real-time state data of the davit after denoising into the early fault diagnosis model, and output the fault diagnosis result;

[0022] Among them, the construction of the early fault diagnosis model includes:

[0023] Obtain the operating data of the early failure of the davit and remove the early noise that obscures the operating data;

[0024] Extract fault features from the operating data after removing early noise, and obtain the correlation between early time-frequency information and fault features;

[0025] The early time-frequency information with fault characteristics is input into the fault diagnosis model for training to obtain a trained early fault diagnosis model.

[0026] A third object of the present invention is to provide an electronic device, comprising a processor and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the electronic device executes a method for diagnosing early faults of a davit.

[0027] A fourth object of the present invention is to provide a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the davit early fault diagnosis method as in the first object.

[0028] Compared with the prior art, the present invention has the following advantages and positive effects:

[0029] To address the problem that it is difficult to perform early fault diagnosis on the davit in a complex electromechanical and hydraulic noise environment, the early operating status of the davit is obtained. After denoising and decomposition, the long-term dependency of long sequence signals is obtained using a feature extraction network. The proxy attention mechanism is introduced to improve the Transformer network. While obtaining the global position information of early weak signals, the amount of model calculation is reduced, thus achieving early fault diagnosis of the davit complex system under early strong noise interference conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0031] Figure 1 Flow chart of the method for diagnosing early faults of the davit in Embodiments 1-5 of the present invention.

[0032] Figure 2 Schematic diagram of the genetic particle swarm optimization process in Examples 1-5 of the present invention.

[0033] Figure 3 Schematic diagram of the LSTM network in Examples 1-5 of the present invention.

[0034] Figure 4 Schematic diagram of the Transformer network structure in Examples 1-5 of the present invention.

[0035] Figure 5 Schematic diagram of the agent attention calculation process in Examples 1-5 of the present invention. DETAILED DESCRIPTION

[0036] Example 1

[0037] In a typical embodiment of the present invention, Figure 1-Figure 5 As shown, a method for early fault diagnosis of davits is given.

[0038] A method for diagnosing early faults of a davit, comprising:

[0039] Get real-time status data of davits;

[0040] The real-time status data of the davit is input into the early fault diagnosis model, and the fault diagnosis results are output;

[0041] Among them, the construction of the early fault diagnosis model includes:

[0042] Obtain the operating data of the early failure of the davit and remove the early noise that obscures the operating data;

[0043] Extract fault features from the operating data after removing early noise, and obtain the correlation between early time-frequency information and fault features;

[0044] The early time-frequency information with fault characteristics is input into the fault diagnosis model for training to obtain a trained early fault diagnosis model.

[0045] like Figure 1 As shown in the figure, early fault diagnosis of the davit is carried out based on the combination of Transformer and improved VMD. The early state information of the davit system is collected by the data acquisition system, and the VMD parameters are optimized by combining the genetic particle swarm optimization algorithm and the original signal is decomposed by VMD to achieve noise reduction of the early signal; then, the IMF component of the VMD decomposition is transformed into the instantaneous frequency and instantaneous amplitude of the signal through Hilbert transform to form the corresponding time-frequency graph; then, the composed time-frequency graph is passed through the LSTM feature extraction network to obtain the long-term dependency of the long sequence signal, and the proxy attention mechanism is introduced to improve the Transformer network. While obtaining the global position information of the early weak signal, the model calculation amount is reduced, and the early fault diagnosis of the complex system of the davit under the condition of early strong noise interference is realized.

[0046] In this embodiment, early faults include: failure of balancing valve, failure of bearing rolling element, failure of bearing inner ring, failure of oil cylinder due to insufficient hydraulic pressure in accumulator, lack of phase in pump station motor, etc. There is no clear definition of so-called early faults. It is just that the normal use of equipment but some relatively minor problems are called early fault diagnosis.

[0047] Early strong noise refers to the harsh working environment of the ocean. Mechanical noise includes vibration and friction noise generated by equipment during operation, fluid noise such as liquid flow and pressure change, and electromagnetic interference and radiation noise generated by the operation of electrical systems. The goal of noise removal is to remove the above noise.

[0048] The data acquisition system is developed based on LabVIEW software. It consists of sensors, acquisition cards, host computers, and power supply systems. The data contained includes fluid pressure, nitrogen pressure, equipment temperature, ambient temperature, wire rope tension, boom displacement, bearing vibration, bearing speed, boom strain and other status data.

[0049] Specifically, Figure 1 As shown, relying on the status data of the davit system, accurate diagnosis of early faults of the complex electromechanical and hydraulic system under strong interference conditions can be achieved.

[0050] Firstly, the davit system status data is collected through the data acquisition system;

[0051] The VMD parameters are optimized by genetic particle swarm optimization algorithm, and then the original signal is decomposed by VMD to obtain the IMF component of the signal, thus realizing the denoising of the early signal.

[0052] Then, the IMF components are transformed into the time-frequency diagram of the signal by combining Hilbert transform as the signal input of the diagnostic model;

[0053] Through the LSTM feature extraction network, the feature extraction of early time series signals is realized;

[0054] Then, the proxy attention mechanism is combined with the Transformer network to improve the expressiveness of the model while reducing the computational complexity;

[0055] Realize the diagnosis of early weak faults.

[0056] like Figure 2 As shown, this embodiment proposes an improved variational mode decomposition denoising network.

[0057] The genetic particle swarm optimization algorithm is added to the VMD algorithm. By optimizing the number of modes, penalty coefficient and other parameters, the optimal decomposition model is obtained. The davit signal is subjected to VMD decomposition so that the decomposed IMF components satisfy some specific properties such as the minimum value of permutation entropy, envelope entropy, information entropy and sample entropy.

[0058] in yes The envelope amplitude of It's a signal A non-monotonic decreasing phase function of .

[0059] The genetic particle swarm optimization algorithm combines the genetic algorithm with the particle swarm optimization algorithm, generates new particles by crossover between particles, and performs an elite retention strategy on particles according to the size of their fitness.

[0060] During the optimization process, select and Update the particle's velocity and position respectively.

[0061] Where k represents the number of iterations, i represents the i-th particle, and j represents the j-th dimension. represents the inertia weight, is the learning factor of the individual particle, represents the group learning factor, and represents a random number in [0,1], represents the individual current extreme value of the i-th particle, Indicates the current global extreme value of the particle swarm.

[0062] After forming a new population, the particles are sorted by fitness value. For particles with low fitness, the formula Conduct crossover experiments;

[0063] Where A and B represent parent particles, represents the i-th dimension information of the particle, Represents a random number between 0 and 1, which is set to 0.3 during the iteration.

[0064] The particles with higher fitness are combined into a new population, and finally the optimal solution of the genetic particle swarm optimization algorithm is found through continuous iteration.

[0065] The collected early davit original signal is used to remove the confusing early noise in the signal through the improved VMD network.

[0066] Then, combined with the Hilbert transform, each eigenmode function Do the Hilbert transform to get , where express The phase difference, represents the time step, and obtains the time-frequency information of the non-stationary signal in the davit.

[0067] The time-frequency graph after Hilbert transformation contains the early fault after denoising. The built LSTM feature extraction network is used to obtain the long-term temporal dependency of the fault features in the time-frequency graph, such as Figure 3 As shown in the figure, the gating mechanism enables the recurrent neural network to retain meaningful historical information and selectively forget some meaningless information, thereby improving the diagnostic accuracy of early time series signals.

[0068] like Figure 3 As shown in the figure, in the LSTM network, the time-frequency graph extracts fault features through two layers of convolution, and the forget gate in the LSTM network is used to retain and forget the features transferred from the previous unit module. The expression is: ,in represents the output of the previous unit, represents the input of the current unit, and Respectively represent the weight coefficient and bias coefficient of the forget gate. The input gate determines which features are stored in the unit module, and its expression is ,in and Represents the weight coefficient and bias coefficient of the input gate sigmoid layer. The tanh layer creates a new candidate value vector , generate candidate memories, where and They represent the weight coefficient and bias coefficient of the input gate tanh layer respectively. Finally, the state of this unit is obtained . The output gate is activated by the activation function Determine the output characteristics and realize the output of the unit module state. The output gate formula is , where the hidden state is calculated as follows .

[0069] like Figure 4 As shown in the figure, the time-frequency graph is passed through the LSTM feature extraction network to obtain early fault features, and early fault diagnosis is achieved through the Transformer module.

[0070] In the Transformer network, the proposed time series features are encoded by the position encoder, so that the long sequence features extracted by the feature extraction module carry the position information in the time series signal through encoding. In the long sequence features, the sine function is used and cosine function Encoding can ensure that the encoding vectors corresponding to any two different positions are not equal in each dimension, allowing the model to better capture the structural information of the sequence.

[0071] Among them, Pe is the position code, pos is its position in the sequence, j is the jth element of the feature vector encoded by the position information, and dmo is the length of the feature vector encoded by the position information. The position-encoded vector is combined with the proxy attention, and the relationship between the query and the key-value pair is balanced by introducing the proxy vector. The formula can be expressed as: . Where A represents the proxy matrix, Q, K, and V represent the Query, Key, and Value matrices respectively. Represents Softmax attention.

[0072] like Figure 5 As shown in the figure, in the proxy attention calculation process, the first step is to use A as the query, calculate the attention between A, K and V, and obtain the proxy feature, thereby obtaining the feature information. Then, use A as the key, As the Value, a second attention calculation is performed with Q to broadcast the global information in the proxy feature back to each feature and obtain the final output O, which reduces the computational complexity while retaining the ability to express features.

[0073] Then, the residual network is used to accelerate the convergence of the training network, avoid the phenomenon of gradient disappearance or explosion during the training process, and improve the performance and generalization ability of model diagnosis. Finally, the output of the residual network is converted into a probability distribution through the softmax layer to realize the early fault classification of the complex system of the davit, thereby realizing the diagnosis of the early fault of the davit.

[0074] Example 2

[0075] In another typical embodiment of the present invention, Figure 1-Figure 5 As shown, a davit early fault diagnosis system is given.

[0076] A davit early fault diagnosis system comprises:

[0077] The acquisition module is configured to: obtain real-time status data of the davit;

[0078] The fault diagnosis module is configured to: input the real-time state data of the davit after denoising into the early fault diagnosis model, and output the fault diagnosis result;

[0079] Among them, the construction of the early fault diagnosis model includes:

[0080] Obtain the operating data of the early failure of the davit and remove the early noise that obscures the operating data;

[0081] Extract fault features from the operating data after removing early noise, and obtain the correlation between early time-frequency information and fault features;

[0082] The early time-frequency information with fault characteristics is input into the fault diagnosis model for training to obtain a trained early fault diagnosis model.

[0083] The working process of the early fault diagnosis system for a davit is shown in the content of Example 1 and will not be repeated here.

[0084] Example 3

[0085] In another embodiment of the present invention, Figure 1-Figure 5 As shown, an electronic device is provided.

[0086] The electronic device is used to implement the davit early fault diagnosis method in the above embodiment 1, and the electronic device includes one or more processors, one or more memories coupled to the processors, and a communication module coupled to the processors.

[0087] The memory may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, at least one of the following: read-only memory (ROM), erasable programmable read only memory (EPROM), flash memory, hard disk, compact disc (CD), digital video disc (DVD) or other magnetic storage and / or optical storage. Examples of volatile memories include, but are not limited to, at least one of the following: random access memory (RAM), or other volatile memories that do not persist during the duration of a power outage. The computer program may be stored in the ROM. When the processor executes the computer program, the above-mentioned davit early fault diagnosis method is implemented.

[0088] Example 4

[0089] In another embodiment of the present invention, Figure 1-Figure 5 As shown, a computer-readable storage medium is provided.

[0090] The program may be tangibly contained in a computer-readable medium, which may be included in a device (such as in a memory) or other storage device accessible by the device. The program may be loaded from the computer-readable medium to the RAM for execution. The computer-readable medium may include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, and the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the davit early fault diagnosis method as in Example 1.

[0091] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a server or terminal, the process or function described in the embodiment of the present application is generated in whole or in part. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a server or terminal or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (such as a floppy disk, a hard disk and a tape, etc.), an optical medium (such as a digital video disk (DVD), etc.), or a semiconductor medium (such as a solid-state hard disk, etc.).

[0092] In addition, although each operation is described in a specific order, this should be understood as requiring such operation to be performed in the specific order shown or in a sequential order, or requiring that all illustrated operations should be performed to obtain desired results. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the application. Some features described in the context of a separate embodiment can also be implemented in a single implementation in combination. On the contrary, the various features described in the context of a single implementation can also be implemented in multiple implementations individually or in any suitable sub-combination mode.

[0093] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A method for diagnosing early faults of a davit, characterized in that: include: Get real-time status data of davits; The real-time status data of the davit is input into the early fault diagnosis model, and the fault diagnosis results are output; Among them, the construction of the early fault diagnosis model includes: Obtain the operating data of the early failure of the davit and remove the early noise that obscures the operating data; Extract the fault features from the operation data after removing the early noise, and obtain the correlation between the early time-frequency information and the fault features; input the early time-frequency information with fault features into the fault diagnosis model for training, and obtain the trained early fault diagnosis model; use the improved variational mode decomposition denoising network to decompose the operation data of the davit at the early stage of failure, and screen the specific properties of the components obtained after constraining the decomposition to remove the early noise; The improved variational mode decomposition denoising network: the genetic particle swarm optimization algorithm is added to the VMD algorithm to optimize the number of modes and the penalty coefficient, and the optimal decomposition model is obtained through u k (t) = A k (t)cosφ k (t) Perform VMD decomposition on the davit signal; where A k (t) is u k (t) envelope amplitude, φ k (t) is the signal u k (t) a non-monotonic decreasing phase function; Encode the time-frequency features corresponding to the time-frequency information so that the long sequence features extracted by the feature extraction module carry the position information in the time series signal through encoding; In long sequence features, use the sine function and cosine function Encode; wherein Pe is the position code, pos is the position in the sequence, j is the jth element of the feature vector encoded by the position information, and dmo is the length of the feature vector encoded by the position information; The position-encoded vector is combined with the proxy attention, and the relationship between the query and the key-value pair is balanced by introducing the proxy vector. The formula is: A =Attn S (Q,A,Attn S (A, K, V)); A represents the proxy matrix, Q, K, V represent the Query, Key, and Value matrices respectively, Attn S Represents Softmax attention.

2. The method for diagnosing early faults of a davit according to claim 1, characterized in that: The operating data after removing the early noise is transformed to obtain the time-frequency information of the non-stationary signal in the davit, and the constructed feature extraction network is used to obtain the temporal long-term dependency of the fault features in the time-frequency information.

3. The method for diagnosing early faults of a davit according to claim 2, characterized in that: The forget gate in the feature extraction network is used to screen the fault features, and the screened fault features are input into the fault diagnosis model.

4. The method for diagnosing early faults of a davit according to claim 1, characterized in that: In long sequence features, the encoding vectors corresponding to any two different positions are not equal in each dimension.

5. The method for diagnosing early faults of a davit according to claim 4, characterized in that: The training process is accelerated to converge through the residual network, and the output of the residual network is converted into probability distribution to achieve early fault classification of the davit.

6. A davit early fault diagnosis system, using the davit early fault diagnosis method according to any one of claims 1 to 5, characterized in that: include: The acquisition module is configured to: obtain real-time status data of the davit; The fault diagnosis module is configured to: input the real-time status data of the davit into an early fault diagnosis model and output a fault diagnosis result; wherein the construction of the early fault diagnosis model includes: Obtain the operating data of the early failure of the davit and remove the early noise that obscures the operating data; The fault features in the operating data after early noise removal are extracted to obtain the correlation between the early time-frequency information and the fault features; the early time-frequency information with the fault features is input into the fault diagnosis model for training to obtain the trained early fault diagnosis model.

7. An electronic device, characterized in that: The electronic device comprises a processor and a memory, wherein the memory stores computer instructions. When the computer instructions are executed by the processor, the electronic device executes the method for diagnosing early faults of the davit according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the davit early fault diagnosis method according to any one of claims 1 to 5.

Images