Mechanical fault diagnosis method and device, equipment and storage medium

Abstract

The invention discloses a mechanical fault diagnosis method and device, equipment and a storage medium, and relates to the technical field of mechanical equipment state monitoring, and the method comprises the steps: carrying out the signal collection of target rotating equipment, so as to obtain a target vibration signal, and carrying out the encoding of the target vibration signal through an original signal encoder, obtaining a feature vector corresponding to the target vibration signal; obtaining fault diagnosis information corresponding to the target vibration signal from a preset diagnosis knowledge base according to the feature vector, and generating a spectrum analysis conclusion corresponding to the target vibration signal by using a target large language model and the target vibration signal; generating a fault diagnosis analysis conclusion corresponding to the target rotating equipment by using the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model; wherein the target large language model is a model obtained by training historical fault diagnosis information based on different types of rotating equipment. By using the large language model and the diagnosis knowledge base, fault diagnosis of multiple devices can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical equipment status monitoring, and in particular to a mechanical fault diagnosis method, device, equipment and storage medium. Background Art

[0002] Mechanical operation status monitoring is an essential part in the manufacturing process using mechanical devices, and is of great significance for understanding the operating status of mechanical parts.

[0003] At present, the most common and accurate method for monitoring the operating status and fault diagnosis of rotating machinery equipment is to rely on diagnostic analysts to use time-frequency domain analysis tools to build mechanism models, observe changes from the spectrum waveform, infer possible faults, and give relevant diagnostic analysis conclusions. The mechanism model analysis method has a relatively high accuracy rate, but is also highly complex. The equipment that a model focuses on is relatively single, and it is difficult to migrate to other equipment. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a mechanical fault diagnosis method, device, equipment and storage medium, which can realize fault diagnosis of multiple devices by using a large language model and a diagnostic knowledge base. The specific scheme is as follows:

[0005] In a first aspect, the present application provides a mechanical fault diagnosis method, comprising:

[0006] When the target rotating device is in a running state, signal acquisition is performed on the target rotating device to obtain a target vibration signal corresponding to the target rotating device, and the target vibration signal is encoded using a target signal encoder to obtain a feature vector corresponding to the target vibration signal;

[0007] Acquiring fault diagnosis information corresponding to the target vibration signal from a preset diagnosis knowledge base according to the feature vector, and generating a spectrum analysis conclusion corresponding to the target vibration signal using a target large language model and the target vibration signal;

[0008] The fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model are used to generate a fault diagnosis analysis conclusion corresponding to the target rotating equipment; wherein the target large language model is a model trained based on historical fault diagnosis information of different types of rotating equipment.

[0009] Optionally, before collecting signals from the target rotating device when the target rotating device is in the running state, the method further includes:

[0010] Acquire historical fault diagnosis information corresponding to different rotating devices, and split the historical fault diagnosis information into different text blocks;

[0011] Each of the text blocks is encoded using a preset text encoder, and the historical fault diagnosis information is stored in the preset diagnosis knowledge base in the form of text block encoding.

[0012] Optionally, before collecting signals from the target rotating device when the target rotating device is in the running state, the method further includes:

[0013] Respectively obtain the historical vibration signal corresponding to each rotating device and the historical spectrum analysis conclusion corresponding to each historical vibration signal, and respectively obtain the first historical feature vector corresponding to each historical spectrum analysis conclusion according to each historical spectrum analysis conclusion;

[0014] Each of the historical vibration signals is segmented into a preset length to obtain a second historical feature vector corresponding to each of the historical vibration signals, and an original signal encoder is trained using each of the first historical feature vectors and each of the second historical feature vectors to obtain the target signal encoder.

[0015] Optionally, respectively acquiring the historical vibration signals corresponding to the rotating devices includes:

[0016] Setting a reference sampling frequency, and collecting signals from each of the rotating devices according to the reference sampling frequency to obtain an initial vibration signal corresponding to each of the rotating devices;

[0017] The initial vibration signal is interpolated and filtered to obtain the historical vibration signal corresponding to each of the rotating devices.

[0018] Optionally, the using each of the first historical feature vectors and each of the second historical feature vectors to train an original signal encoder to obtain the target signal encoder includes:

[0019] The original signal encoder is trained according to the cosine similarity between each of the first historical feature vectors and each of the second historical feature vectors to obtain the target signal encoder.

[0020] Optionally, the training the original signal encoder according to the cosine similarity between each of the first historical feature vectors and each of the second historical feature vectors to obtain the target signal encoder includes:

[0021] The original signal encoder is trained according to the cosine similarity between each of the first historical feature vectors and each of the second historical feature vectors to optimize a preset objective function in the original signal encoder to obtain the target signal encoder.

[0022] Optionally, the using the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model to generate a fault diagnosis analysis conclusion corresponding to the target rotating device includes:

[0023] Generate a target prompt word based on the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the large language model prompt word mechanism;

[0024] The target prompt word and the target large language model are used to generate the fault diagnosis analysis conclusion corresponding to the target rotating equipment.

[0025] In a second aspect, the present application provides a mechanical fault diagnosis device, comprising:

[0026] A feature vector acquisition module, used for collecting signals from the target rotating device when the target rotating device is in operation to obtain a target vibration signal corresponding to the target rotating device, and encoding the target vibration signal using a target signal encoder to obtain a feature vector corresponding to the target vibration signal;

[0027] A diagnostic information acquisition module, used to acquire fault diagnostic information corresponding to the target vibration signal from a preset diagnostic knowledge base according to the feature vector, and generate a spectrum analysis conclusion corresponding to the target vibration signal using a target large language model and the target vibration signal;

[0028] An analysis conclusion acquisition module is used to generate a fault diagnosis analysis conclusion corresponding to the target rotating equipment by using the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model; wherein the target large language model is a model trained based on historical fault diagnosis information of different types of rotating equipment.

[0029] In a third aspect, the present application provides an electronic device, including:

[0030] Memory, used to store computer programs;

[0031] A processor is used to execute the computer program to implement the aforementioned mechanical fault diagnosis method.

[0032] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, wherein the computer program implements the aforementioned mechanical fault diagnosis method when executed by a processor.

[0033] In the present application, first, signal acquisition is performed on the target rotating device when the target rotating device is in operation to obtain a target vibration signal corresponding to the target rotating device, and the target vibration signal is encoded using a target signal encoder to obtain a feature vector corresponding to the target vibration signal, and then fault diagnosis information corresponding to the target vibration signal is obtained from a preset diagnostic knowledge base according to the feature vector, and a target large language model and the target vibration signal are used to generate a spectrum analysis conclusion corresponding to the target vibration signal, and finally, the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model are used to generate a fault diagnosis analysis conclusion corresponding to the target rotating device; wherein, the target large language model is a model trained based on historical fault diagnosis information of different types of rotating equipment. It can be seen that the present application obtains the vibration signal corresponding to the rotating equipment, and uses a large language model that can output fault diagnosis analysis conclusions for different types of rotating equipment to process the fault diagnosis information and spectrum analysis conclusions corresponding to the vibration signal, thereby generating a fault diagnosis analysis conclusion corresponding to the target rotating equipment, so that the fault analysis of the rotating equipment has the ability to generalize, avoiding the problem that the mechanism model analysis method is used, which results in the mechanism model being able to only perform diagnosis and analysis on the current equipment and is difficult to migrate to other equipment for fault diagnosis and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0035] Figure 1 A flow chart of a mechanical fault diagnosis method provided in this application;

[0036] Figure 2 A schematic diagram of a specific mechanical fault diagnosis method provided in this application;

[0037] Figure 3 A flow chart of a method for generating characteristic information provided in this application;

[0038] Figure 4 A schematic diagram of a comparative learning encoding method provided for this application;

[0039] Figure 5 A data schematic diagram of a fine-tuned large model provided for this application;

[0040] Figure 6 A schematic diagram of a signal time-frequency analysis comparison result provided in this application;

[0041] Figure 7 A schematic diagram of the structure of a mechanical fault diagnosis device provided in this application;

[0042] Figure 8 A structural diagram of an electronic device provided for this application. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] At present, the methods related to mechanical fault diagnosis focus on a relatively single device, and it is difficult to migrate the fault analysis model for a certain mechanical device to other devices. To this end, the present application provides a mechanical fault diagnosis method, which processes the fault diagnosis information and spectrum analysis conclusions corresponding to the vibration signal by utilizing a large language model that can output fault diagnosis analysis conclusions for different types of rotating equipment, thereby making the fault analysis of rotating equipment generalizable.

[0045] See also Figure 1 As shown, an embodiment of the present invention discloses a mechanical fault diagnosis method, comprising:

[0046] Step S11, when the target rotating device is in operation, signal acquisition is performed on the target rotating device to obtain a target vibration signal corresponding to the target rotating device, and the target vibration signal is encoded using a target signal encoder to obtain a feature vector corresponding to the target vibration signal.

[0047] In this embodiment, Figure 2 As shown, firstly, it is necessary to obtain the vibration signal of the target rotating equipment when the equipment is in operation;

[0048] Before collecting signals from rotating equipment in operation, the method also includes: obtaining historical fault diagnosis information corresponding to different rotating equipment, and splitting the historical fault diagnosis information into different text blocks; encoding each text block using a preset text encoder, and storing the historical fault diagnosis information in the form of text block encoding in a preset diagnostic knowledge base; specifically, collecting and organizing fault diagnosis-related documents and historical diagnostic reports, that is, historical fault diagnosis information, and splitting it into multiple text blocks, using a pre-trained text encoder to encode the text blocks of the above historical fault diagnosis information into D-dimensional vectors, constructing an index, and finally storing it in a vector database. The construction of the diagnostic knowledge base is completed through the above process; when in use, the knowledge base can be retrieved according to the characteristic vector corresponding to the vibration signal, that is, the encoding vector of the signal, to obtain relevant historical fault diagnosis information.

[0049] It should be noted that the vibration signal collected in this embodiment , the signal length is t, where Indicates The vibration signal obtained by sampling a device is defined as:

[0050] ;

[0051] The time-frequency analysis description of the current signal is defined as , the vibration signal and the time-frequency analysis description constitute a sample pair, and the definition of the sample pair is as follows, where is the device number,

[0052] ;

[0053] A contrastive learning model is constructed based on the sample pairs to train the target signal encoder. The contrastive learning model is shown below:

[0054] ;

[0055] in, Represents the original signal encoder, Represents a text encoder, Represents the encoded feature vector.

[0056] In this embodiment, before performing signal acquisition on the rotating equipment in operation, it also includes: respectively acquiring the historical vibration signal corresponding to each rotating equipment and the historical spectrum analysis conclusion corresponding to each of the historical vibration signals, and respectively acquiring the first historical feature vector corresponding to each of the historical spectrum analysis conclusions according to each of the historical spectrum analysis conclusions; dividing each of the historical vibration signals by a preset length to obtain the second historical feature vectors corresponding to each of the historical vibration signals, and using each of the first historical feature vectors and each of the second historical feature vectors to train the original signal encoder to obtain the target signal encoder; the above process of acquiring the historical vibration signal corresponding to the rotating equipment may specifically include: setting a reference sampling frequency, performing signal acquisition on each of the rotating equipment according to the reference sampling frequency to obtain the initial vibration signal corresponding to each of the rotating equipment; interpolating and filtering the initial vibration signal to obtain the historical vibration signal corresponding to each of the rotating equipment, the above process is to pre-set the reference sampling frequency, and to interpolate and filter the vibration signal data at different sampling frequencies to ensure time alignment, and remove the mean value. The above process of acquiring the second historical feature vector is as follows: Figure 3 As shown, it includes constructing a signal feature mapping network (which is mainly the Transformer encoder part, that is, the original signal encoder), signal slicing, feature extraction, and projection embedding modules. Specifically, the historical vibration signal, that is, the original signal, is divided into a set of multiple sub-segments. Assuming that the sliding window size is w and the sliding step size is w⁄2, the original signal is split into m sub-signal segments of length w, where the calculation formula of m is as follows,

[0057] ;

[0058] Then, the absolute position coding is used to add the order information to the m sub-signal segments of length w. The specific formula is as follows:

[0059] ;

[0060] ;

[0061] Where w represents the length of the sub-signal segment, Indicates the order of the fragments.

[0062] Based on the stacking of 6-layer 8-head attention modules, the time-frequency domain signs of the signal data are extracted, and linear projection is used to map them to a D-dimensional vector space representation, that is, the second historical feature vector corresponding to the historical vibration signal is obtained, and it is mapped to a D-dimensional vector space representation. It should be noted that multiple historical vibration signals can be input into the original signal encoder, and each historical vibration signal will be converted into a corresponding second historical feature vector.

[0063] In this embodiment, the operation of obtaining the historical first feature vector corresponding to the historical spectrum analysis conclusion is implemented by a text encoder: first, a pre-trained embedding model of the signal description text is constructed, that is, a pre-trained text encoder. For example, a BGE (BAAI General Embedding, a vector model)-like Chinese sentence encoder model can be used to project the input signal description text, that is, the historical spectrum analysis conclusion, into a D-dimensional vector space representation, that is, to obtain the first historical feature vector; since a pre-trained model is used, the parameters of the encoder are not updated during training.

[0064] The above process of training the original signal encoder using the first historical feature vectors and the second historical feature vectors specifically includes: training the original signal encoder according to the cosine similarity between the first historical feature vectors and the second historical feature vectors to obtain the target signal encoder. It should be noted that the correlation between the same sample pairs is large, and the correlation between different sample pairs is small, that is, the cosine similarity between the same sample pairs is large, and the cosine similarity between different sample pairs is small. The original signal encoder can be trained based on this characteristic; specifically, Figure 4 As shown, the first historical feature vector and the second historical feature vector are projected into the same feature space, and in the same D-dimensional vector space, according to the cosine similarity between each first historical feature vector and each second historical feature vector, the process of training the original signal encoder specifically includes: according to the cosine similarity between each first historical feature vector and each second historical feature vector, the original signal encoder is trained to optimize the preset objective function in the original signal encoder to obtain the target signal encoder; the above process is to update the model parameters of the original signal encoding by optimizing the objective function, so as to obtain the target signal encoder, wherein the above objective function is as follows,

[0065] ;

[0066] in, Indicates The signal and The cosine similarity between the historical spectrum analysis conclusions, represents the first historical feature vector corresponding to the i-th historical spectrum analysis conclusion, Represents the first historical feature vector corresponding to the j-th signal.

[0067] By converting signals into semantic space representation, data analysis support is provided for retrieving knowledge bases and generating signal descriptions, thereby providing more comprehensive and reliable information support for equipment status assessment and improving the accuracy of equipment status fault diagnosis.

[0068] Step S12: acquiring fault diagnosis information corresponding to the target vibration signal from a preset diagnosis knowledge base according to the feature vector, and generating a spectrum analysis conclusion corresponding to the target vibration signal using a target large language model and the target vibration signal.

[0069] In this embodiment, the feature vector corresponding to the target vibration signal can be used to retrieve fault diagnosis information corresponding to the target vibration signal from the diagnostic knowledge base. It can be understood that the diagnostic knowledge base stores fault diagnosis information corresponding to multiple vibration devices and historical vibration signals, which also includes fault diagnosis information corresponding to the target vibration signal. In this embodiment, before using the target large language model to output the spectrum analysis conclusion, it is also necessary to use LoRA (Long Range Radio, a long-distance communication technology) technology to fine-tune the Llama pre-trained model (a large language model), that is, the target large language model; it should be noted that the original Llama model is a text-to-text generation model that uses the Transformer decoder architecture and uses a causal attention mechanism to generate text; in addition, Figure 5 As shown in FIG. 1 , the feature vector of the signal encoder, i.e., the first historical feature vector, and the encoding vector of the signal description, i.e., the second historical feature vector, can be combined as the input of the fine-tuned Llama model. The fine-tuned model can analyze the following from the signal: Figure 6 The change of the time-frequency trend shown in FIG. 1 is shown. By inputting the vibration signal, the target large language model can output the time-frequency analysis conclusion corresponding to the vibration signal. By using the target large language model obtained after training to output the time-frequency analysis conclusion corresponding to the vibration signal, the accuracy of the time-frequency analysis conclusion is guaranteed.

[0070] Step S13, using the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model to generate a fault diagnosis analysis conclusion corresponding to the target rotating equipment; wherein the target large language model is a model trained based on historical fault diagnosis information of different types of rotating equipment.

[0071] In this embodiment, the process of using the target large language model and generating the fault diagnosis analysis conclusion corresponding to the target rotating equipment according to the fault diagnosis information and spectrum analysis conclusion corresponding to the target vibration signal can specifically include: generating the target prompt word based on the fault diagnosis information, spectrum analysis conclusion and large language model prompt word mechanism corresponding to the target vibration signal; generating the fault diagnosis analysis conclusion corresponding to the target rotating equipment by using the target prompt word and the target large language model; that is, constructing the prompt word based on the retrieved fault diagnosis information and the time-frequency analysis conclusion of the vibration signal by the large model, inputting the large language model, integrating the information by the large language model, and outputting the fault diagnosis analysis conclusion. This embodiment utilizes the prompt word mechanism of the large language model, constructs the prompt word by using the fault diagnosis information and spectrum analysis conclusion corresponding to the vibration signal, and outputs the fault diagnosis analysis conclusion according to the prompt word, thereby ensuring the accuracy of the fault diagnosis reasoning of the large language model.

[0072] It can be seen that the present application retrieves the diagnostic knowledge base by the feature vector corresponding to the target vibration signal, and combines the retrieved fault diagnosis analysis conclusion as a prompt word. The large model trained based on the historical fault diagnosis information of different types of rotating equipment completes the fault diagnosis analysis and outputs the diagnosis conclusion, so that the fault analysis of rotating equipment has the ability to generalize, avoiding the problem that the mechanism model can only perform diagnostic analysis on the current equipment due to the use of the mechanism model analysis method, and it is difficult to migrate to other equipment for fault diagnosis analysis; by converting the signal into a semantic space representation, it provides data analysis support for retrieving the knowledge base and generating signal descriptions, thereby providing more comprehensive and reliable information support for equipment status assessment and improving the accuracy of equipment status fault diagnosis.

[0073] See also Figure 7 As shown, the embodiment of the present application discloses a mechanical fault diagnosis device, comprising:

[0074] A feature vector acquisition module 11 is used to collect signals from the target rotating device when the target rotating device is in operation to obtain a target vibration signal corresponding to the target rotating device, and encode the target vibration signal using a target signal encoder to obtain a feature vector corresponding to the target vibration signal;

[0075] A diagnostic information acquisition module 12, configured to acquire fault diagnostic information corresponding to the target vibration signal from a preset diagnostic knowledge base according to the feature vector, and generate a spectrum analysis conclusion corresponding to the target vibration signal using a target large language model and the target vibration signal;

[0076] The analysis conclusion acquisition module 13 is used to generate a fault diagnosis analysis conclusion corresponding to the target rotating equipment by using the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model; wherein the target large language model is a model trained based on historical fault diagnosis information of different types of rotating equipment.

[0077] It can be seen that the present application obtains the vibration signal corresponding to the rotating equipment, and uses a large language model that can output fault diagnosis analysis conclusions for different types of rotating equipment to process the fault diagnosis information and spectrum analysis conclusions corresponding to the vibration signal, thereby generating a fault diagnosis analysis conclusion corresponding to the target rotating equipment, so that the fault analysis of the rotating equipment has the ability to generalize, avoiding the problem that the mechanism model analysis method is used, which results in the mechanism model being able to only perform diagnosis and analysis on the current equipment and is difficult to migrate to other equipment for fault diagnosis and analysis.

[0078] In some specific implementations, the feature vector acquisition module 11 further includes:

[0079] An information splitting unit, used for obtaining historical fault diagnosis information corresponding to different rotating equipment, and splitting the historical fault diagnosis information into different text blocks;

[0080] The information storage unit is used to encode each of the text blocks using a preset text encoder, and store the historical fault diagnosis information in the form of text block encoding into the preset diagnosis knowledge base.

[0081] In some specific implementations, the feature vector acquisition module 11 further includes:

[0082] A feature vector acquisition submodule, used to respectively acquire the historical vibration signal corresponding to each of the rotating devices, the historical spectrum analysis conclusion corresponding to each of the historical vibration signals, and to respectively acquire the first historical feature vector corresponding to each of the historical spectrum analysis conclusions according to each of the historical spectrum analysis conclusions;

[0083] The signal encoder training submodule is used to divide each of the historical vibration signals into preset lengths to obtain a second historical feature vector corresponding to each of the historical vibration signals, and to train the original signal encoder using each of the first historical feature vectors and each of the second historical feature vectors to obtain the target signal encoder.

[0084] In some specific implementations, the feature vector acquisition submodule specifically includes:

[0085] An initial signal acquisition unit, used to set a reference sampling frequency, and collect signals from each of the rotating devices according to the reference sampling frequency to obtain an initial vibration signal corresponding to each of the rotating devices;

[0086] The historical signal acquisition unit is used to perform interpolation and filtering processing on the initial vibration signal to obtain the historical vibration signal corresponding to each of the rotating devices.

[0087] In some specific implementations, the signal encoder training submodule specifically includes:

[0088] A signal encoder training unit is used to train the original signal encoder according to the cosine similarity between each of the first historical feature vectors and each of the second historical feature vectors to obtain the target signal encoder.

[0089] In some specific embodiments, the signal encoder training unit is specifically used to train the original signal encoder according to the cosine similarity between each of the first historical feature vectors and each of the second historical feature vectors to optimize a preset objective function in the original signal encoder to obtain the target signal encoder.

[0090] In some specific implementations, the analysis conclusion acquisition module 13 specifically includes:

[0091] A prompt word generating unit, configured to generate a target prompt word based on the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and a large language model prompt word mechanism;

[0092] The analysis conclusion generating unit is used to generate the fault diagnosis analysis conclusion corresponding to the target rotating equipment by using the target prompt word and the target large language model.

[0093] Furthermore, the present application also discloses an electronic device. Figure 8 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.

[0094] Figure 8 A schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the mechanical fault diagnosis method disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0095] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0096] In addition, the memory 22 as a carrier for storing resources may be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage method may be temporary storage or permanent storage.

[0097] The operating system 221 is used to manage and control the hardware devices on the electronic device 20 and the computer program 222, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the mechanical fault diagnosis method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.

[0098] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the aforementioned disclosed mechanical fault diagnosis method. The specific steps of the method can refer to the corresponding contents disclosed in the aforementioned embodiments, and will not be repeated here.

[0099] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0100] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0101] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0102] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0103] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A mechanical fault diagnosis method, characterized in that: include: When the target rotating device is in a running state, signal acquisition is performed on the target rotating device to obtain a target vibration signal corresponding to the target rotating device, and the target vibration signal is encoded using a target signal encoder to obtain a feature vector corresponding to the target vibration signal; Acquiring fault diagnosis information corresponding to the target vibration signal from a preset diagnosis knowledge base according to the feature vector, and generating a spectrum analysis conclusion corresponding to the target vibration signal using a target large language model and the target vibration signal; The fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model are used to generate a fault diagnosis analysis conclusion corresponding to the target rotating equipment; wherein the target large language model is a model trained based on historical fault diagnosis information of different types of rotating equipment.

2. The mechanical fault diagnosis method according to claim 1, characterized in that: Before collecting signals from the target rotating device when the target rotating device is in the running state, the method further includes: Acquire historical fault diagnosis information corresponding to different rotating devices, and split the historical fault diagnosis information into different text blocks; Each of the text blocks is encoded using a preset text encoder, and the historical fault diagnosis information is stored in the preset diagnosis knowledge base in the form of text block encoding.

3. The mechanical fault diagnosis method according to claim 2, characterized in that: Before collecting signals from the target rotating device when the target rotating device is in the running state, the method further includes: Respectively obtain the historical vibration signal corresponding to each rotating device and the historical spectrum analysis conclusion corresponding to each historical vibration signal, and respectively obtain the first historical feature vector corresponding to each historical spectrum analysis conclusion according to each historical spectrum analysis conclusion; Each of the historical vibration signals is segmented into a preset length to obtain a second historical feature vector corresponding to each of the historical vibration signals, and an original signal encoder is trained using each of the first historical feature vectors and each of the second historical feature vectors to obtain the target signal encoder.

4. The mechanical fault diagnosis method according to claim 3, characterized in that: The respectively obtaining of the historical vibration signals corresponding to the rotating devices comprises: Setting a reference sampling frequency, and collecting signals from each of the rotating devices according to the reference sampling frequency to obtain an initial vibration signal corresponding to each of the rotating devices; The initial vibration signal is interpolated and filtered to obtain the historical vibration signal corresponding to each of the rotating devices.

5. The mechanical fault diagnosis method according to claim 4, characterized in that: The using each of the first historical feature vectors and each of the second historical feature vectors to train the original signal encoder to obtain the target signal encoder includes: The original signal encoder is trained according to the cosine similarity between each of the first historical feature vectors and each of the second historical feature vectors to obtain the target signal encoder.

6. The mechanical fault diagnosis method according to claim 5, characterized in that: The training of the original signal encoder according to the cosine similarity between each of the first historical feature vectors and each of the second historical feature vectors to obtain the target signal encoder includes: The original signal encoder is trained according to the cosine similarity between each of the first historical feature vectors and each of the second historical feature vectors to optimize a preset objective function in the original signal encoder to obtain the target signal encoder.

7. The mechanical fault diagnosis method according to any one of claims 1 to 6, characterized in that: The method of using the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model to generate a fault diagnosis analysis conclusion corresponding to the target rotating equipment includes: Generate a target prompt word based on the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the large language model prompt word mechanism; The target prompt word and the target large language model are used to generate the fault diagnosis analysis conclusion corresponding to the target rotating equipment.

8. A mechanical fault diagnosis device, characterized in that: include: A feature vector acquisition module is used to collect signals from the target rotating device when the target rotating device is in operation to obtain a target vibration signal corresponding to the target rotating device, and encode the target vibration signal using a target signal encoder to obtain a feature vector corresponding to the target vibration signal; A diagnostic information acquisition module, used to acquire fault diagnostic information corresponding to the target vibration signal from a preset diagnostic knowledge base according to the feature vector, and generate a spectrum analysis conclusion corresponding to the target vibration signal using a target large language model and the target vibration signal; An analysis conclusion acquisition module is used to generate a fault diagnosis analysis conclusion corresponding to the target rotating equipment by using the fault diagnosis information corresponding to the target vibration signal, the spectrum analysis conclusion and the target large language model; wherein the target large language model is a model trained based on historical fault diagnosis information of different types of rotating equipment.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the mechanical fault diagnosis method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, which, when executed by a processor, implements the mechanical fault diagnosis method according to any one of claims 1 to 7.

Images