Hierarchical fault identification method and system for rotating machinery based on representative patterns

Abstract

The present invention belongs to the technical field of rotating machinery fault diagnosis, and discloses a rotating machinery hierarchical fault identification method and system based on representative patterns, including obtaining vibration signals and preprocessing the vibration signals; performing dynamic modeling on the preprocessed vibration signal data of each category to form a large-scale candidate pattern library; selecting representative patterns at multiple levels based on a representative selection algorithm of dynamics to form a hierarchical representative pattern library; designing dynamic pattern similarity measurement methods of different resolutions based on the hierarchical representative pattern library according to the difficulty of different identification tasks, and then constructing a hierarchical fault identification framework; inputting the vibration signal to be diagnosed into the hierarchical fault identification framework for layer-by-layer dynamic matching to obtain a fault identification result. The present invention realizes the step-by-step identification of the source, location and severity of the rotating machinery vibration fault from coarse to fine, and effectively improves the accuracy, robustness and computational efficiency of the detection model.

Description

Technical Field

[0001] The present invention relates to the technical field of rotating machinery fault diagnosis, and in particular to a rotating machinery hierarchical fault identification method and system based on representative patterns. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Rotating machinery plays an important role in key industries such as energy, chemical industry, military industry, and aerospace. With the continuous development of industrial informatization and intelligence, the reliability requirements for rotating machinery are increasing. Bearings and gears are the core components of rotating machinery, and their performance degradation has a significant impact on the reliability and stability of the entire system. Therefore, the development of effective rotating machinery fault diagnosis methods is of great significance to ensure the healthy operation of key equipment.

[0004] Rotating machinery fault diagnosis is essentially a pattern recognition problem, which includes two key steps: feature extraction and fault classification. Since vibration signals are easy to obtain and can be measured accurately, the method of rotating machinery fault identification based on vibration signal analysis has always been the focus of attention in the industry. Early fault identification methods based on signal processing and shallow machine learning generally express the differences between different operating states of the equipment by extracting the time domain and frequency domain features of vibration data, and then derive the fault category through machine learning classifiers such as random forests and support vector machines. This type of method is simple and efficient, but the diagnostic performance is heavily dependent on the quality of manual selection and extraction of data features. In recent years, with the development of artificial intelligence technology, deep learning methods represented by neural networks have been widely used in rotating machinery fault identification. With the deep network structure, deep learning can automatically extract data features to obtain high-level feature expressions, and then establish a complex nonlinear relationship between features and fault categories, which greatly improves the accuracy of fault identification.

[0005] Although deep learning has made significant progress in the field of rotating machinery fault diagnosis, there are still many problems that need to be solved in practical applications:

[0006] 1. The high performance of deep learning models depends on large-scale data training, but in industrial environments, the dramatic increase in data volume often leads to a huge consumption of computing and storage resources, which in turn affects the real-time recognition effect. How to reduce data redundancy and improve recognition and diagnosis efficiency is a key problem.

[0007] 2. The internal structure of rotating machinery is complex, and key components such as bearings and gears are prone to failure. Traditional methods usually install multiple sensors near each component to monitor its health status separately. However, in actual engineering, due to spatial accessibility, the installation location and number of sensors are strictly restricted. How to identify multiple faults based on data from a single measurement point has become a technical bottleneck that needs to be broken through.

[0008] 3. Existing methods perform well in single recognition tasks, but their fault recognition capabilities are poor in complex scenarios and complex tasks. How to further improve the accuracy and robustness of the recognition model remains a major challenge. Summary of the invention

[0009] In order to solve the above problems, the present invention proposes a hierarchical fault identification method and system for rotating machinery based on representative patterns, which uses a deterministic learning algorithm to perform dynamic modeling on the vibration signal of a single measuring point of the rotating machinery, and selects representative dynamic patterns and integrates The dynamic similarity measurement of the order partial derivative information is used to build a hierarchical fast recognition architecture, which can realize the step-by-step recognition of the source, location and severity of the rotating machinery vibration fault from coarse to fine, effectively improving the accuracy, robustness and computational efficiency of the detection model.

[0010] In order to achieve the above object, the present invention adopts the following technical solution:

[0011] In a first aspect, the present invention provides a method for hierarchical fault identification of rotating machinery based on representative patterns, comprising the following steps:

[0012] Acquire vibration signal data of the rotating machinery and pre-process the vibration signal data;

[0013] Perform dynamic modeling on each category of preprocessed vibration signal data to form a large-scale candidate pattern library;

[0014] A representative selection algorithm based on dynamics selects representative patterns at multiple levels to form a hierarchical representative pattern library;

[0015] According to the difficulty of different recognition tasks, dynamic pattern similarity measurement methods with different resolutions are designed based on the hierarchical representative pattern library, and then a hierarchical fault recognition framework is constructed;

[0016] The vibration signal to be diagnosed is input into the hierarchical fault identification framework for layer-by-layer dynamic matching to obtain the fault identification result.

[0017] As an optional implementation, the vibration signal data is preprocessed, specifically:

[0018] First, the single-channel original vibration signal obtained by the sensor is amplified to two channels through the observer, and then the vibration signal is filtered and normalized in turn, and finally the vibration signal is labeled.

[0019] As an optional implementation, a deterministic learning algorithm is used to perform dynamic modeling on each category of preprocessed vibration signal data to obtain a large-scale candidate pattern library composed of neural network weights.

[0020] As an optional implementation, for different dynamic patterns in the large-scale candidate pattern library, a pattern with dynamic representativeness is selected through a representative selection algorithm, specifically:

[0021] First, the two farthest patterns are selected and included in the representative subset by calculating the dynamic distance. Then, the dynamic distances between the remaining patterns and the selected patterns are calculated respectively, and a group of patterns with the shortest distances are selected. The pattern with the largest distance is further selected from them and included in the representative subset. This process is repeated until the number of patterns in the representative subset reaches the predetermined target.

[0022] As an optional implementation, a group of the most dynamically representative patterns are selected at the three levels of fault source, fault location, and fault severity to form a hierarchical representative pattern library, which includes a representative fault source pattern library, a representative fault location pattern library, and a representative severity pattern library.

[0023] As an optional implementation, a hierarchical fault identification framework is constructed based on a hierarchical representative pattern library and a dynamic pattern similarity measurement method with different resolutions. The hierarchical fault identification framework includes a fault source identification layer, a fault area location layer, and a severity assessment layer, specifically:

[0024] Construct a fault source identification layer based on the similarity measurement of representative fault source pattern library and system dynamics differences;

[0025] The fault area location layer is constructed based on the representative fault location pattern library and the similarity measurement of the fusion system dynamics difference and the dynamics first-order partial derivative difference;

[0026] A severity assessment layer is constructed based on a representative severity pattern library and a similarity measure that integrates the system dynamics difference and the dynamics second-order partial derivative difference.

[0027] In a second aspect, the present invention provides a hierarchical fault identification system for rotating machinery based on representative patterns, comprising:

[0028] The data acquisition module is configured to: acquire vibration signal data of the rotating machinery and pre-process the vibration signal data;

[0029] The large-scale candidate pattern library building module is configured to: perform dynamic modeling on the pre-processed vibration signal data of each category to form a large-scale candidate pattern library;

[0030] The representative pattern library building module is configured as follows: a dynamics-based representative selection algorithm to select representative patterns at multiple levels to form a hierarchical representative pattern library;

[0031] The hierarchical fault identification framework building module is configured as follows: according to the difficulty of different identification tasks, dynamic pattern similarity measurement methods with different resolutions are designed based on the hierarchical representative pattern library, and then a hierarchical fault identification framework is constructed;

[0032] The fault identification module is configured to: input the vibration signal to be diagnosed into a hierarchical fault identification framework for layer-by-layer dynamic matching to obtain a fault identification result.

[0033] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in the first aspect is performed.

[0034] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0035] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method described in the first aspect.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. The representative pattern selection method proposed in the present invention significantly reduces data redundancy and storage resource usage by selecting the most dynamically representative dynamic pattern to replace the entire pattern library. Compared with the existing dynamic pattern recognition method that relies on a complete pattern library, the present invention can achieve accurate fault recognition by only using the representative pattern library, significantly reducing the demand for computing resources and greatly improving computing efficiency.

[0038] 2. The present invention uses a deterministic learning algorithm to extract dynamic features that are more sensitive than traditional features from single-point signals with limited information, showing significant discriminability. The hierarchical fault identification framework proposed on this basis decomposes the complex multi-fault classification task from coarse to fine, and by measuring the dynamic similarity at different levels, it can achieve progressive and accurate identification of the fault source, fault area and fault severity on single-point data.

[0039] 3. The fusion system dynamics and The similarity measurement method of partial derivative information reveals the deeper dynamic differences between patterns. For classification tasks of different difficulty levels, by adjusting the order of partial derivatives, the degree of refinement of similarity measurement can be flexibly changed, effectively improving the accuracy and robustness of recognition without sacrificing computational efficiency.

[0040] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] 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.

[0042] Figure 1 It is a flowchart of a rotating machinery hierarchical fault identification method based on representative patterns according to Embodiment 1 of the present invention;

[0043] Figure 2 It is a hierarchical structure diagram of the data set of the present invention;

[0044] Figure 3 A schematic diagram is selected for representative modes of the present invention;

[0045] Figure 4 It is a hierarchical recognition framework diagram of the present invention;

[0046] Figure 5 This is a structural block diagram of a rotating machinery hierarchical fault identification system based on representative patterns according to Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0047] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0048] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0049] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0050] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0051] Terminology explanation:

[0052] Deterministic learning: Deterministic learning is a new method for machine learning in dynamic environments that can accurately and physically interpret the dynamics of nonlinear systems. The core elements of this theory include:

[0053] (1) Using radial basis function (RBF) neural network;

[0054] (2) Using the neural network weight adjustment law based on Lyapunov stability theory;

[0055] (3) Ensure the convergence of neural network weights by satisfying continuous excitation conditions, and achieve accurate modeling of system dynamics;

[0056] (4) The results of dynamic modeling are stored in the constant weights of the neural network.

[0057] Example 1

[0058] like Figure 1 As shown, this embodiment provides a method for hierarchical fault identification of rotating machinery based on representative patterns, comprising the following steps:

[0059] S1. Acquire vibration signal data of a rotating machine and preprocess the vibration signal data;

[0060] S2, perform dynamic modeling on the pre-processed vibration signal data of each category to form a large-scale candidate pattern library;

[0061] S3, a dynamics-based representative selection algorithm, selects representative patterns at multiple levels to form a hierarchical representative pattern library;

[0062] S4. According to the difficulty of different recognition tasks, dynamic pattern similarity measurement methods with different resolutions are designed based on the hierarchical representative pattern library, and then a hierarchical fault recognition framework is constructed;

[0063] S5. Input the vibration signal to be diagnosed into the hierarchical fault identification framework for layer-by-layer dynamic matching to obtain a fault identification result.

[0064] First, S1, obtain vibration signal data of the rotating machinery, and pre-process the vibration signal data.

[0065] Based on a single-point sensor, the vibration signal data of different fault types of the rotating machinery is obtained and the data is preprocessed. The single-point sensor is a single vibration sensor arranged at a certain position of the outer casing or base of the equipment, such as an accelerometer sensor, a displacement sensor, etc.

[0066] Preprocess the vibration signal data, specifically:

[0067] First, the single-channel original vibration signal obtained by the sensor is amplified to two channels through the observer, and then the vibration signal is filtered and normalized in turn, and finally the vibration signal is labeled. The vibration signal must clearly mark the fault source, fault location and fault severity. The fault sources include bearings and gears; the fault locations include bearing inner ring wear, bearing outer ring wear, bearing rolling element wear, gear tooth breakage, gear tooth surface wear, gear tooth root cracks; the fault severity includes mild, moderate and severe. The dataset hierarchy is as follows Figure 2 shown.

[0068] S2. Perform dynamic modeling on the preprocessed vibration signal data of each category to form a large-scale candidate pattern library.

[0069] The deterministic learning algorithm is used to perform dynamic modeling on each category of preprocessed vibration data, and the extracted dynamic information is stored in the form of neural network constant weights to form a large-scale candidate pattern library.

[0070] The method of using a deterministic learning algorithm to perform dynamic modeling on each category of data is specifically to use an RBF neural network to model the intrinsic dynamic characteristics of the vibration signal:

[0071] (1)

[0072] in, is a two-channel vibration signal, is the RBF neural network weight, is the regression vector of the RBF neural network, represents the true intrinsic dynamics function of the vibration signal, is the modeling error.

[0073] Furthermore, by modeling all vibration signals in the dataset separately, a large-scale candidate dynamic pattern library consisting of neural network weights is obtained. ,in Indicates the number of vibration signals.

[0074] S3. Based on the dynamics representative selection algorithm, representative patterns are selected at three levels: fault source, fault location, and fault severity to form a hierarchical representative pattern library. The hierarchical representative pattern library includes a representative fault source pattern library, a representative fault location pattern library, and a representative severity pattern library.

[0075] The representative selection algorithm selects a representative sample subset from a large-scale candidate dynamic pattern library to characterize the original candidate dynamic pattern library. The present invention proposes a method for selecting representative samples from the perspective of nonlinear dynamics, and the specific steps are as follows:

[0076] 1) Calculate the dynamic distance between two dynamic modes:

[0077] (2)

[0078] 2) Select the two dynamic modes with the longest distance ( ) into a representative subset;

[0079] 3) For the remaining ( ) dynamic modes, and calculate their differences with the selected The dynamic distances between the dynamic modes are calculated and the mode with the shortest distance is selected:

[0080] (3)

[0081] 4) Further select the maximum distance from these shortest distance modes The corresponding patterns are included in the representative subset;

[0082] 5) Repeat steps 3)-4) until the number of representative subset patterns reaches a predetermined number.

[0083] Through the above steps, a group of samples with representative dynamics are selected. Figure 3 As shown, these samples are as far away from each other as possible in the dynamic space to ensure that the sample subset can represent the dynamic diversity and coverage of the original dynamic pattern library.

[0084] The construction process of the representative fault source pattern library is to use the representative selection algorithm to screen out the representative patterns of the two types of bearing and gear patterns involved in the fault source level to form a representative fault source pattern library. The number of representative patterns is set to 10% of the sample size of the original pattern library. This is because the difference between bearing faults and gear faults is large, so representative patterns can be sparsely selected.

[0085] The representative fault location pattern library contains two sub-libraries, namely the bearing fault location pattern library and the gear fault location pattern library. The specific construction process of the bearing fault location pattern library is to use the representative selection algorithm to screen out the representative patterns of the three types of patterns, namely the inner ring wear of the bearing, the outer ring wear of the bearing and the rolling element wear of the bearing, so as to form a representative bearing fault location pattern library. The specific construction process of the gear fault location pattern library is to use the representative selection algorithm to screen out the representative patterns of the three types of patterns, namely the gear tooth breakage, the gear tooth surface wear and the gear tooth root crack, so as to form a representative gear fault location pattern library. Since the difference of faults at different positions under the same source is reduced, more representative patterns are needed, so the number of representative patterns is set to 20% of the sample size of the original pattern library.

[0086] The representative fault severity pattern library contains 6 sub-libraries, namely, the bearing inner ring fault severity pattern library, the bearing outer ring fault severity pattern library, the bearing rolling element fault severity pattern library, the gear tooth breakage severity pattern library, the gear tooth surface wear severity pattern library and the gear tooth root crack severity pattern library. Taking the bearing inner ring fault severity pattern library as an example, the specific construction process is to use the representative selection algorithm to screen out the respective representative patterns for the three types of modes, namely, minor faults, moderate faults and severe faults, to form a representative bearing inner ring fault severity pattern library. The construction process of the remaining 5 sub-libraries is similar. Since the difference between faults of different severity at the same source and position is further reduced, the number of representative patterns is set to 50% of the sample size of the original pattern library.

[0087] S4. According to the difficulty of different recognition tasks, dynamic pattern similarity measurement methods with different resolutions are designed based on the hierarchical representative pattern library, and then a hierarchical fault recognition framework is constructed. The hierarchical fault recognition framework includes a fault source identification layer, a fault area location layer, and a severity assessment layer.

[0088] The dynamic pattern similarity measurement method with different resolutions refers to a dynamic pattern recognition method with different resolution capabilities for recognition tasks of different difficulty in order to achieve a balance between computational efficiency and recognition accuracy. A more refined similarity measurement often requires higher computing resources.

[0089] Specifically, the present invention proposes a method that integrates system dynamics and A new similarity measurement method for the partial derivative information of the order dynamics is defined as follows: if the system and system If the dynamics of the two systems are equivalent, then the difference is kept at Under norm In the neighborhood:

[0090] (4)

[0091] in, represents the vector norm, is a system function, represents the difference in the dynamics of the two systems, Represents the system dynamics Difference of partial derivatives.

[0092] The above definition for measuring pattern similarity combines dynamical differences and dynamical partial derivative differences. The partial derivative information describes the rate of change of the system dynamics with respect to the state, which is a deeper expression of dynamical differences.

[0093] Furthermore, for identification tasks of different difficulty, by adjusting the partial derivative order, similarity measurements of different degrees of precision can be performed. For the fault source identification layer, fault area location layer and severity assessment layer, the partial derivative orders used are zero order, first order and second order respectively.

[0094] The system function or The intrinsic dynamics of can be expressed by the neural network model in step (2):

[0095] (5)

[0096] The system function dynamics Partial derivative or This is achieved by introducing the directional derivative along the trajectory:

[0097] (6)

[0098] in, Represents the unit direction vector along the trajectory, at each moment The partial derivative direction of can be obtained by forward difference, that is, ; The system trajectory state components, Indicates the trajectory The direction cosines of the state components at each moment, is the modeling error, is the binomial coefficient, which means Select from the elements The number of combinations of elements is calculated as .

[0099] S5. Input the vibration signal to be diagnosed into the hierarchical fault identification framework for layer-by-layer dynamic matching, and obtain the final fault identification result through the minimum residual principle.

[0100] The hierarchical recognition framework is as follows Figure 4 As shown in the figure, the vibration signal passes through the fault source identification layer, the fault area location layer and the severity assessment layer in sequence to achieve dynamic matching and identification. Specifically:

[0101] First, a set of fault source identification layer dynamic estimators are constructed based on the similarity metric of the representative fault source pattern library and the zero-order partial derivative (i.e., based only on the system dynamics difference):

[0102] (7)

[0103] in, It indicates that the dynamic estimator of the fault source identification layer is The state of the moment, It indicates that the dynamic estimator of the fault source identification layer is Input at the moment, is the dynamic estimator gain of the fault source identification layer, is the sampling rate of the vibration signal, It is a representative fault source pattern library. is the number of patterns in the representative fault source pattern library.

[0104] Furthermore, by inputting the vibration signal to be diagnosed into this group of dynamic estimators, the identification error (residual) of the signal to be diagnosed relative to the representative fault source pattern library can be obtained:

[0105] (8)

[0106] in, It represents the identification residual of the signal to be diagnosed relative to the representative fault source pattern library, It is the discrete form of the system dynamics function of the signal to be diagnosed.

[0107] Furthermore, pattern matching is performed based on the minimum residual principle, and it is considered that the vibration signal to be diagnosed is most similar to the dynamic pattern corresponding to the minimum residual in the pattern library, thereby obtaining the identification result of the fault source identification layer.

[0108] Then, a set of fault area location layer dynamic estimators are constructed based on the representative fault location pattern library and the similarity metric integrating system dynamics and first-order partial derivatives:

[0109] (9)

[0110] in, The fault area location layer dynamic estimator is The state of the moment, The fault area location layer dynamic estimator is Input at the moment, The dynamic estimator gain for fault region location layer, is the first-order partial derivative fusion coefficient of the dynamic estimator of the fault area positioning layer, is the weight of the neural network after the signal to be diagnosed is determined and modeled. is a representative fault location pattern library, is the number of patterns in the representative fault location pattern library. There are two representative fault location pattern libraries: bearing fault location pattern library and gear fault location pattern library. The specific selection of the pattern library depends on the recognition result of the previous layer.

[0111] Furthermore, by inputting the vibration signal to be diagnosed into this group of dynamic estimators, the identification error (residual) of the signal to be diagnosed relative to the representative fault location pattern library can be obtained:

[0112] (10)

[0113] in, It represents the identification residual of the signal to be diagnosed relative to the representative fault location pattern library.

[0114] Furthermore, pattern matching is performed based on the minimum residual principle, and it is considered that the vibration signal to be diagnosed is most similar to the dynamic pattern corresponding to the minimum residual in the pattern library, thereby obtaining the fault area location layer identification result.

[0115] Finally, a set of dynamic estimators for fault severity assessment layer is constructed based on the representative fault severity pattern library and the similarity metric integrating system dynamics and second-order partial derivatives:

[0116] (11)

[0117] in, It represents the fault severity assessment layer dynamic estimator in The state of the moment, It represents the fault severity assessment layer dynamic estimator in Input at the moment, is the fault severity assessment layer dynamic estimator gain, is the first-order partial derivative fusion coefficient of the fault severity assessment layer dynamic estimator, is the second-order partial derivative fusion coefficient of the fault severity assessment layer dynamic estimator, is a representative fault severity pattern library, is the number of patterns in the representative fault severity pattern library. There are six representative fault severity pattern libraries: bearing inner ring fault severity pattern library, bearing outer ring fault severity pattern library, bearing rolling element fault severity pattern library, gear tooth breakage severity pattern library, gear tooth surface wear severity pattern library and gear tooth root crack severity pattern library. The specific pattern library to be selected depends on the recognition result of the previous layer.

[0118] Furthermore, by inputting the vibration signal to be diagnosed into this group of dynamic estimators, the identification error (residual) of the signal to be diagnosed relative to the representative fault severity pattern library can be obtained:

[0119] (12)

[0120] in, It represents the identification residual of the signal to be diagnosed relative to the representative fault severity pattern library.

[0121] Furthermore, pattern matching is performed based on the minimum residual principle, and it is considered that the vibration signal to be diagnosed is most similar to the dynamic pattern corresponding to the minimum residual in the pattern library, thereby obtaining the fault severity assessment layer identification result.

[0122] Example 2

[0123] like Figure 5 As shown, this embodiment provides a rotating machinery hierarchical fault identification system based on representative patterns, including:

[0124] The data acquisition module is configured to: acquire vibration signal data of the rotating machinery and pre-process the vibration signal data;

[0125] The large-scale candidate pattern library building module is configured to: perform dynamic modeling on the pre-processed vibration signal data of each category to form a large-scale candidate pattern library;

[0126] The representative pattern library building module is configured as follows: a dynamics-based representative selection algorithm to select representative patterns at multiple levels to form a hierarchical representative pattern library;

[0127] The hierarchical fault identification framework building module is configured as follows: according to the difficulty of different identification tasks, dynamic pattern similarity measurement methods with different resolutions are designed based on the hierarchical representative pattern library, and then a hierarchical fault identification framework is constructed;

[0128] The fault identification module is configured to: input the vibration signal to be diagnosed into a hierarchical fault identification framework for layer-by-layer dynamic matching to obtain a fault identification result.

[0129] The fault identification module includes a fault source identification layer module, a fault area location layer module and a severity assessment layer module.

[0130] The fault source identification layer module is configured to: dynamically compare the vibration signal to be diagnosed with the representative fault source pattern library, perform dynamic pattern matching based on the similarity measurement of system dynamics differences, and obtain the fault source identification result.

[0131] The fault area location layer module is configured as follows: first, a corresponding representative fault location pattern library is assigned according to the fault source identification result, and then a dynamic comparison is performed between the vibration signal to be diagnosed and the assigned representative fault location pattern library, and a dynamic pattern matching is performed based on a similarity measure based on the fusion system dynamics difference and the dynamics first-order partial derivative difference to obtain the fault location identification result.

[0132] The severity assessment layer module is configured as follows: first, a corresponding representative severity pattern library is assigned according to the fault location identification result, then a dynamic comparison is performed between the vibration signal to be diagnosed and the assigned representative severity pattern library, and a dynamic pattern matching is performed based on a similarity measure based on the fusion system dynamics difference and the dynamics second-order partial derivative difference to obtain the fault severity identification result.

[0133] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer executable instructions.

[0134] In further embodiments, there is also provided:

[0135] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method described in Embodiment 1 is performed. For the sake of brevity, it will not be described in detail here.

[0136] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0137] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0138] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in Example 1 is completed.

[0139] The method in Example 1 can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it is not described in detail here.

[0140] A computer program product includes a computer program, and when the computer program is executed by a processor, the method described in embodiment 1 is implemented.

[0141] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer executable instructions, such as instructions included in a program module, which are executed in a device on a real or virtual processor of the target to perform the process / method as described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided between program modules as needed. Machine executable instructions for program modules can be executed in local or distributed devices. In distributed devices, program modules can be located in local and remote storage media.

[0142] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the computer or other programmable data processing device, causes the function / operation specified in the flow chart and / or block diagram to be implemented. The program code can be executed completely on a computer, partially on a computer, as an independent software package, partially on a computer and partially on a remote computer or completely on a remote computer or server.

[0143] In the context of the present invention, computer program codes or related data may be carried by any appropriate carrier to enable a device, apparatus or processor to perform the various processes and operations described above. Examples of carriers include signals, computer readable media, and the like. Examples of signals may include electrical, optical, radio, acoustic or other forms of propagation signals, such as carrier waves, infrared signals, and the like.

[0144] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can 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.

[0145] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A hierarchical fault identification method for rotating machinery based on representative patterns, characterized in that: The following steps are involved: Acquire vibration signal data of the rotating machinery and pre-process the vibration signal data; Perform dynamic modeling on each category of preprocessed vibration signal data to form a large-scale candidate pattern library; A dynamics-based representative selection algorithm selects representative patterns at multiple levels to form a hierarchical representative pattern library; According to the difficulty of different recognition tasks, dynamic pattern similarity measurement methods with different resolutions are designed based on the hierarchical representative pattern library, and then a hierarchical fault recognition framework is constructed; The vibration signal to be diagnosed is input into the hierarchical fault identification framework for layer-by-layer dynamic matching to obtain the fault identification result; For different dynamic modes in the large-scale candidate mode library, the representative modes are selected through the representative selection algorithm, specifically: First, the two furthest patterns are selected by calculating the dynamic distance and included in the representative subset. Then, the dynamic distances between the remaining patterns and the selected patterns are calculated respectively, and a group of patterns with the shortest distances are selected. The pattern with the largest distance is further selected from them and included in the representative subset. This process is repeated until the number of patterns in the representative subset reaches the predetermined target. Based on the hierarchical representative pattern library and the dynamic pattern similarity measurement method with different resolutions, a hierarchical fault identification framework is constructed. The hierarchical fault identification framework includes a fault source identification layer, a fault area location layer, and a severity assessment layer, specifically: Construct a fault source identification layer based on the similarity measurement of representative fault source pattern library and system dynamics differences; The fault area location layer is constructed based on the representative fault location pattern library and the similarity measurement of the fusion system dynamics difference and the dynamics first-order partial derivative difference; A severity assessment layer is constructed based on a representative severity pattern library and a similarity measure that integrates the system dynamics difference and the dynamics second-order partial derivative difference.

2. The method for hierarchical fault identification of rotating machinery based on representative patterns according to claim 1, characterized in that: Preprocess the vibration signal data, specifically: First, the single-channel original vibration signal obtained by the sensor is amplified to two channels through the observer, and then the vibration signal is filtered and normalized in turn, and finally the vibration signal is labeled.

3. The method for hierarchical fault identification of rotating machinery based on representative patterns according to claim 1, characterized in that: The preprocessed vibration signal data of each category are dynamically modeled using a deterministic learning algorithm to obtain a large-scale candidate pattern library consisting of neural network weights.

4. The method for hierarchical fault identification of rotating machinery based on representative patterns according to claim 1, characterized in that: A group of the most dynamically representative patterns are selected at the three levels of fault source, fault location, and fault severity to form a hierarchical representative pattern library, which includes a representative fault source pattern library, a representative fault location pattern library, and a representative severity pattern library.

5. A rotating machinery hierarchical fault identification system based on representative patterns, adopting a rotating machinery hierarchical fault identification method based on representative patterns as claimed in any one of claims 1 to 4, characterized in that: include: The data acquisition module is configured to: acquire vibration signal data of the rotating machinery and pre-process the vibration signal data; The large-scale candidate pattern library building module is configured to: perform dynamic modeling on the pre-processed vibration signal data of each category to form a large-scale candidate pattern library; The representative pattern library building module is configured as follows: a dynamics-based representative selection algorithm to select representative patterns at multiple levels to form a hierarchical representative pattern library; The hierarchical fault identification framework building module is configured as follows: according to the difficulty of different identification tasks, dynamic pattern similarity measurement methods with different resolutions are designed based on the hierarchical representative pattern library, and then a hierarchical fault identification framework is constructed; The fault identification module is configured to: input the vibration signal to be diagnosed into a hierarchical fault identification framework for layer-by-layer dynamic matching to obtain a fault identification result.

6. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 4 is completed.

7. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 4.

8. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 4.

Images