Equipment real-time fault diagnosis method

Abstract

The invention provides an equipment real-time fault diagnosis method, and belongs to the technical field of equipment operation and maintenance diagnosis, and the method comprises the steps: obtaining real-time multi-source data in an equipment operation state; preprocessing edge end data in the real-time multi-source data to obtain preprocessed data; extracting a spatio-temporal feature vector of the preprocessed data; constructing a knowledge base including historical fault cases and a fault mechanism base; based on the knowledge base, the spatial feature vector and the time sequence feature vector of the preprocessed data, performing fault recognition, and according to a fault recognition result, performing fault positioning and / or updating the knowledge base; and carrying out fault level evaluation based on the fault identification result and the fault positioning result, and carrying out fault early warning. The method not only improves the efficiency of fault recognition and positioning, but also can continuously enrich the content of the knowledge base, improves the accuracy of subsequent diagnosis, and avoids the problems of diagnosis delay and misjudgment and missed judgment caused by insufficient experience or limited number of engineers.

Description

Technical Field

[0001] This invention belongs to the technical field of equipment operation and maintenance diagnosis, and more specifically, it relates to a method for real-time equipment fault diagnosis. Background Technology

[0002] In core mission execution scenarios, the stable operation of critical equipment directly determines the successful completion of the mission, operational safety, and overall efficiency. Therefore, equipment maintenance is of irreplaceable importance. As the technical complexity of equipment continues to increase, its internal structure and operating mechanism become more sophisticated, and the manifestations of faults also exhibit diverse and concealed characteristics. Currently, the diagnosis and troubleshooting of such equipment faults mainly rely on senior engineers' long-accumulated technical experience, combined with on-site observation and limited historical records for judgment.

[0003] Existing equipment maintenance methods suffer from the following drawbacks: Fault diagnosis relies heavily on engineers' technical experience, requiring on-site inspections, parameter comparisons, and experience reviews—a time-consuming process that results in low efficiency. The accuracy of fault diagnosis is highly dependent on the engineers' technical skills. Engineers with strong technical expertise and extensive experience can quickly identify critical fault signals, while less skilled engineers are prone to misdiagnosis or missed diagnosis, compromising the reliability of the results. Furthermore, the overall efficiency of fault diagnosis is limited by the number of engineers. When multiple pieces of equipment simultaneously issue fault warnings, an insufficient number of engineers can lead to delayed diagnostic responses, extending equipment downtime and impacting the overall maintenance schedule and task execution rhythm. Summary of the Invention

[0004] The purpose of this invention is to provide a real-time fault diagnosis method for equipment, which aims to solve the problem that fault diagnosis in the equipment operation and maintenance phase depends on the level and number of engineers, and the accuracy and efficiency of fault diagnosis are unstable.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is: to provide a real-time equipment fault diagnosis method, comprising: Acquire real-time multi-source data on the equipment's operational status; The edge data in the real-time multi-source data is preprocessed to obtain preprocessed data; Extract the spatiotemporal feature vectors from the preprocessed data; Build a knowledge base that includes historical failure cases and failure mechanism databases; Fault identification is performed based on the knowledge base, the spatial feature vector and time series feature vector of the preprocessed data, and fault location and / or updating of the knowledge base are performed based on the fault identification results. Based on the fault identification and fault location results, a fault level assessment is performed, and a fault warning is issued.

[0006] In one possible implementation, the real-time multi-source information includes equipment operating parameter information, timestamps, and positioning information.

[0007] In one possible implementation, the preprocessing of edge data in the real-time multi-source data to obtain preprocessed data includes: The edge data is filtered to remove high-frequency noise and random noise, resulting in filtered edge data. The filtered edge data is normalized to eliminate scale differences between data of different latitudes, thereby obtaining preprocessed data.

[0008] In one possible implementation, extracting the spatiotemporal feature vector of the preprocessed data includes: Extract the spatial feature vector of the preprocessed data; The spatiotemporal feature vectors of the preprocessed data are extracted based on the spatial feature vectors.

[0009] In one possible implementation, extracting the spatial feature vector of the preprocessed data includes: The preprocessed data is reconstructed into a two-dimensional data matrix; Local spatial features are extracted from the two-dimensional data matrix and then dimensionality is reduced to obtain a local spatial feature map. Deep spatial features are extracted from the local spatial feature map and then dimensionality is reduced to obtain a deep spatial feature map. The deep spatial feature map is mapped to a one-dimensional feature vector, thereby obtaining the spatial feature vector.

[0010] In one possible implementation, the step of extracting the spatiotemporal feature vector of the preprocessed data based on the spatial feature vector includes: The spatial feature vectors are concatenated in chronological order to construct time series data; Preliminary time series features are extracted from the time series data; Deep temporal correlation features are extracted based on the preliminary time series features; The deep temporal correlation features are mapped to spatiotemporal feature vectors with fixed dimensions.

[0011] In one possible implementation, the step of performing fault identification based on the knowledge base, the spatial feature vectors and time series feature vectors of the preprocessed data, and performing fault location and / or updating the knowledge base based on the fault identification results includes: Extract feature vectors from historical fault cases in the knowledge base to construct a historical fault feature vector library; Calculate the similarity between the spatiotemporal feature vector and each vector in the historical fault feature vector library; Fault types are matched based on the similarity; Based on the matching results, fault location and / or updating of the knowledge base are performed.

[0012] In one possible implementation, the step of locating the fault and / or updating the knowledge base based on the matching result includes: If the matching result is a known fault type, the fault mechanism library is called to locate the fault location, and the fault type and location information are output. If the matching result is an unknown fault, then the fault is marked as a new fault, the fault mechanism library is called to locate the fault location, and the new fault and location information are output, while the knowledge base is updated.

[0013] In one possible implementation, updating the knowledge base includes: The aforementioned fault is recorded as a new fault case in the historical fault case library, resulting in a new historical fault case library; A new historical fault feature vector library is constructed based on the new historical fault case library.

[0014] In one possible implementation, the step of assessing the fault level based on the fault identification result and the fault location result, and then providing a fault warning, includes: Calculate the severity score of the fault and determine the warning level; Output the fault type, location information, and warning level.

[0015] The beneficial effects of the real-time equipment fault diagnosis method provided by this invention are as follows: Compared with the prior art, this invention's real-time equipment fault diagnosis method, by acquiring real-time multi-source data during equipment operation, can comprehensively capture various key information during equipment operation, providing rich and timely data support for subsequent diagnosis, and avoiding the problem of incomplete information caused by relying on limited historical records and on-site observations. Preprocessing the edge data in the real-time multi-source data can effectively remove noise interference, eliminate scale differences between data of different dimensions, ensure data quality, lay a reliable foundation for subsequent feature extraction and fault identification, and reduce diagnostic bias caused by data quality issues.

[0016] Extracting spatiotemporal feature vectors from preprocessed data allows for in-depth mining of fault-related information from both spatial and temporal dimensions. It not only focuses on the spatial distribution characteristics of data at a specific moment but also captures the patterns of data change over time, providing a more comprehensive reflection of the changing trends in equipment operating status. Compared to previous diagnostic methods that relied solely on single-dimensional information, this significantly improves the ability to capture fault characteristics. Constructing a knowledge base that includes historical fault cases and fault mechanism libraries integrates past fault experience and knowledge of equipment fault mechanisms, providing rich reference data for fault identification. This reduces over-reliance on individual engineers' experience, making the diagnostic process more standardized and scientific.

[0017] Fault identification based on a knowledge base, spatial feature vectors, and time-series feature vectors can quickly and accurately match fault types. The fault identification results are then combined with fault location and knowledge base updates. This not only improves the efficiency of fault identification and location but also continuously enriches the knowledge base, enhancing the accuracy of subsequent diagnoses and avoiding diagnostic delays, misdiagnoses, and missed diagnoses caused by insufficient engineer experience or limited knowledge base. Fault level assessment and early warning based on fault identification and location results can promptly identify potential equipment fault risks and issue corresponding warnings according to the severity of the fault. This provides maintenance personnel with more processing time, reduces equipment downtime, ensures stable equipment operation, and guarantees that the overall maintenance progress and task execution rhythm are not affected. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A schematic diagram illustrating the main steps of the real-time equipment fault diagnosis method provided in this embodiment of the invention; Figure 2 This is a flowchart illustrating the real-time equipment fault diagnosis method provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0021] Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this application.

[0022] It should be further noted that the accompanying drawings and embodiments of the present invention mainly describe the concept of the present invention. Based on this concept, some specific forms and arrangements of connection relationships, positional relationships, power mechanisms, power supply systems, hydraulic systems and control systems may not be fully described. However, under the premise that those skilled in the art understand the concept of the present invention, they can implement the above-mentioned specific forms and arrangements in a well-known manner.

[0023] When a component is referred to as being "fixed to" or "set on" another component, it can be directly on or indirectly on that other component. When a component is referred to as being "connected to" another component, it can be directly connected to or indirectly connected to that other component.

[0024] In the description of this invention, "a plurality of" means two or more, and "several" means one or more, unless otherwise explicitly specified.

[0025] The directional terms "inner" and "outer" refer to the inner and outer contours relative to the outline of each component itself; the term "length"... "Width", "Top", "Bottom", "Front", "Back", "Left", "Right", "Vertical" The terms "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the purpose of facilitating the description of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the present invention.

[0026] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," and "above" are used here to describe the spatial positional relationship between a device or feature and other devices or features, as shown in the figure. It should be understood that spatial relative terms are intended to... The invention includes different orientations of the device in use or operation, in addition to those described in the figures. For example, if a device in the figures is inverted, a device described as "above" or "on top of" other devices or structures will be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below". The device may also be positioned in other different ways, and the spatial relative descriptions used herein are interpreted accordingly. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the invention, "a plurality of" means two or more, and "a number" means one or more, unless otherwise explicitly specified.

[0027] Reference Figures 1 to 2 The real-time equipment fault diagnosis method provided by the present invention will now be described. The real-time equipment fault diagnosis method includes...

[0028] S100. Acquire real-time multi-source data on the equipment's operational status.

[0029] In one possible implementation, the real-time multi-source information in step S100 includes equipment operating parameter information, timestamps, and positioning information.

[0030] S200. Preprocess the edge data in the real-time multi-source data to obtain the preprocessed data.

[0031] In one possible implementation, step S200 involves preprocessing the edge data in the real-time multi-source data to obtain preprocessed data, including: S210. Filter the edge data to remove high-frequency noise and random noise, and obtain the filtered edge data.

[0032] To address high-frequency noise in vibration data and random noise in temperature data, a wavelet thresholding denoising algorithm is employed. First, the original data is decomposed using wavelet decomposition, selecting the db4 wavelet as the base wavelet. The decomposition is performed at three levels, yielding approximate and detail components. Then, the threshold for the detail components is calculated, and a soft thresholding method is used for thresholding. Finally, the denoised data is reconstructed using inverse wavelet transform.

[0033] Wavelet decomposition is performed using the following formula:

[0034] in, The original data signal, The minimum number of decomposition levels. These are approximate component coefficients. For scaling function, For detail component coefficients, For wavelet functions, This is the translation coefficient.

[0035] Soft thresholding is performed using the following formula:

[0036] in, These are the original detail component coefficients. For the processed detail component coefficients, For the threshold, The standard deviation of noise. This represents the data length.

[0037] S220. Normalize the filtered edge data to eliminate scale differences between data of different latitudes, thereby obtaining preprocessed data.

[0038] S300. Extract the spatiotemporal feature vectors of the preprocessed data.

[0039] In one possible implementation, step S300, extracting the spatiotemporal feature vector of the preprocessed data, includes: S310. Extract the spatial feature vector of the preprocessed data.

[0040] In one possible implementation, step S310, extracting the spatial feature vector of the preprocessed data, includes: S311. Reconstruct the preprocessed data into a two-dimensional data matrix.

[0041] S312. Extract local spatial features from the two-dimensional data matrix and perform dimensionality reduction to obtain a local spatial feature map.

[0042] S313. Extract deep spatial features from the local spatial feature map and perform dimensionality reduction to obtain the deep spatial feature map.

[0043] S314. Map the deep spatial feature map to a one-dimensional feature vector to obtain the spatial feature vector.

[0044] A CNN-LSTM hybrid deep learning model is constructed. Preprocessed multi-source data is input into the model, and spatial features are extracted from the data through the CNN network. The CNN network structure includes two convolutional layers, two pooling layers, and one flattening layer. First, the normalized multi-source data is reconstructed into a two-dimensional data matrix (dimension [100, 2], where 100 is the time dimension and 2 is the feature dimension) according to a time window (window size of 100 data points and stride of 50 data points). Then, the first convolutional layer (3×1 kernel size, 32 kernels, ReLU activation function) extracts local spatial features, which are then reduced in dimensionality by a max pooling layer (2×1 kernel size, stride of 2). Next, the second convolutional layer (3×1 kernel size, 64 kernels, ReLU activation function) further extracts deep spatial features, which are then reduced in dimensionality by a max pooling layer (2×1 kernel size, stride of 2). Finally, the flattening layer maps the two-dimensional features into a one-dimensional feature vector.

[0045] When extracting local spatial features through the first convolutional layer and further extracting deep spatial features through the second convolutional layer, convolution operations are performed using the following formula:

[0046] in, Let (i,j) be the (i,j)th element of the convolutional feature map. For the input data matrix, For convolution kernel, For bias terms, This is the location index of the feature map.

[0047] S320. Extract the spatiotemporal feature vectors of the preprocessed data based on the spatial feature vectors.

[0048] In one possible implementation, step S320, extracting the spatiotemporal feature vector of the preprocessed data based on the spatial feature vector, includes: S321. Concatenate the spatial feature vectors in chronological order to construct time series data.

[0049] S322. Extract preliminary time series features based on time series data.

[0050] S323. Extract deep temporal correlation features based on preliminary time series features.

[0051] S324. Map deep temporal correlation features to spatiotemporal feature vectors with fixed dimensions.

[0052] This paper describes a method for automatically mining deep spatiotemporal features from data by using an LSTM network to extract time-series features. The LSTM network structure consists of two LSTM layers and one fully connected layer. The spatial feature vectors extracted by the CNN are concatenated chronologically to construct a time-series data (sequence length 20), which is then input into the first LSTM layer (64 hidden units, returning True) to extract preliminary time-series features. This is then input into the second LSTM layer (32 hidden units, returning False) to extract deep temporal correlation features. Finally, the LSTM output is mapped to a fixed-dimensional spatiotemporal feature vector through a fully connected layer (output dimension 16). The model uses the Adam optimizer with cross-entropy loss and employs an early stopping strategy during training (training stops if the validation set loss does not decrease after 3 epochs).

[0053] S400. Build a knowledge base that includes historical failure cases and a failure mechanism library.

[0054] S500. Based on the knowledge base, spatial feature vectors and time series feature vectors of preprocessed data, fault identification is performed, and fault location and / or knowledge base is updated according to the fault identification results.

[0055] In one possible implementation, step S500 involves fault identification based on a knowledge base, spatial feature vectors of preprocessed data, and time-series feature vectors, and fault location and / or knowledge base updating based on the fault identification results, including: S510. Extract feature vectors from historical fault cases in the knowledge base and construct a historical fault feature vector library.

[0056] S520. Calculate the similarity between the spatiotemporal feature vector and each vector in the historical fault feature vector library.

[0057] Vector similarity is calculated using the following formula:

[0058] in, Let cosine similarity be the similarity between two vectors. The input is the spatiotemporal feature vector. It is a vector in the historical fault feature vector library. The dimension of the feature vector. for The i-th element, for The i-th element.

[0059] S530. Match fault types based on similarity.

[0060] A similarity threshold is set. When a historical fault feature vector with a similarity not lower than the similarity threshold exists, it is matched as the corresponding fault type. When the similarity of all historical fault feature vectors is lower than the similarity threshold, it is determined as a new fault case, and the feature vector, equipment operating parameters and field conditions of the case are recorded.

[0061] S540. Based on the matching results, locate the fault and / or update the knowledge base.

[0062] In one possible implementation, step S540, fault location and / or knowledge base update based on the matching results, includes: S541. If the matching result is a known fault type, call the fault mechanism library to locate the fault location and output the fault type and location information.

[0063] S542. If the matching result is an unknown fault, mark the fault as a new fault, call the fault mechanism library to locate the fault location, and output the new fault and location information, while updating the knowledge base.

[0064] In one possible implementation, updating the knowledge base in step S542 includes: S542a. Enter the fault as a new fault case into the historical fault case library to obtain a new historical fault case library.

[0065] S542b. Construct a new historical fault feature vector library based on the new historical fault case library.

[0066] For a known fault type that is matched, the fault mechanism library is invoked. This library stores the association between different fault types and equipment components (e.g., bearing wear faults correspond to bearing components, motor overheating faults correspond to motor stator components). Combined with sensor deployment location information, the specific component where the fault occurred is accurately located. For a newly identified fault case, it is automatically entered into the historical fault case library. At the same time, the incremental training of the CNN-LSTM model is triggered (training is performed using new case data and similar case data, with 10 training rounds). The model parameters and the historical fault feature vector library are updated to complete the adaptive update of the knowledge base.

[0067] S600. Based on the fault identification and fault location results, the fault level is assessed and a fault warning is issued.

[0068] In one possible implementation, step S600 involves assessing the fault level based on the fault identification and fault location results, and issuing a fault warning, including: S610. Calculate the severity score of the fault and determine the warning level.

[0069] The severity score of the fault is calculated using the following formula:

[0070] in, The severity of the fault is scored. These are the weighting coefficients. Score the extent of the fault's impact. Score based on the speed of fault development. Points are awarded for downtime losses.

[0071] A score of S < 30 indicates a Level 1 warning (potential fault), a score of 30 ≤ S < 60 indicates a Level 2 warning (minor fault), and a score of S ≥ 60 indicates a Level 3 warning (serious fault). Once a warning is triggered, the fault type and location information are automatically associated to generate the warning content.

[0072] S620. Outputs fault type, location information, and warning level.

[0073] An industrial internet communication link is constructed using the MQTT protocol (Message Queuing Telemetry Transport Protocol). Information such as the warning level, fault type, fault location, and fault occurrence time, combined with the corresponding fault handling measures in the emergency suggestion database (such as strengthening monitoring frequency for level 1 warnings, planned maintenance for level 2 warnings, and immediate shutdown for maintenance for level 3 warnings), is packaged into a warning message and pushed to the computer terminals and mobile apps of maintenance personnel. At the same time, the warning information and handling results (subsequent supplementary entries) are entered into the warning log database, and the log information is retained for one year for traceability and analysis.

[0074] The beneficial effects of the real-time equipment fault diagnosis method provided by this invention are as follows: Compared with the prior art, this invention's real-time equipment fault diagnosis method, by acquiring real-time multi-source data during equipment operation, can comprehensively capture various key information during equipment operation, providing rich and timely data support for subsequent diagnosis, and avoiding the problem of incomplete information caused by relying on limited historical records and on-site observations. Preprocessing the edge data in the real-time multi-source data can effectively remove noise interference, eliminate scale differences between data of different dimensions, ensure data quality, lay a reliable foundation for subsequent feature extraction and fault identification, and reduce diagnostic bias caused by data quality issues.

[0075] Extracting spatiotemporal feature vectors from preprocessed data allows for in-depth mining of fault-related information from both spatial and temporal dimensions. It not only focuses on the spatial distribution characteristics of data at a specific moment but also captures the patterns of data change over time, providing a more comprehensive reflection of the changing trends in equipment operating status. Compared to previous diagnostic methods that relied solely on single-dimensional information, this significantly improves the ability to capture fault characteristics. Constructing a knowledge base that includes historical fault cases and fault mechanism libraries integrates past fault experience and knowledge of equipment fault mechanisms, providing rich reference data for fault identification. This reduces over-reliance on individual engineers' experience, making the diagnostic process more standardized and scientific.

[0076] Fault identification based on knowledge base, spatial feature vectors, and time-series feature vectors can quickly and accurately match fault types. Combining fault identification results with fault location and knowledge base updates not only improves the efficiency of fault identification and location but also continuously enriches the knowledge base, enhancing the accuracy of subsequent diagnoses and avoiding diagnostic delays, misjudgments, and omissions caused by insufficient engineer experience or limited knowledge base. Fault level assessment and early warning based on fault identification and location results can promptly identify potential equipment fault risks and issue corresponding warnings according to the severity of the fault. This provides maintenance personnel with more processing time, reduces equipment downtime, ensures stable equipment operation, and guarantees that the overall maintenance progress and task execution rhythm are not affected. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0077] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0078] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following drawings denote similar items; therefore, once an item is defined in one drawing, it need not be further discussed in subsequent drawings.

Claims

1. A real-time fault diagnosis method for equipment, characterized in that, include: Acquire real-time multi-source data on the equipment's operational status; The edge data in the real-time multi-source data is preprocessed to obtain preprocessed data; Extract the spatiotemporal feature vectors from the preprocessed data; Build a knowledge base that includes historical failure cases and failure mechanism databases; Fault identification is performed based on the knowledge base, the spatial feature vector and time series feature vector of the preprocessed data, and fault location and / or updating of the knowledge base are performed based on the fault identification results. Based on the fault identification and fault location results, a fault level assessment is performed, and a fault warning is issued.

2. The real-time equipment fault diagnosis method as described in claim 1, characterized in that, The real-time multi-source information includes equipment operating parameters, timestamps, and location information.

3. The real-time equipment fault diagnosis method as described in claim 1, characterized in that, The step of preprocessing the edge data in the real-time multi-source data to obtain preprocessed data includes: The edge data is filtered to remove high-frequency noise and random noise, resulting in filtered edge data. The filtered edge data is normalized to eliminate scale differences between data of different latitudes, thereby obtaining preprocessed data.

4. The real-time equipment fault diagnosis method as described in claim 1, characterized in that, The extraction of the spatiotemporal feature vector from the preprocessed data includes: Extract the spatial feature vector of the preprocessed data; The spatiotemporal feature vectors of the preprocessed data are extracted based on the spatial feature vectors.

5. The real-time equipment fault diagnosis method as described in claim 4, characterized in that, The extraction of spatial feature vectors from the preprocessed data includes: The preprocessed data is reconstructed into a two-dimensional data matrix; Local spatial features are extracted from the two-dimensional data matrix and then dimensionality is reduced to obtain a local spatial feature map. Deep spatial features are extracted from the local spatial feature map and then dimensionality is reduced to obtain a deep spatial feature map. The deep spatial feature map is mapped to a one-dimensional feature vector, thereby obtaining the spatial feature vector.

6. The real-time equipment fault diagnosis method as described in claim 4, characterized in that, The step of extracting the spatiotemporal feature vector of the preprocessed data based on the spatial feature vector includes: The spatial feature vectors are concatenated in chronological order to construct time series data; Preliminary time series features are extracted from the time series data; Deep temporal correlation features are extracted based on the preliminary time series features; The deep temporal correlation features are mapped to spatiotemporal feature vectors with fixed dimensions.

7. The real-time equipment fault diagnosis method as described in claim 1, characterized in that, The step of identifying faults based on the knowledge base, the spatial feature vectors and time series feature vectors of the preprocessed data, and locating faults and / or updating the knowledge base based on the fault identification results includes: Extract feature vectors from historical fault cases in the knowledge base to construct a historical fault feature vector library; Calculate the similarity between the spatiotemporal feature vector and each vector in the historical fault feature vector library; Fault types are matched based on the similarity; Based on the matching results, fault location and / or updating of the knowledge base are performed.

8. The real-time equipment fault diagnosis method as described in claim 7, characterized in that, The step of locating the fault and / or updating the knowledge base based on the matching results includes: If the matching result is a known fault type, the fault mechanism library is called to locate the fault location, and the fault type and location information are output. If the matching result is an unknown fault, then the fault is marked as a new fault, the fault mechanism library is called to locate the fault location, and the new fault and location information are output, while the knowledge base is updated.

9. The real-time equipment fault diagnosis method as described in claim 8, characterized in that, The updating of the knowledge base includes: The aforementioned fault is recorded as a new fault case in the historical fault case library, resulting in a new historical fault case library; A new historical fault feature vector library is constructed based on the new historical fault case library.

10. The real-time equipment fault diagnosis method as described in claim 1, characterized in that, The process of assessing the fault level and issuing a fault warning based on the fault identification and fault location results includes: Calculate the severity score of the fault and determine the warning level; Output the fault type, location information, and warning level.

Images