Equipment fault diagnosis method and device, electronic equipment and storage medium

Abstract

The invention discloses an equipment fault diagnosis method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence. The pre-training joint reasoning model is used for comprehensively analyzing the multi-dimensional data to generate an equipment fault probability prediction result, a composite fault mode can be effectively recognized, dynamic working condition changes can be adapted, meanwhile, tracing analysis is conducted on a fault propagation path in combination with the graph neural network, the effectiveness of feature extraction and modeling is improved, and therefore the fault probability prediction result is obtained. The technical problems that in an existing equipment fault diagnosis system, a threshold model is difficult to recognize a composite fault, a physical model cannot adapt to a dynamic working condition, and the fault traceability of a correlation analysis model is limited can be solved, and the purposes of improving the accuracy and adaptability of equipment fault diagnosis, enhancing the fault traceability and improving the reliability of equipment fault diagnosis are achieved. And the technical effect of promoting predictive maintenance to develop to higher-level intellectualization is achieved.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to a method and apparatus for diagnosing equipment failures, electronic devices, and storage media. Background Technology

[0002] Intelligent equipment monitoring and predictive maintenance systems are key technologies for ensuring efficient operation and extending the service life of industrial equipment, and are widely used in various industrial fields such as power, metallurgy, chemical, and coal. Related technologies, through the collaborative operation of the data acquisition layer, transmission layer, analysis and processing layer, and application layer, construct a complete technical system from equipment status perception to fault prediction and maintenance decision-making.

[0003] In existing equipment fault diagnosis systems, threshold models can only handle single-dimensional anomalies and struggle to identify complex fault modes; physical models rely on static parameters and cannot adapt to dynamic changes in operating conditions; and correlation analysis models lack effective feature extraction and modeling methods, limiting fault tracing capabilities. These problems are particularly prominent in real-world industrial scenarios, restricting the development of predictive maintenance towards higher levels of intelligence. Summary of the Invention

[0004] This disclosure provides a method, apparatus, electronic device, and storage medium for diagnosing equipment faults. Its main purpose is to address the problem that the lack of effective feature extraction and modeling methods in correlation analysis models limits fault tracing capabilities.

[0005] According to a first aspect of this disclosure, a method for diagnosing equipment faults is provided, comprising:

[0006] Acquire process parameter data and status monitoring data of sensing equipment in the distributed control system;

[0007] Based on the pre-trained joint inference model, the process parameter data and the status monitoring data are analyzed to generate equipment failure probability prediction results.

[0008] Based on the graph neural network, the fault propagation path of the equipment fault probability prediction result is traced and analyzed, and a diagnostic report is outputting the fault type, occurrence location and impact range.

[0009] Optionally, before analyzing the process parameter data and the condition monitoring data based on the pre-trained joint inference model to generate equipment failure probability prediction results, the method further includes:

[0010] The process parameter data and the status monitoring data are time-stamped and spatially mapped to achieve spatiotemporal alignment.

[0011] Data cleaning, format conversion, and feature extraction are performed on the spatiotemporally aligned process parameter data and status monitoring data, and the data format is unified.

[0012] Optionally, the step of synchronizing the process parameter data and the status monitoring data with timestamps and mapping spatial coordinates to achieve spatiotemporal alignment includes:

[0013] The process parameter data and the status monitoring data are respectively timestamped and calibrated.

[0014] By mapping three-dimensional coordinate information to the identifiers of corresponding device nodes in the distributed control system, a unified device space topology is constructed.

[0015] Optionally, the process parameter data and status monitoring data after spatiotemporal alignment are subjected to data cleaning, format conversion, and feature extraction, and the data format is unified, including:

[0016] The process parameter data and the status monitoring data are cleaned based on an outlier detection algorithm;

[0017] The process parameter data and the status monitoring data after cleaning are subjected to feature engineering processing, including time-domain statistical feature extraction, frequency-domain transform analysis and wavelet decomposition.

[0018] Optionally, the pre-trained joint inference model includes a physical model and a machine learning model; the step of analyzing the process parameter data and the condition monitoring data based on the pre-trained joint inference model to generate equipment failure probability prediction results includes:

[0019] The simulation results of the physical model are used as input features of the machine learning model to improve the machine learning model's ability to understand the operating mechanism of the equipment.

[0020] The pre-trained joint inference model is optimized using an online learning mechanism.

[0021] Optionally, the fault propagation path analysis based on the fault probability prediction results of the equipment using graph neural networks, and the output of a diagnostic report including fault type, location of occurrence, and scope of impact, includes:

[0022] Construct a dynamic correlation weight matrix between the process parameter data and the status monitoring data;

[0023] The graph neural network employs an attention mechanism to enhance its focus on key parameter nodes, thereby improving the accuracy of fault tracing.

[0024] According to a second aspect of this disclosure, a diagnostic apparatus for equipment faults is provided, comprising:

[0025] The acquisition unit is used to acquire process parameter data of the distributed control system and status monitoring data of sensing equipment;

[0026] The analysis unit is used to analyze the process parameter data and the status monitoring data based on a pre-trained joint inference model to generate equipment failure probability prediction results.

[0027] The output unit is used to perform source analysis on the fault propagation path of the fault probability prediction result of the equipment based on the graph neural network, and output a diagnostic report on the fault type, occurrence location and impact range.

[0028] Optionally, the device further includes:

[0029] The processing unit is used to perform time stamp synchronization and spatial coordinate mapping on the process parameter data and the status monitoring data before the analysis unit analyzes the process parameter data and the status monitoring data based on the pre-trained joint inference model and generates equipment failure probability prediction results, so as to achieve spatiotemporal alignment.

[0030] The processing unit is also used to perform data cleaning, format conversion and feature extraction on the spatiotemporally aligned process parameter data and the status monitoring data, and to unify the data format.

[0031] Optionally, the processing unit is further configured to:

[0032] The process parameter data and the status monitoring data are respectively timestamped and calibrated.

[0033] By mapping three-dimensional coordinate information to the identifiers of corresponding device nodes in the distributed control system, a unified device space topology is constructed.

[0034] Optionally, the processing unit is further configured to:

[0035] The process parameter data and the status monitoring data are cleaned based on an outlier detection algorithm;

[0036] The process parameter data and the status monitoring data after cleaning are subjected to feature engineering processing, including time-domain statistical feature extraction, frequency-domain transform analysis and wavelet decomposition.

[0037] Optionally, the pre-trained joint inference model includes a physical model and a machine learning model; the analysis unit is further used for:

[0038] The simulation results of the physical model are used as input features of the machine learning model to improve the machine learning model's ability to understand the operating mechanism of the equipment.

[0039] The pre-trained joint inference model is optimized using an online learning mechanism.

[0040] Optionally, the output unit is further configured to:

[0041] Construct a dynamic correlation weight matrix between the process parameter data and the status monitoring data;

[0042] The graph neural network employs an attention mechanism to enhance its focus on key parameter nodes, thereby improving the accuracy of fault tracing.

[0043] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0044] At least one processor; and

[0045] A memory communicatively connected to the at least one processor; wherein,

[0046] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0047] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0048] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0049] The equipment fault diagnosis method, apparatus, electronic device, and storage medium disclosed herein mainly include the following technical solutions: acquiring process parameter data of a distributed control system and status monitoring data of sensing equipment; analyzing the process parameter data and status monitoring data based on a pre-trained joint inference model to generate equipment fault probability prediction results; and performing source tracing analysis on the fault propagation path of the equipment fault probability prediction results based on a graph neural network, outputting a diagnostic report on the fault type, location, and scope of impact. Compared with related technologies, this application, by acquiring process parameter data of a distributed control system and status monitoring data of sensing equipment, and using a pre-trained joint inference model to comprehensively analyze multi-dimensional data to generate equipment fault probability prediction results, can effectively identify complex fault modes and adapt to dynamic operating condition changes. Simultaneously, by combining graph neural networks to perform source tracing analysis on fault propagation paths, the effectiveness of feature extraction and modeling is improved. Therefore, it can solve the technical problems in existing equipment fault diagnosis systems where threshold models are unable to identify complex faults, physical models cannot adapt to dynamic operating conditions, and correlation analysis models have limited fault tracing capabilities. This achieves the technical effect of improving the accuracy and adaptability of equipment fault diagnosis, enhancing fault tracing capabilities, and promoting predictive maintenance towards a higher level of intelligence.

[0050] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0051] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0052] Figure 1 A schematic flowchart illustrating a method for diagnosing equipment faults provided in an embodiment of this disclosure;

[0053] Figure 2 A schematic diagram of the structure of a device for diagnosing equipment faults provided in an embodiment of this disclosure;

[0054] Figure 3 A schematic diagram of the structure of a device for diagnosing equipment faults provided in an embodiment of this disclosure;

[0055] Figure 4 A schematic block diagram of an example electronic device provided for embodiments of this disclosure. Detailed Implementation

[0056] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0057] The following description, with reference to the accompanying drawings, outlines a method, apparatus, electronic device, and storage medium for diagnosing device malfunctions according to embodiments of the present disclosure.

[0058] Figure 1 This is a schematic flowchart illustrating a method for diagnosing equipment faults provided in an embodiment of this disclosure.

[0059] like Figure 1 As shown, the method includes the following steps:

[0060] Step 101: Obtain process parameter data of the distributed control system and status monitoring data of sensing equipment;

[0061] The acquisition of process parameter data in a Distributed Control System (DCS) is achieved through the establishment of a secure data interface. This interface enables secure and stable data exchange with the enterprise's existing DCS system, collecting various process parameters of equipment in real time during production. These process parameters include, but are not limited to, temperature, pressure, flow rate, speed, current, and voltage during equipment operation. They directly reflect the core operating status of the equipment under actual production conditions and are crucial for determining whether the equipment is within the normal process range. Meanwhile, the acquisition of status monitoring data from sensing equipment is achieved using various specialized sensors and intelligent monitoring devices. These sensing devices are deployed at key parts or monitoring points of the equipment according to monitoring needs. For example, vibration sensors can monitor the vibration of rotating equipment in real time, thereby capturing potential fault characteristics such as imbalance, misalignment, and bearing wear; infrared temperature sensors and thermocouples can accurately collect temperature data of key parts on the surface and inside of the equipment, promptly detecting abnormal temperature phenomena caused by poor contact, overload, etc.; oil sensors are used to detect key indicators such as viscosity, water content, and metal particle concentration of the equipment's lubricating oil, thereby assessing the wear degree of the equipment and the operating status of the lubrication system; intelligent cameras combined with AI visual recognition algorithms can visually monitor the appearance and operating status of the equipment, effectively capturing visually identifiable status information such as pipeline leaks, loose components, conveyor belt misalignment, and abnormal valve opening and closing states. Through these two data acquisition methods, the system can comprehensively and in real time collect process parameter data and status monitoring data during equipment operation, ensuring the completeness and timeliness of the raw data required for subsequent diagnostic analysis.

[0062] Step 102: Based on the pre-trained joint inference model, analyze the process parameter data and the status monitoring data to generate equipment failure probability prediction results;

[0063] In the specific analysis process, the pre-trained joint inference model first receives process parameter data (such as temperature, pressure, flow rate, and speed) from the distributed control system and status monitoring data from various sensing devices (such as vibration characteristics captured by vibration sensors, temperature data collected by infrared thermometers, lubricating oil indicators detected by oil sensors, and visual status information acquired by smart cameras). Next, the model performs targeted feature extraction and correlation analysis on these input data. For example, it matches the real-time trends of process parameters with abnormal features in the status monitoring data to identify potential correlations and abnormal patterns between the data. Because the model has been pre-trained, it has learned the correspondence between data characteristics and failure probabilities under different operating conditions. Therefore, based on the input real-time data and the knowledge graph and inference rules formed during historical training, it can calculate the probability of various failures (such as imbalance, misalignment, bearing wear, abnormal temperature, and lubrication failure) that may occur in the equipment now and in the future. Finally, through the model's joint inference calculation, it outputs specific equipment failure probability prediction results, providing a quantitative analytical basis for subsequent equipment status assessment and early warning.

[0064] Step 103: Based on the graph neural network, perform source analysis on the fault propagation path of the equipment fault probability prediction result, and output a diagnostic report on the fault type, location of occurrence and scope of impact.

[0065] Graph neural networks, as a deep learning model that can effectively process non-Euclidean data, are used in this step to construct a graph structure connecting various components, system modules, and parameters of the equipment. This graph structure is based on the physical connections, functional dependencies, and historical fault propagation patterns of the equipment. It abstracts key components, sensor monitoring points, and process parameter nodes of the equipment into nodes in the graph, while the edges between nodes represent the correlation strength between various elements (such as signal transmission relationships, energy flow relationships, fault impact weights, etc.).

[0066] In the specific analysis process, the graph neural network first receives the information of high-probability fault-related nodes marked in the equipment failure probability prediction results. Then, through its unique node aggregation and message passing mechanism, it iteratively learns and infers the state characteristics and relationships of each node in the graph structure. This process can effectively capture the dynamic path of fault propagation from the initial occurrence node to other related nodes. For example, when the bearing wear failure probability of rotating equipment is high, the graph neural network can trace the propagation chain from bearing wear to how mechanical vibration affects the balance of the drive shaft, leading to abnormal motor load, by analyzing the connection relationship between the bearing and components such as the drive shaft and motor, as well as the transmission and changes of parameters such as vibration and temperature among these components.

[0067] Through the aforementioned source tracing analysis, the graph neural network can accurately identify the root location of the fault (i.e., the initial fault node), clarify the specific type of fault (such as bearing wear, abnormal temperature, lubrication failure, etc., corresponding to the fault type in the prediction results), and assess the impact of the fault on surrounding components, related system modules, and the overall production process based on the correlation strength and propagation range between nodes in the graph structure (e.g., whether it only affects a single component or a functional subsystem, or whether it may spread to the entire production equipment chain). Finally, based on the analysis results of the graph neural network, the system automatically generates a diagnostic report containing the fault type, location, and impact range, providing a clear and traceable basis for subsequent maintenance decisions.

[0068] In some embodiments, before analyzing the process parameter data and the condition monitoring data based on a pre-trained joint inference model to generate equipment failure probability prediction results, the method further includes:

[0069] The process parameter data and the status monitoring data are time-stamped and spatially mapped to achieve spatiotemporal alignment.

[0070] Data cleaning, format conversion, and feature extraction are performed on the spatiotemporally aligned process parameter data and status monitoring data, and the data format is unified.

[0071] In this method of the intelligent equipment preventive diagnostic management system, before analyzing process parameter data and condition monitoring data based on a pre-trained joint inference model to generate equipment failure probability prediction results, data preprocessing operations must be completed first. These operations include spatiotemporal alignment and data standardization. Spatiotemporal alignment is the foundation for ensuring effective fusion and analysis of multi-source data. Timestamp synchronization aims to adjust process parameter data from the distributed control system (DCS) and condition monitoring data from various sensing devices (such as vibration sensors, infrared temperature sensors, oil sensors, and smart cameras) to a unified time reference, eliminating time deviations caused by factors such as the response speed of data acquisition devices and transmission delays, so that the equipment operating status data at the same moment can be accurately matched. Spatial coordinate mapping, on the other hand, accurately associates the condition monitoring data collected by different sensing devices with the specific components and monitoring points of the equipment in space according to the physical layout of the equipment and the deployment location of the sensors. This clarifies the physical location of the equipment corresponding to the data. For example, the monitoring data of the vibration sensor is mapped to the bearing position of the rotating equipment where it is installed, and the infrared temperature measurement data is mapped to the specific heat-generating component of the equipment, ensuring the consistency of the data in the spatial dimension.

[0072] After completing spatiotemporal alignment, further standardization processing of the data is required. The data cleaning stage primarily utilizes specialized data processing tools to identify and remove noisy data (such as abnormal fluctuations caused by transient sensor interference), invalid data (such as null values ​​or erroneous codes resulting from failed data acquisition), and duplicate data, ensuring data accuracy and purity. Format conversion transforms raw data from different sources and formats (such as structured process parameters output by DCS systems, analog signal data output by sensors, and image data output by smart cameras) into a unified data format recognizable by the pre-trained joint inference model. For example, analog signals are converted to digital quantities, and image data is extracted as feature vectors. The feature extraction stage extracts key features related to equipment failure from the cleaned and converted standardized data. This includes extracting vibration frequency and amplitude from vibration data, temperature change rate and peak temperature from temperature data, and the degree to which process parameters deviate from normal ranges from process parameters. These features provide high-quality input for subsequent analysis by the pre-trained joint inference model, ensuring that the model can generate accurate equipment failure probability predictions based on standardized and effective data.

[0073] In some embodiments, the step of synchronizing the process parameter data and the status monitoring data with timestamps and mapping spatial coordinates to achieve spatiotemporal alignment includes:

[0074] The process parameter data and the status monitoring data are respectively timestamped and calibrated.

[0075] By mapping three-dimensional coordinate information to the identifiers of corresponding device nodes in the distributed control system, a unified device space topology is constructed.

[0076] Timestamp calibration is performed based on the characteristics of data from different acquisition sources: Due to differences in the hardware acquisition frequency and data transmission path between the distributed control system (DCS) and various sensing devices (such as vibration sensors, infrared temperature sensors, oil sensors, smart cameras, etc.), the timestamps of the original data may be biased. Therefore, it is necessary to calibrate the timestamps of process parameter data and status monitoring data separately. By introducing a unified system time base (such as based on a high-precision clock synchronization protocol), the time offset caused by equipment response delays and transmission time during data acquisition is corrected, ensuring that data from different sources can accurately correspond in the time dimension, laying a foundation for time consistency for subsequent data correlation analysis.

[0077] Spatial coordinate mapping focuses on the precise correlation between data and the physical location of equipment. First, based on the actual installation layout and 3D modeling data of the equipment, the 3D coordinate information of the deployment locations of each sensing device and the objects monitored by the DCS system is obtained. This coordinate information accurately reflects the installation points of the sensors on the equipment (e.g., the 3D coordinates of a vibration sensor corresponding to the bearing of a rotating device, the 3D coordinates of an infrared temperature sensor corresponding to the motor winding, etc.) and the physical location of the equipment associated with each process parameter in the DCS system. Then, this 3D coordinate information is mapped one-to-one with the preset equipment node identifiers (e.g., equipment numbers, module codes, etc.) in the distributed control system. By establishing the correspondence between coordinates and identifiers, the scattered monitoring data is anchored to specific equipment entities or components, thereby constructing a unified equipment spatial topology. This topology clearly presents the spatial distribution, connection relationships, and locational associations of the various components of the equipment, ensuring that the data has a clear spatial dimension and effectively eliminating the spatial ambiguity caused by the dispersed data sources. This provides spatial consistency assurance for subsequent multi-source data fusion analysis.

[0078] In some embodiments, the process parameter data and status monitoring data after spatiotemporal alignment are subjected to data cleaning, format conversion, and feature extraction, and the data format is unified as follows:

[0079] The process parameter data and the status monitoring data are cleaned based on an outlier detection algorithm;

[0080] The process parameter data and the status monitoring data after cleaning are subjected to feature engineering processing, including time-domain statistical feature extraction, frequency-domain transform analysis and wavelet decomposition.

[0081] The data cleaning process is based on outlier detection algorithms. Considering that process parameter data (such as temperature, pressure, and flow rate) and condition monitoring data (such as vibration signals from vibration sensors, temperature values ​​from infrared thermography, and detection indicators from oil sensors) may generate outliers during acquisition and transmission due to factors such as momentary sensor failures, electromagnetic interference, and packet loss, the system employs outlier detection algorithms (such as statistical Z-score methods, IQR methods, or machine learning-based isolated forest algorithms) to scan the data. This accurately identifies and removes noise data (such as sudden spikes in vibration signals) and invalid data (such as continuous null values ​​caused by sensor offline). Simultaneously, some repairable outlier data (such as deviations caused by momentary fluctuations) is smoothly corrected to ensure that the cleaned data accurately reflects the actual operating status of the equipment, providing a reliable foundation for subsequent processing.

[0082] After data cleaning, the system performs feature engineering to extract key information related to equipment failures. This includes time-domain statistical feature extraction, frequency-domain transform analysis, and wavelet decomposition. Time-domain statistical feature extraction targets the time-varying characteristics of the data, extracting statistical quantities such as mean, variance, peak value, kurtosis, and root mean square from time-series data like vibration and temperature. For example, extracting the peak factor from the vibration time-domain signal of rotating equipment can effectively reflect whether the equipment has an impact failure. Frequency-domain transform analysis uses methods such as Fourier transform to convert the time-domain signal to the frequency domain, analyzing the frequency components and amplitude distribution of the signal. For example, identifying the resonance peak at a specific frequency from the vibration frequency spectrum can locate the characteristic frequencies corresponding to equipment imbalances and misalignments. Wavelet decomposition is suitable for processing non-stationary signals. Through multi-scale decomposition, it breaks down complex original signals into sub-signals of different frequency bands, effectively separating background noise from fault characteristic signals during normal equipment operation. For example, wavelet decomposition of viscosity change signals collected by oil sensors can extract subtle viscosity fluctuations caused by equipment wear. Through the aforementioned feature engineering process, the system transforms the raw data into feature vectors with clear physical meaning, and unifies the format to a data format that can be directly input into the pre-trained joint inference model, laying a data foundation for the model to accurately analyze equipment faults.

[0083] In some embodiments, the pre-trained joint inference model includes a physical model and a machine learning model; the step of analyzing the process parameter data and the condition monitoring data based on the pre-trained joint inference model to generate equipment failure probability prediction results includes:

[0084] The simulation results of the physical model are used as input features of the machine learning model to improve the machine learning model's ability to understand the operating mechanism of the equipment.

[0085] The pre-trained joint inference model is optimized using an online learning mechanism.

[0086] A physical model is a simulation model built upon professional knowledge of the equipment's operating principles, mechanical structure, and physical characteristics. It combines equipment design parameters, material properties, and operating conditions to simulate the equipment's operation under different process parameters (such as temperature, pressure, and speed) and condition monitoring data (such as vibration and lubrication conditions) through numerical calculations, mechanical analysis, and heat conduction simulations. It outputs simulation results such as stress distribution, wear evolution, and performance degradation of key components, intuitively reflecting the equipment's inherent operating mechanisms and physical laws. A machine learning model, on the other hand, is a model trained using algorithms such as random forests, support vector machines, and LSTMs, based on historical fault data, normal operating data, and labeled fault features. It possesses the ability to mine hidden patterns and identify fault features from massive amounts of data.

[0087] When analyzing process parameter data and condition monitoring data based on a pre-trained joint inference model, the simulation results of the physical model are first incorporated into the analysis process as input features of the machine learning model. The simulation results of the physical model provide the machine learning model with prior knowledge based on physical laws. For example, the "relationship curve between bearing temperature and wear rate at different speeds" and the "theoretical threshold of pipeline vibration frequency under specific pressure" obtained through physical model simulation can help the machine learning model overcome the limitations of simply relying on statistical data, and gain a deeper understanding of the internal mechanism of equipment operation. Thus, when facing complex operating conditions and multi-parameter coupling scenarios, it can accurately identify fault characteristics caused by physical performance degradation and improve the accuracy of judging abnormal equipment states.

[0088] Meanwhile, to ensure the model's adaptability and predictive accuracy during long-term equipment operation, this method continuously optimizes the pre-trained joint inference model through an online learning mechanism. This online learning mechanism receives new process parameter data, condition monitoring data, and actual fault feedback information collected by the system in real time. This new data serves as incremental training samples for the model, continuously adjusting the simulation parameters of the physical model (e.g., correcting changes in physical performance parameters due to equipment aging) and the algorithm weights of the machine learning model (e.g., updating the mapping relationship between fault features and probabilities). This allows the model to dynamically adapt to changes in operating conditions, component aging, environmental interference, and other factors during equipment operation, maintaining high accuracy in predicting equipment fault probabilities and providing reliable model support for preventative equipment diagnosis.

[0089] In some embodiments, the fault propagation path analysis based on the graph neural network-based prediction of the equipment fault probability, and the output of a diagnostic report including fault type, location of occurrence, and scope of impact, includes:

[0090] Construct a dynamic correlation weight matrix between the process parameter data and the status monitoring data;

[0091] The graph neural network employs an attention mechanism to enhance its focus on key parameter nodes, thereby improving the accuracy of fault tracing.

[0092] Constructing a dynamic correlation weight matrix between process parameter data and condition monitoring data is the foundation of source tracing analysis. This matrix focuses on the correlation of multi-source data during equipment operation, forming structured data support by quantifying the influence relationships between different data. Specifically, the element values ​​in the matrix represent the dynamic correlation strength between process parameters (such as temperature, pressure, flow rate, and speed) and condition monitoring data (such as vibration signal characteristics, abnormal temperature values, oil indicators, and visual monitoring results). The values ​​are dynamically updated according to changes in real-time collected data. For example, when a vibration sensor detects an abnormal vibration amplitude in the bearing of a rotating equipment, the correlation weight between "vibration amplitude" and related parameters such as "bearing temperature" and "motor current" in the matrix will increase accordingly to reflect the strong coupling relationship between data when a fault occurs, thus clearly presenting the correlation network at the data level.

[0093] Meanwhile, graph neural networks employ an attention mechanism in source tracing analysis to enhance the focus on key parameter nodes. This attention mechanism enables the graph neural network to automatically assign different attention weights based on the degree of influence of each node on fault propagation when processing the graph structure composed of equipment parameter nodes. This allows the model to focus more on key nodes highly relevant to the fault. For example, when the fault probability prediction result indicates a high risk of bearing wear, the attention mechanism will increase the weights of vibration sensor data nodes, bearing temperature monitoring nodes, and related speed process parameter nodes, strengthening the influence of these nodes in message passing and feature aggregation processes, and effectively filtering out interference from secondary parameters. In this way, the graph neural network can more accurately capture the critical path of fault propagation from the initial node to other related nodes, accurately identify fault types (such as bearing wear, abnormal temperature, etc.), locate the specific location of the fault (such as a specific bearing, motor component, etc.), and assess the impact range of the fault on surrounding equipment and system modules based on the coverage of the correlation strength in the dynamic correlation weight matrix. Finally, it outputs a diagnostic report containing the above information, providing accurate analytical basis for fault source tracing.

[0094] Corresponding to the above-described equipment fault diagnosis method, the present invention also proposes an equipment fault diagnosis device. Since the device embodiments of the present invention correspond to the above-described method embodiments, details not disclosed in the device embodiments can be referred to the above-described method embodiments, and will not be repeated here.

[0095] Figure 2This is a schematic diagram of the structure of a device for diagnosing equipment faults provided in an embodiment of this disclosure, as shown below. Figure 2 As shown, it includes:

[0096] Acquisition unit 21 is used to acquire process parameter data of the distributed control system and status monitoring data of sensing equipment;

[0097] Analysis unit 22 is used to analyze the process parameter data and the status monitoring data based on a pre-trained joint inference model to generate equipment failure probability prediction results.

[0098] Output unit 23 is used to perform source analysis on the fault propagation path of the fault probability prediction result of the equipment based on graph neural network, and output a diagnostic report on fault type, occurrence location and impact range.

[0099] Furthermore, in one possible implementation of the embodiments of this disclosure, such as Figure 3 As shown, the device further includes:

[0100] Processing unit 24 is used to perform time stamp synchronization and spatial coordinate mapping on the process parameter data and the status monitoring data before the analysis unit 22 analyzes the process parameter data and the status monitoring data based on the pre-trained joint inference model and generates equipment failure probability prediction results, so as to achieve spatiotemporal alignment.

[0101] The processing unit 24 is also used to perform data cleaning, format conversion and feature extraction on the spatiotemporally aligned process parameter data and the status monitoring data, and to unify the format of the data.

[0102] Furthermore, in one possible implementation of this disclosure, the processing unit 24 is further configured to:

[0103] The process parameter data and the status monitoring data are respectively timestamped and calibrated.

[0104] By mapping three-dimensional coordinate information to the identifiers of corresponding device nodes in the distributed control system, a unified device space topology is constructed.

[0105] Furthermore, in one possible implementation of this disclosure, the processing unit 24 is further configured to:

[0106] The process parameter data and the status monitoring data are cleaned based on an outlier detection algorithm;

[0107] The process parameter data and the status monitoring data after cleaning are subjected to feature engineering processing, including time-domain statistical feature extraction, frequency-domain transform analysis and wavelet decomposition.

[0108] Furthermore, in one possible implementation of this disclosure, the pre-trained joint inference model includes a physical model and a machine learning model; the analysis unit 22 is further configured to:

[0109] The simulation results of the physical model are used as input features of the machine learning model to improve the machine learning model's ability to understand the operating mechanism of the equipment.

[0110] The pre-trained joint inference model is optimized using an online learning mechanism.

[0111] Furthermore, in one possible implementation of this disclosure, the output unit 23 is further configured to:

[0112] Construct a dynamic correlation weight matrix between the process parameter data and the status monitoring data;

[0113] The graph neural network employs an attention mechanism to enhance its focus on key parameter nodes, thereby improving the accuracy of fault tracing.

[0114] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of the embodiments of this disclosure, and the principle is the same. Therefore, the embodiments of this disclosure are not limited thereto.

[0115] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0116] Figure 4 A schematic block diagram of an example electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0117] like Figure 4As shown, device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 302 or a computer program loaded from storage unit 308 into RAM (Random Access Memory) 303. RAM 303 can also store various programs and data required for the operation of device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via bus 304. I / O (Input / Output) interface 305 is also connected to bus 304.

[0118] Multiple components in device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of monitors, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0119] The computing unit 301 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as methods for diagnosing device faults. For example, in some embodiments, the methods for diagnosing device faults may be implemented as computer software programs tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, the computing unit 301 may be configured to perform the aforementioned diagnostic method for device faults by any other suitable means (e.g., by means of firmware).

[0120] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0121] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0122] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0123] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0124] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0125] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server, or simply "VPS") services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0126] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0127] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0128] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for diagnosing equipment faults, characterized in that, include: Acquire process parameter data and status monitoring data of sensing equipment in the distributed control system; Based on the pre-trained joint inference model, the process parameter data and the status monitoring data are analyzed to generate equipment failure probability prediction results. Based on the graph neural network, the fault propagation path of the equipment fault probability prediction result is traced and analyzed, and a diagnostic report is outputting the fault type, occurrence location and impact range.

2. The method according to claim 1, characterized in that, Before analyzing the process parameter data and the condition monitoring data based on the pre-trained joint inference model to generate equipment failure probability prediction results, the method further includes: The process parameter data and the status monitoring data are time-stamped and spatially mapped to achieve spatiotemporal alignment. Data cleaning, format conversion, and feature extraction are performed on the spatiotemporally aligned process parameter data and status monitoring data, and the data format is unified.

3. The method according to claim 2, characterized in that, The step of synchronizing the process parameter data and the status monitoring data with timestamps and mapping spatial coordinates to achieve spatiotemporal alignment includes: The process parameter data and the status monitoring data are respectively timestamped and calibrated. By mapping three-dimensional coordinate information to the identifiers of corresponding device nodes in the distributed control system, a unified device space topology is constructed.

4. The method according to claim 2, characterized in that, The process parameter data and status monitoring data after spatiotemporal alignment are subjected to data cleaning, format conversion, and feature extraction, and the data format is unified as follows: The process parameter data and the status monitoring data are cleaned based on an outlier detection algorithm; The process parameter data and the status monitoring data after cleaning are subjected to feature engineering processing, including time-domain statistical feature extraction, frequency-domain transform analysis and wavelet decomposition.

5. The method according to claim 1, characterized in that, The pre-trained joint inference model includes a physical model and a machine learning model; the analysis of the process parameter data and the condition monitoring data based on the pre-trained joint inference model to generate equipment failure probability prediction results includes: The simulation results of the physical model are used as input features of the machine learning model to improve the machine learning model's ability to understand the operating mechanism of the equipment. The pre-trained joint inference model is optimized using an online learning mechanism.

6. The method according to claim 1, characterized in that, The fault propagation path analysis based on the fault probability prediction results of the equipment using graph neural networks, and the output of a diagnostic report including fault type, location of occurrence, and scope of impact, include: Construct a dynamic correlation weight matrix between the process parameter data and the status monitoring data; The graph neural network employs an attention mechanism to enhance its focus on key parameter nodes, thereby improving the accuracy of fault tracing.

7. A diagnostic device for equipment faults, characterized in that, include: The acquisition unit is used to acquire process parameter data of the distributed control system and status monitoring data of sensing equipment; The analysis unit is used to analyze the process parameter data and the status monitoring data based on a pre-trained joint inference model to generate equipment failure probability prediction results. The output unit is used to perform source analysis on the fault propagation path of the fault probability prediction result of the equipment based on the graph neural network, and output a diagnostic report on the fault type, occurrence location and impact range.

8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images