AR-based thermoelectric unit fault visual diagnosis method and system

Abstract

The invention provides an AR-based thermoelectric unit fault visual diagnosis method and system, and the method comprises the steps: obtaining a thermoelectric unit operation data set, constructing an AR three-dimensional scene model containing a virtual part model and a spatial position association relation, carrying out the fault feature extraction of the thermoelectric unit operation data set, and carrying out the fault feature extraction of the AR three-dimensional scene model; the method comprises the steps of obtaining a fault feature set reflecting equipment operation state abnormity, performing space mapping on the fault feature set and an AR three-dimensional scene model, and generating fault visualization information which is superposed on the surface of a virtual component model and comprises abnormity type identification, influence range and development trend description; according to the method, the AR interaction device performs interaction diagnosis operation on the fault visualization information, including checking of abnormal parameters, simulation of fault isolation and generation of processing suggestions, so that visual and accurate diagnosis of the faults of the thermoelectric unit is realized, and the fault processing efficiency and convenience are improved.

Description

Technical Field

[0001] This invention relates to the field of augmented reality technology, and more specifically, to an AR-based method and system for visual diagnosis of thermal power unit faults. Background Technology

[0002] In the operation and maintenance of thermal power units, the accuracy and timeliness of fault diagnosis are crucial to ensuring the safe and stable operation of the units. Traditional fault diagnosis methods for thermal power units mainly rely on manual inspections and experience-based judgment. This approach is not only inefficient but also easily affected by human factors, making it difficult to guarantee the accuracy and timeliness of fault diagnosis. With the rapid development of information technology, although some data analysis-based fault diagnosis methods have been gradually applied to thermal power units, these methods often only provide abstract data reports and cannot intuitively show the specific location and scope of the fault, causing great inconvenience to maintenance personnel in handling faults. Therefore, a visual diagnostic method that can intuitively and accurately display fault information of thermal power units is needed to improve the efficiency and accuracy of fault diagnosis. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an AR-based method for visual diagnosis of thermal power unit faults, the method comprising: Acquire a set of thermal power unit operating data, which includes real-time equipment monitoring data, historical fault record data, and component structural parameter data; Based on the thermal power unit's operating data set, an AR 3D scene model of the thermal power unit is constructed. The AR 3D scene model includes virtual component models corresponding to the actual physical structure of the unit and the spatial positional relationships of each component. The thermal power unit's operating data set is processed by fault feature extraction to obtain a fault feature set reflecting abnormal equipment operating status. The fault feature set includes abnormal component temperature features, abnormal vibration features, and abnormal pressure fluctuation features. The fault feature set is spatially mapped to the AR 3D scene model to generate fault visualization information superimposed on the surface of the virtual component model. The fault visualization information includes anomaly type identifier, anomaly impact range, and anomaly development trend description. The AR interactive device performs interactive diagnostic operations on the visualized fault information. These operations include viewing detailed abnormal parameters, simulating fault isolation operations, and generating fault handling suggestions.

[0004] In another aspect, embodiments of the present invention also provide an AR-based thermal power unit fault visualization diagnostic system, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0005] Based on the above, this embodiment of the invention constructs an AR three-dimensional scene model of the thermal power unit, realizing a virtual reproduction of the unit's physical structure and providing maintenance personnel with an intuitive and three-dimensional view of the unit. By extracting fault features from the thermal power unit's operating data set and spatially mapping the fault feature set with the AR three-dimensional scene model, fault visualization information superimposed on the surface of the virtual component model is generated. This allows maintenance personnel to intuitively see the specific location of the fault, the type of anomaly, the scope of impact, and the development trend, greatly improving the accuracy and timeliness of fault diagnosis. In addition, by performing interactive diagnostic operations on the fault visualization information through AR interactive devices, maintenance personnel can easily view detailed anomaly parameters, simulate fault isolation operations, and generate fault handling suggestions, further improving the efficiency and convenience of fault handling. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of the execution flow of the AR-based thermal power unit fault visualization diagnosis method provided in the embodiments of the present invention.

[0007] Figure 2 This is a schematic diagram of exemplary hardware and software components of the AR-based thermal power unit fault visualization diagnostic system provided in an embodiment of the present invention. Detailed Implementation

[0008] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating an AR-based thermal power unit fault visualization diagnosis method according to an embodiment of the present invention. The following is a detailed description of the AR-based thermal power unit fault visualization diagnosis method.

[0009] Step S110: Obtain the thermal power unit operation data set, which includes real-time equipment monitoring data, historical fault record data, and component structural parameter data.

[0010] Step S111: Collect real-time monitoring data through a sensor network deployed on key components of the thermal power unit. The real-time monitoring data includes temperature sensing data, vibration frequency data, and pressure transmission data of each component.

[0011] In the operation of a thermal power unit, the stable operation of key components is crucial to ensuring the overall efficiency of the unit. To monitor the operating status of these key components in real time and with high accuracy, a dedicated sensor network needs to be deployed on them. This sensor network consists of multiple types of sensors working together, each collecting data for specific operating parameters.

[0012] Temperature data acquisition relies on temperature sensors installed on key components of the thermal power unit. Taking the boiler as an example, a crucial component, the boiler generates a large amount of heat during operation due to internal combustion, and its internal temperature is constantly changing. Temperature sensors are precisely installed in different parts of the boiler, such as the furnace wall and furnace chamber, to continuously monitor temperature changes in these areas. These sensors convert the sensed temperature information into electrical signals, which are then processed through analog-to-digital conversion and transmitted digitally to the data acquisition system. Assuming the boiler is operating normally, its furnace wall temperature should be maintained within a set, reasonable range. When the furnace wall temperature collected by the temperature sensors exceeds this range, it may indicate potential boiler malfunctions such as abnormal combustion or poor heat dissipation.

[0013] The acquisition of vibration frequency data relies on vibration sensors installed on the rotating components of the thermal power unit. For example, during high-speed rotation, the vibration state of a steam turbine rotor reflects its balance and mechanical performance. Vibration sensors can accurately detect parameters such as the rotor's vibration frequency and amplitude during operation and convert these vibration signals into digital signals. Under normal circumstances, the vibration frequency of the steam turbine rotor should be within a stable range. If the vibration frequency acquired by the vibration sensor shows abnormal fluctuations, such as a sudden increase in frequency or amplitude, it may indicate problems such as rotor imbalance, bearing wear, or loose connections. If these problems are not addressed promptly, they may lead to more serious equipment damage.

[0014] Pressure transmission data is acquired using pressure sensors installed in the piping system and pressure vessels of the thermal power unit. In the steam pipelines of the thermal power unit, stable steam pressure is crucial for the unit's normal operation. Pressure sensors monitor real-time pressure changes within the pipelines and transmit this pressure information as electrical signals to the data acquisition system. If the pressure value collected by the pressure sensor deviates from the normal pressure range, it may indicate leaks, blockages, or abnormal valve regulation in the pipeline. These problems can affect the normal delivery of steam and the overall performance of the thermal power unit.

[0015] Step S112: Retrieve historical fault record data from the thermal power unit operation and maintenance management system. The historical fault record data includes the fault occurrence time, fault phenomenon description and corresponding handling solution information of each component within a preset time period in the past.

[0016] The thermal power unit operation and maintenance management system is an integrated data management platform that centrally stores various information from the long-term operation of thermal power units, among which historical fault records are an important component. Accessing this system allows for the retrieval of historical fault records within a preset time period. The preset time period can be flexibly adjusted according to the operating characteristics of the thermal power unit and the needs of fault analysis; typically, a relatively long period is chosen to cover a sufficient number of fault cases, thereby enabling a more comprehensive analysis of fault patterns.

[0017] Historical fault occurrence time information in historical fault records has significant analytical value, helping to understand the temporal distribution patterns of faults in various components of a thermal power unit. For example, certain components may be prone to failure during specific seasons or operating cycles. Statistical analysis of fault occurrence times allows for the development of corresponding maintenance plans in advance, reducing the probability of faults.

[0018] The fault description section details the specific behavior of each component of the thermal power unit when a fault occurs. Taking the generator as an example, when a generator malfunctions, the fault description may include abnormal fluctuations in the generator's output power, abnormal noises, and localized temperature increases. This fault description provides direct evidence for accurate fault diagnosis; by comparing the current fault symptoms with historical fault descriptions, possible causes of the fault can be quickly identified.

[0019] The corresponding handling plan information records the specific handling measures and solutions taken for each fault. For example, in past fault cases, if the generator experienced unstable output power and the fault was successfully resolved by adjusting the excitation current and checking the winding connections, these handling plans can help maintenance personnel quickly formulate effective maintenance strategies when encountering similar faults.

[0020] Step S113: Access the thermal power unit design document database to obtain component structural parameter data. The component structural parameter data includes the geometric dimensions, material properties, and connection methods of each component with adjacent components.

[0021] The thermal power unit design document database is a collection of various documents and data generated during the design and manufacturing process of thermal power units. Accessing this database allows users to obtain detailed structural parameter data for each component of the thermal power unit.

[0022] Geometric dimensional information describes the external dimensions, shape, and other characteristics of various components in a thermal power unit. Taking turbine blades as an example, their length, width, thickness, and other geometric dimensions directly affect their aerodynamic performance and mechanical strength. Accurate geometric dimensional information helps in determining whether components have undergone deformation, wear, or other defects during fault diagnosis. For instance, if a slight change in blade length is found during inspection, it may indicate that the blade has been subjected to external impact or has fatigue damage.

[0023] Material property information records the types, physical properties, and chemical properties of materials used in various components of a thermal power unit. Different materials possess different characteristics, such as strength, hardness, heat resistance, and corrosion resistance. Understanding the material properties of components helps analyze their performance under different operating conditions and the types of failures that may occur. For example, components operating in high-temperature environments may experience problems such as thermal deformation and oxidation if their materials have insufficient heat resistance.

[0024] The connection information describes the connection relationships and methods between the various components of the thermal power unit. These connection methods include bolted connections, welding, and flanged connections. Different connection methods have different characteristics and applicable scopes. For example, bolted connections offer advantages such as detachability and ease of maintenance, but may carry the risk of loosening; welded connections offer high connection strength and good sealing, but are more difficult to repair if problems arise. Understanding the connection information between components can help determine whether a fault is related to the connection points during fault diagnosis and how to take appropriate maintenance measures.

[0025] Step S114: Perform format unification processing on the real-time monitoring data, historical fault record data and component structural parameter data of the equipment to generate a standardized data set with the same timestamp and component identification code.

[0026] After acquiring real-time monitoring data, historical fault record data, and component structural parameter data, it's important to understand that these data come from different data sources, and their formats and structures may vary. To facilitate subsequent data analysis and processing, it's necessary to standardize the format of this data.

[0027] First, the real-time monitoring data from the equipment is processed. This data is collected in real-time, and its timestamps reflect the specific moment of data acquisition. During the format standardization process, the data format needs to be standardized to ensure that all data such as temperature, vibration frequency, and pressure are represented using the same units and precision. Simultaneously, a unique component identification code is added to each data point to clearly identify the corresponding thermal power unit component.

[0028] Historical fault record data also needs to be formatted. The fault occurrence time should be converted to a uniform time format, and a corresponding component identification code should be added to each fault record. In addition, the fault phenomenon description and handling solution information should be standardized in text processing, such as standardizing terminology and removing redundant information, to facilitate subsequent text analysis and matching.

[0029] The component structural parameter data also needs to be formatted. Geometric dimensions, material properties, and connection methods are organized according to a unified structure, and a unique identifier is added to each component. After completing the above format specifications, the same timestamp is added to all data; this timestamp can be the start time of data acquisition or a unified reference time. Through this process, a standardized dataset with identical timestamps and component identifiers is generated.

[0030] Step S115: Perform missing value imputation and noise filtering on the standardized data set to obtain the thermal power unit operation data set.

[0031] The standardized dataset obtained after format unification may contain missing values ​​and noisy data. Missing values ​​may be due to sensor malfunctions, data transmission interruptions, etc., while noisy data may be caused by external interference, sensor errors, and other factors. To ensure data quality and accuracy, the standardized dataset needs to undergo missing value imputation and noise filtering.

[0032] There are several methods for imputing missing values. One common method is based on statistical analysis, such as using the mean, median, or mode to fill in the missing values. For example, if a temperature sensor in a device's real-time monitoring data set has missing data at a certain moment, the mean of the temperature data collected by that sensor at other moments can be calculated and used to fill in the missing temperature value. Another method is based on model prediction, such as using a machine learning model to predict missing values ​​based on other relevant data.

[0033] The purpose of noise filtering is to remove interfering information from data, making the data smoother and more accurate. Noise filtering can be achieved using filtering algorithms, such as moving average filtering and median filtering. Taking moving average filtering as an example, for vibration frequency data in real-time equipment monitoring data, its average value within a fixed window is calculated. This average value is then used to replace the original data within the window, effectively smoothing the data and reducing the impact of noise.

[0034] By imputing missing values ​​and filtering noise from the standardized dataset, the final set of thermal power unit operating data was obtained, which is of higher quality and more accurate.

[0035] Step S120: Construct an AR three-dimensional scene model of the thermal power unit based on the thermal power unit operation data set. The AR three-dimensional scene model includes virtual component models corresponding to the actual physical structure of the unit and the spatial position association of each component.

[0036] After acquiring a high-quality set of operating data for thermal power units, in order to achieve visualized diagnosis of thermal power unit faults, it is necessary to construct an AR 3D scene model of the thermal power unit based on this operating data set. This AR 3D scene model can intuitively and three-dimensionally display the physical structure and operating status of the thermal power unit, providing a visual auxiliary tool for fault diagnosis.

[0037] Step S121: Extract the component structural parameter data from the thermal power unit operation data set, and construct the basic three-dimensional model of each component based on the geometric dimension information and connection method information.

[0038] For geometric dimensional information, taking the steam turbine of a thermal power unit as an example, based on the geometric dimensional data such as the length, width, and thickness of its blades, and the diameter and height of its impeller, 3D modeling software can be used to accurately construct 3D models of the turbine blades and impeller. These models are presented in a virtual 3D form in the software, and their dimensions are completely consistent with the actual components. By analyzing and processing the geometric dimensional information of different components, the 3D models of each component are constructed one by one.

[0039] The connection method information determines how the connection relationships between components are represented in the 3D model. For example, when it is known that the turbine blades and impeller are connected by tenons, the 3D model can accurately simulate the structure of the tenons and mortises at the corresponding positions on the blades and impeller to reflect this connection method. Similarly, the connection methods between other components, such as bolted connections and welding, will also be accurately simulated and displayed in the 3D model.

[0040] By comprehensively utilizing the geometric dimensions and connection information in the component structural parameter data, the basic three-dimensional models of each component of the thermal power unit are gradually constructed.

[0041] Step S122: Match and associate the basic 3D model with the component spatial coordinate information in the real-time monitoring data of the equipment, and adjust the spatial coordinate parameters of each basic 3D model to restore the physical layout of the actual unit.

[0042] After constructing the basic 3D models of each component, the basic 3D models need to be accurately placed in the virtual space to restore the actual physical layout of the thermal power unit. This requires matching and associating the basic 3D models with the spatial coordinate information of the components in the real-time monitoring data of the equipment.

[0043] The real-time monitoring data of the equipment includes the spatial coordinates of each component of the thermal power unit, recording the specific location of each component in the actual physical space. For example, the relative position of the boiler within the thermal power unit, the distance and angle between the turbine and the generator, etc. By aligning and matching the coordinate system of the basic 3D model with the spatial coordinate system in the real-time monitoring data, the initial position of each basic 3D model in virtual space is determined.

[0044] Then, based on the spatial coordinate information of the components, the spatial coordinate parameters of each basic 3D model are adjusted. Assuming that in the actual physical layout, the turbine is located to the left of the generator, with a certain distance between them, in the virtual space, the coordinate parameters of the turbine's basic 3D model need to be adjusted accordingly to ensure it is accurately located to the left of the generator's basic 3D model, maintaining the same distance as in reality. Through these adjustments, the positional relationship of each basic 3D model in the virtual space is consistent with the physical layout of the actual unit.

[0045] Step S123: Extract component deformation information when typical faults occur from the historical fault record data, perform morphological correction processing on the basic three-dimensional model of the corresponding component, and generate a multi-morphological virtual component model that includes normal state and typical fault state.

[0046] Historical fault records contain a wealth of fault information, among which component deformation information during typical faults is of great value for constructing multi-morphological virtual component models. By extracting and utilizing this deformation information, virtual component models can be made to represent different forms under normal and typical fault conditions.

[0047] First, typical failure cases are selected from historical failure records, and information related to component deformation is extracted. For example, in one failure, the turbine blades bent and deformed due to prolonged high-speed rotation and high temperatures. Historical failure records may contain detailed descriptions of the degree of bending and the location of the deformation.

[0048] Then, based on the extracted component deformation information, the basic 3D model of the corresponding component is subjected to morphological correction processing. For the turbine blade example above, the deformation tool in the 3D modeling software is used to perform bending deformation operations on the basic 3D model of the blade according to the degree of bending and deformation location in the fault record, generating a 3D model of the blade under fault conditions.

[0049] By processing the deformation information of different components under different typical faults, a three-dimensional model of each component under normal state and multiple typical fault states is generated, thereby forming a multi-form virtual component model that includes normal state and typical fault states. The multi-form virtual component model can more realistically reflect the actual state of thermal power unit components under different operating conditions.

[0050] Step S124: Analyze the connection relationship between each virtual component model, and add virtual interface identifiers representing the connection relationship between adjacent virtual component models based on the connection method information. The virtual interface identifiers include interface type and connection strength description.

[0051] After constructing various virtual component models, it is necessary to further clarify the connection relationships between the virtual component models and intuitively display these connection relationships by adding virtual interface identifiers.

[0052] First, based on the connection information in the component structural parameter data, analyze the connection methods between the virtual component models. For example, the boiler and steam pipes might be connected via flanges, while the turbine and generator might be connected via couplings. Different connection methods have different characteristics and performance.

[0053] Then, based on this connection information, virtual interface identifiers are added between adjacent virtual component models. These virtual interface identifiers include the interface type and a connection strength description. The interface type specifies the exact connection method, such as flange connection, bolted connection, or welding. The connection strength description reflects the strength and reliability of the connection. For example, for a bolted connection, the connection strength description could include information such as bolt specifications and tightening torque. In the virtual space, these virtual interface identifiers are displayed in specific graphic or symbolic form at the connection points of adjacent virtual component models, allowing operators to intuitively understand the connection relationships and characteristics between components.

[0054] Adding virtual interface identifiers not only enhances the visualization of AR 3D scene models but also provides more detailed information for fault diagnosis. For example, when a component malfunctions, the virtual interface identifiers can be used to determine whether the fault is related to the connection points and whether the connection strength is sufficient.

[0055] Step S125: Map the multi-form virtual component model and virtual interface identifier to the physical space of the real unit using the AR coordinate system calibration algorithm to generate an AR three-dimensional scene model of the thermal power unit that is aligned with the actual scene location.

[0056] In order to ensure that the constructed multi-form virtual component models and virtual interface identifiers can be accurately aligned with the physical space of the real thermal power unit, it is necessary to use the AR coordinate system calibration algorithm for mapping processing.

[0057] The core of the AR coordinate system calibration algorithm is to accurately match the coordinate system in virtual space with the coordinate system in real physical space. First, multiple reference points are selected at the actual thermal power unit site; the locations of these reference points are known in physical space. Simultaneously, the same reference points are set at corresponding locations in virtual space. By measuring the coordinates of these reference points in both physical and virtual spaces, the transformation relationship between the two coordinate systems is calculated, including parameters such as translation, rotation, and scaling.

[0058] Then, based on the calculated transformation relationship, the multi-form virtual component models and virtual interface identifiers are mapped from the virtual space to the physical space of the actual unit. During the mapping process, it is ensured that the position and orientation of each virtual component model and virtual interface identifier are completely consistent with the corresponding components and connection points in the actual unit.

[0059] Through AR coordinate system calibration algorithms, an AR 3D scene model of the thermal power unit aligned with the actual scene location was generated. When operators observe this AR 3D scene model through AR devices, they can see that the virtual model is perfectly integrated with the real unit, as if the virtual model truly exists in physical space. This provides an intuitive and accurate virtual scene for the visual diagnosis of thermal power unit faults, enabling operators to more easily perform fault diagnosis and analysis.

[0060] Step S130: Perform fault feature extraction processing on the thermal power unit operation data set to obtain a fault feature set reflecting abnormal equipment operation status. The fault feature set includes abnormal component temperature features, abnormal vibration features, and abnormal pressure fluctuation features.

[0061] In order to accurately identify abnormal conditions in the operation of equipment, it is necessary to perform fault feature extraction processing on the thermal power unit operation data set, extract the fault-related feature information, and form a fault feature set.

[0062] Step S131: Perform time series analysis on the temperature sensing data in the real-time monitoring data of the device, extract the temperature change rate parameters of each component at continuous time points, and generate component temperature anomaly characteristics based on the comparison results of the temperature change rate parameters and the preset temperature threshold.

[0063] The temperature sensing data in the real-time monitoring data of the equipment records the temperature values ​​of various components of the thermal power unit at different time points. In order to extract fault-related features from this temperature data, time series analysis is required. Time series analysis is a method used to study the changing patterns of data over time. By performing time series analysis on the temperature sensing data, the rate of temperature change parameters of each component at continuous time points can be extracted.

[0064] Specifically, the temperature change rate parameter is calculated based on the ratio of the temperature difference between adjacent time points to the time interval. Assuming that at time points t1 and t2, the temperatures of a component are T1 and T2 respectively, and the time interval is Δt, the temperature change rate of that component during this period can be obtained by calculating (T2-T1) / Δt. By performing this calculation on the temperature values ​​of each component at different time points in the real-time monitoring data of the equipment, a sequence of temperature change rate parameters for each component at continuous time points can be obtained.

[0065] After obtaining the temperature change rate parameter, it needs to be compared with a preset temperature threshold. The preset temperature threshold is a reasonable range determined based on the design requirements and historical operating experience of the thermal power unit. When the temperature change rate parameter of a component exceeds this preset temperature threshold, it indicates that the temperature change of that component is abnormal and may indicate a potential malfunction. For example, if the temperature of a component rises rapidly in a short period of time, and its temperature change rate exceeds the preset temperature threshold, this may mean that the component has an overheating problem, such as an internal short circuit or poor heat dissipation. Based on this comparison result, a component temperature anomaly characteristic can be generated. This component temperature anomaly characteristic can be represented by a vector, where each element in the vector corresponds to a temperature anomaly of a component; for example, 1 represents a temperature anomaly, and 0 represents a normal temperature.

[0066] Step S132: Perform Fourier transform processing on the vibration frequency data in the real-time monitoring data of the equipment, extract the main frequency components and secondary frequency components of the vibration signals of each component, and analyze the degree of matching between the main frequency components and the natural frequency of the components to generate vibration anomaly characteristics.

[0067] The vibration frequency data in the real-time monitoring data of the equipment is a time-varying signal that contains vibration information of various components of the thermal power unit during operation. To extract fault-related features from these vibration signals, Fourier transform processing is required. Fourier transform is a method for converting time-domain signals to frequency-domain signals. Through Fourier transform, vibration frequency data can be transformed from the time domain to the frequency domain, thereby obtaining the frequency component distribution of the vibration signals of each component.

[0068] After performing a Fourier transform, the resulting frequency domain signal contains multiple frequency components. The dominant frequency component refers to the frequency component with the largest amplitude in the frequency domain signal, reflecting the main characteristics of the component's vibration. Secondary frequency components refer to other frequency components besides the dominant frequency component, which may be caused by some minor vibration factors of the component.

[0069] After extracting the dominant and secondary frequency components of the vibration signals from each component, it is necessary to analyze the degree of matching between the dominant frequency components and the component's natural frequency. The component's natural frequency refers to the vibration frequency of the component in a free vibration state, which is determined by the component's physical structure and material properties. Under normal operating conditions, the dominant frequency component of the component should match its natural frequency. If the dominant frequency component does not match the component's natural frequency, or if additional frequency components appear, it may indicate an abnormal vibration in the component. For example, when the component's dominant frequency component deviates from its natural frequency, it may be due to factors such as component imbalance, looseness, or wear.

[0070] Vibration anomaly characteristics can be generated based on the degree of matching between the dominant frequency component and the natural frequency of the component. These characteristics can also be represented by a vector, where each element corresponds to a vibration anomaly in the component. For example, based on the degree of deviation between the dominant frequency component and the component's natural frequency, vibration anomalies can be categorized into different levels, each represented by a different numerical value.

[0071] Step S133: Perform window sliding statistical processing on the pressure transmission data in the real-time monitoring data of the device, calculate the maximum value, minimum value and variance parameter of the pressure value within the preset time window, and generate pressure fluctuation abnormal characteristics based on the degree of deviation between the variance parameter and the standard pressure fluctuation range.

[0072] The pressure transmission data in the real-time monitoring data of the equipment records the pressure values ​​of various components of the thermal power unit at different time points. In order to extract fault-related features from these pressure data, window sliding statistical processing is required. Window sliding statistical processing is a method for performing local statistical analysis on time series data. By setting a preset time window, the window is slid across the data series, and statistical calculations are performed on the data within the window.

[0073] Specifically, for pressure transmission data, at each time point, pressure values ​​within a preset time window are selected for statistical analysis. The maximum, minimum, and variance parameters of the pressure values ​​within this time window are calculated. The maximum value reflects the highest pressure value within the time window, the minimum value reflects the lowest value, and the variance parameter reflects the degree of fluctuation of the pressure value within the time window.

[0074] After obtaining the variance parameter, it needs to be compared with the standard pressure fluctuation range. The standard pressure fluctuation range is a reasonable pressure fluctuation interval determined based on the design requirements and historical operating experience of the thermal power unit. When the variance parameter exceeds this standard pressure fluctuation range, it indicates that the pressure fluctuation of the component is abnormal. For example, if the pressure of a component fluctuates drastically in a short period of time, and its variance parameter is much larger than the standard pressure fluctuation range, this may mean that the component has a pressure instability problem, such as pipeline blockage or valve failure.

[0075] Pressure fluctuation anomaly characteristics can be generated based on the degree of deviation of the variance parameter from the standard pressure fluctuation range. These characteristics can also be represented by a vector, where each element corresponds to a pressure fluctuation anomaly in a component. For example, pressure fluctuation anomalies can be categorized into different levels based on the magnitude of the deviation, and these levels can be represented by different numerical values.

[0076] Step S134: Perform similarity matching processing on the abnormal temperature characteristics, abnormal vibration characteristics, and abnormal pressure fluctuation characteristics of the component with the typical fault characteristics in the historical fault record data, and filter out the effective fault characteristics related to the current operating state.

[0077] After generating abnormal component temperature, vibration, and pressure fluctuation characteristics, these characteristics need to be matched with typical fault characteristics in historical fault records. Historical fault records document various typical faults that have occurred in the thermal power unit in the past, along with their corresponding characteristic information. By matching the current fault characteristics with historical typical fault characteristics, valid fault characteristics relevant to the current operating state can be filtered out.

[0078] Similarity matching can employ various methods, such as cosine similarity and Euclidean distance. Taking cosine similarity as an example, it measures the cosine of the angle between two vectors, indicating their degree of similarity. The cosine similarity of abnormal component temperature, vibration, and pressure fluctuation features with typical fault feature vectors from historical fault records is calculated. If a current fault feature vector has a high cosine similarity to a historical typical fault feature vector, it indicates that the two feature vectors are similar and may correspond to the same fault type.

[0079] By performing similarity matching calculations on all current fault feature vectors and historical typical fault feature vectors, fault features with high similarity are selected. The selected fault features are valid fault features related to the current operating state.

[0080] Step S135: Perform causal relationship analysis on the effective fault features, identify the correlation between abnormal features, and generate a fault feature set with a hierarchical structure.

[0081] After obtaining valid fault characteristics, causal relationship analysis is needed to gain a deeper understanding of the nature and propagation mechanism of the fault. Causal relationship analysis is a method used to study causal connections between things. By performing causal relationship analysis on valid fault characteristics, the correlation between abnormal characteristics can be identified.

[0082] There may be various relationships between abnormal features, such as causal relationships, parallel relationships, and progressive relationships. For example, an abnormal temperature in a component may lead to abnormal vibration in that component; in this case, there is a causal relationship between the temperature abnormality and the vibration abnormality. By performing causal relationship analysis on valid fault features, these relationships can be identified, and a hierarchical set of fault features can be constructed.

[0083] When constructing a hierarchical fault feature set, causal features in a cause-effect relationship can be used as upper-level features, and result features as lower-level features. For example, if temperature anomaly is the cause of vibration anomaly, then the temperature anomaly feature is at the upper level of the hierarchy, and the vibration anomaly feature is at the lower level. In this way, all valid fault features can be organized into a hierarchical structure according to their relationships. This hierarchical fault feature set can be represented by a tree structure, where the root node represents the most fundamental fault cause feature, branch nodes represent intermediate fault features, and leaf nodes represent the final fault manifestation features. This hierarchical fault feature set can more clearly demonstrate the development process and propagation path of the fault.

[0084] Step S140: Perform spatial mapping processing on the fault feature set and the AR three-dimensional scene model to generate fault visualization information superimposed on the surface of the virtual component model. The fault visualization information includes anomaly type identifier, anomaly impact range, and anomaly development trend description.

[0085] After obtaining the fault feature set and constructing the AR 3D scene model, in order to intuitively display the fault information in the virtual scene, it is necessary to perform spatial mapping processing between the fault feature set and the AR 3D scene model. Through the above mapping processing, fault visualization information superimposed on the surface of the virtual component model can be generated, enabling operators to more intuitively understand the fault status of the thermal power unit.

[0086] Step S141: Extract the component identification code corresponding to each abnormal feature in the fault feature set, and locate the corresponding virtual component model in the AR three-dimensional scene model according to the component identification code.

[0087] Each abnormal feature in the fault feature set corresponds to a component of the thermal power unit, and this correspondence is represented by a component identification code. First, the component identification code corresponding to each abnormal feature is extracted from the fault feature set. The component identification code is a unique identifier assigned to each component when acquiring the thermal power unit's operating data set, used to distinguish different components.

[0088] Then, based on the component identification code, the corresponding virtual component model is located in the AR 3D scene model. Each virtual component model in the AR 3D scene model has a corresponding component identification code. By matching the component identification codes, the virtual component model corresponding to the abnormal feature can be accurately found. For example, if the component identification code corresponding to a certain temperature abnormality feature in the fault feature set is A, then in the AR 3D scene model, the location of that component can be located in the virtual scene by searching for the virtual component model with component identification code A.

[0089] Step S142: Assign a visual identifier corresponding to the anomaly type to each located virtual component model. The visual identifier includes color coding and graphic symbols, wherein the color coding is used to distinguish the severity of the anomaly, and the graphic symbols are used to identify the anomaly type.

[0090] After locating the virtual component model corresponding to the abnormal characteristics, a visual identifier corresponding to the abnormality type needs to be assigned to the virtual component model. The visual identifier is an intuitive representation that conveys fault information through color coding and graphic symbols.

[0091] Color coding is used to distinguish the severity of anomalies. Different colors can represent different levels of anomaly severity; for example, red might indicate a severe anomaly, yellow a moderate anomaly, and green normal or mild anomaly. Based on the severity of the anomaly feature, a corresponding color code is assigned to each located virtual component model. For example, if a component's temperature anomaly is very severe, exceeding a preset severity threshold, then the virtual component model for that component is assigned a red color code.

[0092] Graphical symbols are used to identify anomaly types. Different symbols represent different anomaly types; for example, a triangle might represent a temperature anomaly, a circle a vibration anomaly, and a square a pressure fluctuation anomaly. Based on the type of anomaly, a corresponding graphic symbol is assigned to each located virtual component model. For example, if a component's anomaly is vibration, a circular graphic symbol is displayed on the surface of its virtual component model. In this way, operators can quickly understand the type and severity of the component's anomaly by observing the color coding and graphic symbols on the surface of the virtual component model.

[0093] Step S143: Analyze the spatial propagation characteristics of the abnormal features, combine the spatial positional relationships of the virtual component models in the AR 3D scene model, calculate the range of adjacent components affected by the abnormality, and generate a description of the range of abnormality impact containing the identifiers of the affected components.

[0094] Anomalies may propagate spatially within the thermal power unit, affecting adjacent components. To accurately understand the extent of the anomaly's impact, it is necessary to analyze the spatial propagation characteristics of the anomaly and perform calculations by combining this with the spatial positional relationships of virtual component models within the AR 3D scene model.

[0095] Step S1431: Extract the connection method information of the component corresponding to the abnormal feature, and determine its physical connection type with adjacent components. The physical connection type includes rigid connection, elastic connection and fluid conduction connection.

[0096] From the component structural parameter data of the thermal power unit's operating data set, information on the connection methods of components corresponding to abnormal characteristics is extracted. For example, for a component exhibiting an abnormal temperature, its connection with adjacent components is examined. If the component is tightly fixed to adjacent components with bolts and there is no relative elastic movement, then it can be determined that they are rigidly connected; if they are connected by elastic elements such as springs, then it is an elastic connection; if they are connected through pipes or other means to facilitate fluid transmission, then it is a fluid conduction connection.

[0097] Step S1432: Determine the main path of anomaly propagation based on the physical connection type and the physical properties of the anomaly characteristics. The physical properties include heat conduction properties, vibration transmission properties, and pressure diffusion properties.

[0098] Different types of physical connections and the physical properties of the anomalies determine the primary propagation path. For temperature anomalies, in rigid connections, heat conduction is likely the main propagation path, with heat directly transferred to adjacent components through the connection. In elastic connections, while heat conduction also occurs, it may be relatively weak, while vibration transmission may be accompanied by some heat transfer. In fluid conduction connections, heat may propagate with the flow of the fluid. For vibration anomalies, both rigid and elastic connections are likely to be the primary propagation path, while in fluid conduction connections, if the flow of the fluid causes vibration in the components, it may also become a propagation path. For pressure fluctuation anomalies, fluid conduction connections are the primary path for pressure diffusion, while pressure propagation is relatively limited in rigid and elastic connections.

[0099] Step S1433: Traverse the virtual component models in the AR 3D scene model along the main path and extract the set of components that have a direct or indirect connection with the initial abnormal component.

[0100] After determining the main path of anomaly propagation, the system traverses the AR 3D scene model along this main path, starting from the initial anomalous component. During the traversal, components directly connected to the initial anomalous component are identified, such as those rigidly connected. Indirect connections are also considered, such as those linked through other components. For example, if component A is rigidly connected to component B, and component B is elastically connected to component C, then component C is indirectly connected to component A. Components with direct or indirect connections to the initial anomalous component are extracted to form a component set.

[0101] Step S1434: For each component in the component set, calculate its connection strength parameter with the initial abnormal component. The connection strength parameter is determined based on the connection material strength and contact area parameters in the connection method information.

[0102] For each component in the component set, its connection strength parameter with the initial anomalous component needs to be calculated. The calculation of the connection strength parameter is based on the connection material strength and contact area parameters in the connection method information. Taking a rigid connection as an example, if the connection material has high strength, such as using high-strength steel, and the contact area is large, then the connection strength parameter will be relatively high; conversely, if the connection material has low strength and the contact area is small, the connection strength parameter will be low. For elastic connections, factors such as the elastic modulus of the connection material and the stiffness of the spring will affect the connection strength parameter; for fluid conduction connections, factors such as the pipe material, pipe diameter, and fluid velocity will affect the connection strength parameter. By comprehensively considering these factors, the connection strength parameter between each component and the initial anomalous component is calculated.

[0103] Step S1435: Select components whose connection strength parameters exceed a preset threshold as affected components, and generate an abnormal impact range description including the affected component identifier and connection strength parameters.

[0104] The preset threshold is a reasonable range determined based on the design requirements and historical experience of the thermal power unit. The connection strength parameters of each component in the component set are compared with the preset threshold, and components whose connection strength parameters exceed the preset threshold are identified. These components are considered to be affected by the abnormal characteristics. The identifiers of these affected components and their corresponding connection strength parameters are compiled to generate an abnormal impact range description containing the identifiers of the affected components and their connection strength parameters. This abnormal impact range description can be represented by a list, where each element contains the identifier of an affected component and its connection strength parameters to the originating abnormal component.

[0105] Step S144: Perform trend prediction processing on the time series data in the fault feature set, and generate a numerical change curve of the abnormal features in the future preset time period as a description of the abnormal development trend based on the abnormal development pattern in the historical fault record data.

[0106] The time-series data in the fault feature set records the changes of each anomalous feature over time. To understand the future development trend of these anomalous features, trend prediction processing of the time-series data is required. Various methods can be used for trend prediction processing, such as time-series analysis models and machine learning models.

[0107] In this embodiment, trend prediction is based on the abnormal development patterns in historical fault record data. Historical fault record data documents various abnormal situations that have occurred in the thermal power unit in the past and their development processes. By analyzing this data, patterns in abnormal development can be summarized. For example, some abnormal characteristics may rise slowly in the early stages, then increase rapidly at a certain point in time, and finally stabilize. Based on these patterns, a predictive model can be constructed to predict the time series data in the current fault characteristic set.

[0108] Specifically, current time-series data is used as input to the prediction model, which predicts the numerical values ​​of abnormal characteristics within a preset future time period based on historical anomaly development patterns. The prediction result can be represented by a numerical change curve, reflecting the development trend of the abnormal characteristics over the preset future time period. For example, if the temperature anomaly of a certain component is in an upward phase at the current time, the numerical change curve obtained through trend prediction processing can show whether the temperature anomaly will continue to rise, stabilize, or decline in the future. This anomaly trend description provides operators with information about the future development of the anomaly, helping to formulate countermeasures in advance.

[0109] Step S145: Render the visualization identifier, the description of the abnormal impact range, and the description of the abnormal development trend onto the surface of the corresponding virtual component model in a semi-transparent overlay manner to generate fault visualization information that can dynamically adjust the display position according to the AR viewpoint.

[0110] After obtaining the visual identifiers, descriptions of the anomaly's impact range, and descriptions of its development trend, this information needs to be rendered onto the surface of the corresponding virtual component model using a semi-transparent overlay method to generate fault visualization information. This semi-transparent overlay method allows the fault visualization information to be clearly displayed on the surface of the virtual component model without completely obscuring the model itself, facilitating observation and operation by the operator.

[0111] During the rendering process, leveraging the characteristics of AR technology, the display position of fault visualization information can dynamically adjust with changes in the AR viewing angle. When an operator observes an AR 3D scene model through an AR interactive device, the viewing angle changes. The fault visualization information automatically adjusts its display position on the surface of the virtual component model according to the change in viewing angle, always maintaining a correspondence with the virtual component model. For example, when an operator observes a virtual component model from different angles, the fault visualization information on the surface of that component will move and rotate accordingly with the rotation of the viewing angle, ensuring that the operator can clearly see the fault visualization information from any angle. Through this method, the generated fault visualization information provides operators with an intuitive and dynamic fault display interface, facilitating their fault diagnosis and analysis.

[0112] Step S150: Perform interactive diagnostic operations on the fault visualization information through an AR interactive device. The interactive diagnostic operations include viewing detailed abnormal parameters, simulating fault isolation operations, and generating fault handling suggestions.

[0113] After generating the fault visualization information, operators can use AR interactive devices to perform interactive diagnostic operations on the fault visualization information in order to gain a deeper understanding of the fault situation and formulate corresponding handling solutions.

[0114] Step S151: Select the target fault visualization information through the gesture recognition function of the AR interactive device, and trigger the detailed parameter display operation of the target fault visualization information. The detailed parameter display operation includes calling the original monitoring data and historical fault record data related to the abnormal characteristics in the thermal power unit operation data set.

[0115] AR interactive devices have gesture recognition capabilities, allowing operators to select target fault visualization information using gestures. For example, an operator can tap or swipe their finger on the screen of the AR interactive device to select fault visualization information on the surface of a virtual component model. Once the target fault visualization information is selected, the detailed parameters of that target fault visualization information will be displayed.

[0116] Detailed parameter display is achieved by accessing raw monitoring data and historical fault records related to the abnormal characteristics from the thermal power unit's operating data set. The raw monitoring data includes specific data corresponding to the abnormal characteristic from the equipment's real-time monitoring data, such as temperature sensing data, vibration frequency data, and pressure transmission data. Historical fault records contain relevant information about similar past faults, such as the fault occurrence time, fault phenomenon description, and corresponding handling solutions. By accessing this data, detailed parameters of the target fault can be displayed on the AR interactive device's screen, such as the specific values ​​of the abnormal characteristics and the handling of similar faults in the past. Operators can gain a more comprehensive understanding of the fault by viewing these detailed parameters.

[0117] Step S152: Perform a virtual isolation operation on the target virtual component model in the AR 3D scene model. The virtual isolation operation includes marking the isolation area through the spatial positioning function of the AR interactive device and simulating disconnecting the target component from the adjacent components.

[0118] After understanding the detailed parameters of the target fault visualization information, in order to further analyze the scope of the fault's impact and isolate the fault source, a virtual isolation operation can be performed on the target virtual component model in the AR 3D scene model. The virtual isolation operation first marks the isolation area using the spatial positioning function of the AR interactive device. The operator can specify an area on the AR interactive device screen using gestures or other operations; this area is the scope to be isolated.

[0119] Next, the connection between the target component and its adjacent components is simulated. In the AR 3D scene model, the connection between virtual component models is represented by virtual interface identifiers. During the virtual isolation operation, the connection between the target component and its adjacent components is simulated by modifying the state of the virtual interface identifiers. For example, if the target component and its adjacent components are connected by bolts, the virtual interface identifier representing the bolt connection can be set to the disconnected state during the virtual isolation operation, thereby simulating the actual isolation effect. Through the above virtual isolation operation, the propagation of the fault in the AR 3D scene model can be observed to determine whether the fault has been effectively isolated.

[0120] Step S153: Monitor the changes in the fault visualization information of other components in the AR 3D scene model after the virtual isolation operation is executed, and evaluate the effectiveness of the fault isolation measures based on the changes.

[0121] After performing virtual isolation, it is necessary to closely monitor the changes in the fault visualization information of other components in the AR 3D scene model in order to evaluate the effectiveness of the fault isolation measures.

[0122] Step S1531: Before performing the virtual isolation operation, record the initial state of all fault visualization information in the AR three-dimensional scene model. The initial state includes the color code of the anomaly type identifier, the number of affected components described by the anomaly impact range, and the slope of the numerical change curve described by the anomaly development trend.

[0123] Before performing virtual isolation operations, the initial state of all fault visualization information in the AR 3D scene model is recorded in detail. For the color coding of anomaly type identifiers, different colors represent different anomaly severity, and the color code corresponding to each component is recorded; for the description of the anomaly's impact range, the number of affected components is counted; for the description of the anomaly's development trend, the slope of the numerical change curve is obtained, which reflects the changing trend of the anomaly characteristics over time.

[0124] Step S1532: After performing the virtual isolation operation, update the fault visualization information in the AR three-dimensional scene model in real time, and obtain the current status of the updated fault visualization information. The current status includes the color code change of the anomaly type identifier, the change of the number of affected components describing the anomaly impact range, and the change of the slope of the numerical change curve describing the anomaly development trend.

[0125] After performing virtual isolation, the fault visualization information in the AR 3D scene model will change accordingly. The current state of the updated fault visualization information is obtained by updating it in real time. Corresponding to the initial state, the color coding change of the anomaly type identifier is recorded; for example, a change from red to yellow indicates a decrease in anomaly severity. The change in the number of affected components described by the anomaly's impact range is statistically analyzed; a decrease in the number of affected components indicates that the propagation of the fault has been somewhat controlled. The slope change of the numerical curve describing the anomaly's development trend is also obtained; a decrease in the slope may indicate that the development trend of the anomaly characteristics has been suppressed.

[0126] Step S1533: Calculate the difference parameters between the current state and the initial state. The difference parameters include the reduction in the severity of the color coding change, the reduction in the number of affected components, and the decrease in the slope of the numerical change curve.

[0127] Based on the recorded initial and current state information, the difference parameters between the two are calculated. For the reduction in the severity of the color code change, the severity level represented by the color code is quantified; for example, changing from the highest severity level color code to a lower severity level color code yields the numerical reduction in severity. For the reduction in the number of affected parts, the number of affected parts in the initial state is subtracted from the number of affected parts in the current state, and then divided by the number of affected parts in the initial state. For the decrease in the slope of the numerical change curve, the slope in the initial state is subtracted from the slope in the current state.

[0128] Step S1534: Based on the comparison results between the difference parameters and the preset effectiveness assessment threshold, generate an effectiveness assessment conclusion that includes effective, partially effective, or ineffective isolation measures.

[0129] The preset effectiveness assessment threshold is a reasonable range determined based on the design requirements and historical experience of the thermal power unit. The calculated difference parameters are compared with the preset effectiveness assessment threshold. If the reduction in the severity of color coding changes, the reduction in the number of affected components, and the decrease in the slope of the numerical change curve all exceed the preset effectiveness assessment threshold, then the fault isolation measures can be considered effective. If some parameters exceed the preset effectiveness assessment threshold while others do not, the fault isolation measures are considered partially effective. If all parameters do not exceed the preset effectiveness assessment threshold, the fault isolation measures are considered ineffective. Based on the comparison results, a corresponding effectiveness assessment conclusion is generated.

[0130] Step S1535: Associate and store the difference parameters and effectiveness evaluation conclusions with the specific execution steps of the virtual isolation operation to form a traceable record of the isolation operation effect.

[0131] The calculated difference parameters and generated effectiveness assessment conclusions are linked and stored with the specific execution steps of the virtual isolation operation. A database or log file can be created to record the difference parameters, effectiveness assessment conclusions, and detailed steps of the virtual isolation operation. This allows for easy reference during subsequent fault analysis and experience summarization, providing insights into the effectiveness and execution of each virtual isolation operation and offering a basis for improving fault isolation measures.

[0132] Step S154: Input the set of abnormal features, the description of the scope of abnormal impact, and the effectiveness evaluation results of the virtual isolation operation into the fault handling suggestion generation module, and then perform matching analysis based on successful handling cases in historical fault record data through the fault handling suggestion generation module.

[0133] After evaluating the effectiveness of fault isolation measures, relevant information is input into the fault handling suggestion generation module. This module then performs matching analysis based on successful handling cases in historical fault record data to obtain appropriate fault handling suggestions.

[0134] Step S1541: Perform feature vectorization processing on the abnormal feature set to generate a composite feature vector containing temperature abnormal feature vector, vibration abnormal feature vector and pressure fluctuation abnormal feature vector.

[0135] The set of abnormal features includes various types of abnormal features, such as component temperature anomalies, vibration anomalies, and pressure fluctuation anomalies. These abnormal features are vectorized, converting each type into a vector form. For temperature anomalies, a vector is formed by combining the temperature change rate and other characteristic values ​​of different components at different time points; for vibration anomalies, a vector is formed by combining the dominant and secondary frequency components of the vibration signals of each component; and for pressure fluctuation anomalies, a vector is formed by combining the maximum, minimum, and variance parameters of the pressure values ​​within a preset time window. These three feature vectors are then concatenated to generate a composite feature vector containing the temperature, vibration, and pressure fluctuation feature vectors.

[0136] Step S1542: Extract the set of affected component identifiers and the set of connection strength parameters from the description of the abnormal impact range, and generate a range feature vector containing the component identifier sequence and the connection strength value sequence.

[0137] The affected component identifier set and connection strength parameter set are extracted from the description of the anomaly's impact range. The affected component identifier set is a list containing all affected component identifiers, and the connection strength parameter set is a list containing the connection strength parameters between each affected component and the originating anomaly component. The identifiers in the affected component identifier set are arranged in a predetermined order to form a component identifier sequence; the connection strength parameters in the connection strength parameter set are arranged in the corresponding component identifier order to form a connection strength value sequence. The component identifier sequence and the connection strength value sequence are combined to generate a range feature vector containing both the component identifier sequence and the connection strength value sequence.

[0138] Step S1543: Convert the validity assessment results into an assessment feature vector containing validity level and difference parameters.

[0139] The effectiveness assessment results include an evaluation of the effectiveness of fault isolation measures, such as effective, partially effective, or ineffective, as well as relevant difference parameters, such as the reduction in the severity of color coding changes, the percentage reduction in the number of affected components, and the magnitude of the decrease in the slope of the numerical change curve. The effectiveness level is quantified, for example, 1 represents effective, 0.5 represents partially effective, and 0 represents ineffective; the difference parameters are arranged in a predetermined order. Then, the quantified effectiveness level and difference parameters are combined to form an assessment feature vector containing both the effectiveness level and the difference parameters.

[0140] Step S1544: Extract the feature vector set of all successfully processed cases from the historical fault record data. The feature vector set includes case anomaly feature vectors, case range feature vectors, and case evaluation feature vectors.

[0141] Relevant feature vectors for all successfully handled cases are extracted from historical fault record data. For each successfully handled case, there is a corresponding case anomaly feature vector, which is obtained by vectorizing the anomaly features in the case, similar to the current composite feature vector; the case range feature vector is obtained by processing the description of the anomaly's impact range in the case, similar to the current range feature vector; the case evaluation feature vector is obtained by processing the evaluation results of the effectiveness of fault isolation measures in the case, similar to the current evaluation feature vector. These feature vectors from all successfully handled cases are combined to form a feature vector set.

[0142] Step S1545: Calculate the cosine similarity between the composite feature vector and the abnormal feature vectors of each case, calculate the Manhattan distance between the range feature vector and the range feature vectors of each case, and calculate the Pearson correlation coefficient between the evaluation feature vector and the evaluation feature vectors of each case.

[0143] To measure the similarity between the current diagnostic scenario and historically successful cases, it is necessary to calculate the similarity or distance between different feature vectors. For composite feature vectors and anomalous feature vectors of individual cases, cosine similarity is used. Cosine similarity measures the directional similarity between two vectors; the closer the value is to 1, the more similar the two vectors are. For range feature vectors and range feature vectors of individual cases, Manhattan distance is used. Manhattan distance is the sum of the absolute differences between corresponding elements of two vectors; the smaller the distance, the closer the two vectors are. For evaluation feature vectors and evaluation feature vectors of individual cases, Pearson correlation coefficient is used. Pearson correlation coefficient measures the degree of linear correlation between two vectors; the closer the value is to 1 or -1, the stronger the correlation.

[0144] Step S1546: Perform standardized weighted summation on the cosine similarity, Manhattan distance, and Pearson correlation coefficient according to preset weight parameters to generate a matching score between each successful processing case and the current diagnostic scenario.

[0145] The preset weight parameters are determined based on the importance of different feature vectors to the matching degree. The calculated cosine similarity, Manhattan distance, and Pearson correlation coefficient are standardized to ensure they fall within the same numerical range. Then, the standardized cosine similarity, Manhattan distance, and Pearson correlation coefficient are weighted and summed according to the preset weight parameters. For example, assuming the weight of cosine similarity is w1, the weight of Manhattan distance is w2, and the weight of Pearson correlation coefficient is w3, and the standardized cosine similarity is s1, the Manhattan distance is d1, and the Pearson correlation coefficient is r1, then the matching degree score can be calculated using the formula w1*s1 + w2*d1 + w3*r1. This calculation is performed for each successfully processed case to generate a matching degree score between each successfully processed case and the current diagnostic scenario.

[0146] Step S1547: Select the successful processing case with the highest matching score as a reference case, and extract the processing step information and precaution information from the reference case.

[0147] The matching scores of all successful cases were compared, and the case with the highest matching score was selected as the reference case. This reference case is most similar to the current diagnostic scenario, and its handling method may have high reference value for handling the current fault. Handling step information was extracted from the reference case, which details the specific steps to handle the fault in that case; at the same time, precautions information was extracted, including safety regulations and operational taboos that need to be observed during the handling process.

[0148] Step S155: Extract operation step information and precaution information from the matched successful processing cases, generate fault handling suggestions including operation sequence, tool requirements and safety tips, and output the fault handling suggestions synchronously through the voice broadcast function of the AR interactive device.

[0149] After selecting the highest-scoring successful case as a reference, the operational steps and precautions information were extracted from it. The operational steps detailed the specific steps for troubleshooting, such as which checks to perform first and which repair measures to take. The precautions information included various issues that needed to be considered during the troubleshooting process, such as safety regulations and operational taboos.

[0150] Based on the extracted operation steps and precautions information, troubleshooting suggestions are generated, including the operation sequence, tool requirements, and safety tips. The operation sequence clarifies the order in which to handle the fault, the tool requirements list the necessary tools, and the safety tips emphasize the safety rules that must be followed during the troubleshooting process.

[0151] Finally, troubleshooting suggestions are simultaneously output via the voice broadcast function of the AR interactive device. Operators can directly hear the troubleshooting suggestions through the voice function of the AR interactive device without manually viewing them, improving the convenience and efficiency of the operation. At the same time, operators can also view the text content of the troubleshooting suggestions on the screen of the AR interactive device to understand the handling steps and precautions in more detail.

[0152] For example, the method may also include: Step S210: Record the operation screen and voice interaction content during the diagnosis process using the built-in camera of the AR interactive device, and generate diagnostic process video data containing timestamps.

[0153] Throughout the entire process of visualizing and diagnosing faults in thermal power units, to facilitate subsequent review and analysis, the built-in camera of the AR interactive device needs to record the operational screens and voice interactions during the diagnostic process. The built-in camera of the AR interactive device has high-definition shooting capabilities, clearly capturing various operations performed by the operator within the AR 3D scene model, such as selecting target fault visualization information and performing virtual isolation operations. Simultaneously, the microphone of the AR interactive device can record the operator's voice interactions during the diagnostic process, such as communication with other personnel and discussions about the analysis of the fault situation.

[0154] During the recording process, timestamps are added to each operation screen and voice interaction. These timestamps accurately record the time of each operation and voice interaction, providing a precise time reference for subsequent analysis. By integrating the operation screens, voice interactions, and timestamps, diagnostic process video data containing timestamps is generated.

[0155] Step S220: Extract key operation nodes from the diagnostic process video data. The key operation nodes include the fault visualization information selection time, the virtual isolation operation execution time, and the fault handling suggestion generation time.

[0156] After generating video data of the diagnostic process, key operational nodes need to be extracted. Key operational nodes are important time points in the diagnostic process, marking crucial steps and decision points. These include the moments when fault visualization information is selected, virtual isolation operations are executed, and fault handling suggestions are generated.

[0157] To extract these key operational nodes, video analytics can be employed. By analyzing the diagnostic process video data frame by frame, visual and audio information related to key operational nodes can be identified. For example, when the video shows an operator selecting target fault visualization information using gestures, that moment is recorded as the fault visualization information selection moment; when the video shows an operator performing virtual isolation operations, that moment is recorded as the virtual isolation operation execution moment; and when visual or audio prompts related to fault handling suggestion generation appear, that moment is recorded as the fault handling suggestion generation moment.

[0158] By extracting key operational nodes, the video data of the diagnostic process can be structured, breaking down the complex diagnostic process into multiple key steps, which facilitates subsequent analysis and management.

[0159] Step S230: Associate and store the key operation nodes with the corresponding fault feature set, abnormal visualization information and fault handling suggestions to generate a traceable diagnostic process record file.

[0160] After extracting the key operational nodes, these nodes need to be associated and stored with their corresponding fault feature sets, anomaly visualization information, and fault handling suggestions. The fault feature sets record various fault characteristics that occur in the thermal power unit during the diagnostic process; the anomaly visualization information intuitively displays the manifestation of the fault in the AR 3D scene model; and the fault handling suggestions provide solutions for handling the fault.

[0161] By associating key operational nodes with this relevant information, a complete diagnostic process record can be established. For example, the selection time of fault visualization information can be associated with the fault feature set and anomaly visualization information at that time; the execution time of virtual isolation operation can be associated with the fault feature set, anomaly visualization information, and effectiveness evaluation results of fault isolation measures before and after the operation; and the generation time of fault handling suggestions can be associated with the final fault feature set, anomaly visualization information, and fault handling suggestions.

[0162] The aforementioned associated storage method generates a traceable diagnostic process log file. This log file presents the entire diagnostic process in chronological order and according to operational steps, facilitating subsequent fault analysis, experience summarization, and quality traceability. Operators can review the diagnostic process at any time as needed, reviewing the fault conditions and handling measures corresponding to each key operational node, learning from experience, and improving their fault diagnosis and handling capabilities.

[0163] Step S240: Feed back the diagnostic process record file and the final fault handling result information to the thermal power unit operation and maintenance management system. The final fault handling result information includes the actual handling steps, the handling effect evaluation and the operation data after component repair.

[0164] After generating a traceable diagnostic process record file, it needs to be fed back to the thermal power unit operation and maintenance management system along with the final fault handling result information. The final fault handling result information includes the actual handling steps, the evaluation of the handling effect, and the operation data after the component repair.

[0165] The actual handling steps are recorded in detail, outlining the operational procedures taken during the troubleshooting process, including replacing parts and adjusting parameters. The handling effect evaluation assesses the effectiveness of the troubleshooting, such as whether the fault was completely eliminated and whether equipment performance returned to normal. The post-component repair operational data records the operating data of each component of the thermal power unit after the fault was handled, such as temperature, vibration frequency, and pressure.

[0166] By feeding back diagnostic process records and final fault handling results to the thermal power unit operation and maintenance management system, the system can gain a comprehensive understanding of the unit's fault conditions and handling processes. The system can store and analyze this feedback information, establishing a fault case database to provide a reference for future fault diagnosis and handling. Furthermore, analysis of the feedback information can identify problems in the design, operation, and maintenance of the thermal power unit, allowing for timely improvement measures to enhance its reliability and safety.

[0167] Step S250: Update the historical fault record data using the feedback diagnostic process record file and the final fault handling result information. The operation of updating the historical fault record data includes adding new successful handling cases and correcting the feature vector parameters of existing cases.

[0168] After receiving the diagnostic process log files and final fault handling results, the thermal power unit operation and maintenance management system needs to use this feedback information to update the historical fault record data. Updating the historical fault record data mainly involves adding new successful handling cases and correcting the feature vector parameters of existing cases.

[0169] If the diagnostic process log and final fault handling results show that a fault was successfully handled, and the handling method is representative and has reference value, then this case can be added as a new successful handling case to the historical fault record data. The new successful handling case includes detailed information such as fault characteristics, handling steps, and handling effects, providing a new reference for handling similar faults in the future.

[0170] Simultaneously, the feature vector parameters of existing cases are corrected based on feedback information. As experience in operating and troubleshooting thermal power units accumulates, discrepancies may arise between the feature vector parameters of some existing cases and the actual situation. By analyzing the feedback diagnostic process log files and final fault handling results, these discrepancies can be corrected, making the feature vector parameters of existing cases more accurately reflect the actual fault conditions. This improves the quality and usability of historical fault record data, providing a more reliable reference for the fault handling suggestion generation module, thereby enhancing the efficiency and accuracy of fault diagnosis and handling.

[0171] Figure 2 The illustration shows exemplary hardware and software components of an AR-based thermal power unit fault visualization diagnostic system 100 that can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the AR-based thermal power unit fault visualization diagnostic system 100 and to perform the functions in this application.

[0172] The AR-based thermal power unit fault visualization and diagnosis system 100 can be a general-purpose server or a special-purpose server; both can be used to implement the AR-based thermal power unit fault visualization and diagnosis method of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0173] For example, the AR-based thermal power unit fault visualization and diagnostic system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the AR-based thermal power unit fault visualization and diagnostic system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The AR-based thermal power unit fault visualization and diagnostic system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0174] For ease of explanation, only one processor is described in the AR-based thermal power unit fault visualization and diagnostic system 100. However, it should be noted that the AR-based thermal power unit fault visualization and diagnostic system 100 of this application may also include multiple processors. Therefore, the steps performed by one processor described in this application may also be performed jointly by multiple processors or individually. For example, if the processor of the AR-based thermal power unit fault visualization and diagnostic system 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0175] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned AR-based thermal power unit fault visualization diagnosis method is implemented.

[0176] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for visual diagnosis of thermal power unit faults based on AR, characterized in that, The method includes: Acquire a set of thermal power unit operating data, which includes real-time equipment monitoring data, historical fault record data, and component structural parameter data; Based on the thermal power unit's operating data set, an AR 3D scene model of the thermal power unit is constructed. The AR 3D scene model includes virtual component models corresponding to the actual physical structure of the unit and the spatial positional relationships of each component. The thermal power unit's operating data set is processed by fault feature extraction to obtain a fault feature set reflecting abnormal equipment operating status. The fault feature set includes abnormal component temperature features, abnormal vibration features, and abnormal pressure fluctuation features. The fault feature set is spatially mapped to the AR 3D scene model to generate fault visualization information superimposed on the surface of the virtual component model. The fault visualization information includes anomaly type identifier, anomaly impact range, and anomaly development trend description. The AR interactive device performs interactive diagnostic operations on the visualized fault information. These operations include viewing detailed abnormal parameters, simulating fault isolation operations, and generating fault handling suggestions.

2. The AR-based thermal power unit fault visualization diagnosis method according to claim 1, characterized in that, The acquisition of the thermal power unit operating data set includes: The sensor network deployed on key components of the thermal power unit collects real-time monitoring data, which includes temperature sensing data, vibration frequency data, and pressure transmission data of each component. Historical fault record data is retrieved from the thermal power unit operation and maintenance management system. The historical fault record data includes the fault occurrence time, fault phenomenon description and corresponding handling solution information of each component within a preset time period in the past. Access the thermal power unit design document database to obtain component structural parameter data, which includes the geometric dimensions, material properties, and connection methods of each component with adjacent components; The real-time monitoring data, historical fault record data, and component structural parameter data of the equipment are processed in a unified format to generate a standardized data set with the same timestamp and component identification code; The standardized dataset is then subjected to missing value imputation and noise filtering to obtain the thermal power unit operation dataset.

3. The AR-based thermal power unit fault visualization diagnosis method according to claim 1, characterized in that, The construction of the AR 3D scene model of the thermal power unit based on the thermal power unit operation data set includes: Extract the component structural parameter data from the thermal power unit's operating data set, and construct a basic three-dimensional model of each component based on the geometric dimension information and connection method information; The basic 3D model is matched and associated with the spatial coordinate information of the components in the real-time monitoring data of the equipment, and the spatial coordinate parameters of each basic 3D model are adjusted to restore the physical layout of the actual unit. Extract component deformation information when typical faults occur from the historical fault record data, perform morphological correction processing on the basic three-dimensional model of the corresponding component, and generate a multi-morphological virtual component model that includes normal state and typical fault state. Analyze the connection relationships between each virtual component model, and add virtual interface identifiers representing the connection relationships between adjacent virtual component models based on the connection method information. The virtual interface identifiers include interface type and connection strength description. The multi-form virtual component model and virtual interface identifier are mapped to the physical space of the real unit using an AR coordinate system calibration algorithm, generating an AR three-dimensional scene model of the thermal power unit that is aligned with the actual scene location.

4. The AR-based thermal power unit fault visualization diagnosis method according to claim 1, characterized in that, The process of extracting fault features from the thermal power unit's operating data set yields a set of fault features reflecting abnormal equipment operating conditions, including: Time series analysis is performed on the temperature sensing data in the real-time monitoring data of the device to extract the temperature change rate parameters of each component at continuous time points, and the component temperature anomaly characteristics are generated based on the comparison results of the temperature change rate parameters and the preset temperature threshold. Fourier transform is performed on the vibration frequency data in the real-time monitoring data of the device to extract the main frequency components and secondary frequency components of the vibration signals of each component, and the degree of matching between the main frequency components and the natural frequency of the component is analyzed to generate vibration anomaly characteristics. The pressure transmission data in the real-time monitoring data of the device is subjected to window sliding statistical processing to calculate the maximum value, minimum value and variance parameter of the pressure value within a preset time window. Based on the degree of deviation of the variance parameter from the standard pressure fluctuation range, pressure fluctuation anomaly characteristics are generated. The abnormal temperature characteristics, abnormal vibration characteristics, and abnormal pressure fluctuation characteristics of the components are matched with the typical fault characteristics in the historical fault record data to filter out the effective fault characteristics related to the current operating state. A causal relationship analysis is performed on the effective fault features to identify the correlation between abnormal features and generate a fault feature set with a hierarchical structure.

5. The AR-based thermal power unit fault visualization diagnosis method according to claim 1, characterized in that, The step of spatially mapping the fault feature set with the AR 3D scene model to generate fault visualization information superimposed on the surface of the virtual component model includes: Extract the component identification code corresponding to each abnormal feature in the fault feature set, and locate the corresponding virtual component model in the AR three-dimensional scene model according to the component identification code; Each located virtual component model is assigned a visual identifier corresponding to the anomaly type. The visual identifier includes color coding and graphic symbols, where color coding is used to distinguish the severity of the anomaly and graphic symbols are used to identify the anomaly type. Analyze the spatial propagation characteristics of the abnormal features, combine the spatial positional relationships of the virtual component models in the AR 3D scene model, calculate the range of adjacent components affected by the abnormality, and generate a description of the range of abnormality impact containing the identifiers of the affected components; The time series data in the fault feature set is subjected to trend prediction processing. Based on the abnormal development patterns in historical fault record data, the numerical change curves of abnormal features in the future preset time period are generated as a description of the abnormal development trend. The visual identifiers, descriptions of the scope of the abnormality, and descriptions of the abnormality's development trend are rendered onto the surface of the corresponding virtual component model in a semi-transparent overlay manner, generating fault visualization information that can dynamically adjust its display position according to changes in the AR viewpoint.

6. The AR-based thermal power unit fault visualization diagnosis method according to claim 5, characterized in that, The analysis of the spatial propagation characteristics of the abnormal features, combined with the spatial positional relationships of the virtual component models in the AR 3D scene model, calculates the range of adjacent components affected by the abnormality, and generates a description of the abnormality's impact range containing the identifiers of the affected components, including: Extract the connection information of the components corresponding to the abnormal features, and determine the physical connection type between them and adjacent components. The physical connection type includes rigid connection, elastic connection and fluid conduction connection. Based on the physical connection type and the physical properties of the anomaly characteristics, the main path of anomaly propagation is determined, and the physical properties include heat conduction properties, vibration transmission properties, and pressure diffusion properties. Traverse the virtual component models in the AR 3D scene model along the main path and extract the set of components that have a direct or indirect connection relationship with the initial abnormal component; For each component in the component set, calculate its connection strength parameter with the initial abnormal component. The connection strength parameter is determined based on the connection material strength and contact area parameters in the connection method information. Components whose connection strength parameters exceed a preset threshold are selected as affected components, and an abnormal impact range description containing the affected component identifier and connection strength parameters is generated.

7. The AR-based thermal power unit fault visualization diagnosis method according to claim 1, characterized in that, The step of performing interactive diagnostic operations on the fault visualization information through an AR interactive device includes: The target fault visualization information is selected by the gesture recognition function of the AR interactive device, and the detailed parameter display operation of the target fault visualization information is triggered. The detailed parameter display operation includes calling the original monitoring data and historical fault record data related to the abnormal characteristics in the thermal power unit operation data set. In the AR 3D scene model, a virtual isolation operation is performed on the target virtual component model. The virtual isolation operation includes marking the isolation area through the spatial positioning function of the AR interactive device and simulating the disconnection of the target component from the adjacent components. Monitor the changes in the fault visualization information of other components in the AR 3D scene model after the virtual isolation operation is executed, and evaluate the effectiveness of the fault isolation measures based on the changes. The abnormal feature set, the description of the abnormal impact range, and the effectiveness evaluation results of the virtual isolation operation are input into the fault handling suggestion generation module, which then performs matching analysis based on successful handling cases in historical fault record data. Operation steps and precautions are extracted from the successfully matched cases to generate troubleshooting suggestions that include the operation sequence, tool requirements, and safety tips. These troubleshooting suggestions are then output synchronously through the voice broadcast function of the AR interactive device.

8. The AR-based thermal power unit fault visualization diagnosis method according to claim 7, characterized in that, The monitoring of changes in the fault visualization information of other components in the AR 3D scene model after the virtual isolation operation is executed, and the evaluation of the effectiveness of the fault isolation measures based on the changes, includes: Before performing virtual isolation operations, record the initial state of all fault visualization information in the AR 3D scene model. The initial state includes the color coding of the anomaly type identifier, the number of affected components describing the anomaly's impact range, and the slope of the numerical change curve describing the anomaly's development trend. After performing the virtual isolation operation, the fault visualization information in the AR three-dimensional scene model is updated in real time, and the current status of the updated fault visualization information is obtained. The current status includes the color code change of the anomaly type identifier, the change of the number of affected components describing the anomaly impact range, and the change of the slope of the numerical change curve describing the anomaly development trend. Calculate the difference parameters between the current state and the initial state. The difference parameters include the reduction in the severity of the color coding change, the reduction in the number of affected components, and the decrease in the slope of the numerical change curve. Based on the comparison results between the difference parameters and the preset effectiveness assessment threshold, an effectiveness assessment conclusion is generated, which includes effective, partially effective, or ineffective isolation measures. The difference parameters and effectiveness evaluation conclusions are associated with and stored in relation to the specific execution steps of the virtual isolation operation, forming a traceable record of the isolation operation effect.

9. The AR-based thermal power unit fault visualization diagnosis method according to claim 7, characterized in that, The step involves inputting the set of abnormal features, the description of the scope of abnormal impact, and the effectiveness evaluation results of the virtual isolation operation into the fault handling suggestion generation module. The fault handling suggestion generation module then performs a matching analysis based on successful handling cases in historical fault record data, including: The abnormal feature set is processed into feature vectorization to generate a composite feature vector containing temperature anomaly feature vector, vibration anomaly feature vector, and pressure fluctuation anomaly feature vector. Extract the set of affected component identifiers and the set of connection strength parameters from the description of the abnormal impact range, and generate a range feature vector containing a sequence of component identifiers and a sequence of connection strength values; The effectiveness assessment results are converted into an assessment feature vector containing effectiveness level and difference parameters; Extract a set of feature vectors for all successfully handled cases from the historical fault record data. The set of feature vectors includes case anomaly feature vectors, case range feature vectors, and case evaluation feature vectors. Calculate the cosine similarity between the composite feature vector and the abnormal feature vector of each case; calculate the Manhattan distance between the range feature vector and the range feature vector of each case; calculate the Pearson correlation coefficient between the evaluation feature vector and the evaluation feature vector of each case. The cosine similarity, Manhattan distance, and Pearson correlation coefficient are standardized and weighted according to preset weight parameters to generate a matching score between each successful processing case and the current diagnostic scenario. The successful case with the highest matching score was selected as a reference case, and the processing steps and precautions information were extracted from the reference case.

10. An AR-based thermal power unit fault visualization diagnostic system, characterized in that, The device includes a processor and a memory, the memory being connected to the processor. The memory is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the memory to implement the AR-based thermal power unit fault visualization diagnosis method according to any one of claims 1-9.

Images