Data inventory updating method and system based on multi-source data fusion

Abstract

The application discloses a data list updating method and system based on multi-source data fusion, relates to the technical field of data processing, and comprises the following steps: collecting multi-source heterogeneous data of infrastructure assets; inputting the fusion data set into an infrastructure asset digital twin model for simulation calculation to obtain simulation output values of asset states; constructing a deviation fingerprint library, calculating deviation values of the simulation output values and measured data, and matching and identifying data anomaly patterns; performing traceability analysis based on a causal reasoning network to generate a data repair strategy; generating a new data list version snapshot, recording repair basis and identification information of the deviation fingerprint library, and performing dynamic updating of the data list. The technical problems of existing technologies, such as difficulty in multi-source heterogeneous data fusion, insufficient deviation pattern identification and traceability, and dependence of data list updating on manual work and untraceability, are solved, and the technical effects of improving the automation degree, accuracy, interpretability and auditability of dynamic updating of the data list are achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a data list update method and system based on multi-source data fusion. Background Technology

[0002] In the full lifecycle management of infrastructure assets, the accuracy and timeliness of data lists such as asset registers, equipment ledgers, or digital records directly affect the effectiveness of operation monitoring, maintenance decisions, and risk assessment. Currently, updating and maintaining these data lists still faces numerous challenges. Data sources for infrastructure assets include real-time monitoring data from IoT sensors, manual inspection records, historical operation and maintenance logs, Geographic Information System (GIS) data, Building Information Modeling (BIM) data, and external environmental data. These sources exhibit significant differences in spatiotemporal scale, sampling frequency, semantic expression, and quality, making it difficult to directly integrate the data for accurate asset status assessment. Existing methods typically employ fixed thresholds or... The existing data list identification rules ignore the complex relationship between dynamic system behavior and operational constraints, resulting in many deviations being misjudged as measurement noise or random errors. This fails to reveal potential asset degradation or data recording errors. Furthermore, the correction of existing data lists relies on the experience and judgment of domain experts, and the update results are difficult to trace. When data anomalies occur, they cannot be effectively linked to specific business rule constraints, and it is difficult to locate the source of the anomaly. In addition, the lack of a dynamic update mechanism based on multi-source data fusion, anomaly pattern identification, causal tracing, and automated repair leads to chaotic version management of data lists. It is impossible to record the business basis and deviation fingerprint of each update, which severely limits the credibility and auditability of data assets.

[0003] Therefore, current technologies face challenges such as difficulty in fusing multi-source heterogeneous data, insufficient identification and traceability of deviation patterns, and reliance on manual updates for data lists that are not traceable. Summary of the Invention

[0004] This application provides a data list update method and system based on multi-source data fusion, which solves the technical problems existing in the prior art, such as difficulty in multi-source heterogeneous data fusion, insufficient deviation pattern recognition and traceability, and reliance on manual and untraceable data list updates. It achieves the technical effect of improving the automation, accuracy, interpretability and auditability of dynamic data list updates.

[0005] This application provides a data inventory update method based on multi-source data fusion. The method includes: collecting multi-source heterogeneous data of infrastructure assets and performing spatiotemporal alignment and normalization processing to generate a fused dataset; inputting the fused dataset into a pre-constructed digital twin model of infrastructure assets, running simulation calculations based on the asset status model, and obtaining simulation output values ​​of the asset status; constructing a deviation fingerprint library containing various business constraint rules, calculating the deviation value between the simulation output value and the measured data in the fused dataset, and matching the deviation value with the business constraint rules in the deviation fingerprint library to identify data anomaly patterns; performing source tracing analysis on the error sources corresponding to the data anomaly patterns based on a causal inference network to generate a data repair strategy; updating the corresponding fields in the data inventory according to the data repair strategy, generating a new data inventory version snapshot, and recording the repair basis and the identification information of the deviation fingerprint library in the data inventory version snapshot to dynamically update the data inventory.

[0006] In possible implementations, the multi-source heterogeneous data includes at least asset online runtime sequence data, asset procurement and maintenance record parsing results, and asset deployment environment parameters.

[0007] In a possible implementation, the fused dataset is input into a pre-built digital twin model of infrastructure assets, and simulation calculations based on the asset state model are run to obtain simulation output values ​​of the asset state. This includes: extracting the operating conditions and attribute parameters of the infrastructure assets at a specific time segment from the fused dataset; loading the operating conditions and attribute parameters as initial and boundary conditions into the initial value module and boundary condition module of the pre-built digital twin model of infrastructure assets, respectively; calling the asset state model built into the digital twin model of infrastructure assets to perform multi-state coupled calculations on the operating conditions and attribute parameters to generate internal state evaluation results of the infrastructure assets at the specific time segment; and parsing the asset internal state quantities that cannot be directly measured from the internal state evaluation results as simulation output values ​​of the asset state.

[0008] In one possible implementation, a deviation fingerprint library containing multiple business constraint rules is constructed, including: defining a set of business constraint rules characterizing the reasonable boundaries of the asset's state based on the construction safety specifications and material performance standards of the infrastructure asset, wherein the business constraint rules describe state quantity constraint relationships that cannot be simultaneously satisfied under the normal service state of the asset; encoding each business constraint rule into a computable logical judgment statement containing logical operators and threshold judgments, and assigning a unique deviation fingerprint identifier to each logical judgment statement; aggregating all logical judgment statements and their corresponding deviation fingerprint identifiers to construct a deviation fingerprint library containing multiple business constraint rules.

[0009] In a possible implementation, the deviation between the simulation output value and the measured data in the fused dataset is calculated, and the deviation value is matched with the business constraint rules in the deviation fingerprint database to identify data anomaly patterns. This includes: dynamically determining the business tolerance range threshold under the current operating condition based on the current operating condition of the measured data in the fused dataset and combining the statistical characteristics of historical data under the same operating condition; calculating the deviation between the simulation output value and the measured data in the fused dataset, and determining whether the deviation value exceeds the business tolerance range threshold; if it exceeds, extracting state variables related to the deviation value from the fused dataset to form a current state variable combination; matching the current state variable combination with each computable logical judgment statement in the deviation fingerprint database; when the current state variable combination satisfies the triggering condition of any computable logical judgment statement in the deviation fingerprint database, determining the corresponding data anomaly pattern based on the deviation fingerprint identifier corresponding to the logical judgment statement.

[0010] In a possible implementation, a causal inference network is used to trace the source of errors corresponding to the data anomaly pattern and generate a data repair strategy. This includes: constructing a causal inference network containing nodes representing sensor measurement errors, communication transmission disturbances, manual input deviations, and the actual degradation state of the equipment, with causal relationships between nodes represented by directed edges; inputting the deviation fingerprint identifier and deviation value corresponding to the data anomaly pattern as observation evidence into the causal inference network; querying a conditional probability table constructed based on a historical fault case database to calculate the posterior probability caused by the sensor measurement error, communication transmission disturbance, and manual input deviation nodes, respectively; selecting the error source type represented by the node with the highest posterior probability, and combining it with the context information of the fused dataset to generate a targeted data repair strategy adapted to the error source type.

[0011] In possible implementations, a targeted data repair strategy adapted to the error source type is generated, including: if the error source type is a sensor measurement error, a spatially correlated multi-sensor data consistency verification and interpolation repair strategy is generated, using valid sensor data from adjacent spatial locations in the fused dataset to interpolate and generate correction values ​​to replace sensor data in the fused dataset that are determined to be abnormal; if the error source type is a communication transmission disturbance, a data reconstruction repair strategy based on a time series prediction model is generated to repair missing or abrupt data in the fused dataset caused by communication anomalies; if the error source type is a manual input deviation, a repair strategy is generated that calls the simulation output value of the asset status to overwrite and replace the corresponding fields in the fused dataset.

[0012] In a possible implementation, the corresponding fields in the data list are updated according to the data repair strategy, a new data list version snapshot is generated, and the repair basis and the identification information of the deviation fingerprint database are recorded in the data list version snapshot to dynamically update the data list. This includes: executing the data repair strategy to generate corrected target data; locating the original data field in the data list corresponding to the corrected target data and overwriting the value of the original data field with the corrected target data; creating a new data list version snapshot and assigning a globally unique version identifier to the new data list version snapshot; serializing and writing the overwritten and updated data list content into the new data list version snapshot; in the metadata area of ​​the new data list version snapshot, structurally recording the repair basis that triggered this update, the identification information of the deviation fingerprint database used, and the version number of the associated infrastructure asset digital twin model; and switching the currently effective data list pointer of the system from pointing to the old version snapshot to pointing to the new data list version snapshot.

[0013] In a possible implementation, after generating a new data list version snapshot, the method further includes: within a preset statistical period, aggregating the identification information of the deviation fingerprint database recorded in all data list version snapshots, identifying and extracting the target deviation fingerprint identifier that is frequently triggered; extracting the features of the historical fusion dataset associated with the target deviation fingerprint identifier and the corresponding data repair strategy execution records; and based on the features of the historical fusion dataset and the data repair strategy execution records, reverse-optimizing and adjusting the model parameters of the asset status model in the digital twin model of the infrastructure assets, as well as the business tolerance range threshold of the corresponding business constraint rules in the deviation fingerprint database.

[0014] This application also provides a data inventory update system based on multi-source data fusion. The system includes: a fused data generation module, used to collect multi-source heterogeneous data of infrastructure assets and perform spatiotemporal alignment and normalization processing to generate a fused dataset; a simulation calculation module, used to input the fused dataset into a pre-built digital twin model of infrastructure assets, run simulation calculations based on the asset status model, and obtain simulation output values ​​of the asset status; a data anomaly pattern recognition module, used to construct a deviation fingerprint library containing multiple business constraint rules, calculate the deviation value between the simulation output value and the measured data in the fused dataset, and match the deviation value with the business constraint rules in the deviation fingerprint library to identify data anomaly patterns; a data repair strategy generation module, used to perform source analysis on the error sources corresponding to the data anomaly patterns based on a causal inference network to generate data repair strategies; and a data inventory update module, used to update the corresponding fields in the data inventory according to the data repair strategy, generate a new data inventory version snapshot, and record the repair basis and the identification information of the deviation fingerprint library in the data inventory version snapshot to dynamically update the data inventory.

[0015] This application proposes a data inventory update method and system based on multi-source data fusion. The method involves collecting multi-source heterogeneous data on infrastructure assets; inputting the fused dataset into a digital twin model of the infrastructure assets for simulation calculations to obtain simulated output values ​​of the asset status; constructing a deviation fingerprint database to calculate the deviation between the simulated output values ​​and the measured data, and matching and identifying abnormal data patterns; performing source tracing analysis based on causal inference networks to generate data repair strategies; generating a new data inventory version snapshot, and recording the repair basis and the identification information of the deviation fingerprint database for dynamic updates. This method solves the technical problems of existing technologies, such as difficulties in multi-source heterogeneous data fusion, insufficient deviation pattern recognition and source tracing, and reliance on manual and untraceable data inventory updates. It achieves the technical effect of improving the automation, accuracy, interpretability, and auditability of dynamic data inventory updates. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a schematic diagram of the data list update method based on multi-source data fusion provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the data list update system structure based on multi-source data fusion provided in an embodiment of this application.

[0019] Figure labeling: 10 for fused data generation module, 20 for simulation calculation module, 30 for data anomaly pattern recognition module, 40 for data repair strategy generation module, and 50 for data list update module. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0021] This application provides a data list update method based on multi-source data fusion, such as... Figure 1 As shown, the method includes:

[0022] Step S100: Collect multi-source heterogeneous data of infrastructure assets and perform spatiotemporal alignment and normalization processing to generate a fused dataset.

[0023] Step S100 further includes that the multi-source heterogeneous data includes at least asset online runtime sequence data, asset procurement and maintenance record parsing results, and asset deployment environment parameters.

[0024] Preferably, asset status parameters that change sequentially over time are collected by IoT sensors installed on or around the asset. These include data on electrical assets such as oil surface temperature, winding temperature, load current, online oil chromatography monitoring data, partial discharge, and vibration amplitude; data on mechanical assets such as rotational speed, bearing temperature, inlet and outlet pressure, flow rate, and motor current; and data on bridge assets such as strain gauge readings, displacement, vibration frequency, crack gauge width, and cable force. Historical descriptive data on asset procurement and health management and maintenance, including procurement records and maintenance records, are extracted and parsed from structured or semi-structured documents and management systems. Environmental parameters that significantly affect asset performance, aging rate, and failure modes, such as meteorological conditions like temperature / humidity and wind speed / direction, as well as geographical location and load conditions, are obtained from environmental monitoring stations, meteorological services, or auxiliary sensors. These parameters significantly affect asset performance, aging rate, and failure modes, such as meteorological conditions like temperature / humidity and wind speed / direction, as well as geographical location and load conditions. This results in multi-source heterogeneous data. Then, spatiotemporal alignment and normalization are performed, including converting the timestamps of all data to the same reference system, resampling data with different sampling frequencies to a unified frequency, identifying and correcting time offsets caused by communication transmission delays, and defining a unified three-dimensional spatial coordinate system for infrastructure assets. The installation location of each sensor and the sampling point location of each offline test are mapped to this coordinate system. When analyzing the state of a local area of ​​the equipment, data corresponding to the spatial location of the area are extracted from multi-source data. Then, multi-source data with different dimensions and different numerical ranges are transformed to a unified numerical range. For example, for each physical quantity, its actual value is linearly mapped to the [0, 1] interval according to its theoretical minimum and maximum values. Finally, a fused dataset is obtained. Any record contains the online monitoring value, offline test value and corresponding environmental operating condition parameters of the spatiotemporal point, and the parameter values ​​are in the same dimension and numerical range.

[0025] Step S200: Input the fused dataset into the pre-built digital twin model of infrastructure assets, run simulation calculations based on the asset status model, and obtain the simulation output value of the asset status.

[0026] Preferably, based on the physical principles and operating mechanisms of infrastructure assets, an asset state model is established. For example, for power transformers, the state model includes sub-models such as thermal balance equations, electromagnetic field equations, and oil-paper insulation aging models. Equipment design drawings, factory test reports, and historical operation and maintenance records are collected. Static attributes such as geometric dimensions, material properties, and rated parameters of the equipment are extracted as fixed parameters for model parameter initialization. Then, historical operating data of the equipment under known operating conditions are collected, including input operating conditions and actual observed internal states, such as temperature distribution and vibration amplitude. The historical operating condition data is input into the model, and simulation calculations are run to obtain simulation output values. The error between the simulation output values ​​and the actual observed values ​​is calculated, and the model parameters are adjusted in reverse using gradient descent or Kalman filtering algorithms until the simulation error is less than a preset threshold. The trained model parameters, solution algorithms, and boundary condition processing logic are encapsulated into a digital twin model of infrastructure assets and deployed to servers or edge computing nodes.

[0027] Furthermore, step S200 also includes: extracting the operating conditions and attribute parameters of the infrastructure asset at a specific time segment from the fused dataset; loading the operating conditions and attribute parameters as initial conditions and boundary conditions into the initial value module and boundary condition module of the pre-built infrastructure asset digital twin model, respectively; calling the asset state model built into the infrastructure asset digital twin model to perform multi-state coupling calculations on the operating conditions and attribute parameters to generate the internal state evaluation result of the infrastructure asset at the specific time segment; and parsing the asset internal state quantities that cannot be directly measured from the internal state evaluation result as the simulation output value of the asset state.

[0028] Preferably, the condition of infrastructure assets, such as the degree of steel corrosion inside bridge concrete, the degree of insulation aging of underground cables, the wall thickness reduction inside gas pipelines, and the hot spot temperature of transformer windings, cannot be directly measured by installing sensors. Instead, operating parameters of the infrastructure assets at a specific time segment are extracted from the fused dataset. These parameters represent the external load or environmental conditions the asset is experiencing at the current moment, along with attribute parameters—the inherent and relatively stable characteristics of the asset. For example, the measured load current at that moment is 800A, the measured ambient temperature is 35℃, and the rated capacity of the equipment extracted from the procurement records is 1000kVA, with oil-immersed self-cooling as the cooling method. These operating conditions and attribute parameters are then used as initial and boundary conditions—the initial state and external constraints before model simulation—and applied. The process involves accessing the initial value module and boundary condition module of the digital twin model of infrastructure assets, and then calling the asset state model built into the digital twin model to perform multi-state coupled calculations on the working conditions and attribute parameters. The asset state model is a mathematical equation or simulation model that describes the physical behavior of the asset. It is usually constructed based on finite element analysis, multibody dynamics, thermodynamics, electromagnetic fields, etc. Then, the mutually coupled equations are solved simultaneously to calculate the internal state assessment results of the infrastructure asset at a specific time section, including the stress, temperature, displacement, etc. of each grid point. Finally, the internal state quantities of the asset that cannot be directly measured, such as the maximum tensile stress of the steel bars and the hottest temperature of the transformer winding, are extracted from the internal state assessment results and used as the simulation output values ​​of the asset state.

[0029] Step S300: Construct a deviation fingerprint library containing various business constraint rules, calculate the deviation value between the simulation output value and the measured data in the fused dataset, and match the deviation value with the business constraint rules in the deviation fingerprint library to identify abnormal data patterns.

[0030] Step S300 further includes defining a set of business constraint rules characterizing the reasonable boundaries of the asset's state based on the construction safety specifications and material performance standards of the infrastructure asset. The business constraint rules describe state constraint relationships that cannot be simultaneously satisfied under the normal service state of the asset. Each business constraint rule is encoded into a computable logical judgment statement containing logical operators and threshold judgments, and a unique deviation fingerprint identifier is assigned to each logical judgment statement. All logical judgment statements and their corresponding deviation fingerprint identifiers are aggregated to construct a deviation fingerprint library containing multiple business constraint rules.

[0031] Preferably, the construction safety specifications and material performance standards of infrastructure assets are obtained, such as "the strength of concrete shall not be less than 70% of the design value within 28 days after pouring" and "stress and strain are proportional within the elastic deformation range of steel (Hooke's Law)". A set of business constraint rules characterizing the reasonable boundary of asset status is defined to describe the state quantity constraint relationships that cannot be simultaneously satisfied under the normal service state of the asset. For example, if the static load strain is greater than 2000 microstrain, the main beam deflection is less than L / 800, and the vehicle load is less than the design load, then an anomaly is determined. Then, each business constraint rule is encoded into a computable logical judgment statement containing logical operators and threshold judgments, usually using the IF-THEN structure, containing logical operators (AND, OR, NOT) and threshold judgments (>, <, =). Each encoded rule is assigned a globally unique ID to determine a unique deviation fingerprint identifier. Finally, the various business constraint rules and corresponding unique deviation fingerprint identifiers of all asset types in all business domains are centrally stored to construct a deviation fingerprint database.

[0032] Furthermore, step S300 also includes: dynamically determining the service tolerance range threshold under the current operating condition based on the current operating condition of the measured data in the fused dataset and combining the statistical characteristics of historical data under the same operating condition; calculating the deviation value between the simulation output value and the measured data in the fused dataset, and determining whether the deviation value exceeds the service tolerance range threshold; if it exceeds, extracting state variables related to the deviation value from the fused dataset to form a current state variable combination; matching the current state variable combination with each computable logical judgment statement in the deviation fingerprint database one by one; when the current state variable combination satisfies the triggering condition of any computable logical judgment statement in the deviation fingerprint database, determining the corresponding data anomaly mode according to the deviation fingerprint identifier corresponding to the logical judgment statement.

[0033] Preferably, the current operating condition of the measured data in the fusion dataset is obtained, such as the output power of the generator, the fluid velocity in the pipeline, and the total weight of vehicles on the bridge. In the historical database, all historical time-section data with the same parameters as the current operating condition or within a preset neighborhood (e.g., power ±2%) are retrieved. For the retrieved historical data with the same operating condition, the deviation between the simulation output value and the measured data in each data set is calculated to form a deviation sample set under the operating condition. Then, the statistical characteristics of the sample set are calculated, including the mean and standard deviation. Based on a preset confidence level (e.g., 95%), the business tolerance range threshold is determined to reflect the normal fluctuation range that is allowed under the current specific operating condition and the normal operation of the measurement system, and that conforms to historical patterns. The simulation output value calculated using the digital twin model at the current moment is algebraically subtracted from the corresponding physical quantity in the measured data at the same moment to obtain the original deviation value. This deviation value is then compared with the business tolerance range threshold. If the deviation value exceeds the business tolerance range threshold, an abnormal deviation is determined. All state variables that are related to the current abnormal deviation in terms of physical mechanism or business logic are extracted from the fused dataset to form the current state variable combination. For example, if the simulation output value is "internal stress", the extracted related state variables could be "external strain" and "displacement". The current state variable combination is then used as input and compared one by one with the deviation fingerprint database. Each computable logical judgment statement in the system is evaluated. If and only if the current combination of state variables makes the evaluation result of any logical judgment statement in the deviation fingerprint database "true", it is determined that the current abnormal deviation matches the specific data abnormality pattern defined by the statement. The unique identifier corresponding to the triggered logical judgment statement is recorded, i.e., the deviation fingerprint identifier. Finally, the corresponding data abnormality pattern is identified and determined. For example, the data abnormality pattern corresponding to fingerprint identifier 01 is "systematic deviation caused by sensor zero-point drift"; the pattern corresponding to fingerprint identifier 02 is "data outlier caused by communication interruption".

[0034] Step S400: Based on the causal reasoning network, perform source analysis on the error sources corresponding to the data anomaly patterns and generate a data repair strategy.

[0035] Step S400 further includes: constructing a causal inference network, which includes nodes representing sensor measurement errors, communication transmission disturbances, manual input deviations, and the actual degradation state of the equipment, with causal relationships between nodes represented by directed edges; inputting the deviation fingerprint identifier and deviation value corresponding to the data anomaly pattern as observation evidence into the causal inference network; querying a conditional probability table constructed based on a historical fault case library to calculate the posterior probability caused by the sensor measurement error, communication transmission disturbance, and manual input deviation nodes, respectively; selecting the error source type represented by the node with the highest posterior probability, and combining it with the context information of the fused dataset to generate a targeted data repair strategy adapted to the error source type.

[0036] Preferably, network nodes are characterized by sensing measurement errors, communication transmission disturbances, manual input deviations, and the actual degradation state of the equipment. Directed edges are represented by causal relationships between nodes to construct a causal inference network, which is used to represent the causal dependencies between multiple variables. The directed edges point from the cause node to the result node, indicating the causal direction of the cause leading to the result. For example, sensing measurement errors cause measured data to deviate from the physical truth, thus triggering a specific deviation fingerprint; communication transmission disturbances cause data packet loss or bit errors, resulting in abnormal measured data and triggering a deviation fingerprint; manual input errors cause field values ​​in the data list to be inconsistent with the actual values, triggering a deviation fingerprint; and the actual degradation state of the equipment causes measured data to reflect the actual physical changes, triggering a deviation fingerprint.

[0037] Preferably, the deviation fingerprint identifier and deviation value corresponding to the data anomaly pattern are used as observational evidence, input into the causal inference network, and set as evidence variables. In the Bayesian network, the probability distribution of the evidence variables is fixed so that the probability of the evidence value is 1, and the probability of other values ​​is 0. The historical fault case library is used to store all data anomaly events that occurred during the operation of the equipment in the past and their true causes. Each case record includes a case number, deviation fingerprint identifier, deviation value, true error source, equipment type, and operating parameters. The conditional probability table constructed based on the historical fault case library is used to quantify the probability of observing specific evidence under different values ​​for each cause node. For each cause node, its probability under different deviation values ​​is statistically analyzed based on the historical fault case library. The frequency of occurrence under fingerprint and deviation conditions is normalized to conditional probability values. Assuming the given deviation fingerprint identifier is TR-TH-001-V1, the deviation level is moderate deviation, the conditional probability of sensor measurement error is 0.65, the conditional probability of communication transmission disturbance is 0.15, the conditional probability of manual input deviation is 0.1, and the conditional probability of actual equipment degradation state is 0.1. Then, the posterior probability caused by the observation evidence nodes of sensor measurement error, communication transmission disturbance, and manual input deviation is calculated. The posterior probabilities of each cause node are compared, and the error source type corresponding to the node with the highest probability value is selected. Combined with the context information of the fused dataset, a targeted data repair strategy adapted to the error source type is generated.

[0038] Furthermore, step S400 also includes: if the error source type is a sensing measurement error, then a spatially correlated multi-sensor data consistency verification and interpolation repair strategy is generated, using effective sensor data from adjacent spatial locations in the fused dataset to interpolate and generate correction values ​​to replace the sensing data in the fused dataset that are determined to be abnormal; if the error source type is a communication transmission disturbance, then a data reconstruction repair strategy based on a time series prediction model is generated to repair missing or abrupt data in the fused dataset caused by communication anomalies; if the error source type is a manual input deviation, then a repair strategy is generated to call the simulation output value of the asset status to overwrite and replace the corresponding fields in the fused dataset.

[0039] Preferably, if the error source type is a sensor measurement error, then it is verified whether the measurement values ​​of multiple spatially related sensors at the same time section satisfy the mathematical relationship determined by the physical structure of the device. Specifically, the type of physical quantity measured and the spatial location of the abnormal sensor are determined. The measurement values ​​of multiple sensors that measure the same physical quantity as the abnormal sensor and are spatially adjacent to it at the same time section are extracted from the fusion dataset. The deviation between the measurement values ​​of adjacent sensors is calculated. If the deviation is less than a preset consistency threshold, the adjacent sensor data is determined to be valid. Otherwise, a subset of valid sensors is further filtered. Then, the measurement values ​​of the valid adjacent sensors are used to calculate the theoretical correct value of the location of the abnormal sensor through an interpolation algorithm, and the sensor data in the fusion dataset that is determined to be abnormal is replaced. If the error source is communication transmission disturbance, then normal historical data before and after the abnormal data point is used to estimate the theoretically correct value at the abnormal time using a time series prediction model. This is to repair missing or abrupt data in the fused dataset caused by communication anomalies. This includes locating the missing record or abrupt value in the fused dataset caused by communication anomalies, and calculating the correction value using the prediction model. For missing data, the record at that time point is added to the data list with the correction value; for abrupt data, the original abnormal value is replaced with the correction value. If the error source is manual input deviation, then the physical quantity value corresponding to the abnormal field is extracted from the asset status simulation output value. This value is used to replace the original value of the abnormal field in the fused dataset. This includes determining the field name and corresponding physical quantity type in the fused dataset that is judged to be a manual input deviation, extracting the value of the physical quantity at the current time segment from the simulation output value, and using this simulation output value to overwrite the original value of the corresponding field in the fused dataset.

[0040] Step S500: Update the corresponding fields in the data list according to the data repair strategy, generate a new data list version snapshot, and record the identification information of the repair basis and deviation fingerprint database in the data list version snapshot to dynamically update the data list.

[0041] Step S500 further includes: executing the data repair strategy to generate corrected target data; locating the original data field corresponding to the corrected target data in the data list, and overwriting the value of the original data field with the corrected target data; creating a new data list version snapshot, and assigning a globally unique version identifier to the new data list version snapshot; serializing and writing the overwritten and updated data list content into the new data list version snapshot; in the metadata area of ​​the new data list version snapshot, structurally recording the repair basis that triggered this update, the identification information of the deviation fingerprint library used, and the version number of the associated infrastructure asset digital twin model; and switching the currently effective data list pointer of the system from pointing to the old version snapshot to pointing to the new data list version snapshot.

[0042] Preferably, a repair strategy is invoked and specific calculation operations are performed, including spatial correlation interpolation repair, time series prediction reconstruction, and simulation output value overlay, to generate corrected target data, i.e., new values ​​used to replace the original abnormal data. Then, path addressing is used to locate the original data field corresponding to the corrected target data in the data list, where the data list is a structured storage unit that stores all multi-source heterogeneous data. The abnormal values ​​in the located original data fields are replaced with corrected target values, such as value replacement, record supplementation, and flag updates. Next, a new data list version snapshot is created, which is a complete copy of the data list at a specific point in time, used to record all the contents of the data list at that point in time and to ensure the immutability of the data history. A unique identifier is generated for the newly created version snapshot throughout the entire system for query, traceability, and switching operations. The version identifier is recorded in the metadata area of ​​the version snapshot, where the metadata area is specifically used to store "key" information. The "data within data" area is stored separately from the main content of the data list to record the context information of this update. Then, the updated data list content in memory is converted into a byte stream or text format that can be persistently stored and written to the new data list version snapshot. In the metadata area of ​​the new data list version snapshot, the repair basis that triggered this update, the identification information of the deviation fingerprint library used, and the version number of the associated infrastructure asset digital twin model are recorded in a structured manner. Then, the currently effective data list pointer of the system is switched from pointing to the old version snapshot to pointing to the new data list version snapshot. The currently effective data list pointer is used to store the storage location or version identifier of the currently used data list version snapshot, and the value of the currently effective data list pointer is modified from the identifier of the old version snapshot to the identifier of the new version snapshot. For example, database view switching, symbolic link switching, configuration variable updates, and memory pointer updates are performed to achieve dynamic updates of the data list.

[0043] Furthermore, step S500 also includes, within a preset statistical period, aggregating the identification information of the deviation fingerprint database recorded in all data list version snapshots, identifying and extracting the target deviation fingerprint identifiers that are frequently triggered; extracting the historical fusion dataset features associated with the target deviation fingerprint identifiers and the corresponding data repair strategy execution records; and, based on the historical fusion dataset features and the data repair strategy execution records, reverse-optimizing and adjusting the model parameters of the asset status model in the digital twin model of the infrastructure assets, as well as the business tolerance range threshold of the corresponding business constraint rules in the deviation fingerprint database.

[0044] Preferably, the statistical period is a preset fixed-length time window used to collect and statistically analyze deviation fingerprint identification information in data list version snapshots. Within the preset statistical period, all data list version snapshots created within that period are traversed. The identification information of the deviation fingerprint database is read from each data list version snapshot and aggregated. From the aggregation results, deviation fingerprints with trigger frequencies exceeding a preset threshold are identified and marked as target deviation fingerprints, such as those with trigger counts in the top 20% or those with trigger counts in the current period being twice those in the previous period. For each target deviation fingerprint, the fused dataset features corresponding to all version snapshots that triggered that fingerprint are extracted from historical data, including operating condition features, deviation features, time series features, spatial features, and data source features. For each version snapshot that triggered the target deviation fingerprint, detailed records of the data repair strategy executed at that time are extracted from the metadata area. The record should include at least the error source type, repair strategy type, original outlier, corrected value, and repair execution time. Then, using the data characteristics accumulated from historical trigger events and repair execution records, the system reverse-engineers the unreasonable aspects of the current model parameters and rule thresholds and automatically adjusts them to better match the actual operating characteristics of the equipment. This includes analyzing the systematic deviation between the simulation output value and the measured value of the digital twin model under specific operating conditions based on the characteristics of the historical fusion dataset, adjusting the boundary condition parameter values ​​in the model under the corresponding operating conditions, and analyzing the triggering frequency and deviation distribution of specific business constraint rules in the deviation fingerprint database based on the characteristics of the historical fusion dataset, dynamically adjusting the business tolerance range threshold used in the rule, such as absolute threshold, relative threshold, or statistical threshold. This achieves the co-evolution of the twin model and data quality assessment, ensuring improved automation, accuracy, and traceability of data governance.

[0045] In the above text, refer to Figure 1 A data inventory update method based on multi-source data fusion according to embodiments of the present invention is described in detail. Next, reference will be made to... Figure 2 A data inventory update system based on multi-source data fusion according to an embodiment of the present invention is described.

[0046] The data inventory update system based on multi-source data fusion according to embodiments of the present invention addresses the technical problems in the prior art, such as difficulties in multi-source heterogeneous data fusion, insufficient deviation pattern recognition and traceability, and reliance on manual and untraceable data inventory updates. It achieves the technical effect of improving the automation, accuracy, interpretability, and auditability of dynamic data inventory updates. Figure 2 As shown, the data list update system based on multi-source data fusion includes: fused data generation module 10, simulation calculation module 20, data anomaly pattern recognition module 30, data repair strategy generation module 40, and data list update module 50.

[0047] The system comprises the following modules: a data fusion generation module 10, which collects multi-source heterogeneous data of infrastructure assets and performs spatiotemporal alignment and normalization to generate a fusion dataset; a simulation calculation module 20, which inputs the fusion dataset into a pre-built digital twin model of infrastructure assets, runs simulation calculations based on the asset status model, and obtains the simulation output value of the asset status; a data anomaly pattern recognition module 30, which constructs a deviation fingerprint library containing various business constraint rules, calculates the deviation value between the simulation output value and the measured data in the fusion dataset, and matches the deviation value with the business constraint rules in the deviation fingerprint library to identify data anomaly patterns; a data repair strategy generation module 40, which performs source analysis on the error sources corresponding to the data anomaly patterns based on a causal inference network to generate a data repair strategy; and a data list update module 50, which updates the corresponding fields in the data list according to the data repair strategy, generates a new data list version snapshot, and records the repair basis and the identification information of the deviation fingerprint library in the data list version snapshot to dynamically update the data list.

[0048] The specific configuration of the fusion data generation module 10 will be described in detail below. The fusion data generation module 10 further includes: the multi-source heterogeneous data includes at least asset online runtime sequence data, asset procurement and maintenance record parsing results, and asset deployment environment parameters.

[0049] The specific configuration of the simulation calculation module 20 will be described in detail below. The simulation calculation module 20 further includes: extracting the operating conditions and attribute parameters of the infrastructure asset at a specific time segment from the fused dataset; loading the operating conditions and attribute parameters as initial and boundary conditions into the initial value module and boundary condition module of the pre-built infrastructure asset digital twin model, respectively; calling the asset state model built into the infrastructure asset digital twin model to perform multi-state coupled calculations on the operating conditions and attribute parameters, generating the internal state evaluation result of the infrastructure asset at the specific time segment; and parsing the asset internal state quantities that cannot be directly measured from the internal state evaluation result as the simulation output value of the asset state.

[0050] The specific configuration of the data anomaly pattern recognition module 30 will be described in detail below. The data anomaly pattern recognition module 30 further includes: defining a set of business constraint rules characterizing the reasonable boundaries of the asset's state based on the construction safety specifications and material performance standards of the infrastructure asset; wherein the business constraint rules describe state quantity constraint relationships that cannot be simultaneously satisfied under the asset's normal service state; encoding each business constraint rule into a computable logical judgment statement containing logical operators and threshold judgments, and assigning a unique deviation fingerprint identifier to each logical judgment statement; aggregating all logical judgment statements and their corresponding deviation fingerprint identifiers to construct a deviation fingerprint library containing multiple business constraint rules.

[0051] The specific configuration of the data anomaly pattern recognition module 30 will be described in detail below. The data anomaly pattern recognition module 30 further includes: dynamically determining the business tolerance range threshold under the current operating condition based on the current operating condition of the measured data in the fused dataset, combined with the statistical characteristics of historical data under the same operating condition; calculating the deviation value between the simulation output value and the measured data in the fused dataset, and determining whether the deviation value exceeds the business tolerance range threshold; if it exceeds, extracting state variables related to the deviation value from the fused dataset to form a current state variable combination; matching the current state variable combination with each computable logical judgment statement in the deviation fingerprint database; when the current state variable combination satisfies the triggering condition of any computable logical judgment statement in the deviation fingerprint database, determining the corresponding data anomaly pattern based on the deviation fingerprint identifier corresponding to the logical judgment statement.

[0052] The specific configuration of the data repair strategy generation module 40 will be described in detail below. The data repair strategy generation module 40 further includes: constructing a causal inference network, which contains nodes representing sensor measurement errors, communication transmission disturbances, manual input deviations, and the actual degradation state of the equipment, with causal relationships between nodes represented by directed edges; inputting the deviation fingerprint identifier and deviation value corresponding to the data anomaly pattern as observation evidence into the causal inference network; querying a conditional probability table constructed based on a historical fault case library to calculate the posterior probability caused by the sensor measurement error, communication transmission disturbance, and manual input deviation nodes, respectively; selecting the error source type represented by the node with the highest posterior probability, and combining it with the context information of the fused dataset to generate a targeted data repair strategy adapted to the error source type.

[0053] The specific configuration of the data repair strategy generation module 40 will be described in detail below. The data repair strategy generation module 40 further includes: if the error source type is a sensor measurement error, a spatially correlated multi-sensor data consistency verification and interpolation repair strategy is generated, using valid sensor data from adjacent spatial locations in the fused dataset to interpolate and generate correction values, replacing the sensor data in the fused dataset that is determined to be abnormal; if the error source type is a communication transmission disturbance, a data reconstruction repair strategy based on a time series prediction model is generated to repair missing or abrupt data in the fused dataset caused by communication anomalies; if the error source type is a manual input deviation, a repair strategy is generated that calls the simulation output value of the asset status to overwrite and replace the corresponding fields in the fused dataset.

[0054] The specific configuration of the data list update module 50 will be described in detail below. The data list update module 50 further includes: executing the data repair strategy to generate corrected target data; locating the original data field corresponding to the corrected target data in the data list, and overwriting the value of the original data field with the corrected target data; creating a new data list version snapshot, and assigning a globally unique version identifier to the new data list version snapshot; serializing and writing the overwritten and updated data list content into the new data list version snapshot; in the metadata area of ​​the new data list version snapshot, structurally recording the repair basis that triggered this update, the identification information of the deviation fingerprint library used, and the version number of the associated infrastructure asset digital twin model; and switching the currently active data list pointer from pointing to the old version snapshot to pointing to the new data list version snapshot.

[0055] The specific configuration of the data list update module 50 will be described in detail below. The data list update module 50 further includes: aggregating the identification information of the deviation fingerprint database recorded in all data list version snapshots within a preset statistical period; identifying and extracting frequently triggered target deviation fingerprint identifiers; extracting historical fusion dataset features associated with the target deviation fingerprint identifiers and corresponding data repair strategy execution records; and, based on the historical fusion dataset features and data repair strategy execution records, reverse-optimizing and adjusting the model parameters of the asset status model in the infrastructure asset digital twin model, as well as the business tolerance range threshold of the corresponding business constraint rules in the deviation fingerprint database.

[0056] The data list update system based on multi-source data fusion provided in this embodiment of the invention can execute the data list update method based on multi-source data fusion provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A data inventory update method based on multi-source data fusion, characterized in that, The method includes: Collect multi-source heterogeneous data of infrastructure assets and perform spatiotemporal alignment and normalization processing to generate a fused dataset; The fused dataset is input into a pre-built digital twin model of infrastructure assets, and simulation calculations based on the asset status model are run to obtain the simulation output value of the asset status. A deviation fingerprint database containing various business constraint rules is constructed. The deviation value between the simulation output value and the measured data in the fused dataset is calculated, and the deviation value is matched with the business constraint rules in the deviation fingerprint database to identify abnormal data patterns. Based on the causal reasoning network, the error sources corresponding to the data anomaly patterns are traced and analyzed to generate data repair strategies. Update the corresponding fields in the data list according to the data repair strategy, generate a new data list version snapshot, and record the repair basis and the identification information of the deviation fingerprint database in the data list version snapshot to dynamically update the data list. Construct a deviation fingerprint library containing various business constraint rules, including: Based on the construction safety specifications and material performance standards of the infrastructure assets, a set of business constraint rules are defined to characterize the reasonable boundaries of the asset state. The business constraint rules describe the state quantity constraint relationships that cannot be satisfied simultaneously under the normal service state of the asset. Each business constraint rule is encoded into a computable logical judgment statement containing logical operators and threshold judgments, and a unique deviation fingerprint is assigned to each logical judgment statement. Aggregate all logical judgment statements and their corresponding deviation fingerprint identifiers to construct a deviation fingerprint library containing business constraint rules; Calculate the deviation between the simulation output value and the measured data in the fused dataset, and match the deviation value with the business constraint rules in the deviation fingerprint database to identify data anomaly patterns, including: Based on the current operating conditions of the measured data in the fused dataset, and combined with the statistical characteristics of historical data under the same operating conditions, the threshold of the business tolerance range under the current operating conditions is dynamically determined. Calculate the deviation between the simulation output value and the measured data in the fused dataset, and determine whether the deviation exceeds the business tolerance range threshold. If the deviation value is exceeded, then the state variables related to the deviation value are extracted from the fused dataset to form the current state variable combination; The current state quantity combination is matched one by one with each computable logical judgment statement in the deviation fingerprint database; When the current state quantity combination satisfies the triggering condition of any computable logical judgment statement in the deviation fingerprint database, the corresponding data anomaly mode is determined according to the deviation fingerprint identifier corresponding to the logical judgment statement.

2. The data list update method based on multi-source data fusion as described in claim 1, characterized in that, The multi-source heterogeneous data includes at least asset online runtime sequence data, asset procurement and maintenance record parsing results, and asset deployment environment parameters.

3. The data list update method based on multi-source data fusion as described in claim 1, characterized in that, The fused dataset is input into a pre-built digital twin model of infrastructure assets, and simulation calculations based on the asset status model are run to obtain simulation output values ​​of the asset status, including: Extract the operating conditions and attribute parameters of the infrastructure assets at a specific time segment from the fused dataset; The operating conditions and attribute parameters are used as initial and boundary conditions, respectively, and loaded into the initial value module and boundary condition module of the pre-built digital twin model of infrastructure assets. The asset status model built into the digital twin model of the infrastructure asset is invoked to perform multi-state coupled calculations on the working conditions and attribute parameters, thereby generating the internal status assessment results of the infrastructure asset at the specific time section. The asset internal state quantities that cannot be directly measured are extracted from the internal state assessment results and used as the simulation output values ​​of the asset state.

4. The data list update method based on multi-source data fusion as described in claim 1, characterized in that, Based on a causal reasoning network, the error sources corresponding to the data anomaly patterns are traced and analyzed to generate data repair strategies, including: Construct a causal reasoning network, which includes nodes representing sensing measurement errors, communication transmission disturbances, manual input deviations, and the actual degradation state of the equipment. The causal relationships between nodes are represented by directed edges. The deviation fingerprint identifier and deviation value corresponding to the data anomaly pattern are used as observation evidence and input into the causal inference network; Query the conditional probability table constructed based on the historical fault case database, and calculate the posterior probability of the observed evidence caused by sensor measurement error, communication transmission disturbance and manual input deviation node respectively; Select the error source type represented by the node with the highest posterior probability, and combine it with the context information of the fused dataset to generate a targeted data repair strategy adapted to the error source type.

5. The data list update method based on multi-source data fusion as described in claim 4, characterized in that, Generate a targeted data repair strategy adapted to the error source type, including: If the error source type is a sensor measurement error, a spatially correlated multi-sensor data consistency verification and interpolation repair strategy is generated. The effective sensor data of adjacent spatial locations in the fused dataset are used to interpolate and generate correction values ​​to replace the sensor data in the fused dataset that are judged to be abnormal. If the error source type is communication transmission disturbance, a data reconstruction and repair strategy based on a time series prediction model is generated to repair the missing or abrupt data in the fused dataset caused by communication anomalies. If the error source type is manual input deviation, a repair strategy is generated to overwrite and replace the corresponding field in the fused dataset with the simulation output value of the asset status.

6. The data list update method based on multi-source data fusion as described in claim 1, characterized in that, According to the data repair strategy, the corresponding fields in the data list are updated, a new data list version snapshot is generated, and the repair basis and the identification information of the deviation fingerprint database are recorded in the data list version snapshot. Dynamic updates to the data list are performed, including: The data repair strategy is executed to generate the corrected target data; Locate the original data field in the data list that corresponds to the corrected target data, and update the value of the original data field by overwriting it with the corrected target data; Create a new data inventory version snapshot and assign a globally unique version identifier to the new data inventory version snapshot; The updated data list content will be serialized and written into the new data list version snapshot. In the metadata area of ​​the new data list version snapshot, the structured records the basis for the repair that triggered this update, the identification information of the deviation fingerprint database used, and the version number of the associated infrastructure asset digital twin model; Switch the system's currently active data list pointer from pointing to the old version snapshot to pointing to the new data list version snapshot.

7. The data list update method based on multi-source data fusion as described in claim 6, characterized in that, After generating a new data manifest version snapshot, the following is also included: Within a preset statistical period, aggregate the identification information of the deviation fingerprint database recorded in all data list version snapshots, and identify and extract the target deviation fingerprint identifiers that are frequently triggered. Extract the features of the historical fusion dataset associated with the target deviation fingerprint identifier and the corresponding data repair strategy execution records; Based on the features of the historical fusion dataset and the execution records of the data repair strategy, the model parameters of the asset status model in the digital twin model of infrastructure assets and the business tolerance range threshold of the corresponding business constraint rules in the deviation fingerprint database are optimized and adjusted in reverse.

8. A data inventory update system based on multi-source data fusion, characterized in that, The system is used to implement the data list update method based on multi-source data fusion as described in any one of claims 1 to 7, and the system comprises: The fusion data generation module is used to collect multi-source heterogeneous data of infrastructure assets and perform spatiotemporal alignment and normalization processing to generate a fusion dataset. The simulation calculation module is used to input the fused dataset into the pre-built digital twin model of infrastructure assets, run simulation calculations based on the asset status model, and obtain the simulation output value of the asset status. The data anomaly pattern recognition module is used to construct a deviation fingerprint database containing various business constraint rules, calculate the deviation value between the simulation output value and the measured data in the fused dataset, and match the deviation value with the business constraint rules in the deviation fingerprint database to identify data anomaly patterns. The data repair strategy generation module is used to perform source analysis on the error sources corresponding to the data anomaly patterns based on the causal inference network, and generate data repair strategies. The data list update module is used to update the corresponding fields in the data list according to the data repair strategy, generate a new data list version snapshot, and record the repair basis and the identification information of the deviation fingerprint database in the data list version snapshot to dynamically update the data list.

Images