An oil sample detection method, system, device and medium based on digital twinning

Abstract

This invention discloses a method, system, equipment, and medium for oil sample testing based on digital twins, belonging to the field of power equipment condition monitoring technology. The method includes: constructing a digital twin system architecture; performing four-dimensional modeling of the testing equipment and task status modeling of the oil sample to be tested, establishing a bidirectional real-time mapping relationship between physical space and information space; constructing a data flow chain using unique identifier codes as indexes after preprocessing multi-source heterogeneous data, and standardizing the data through a unified data model; performing three-level fusion processing on the standardized data—data-level, feature-level, and decision-level—to obtain the oil sample testing fusion result; and performing visual monitoring and traceability management based on the oil sample testing fusion result. This invention enables the real-time representation of the entire oil sample testing process's flow status and equipment operating parameters in the information space through bidirectional real-time mapping, and fully records the entire chain of algorithm parameters through the data flow chain, supporting reverse tracing from the testing fusion result back to the original data.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring technology, specifically to an oil sample testing method, system, equipment, and medium based on digital twins. Background Technology

[0002] Transformer insulating oil sample testing is a crucial method for assessing the insulation condition of transformers. By performing chromatographic analysis, moisture detection, dielectric loss testing, and breakdown voltage testing on the insulating oil, the transformer's operational health can be effectively reflected. However, in existing technologies, each testing instrument operates independently, resulting in a lack of a unified data model for multi-source, heterogeneous testing data. Furthermore, the intermediate processing steps cannot be reconstructed after multiple data processing steps, failing to meet the requirement of full lifecycle traceability for testing data. In addition, existing testing systems exhibit a disconnect between physical and information spaces, lacking real-time mapping and visualization monitoring mechanisms for equipment operating status, sample transfer progress, and laboratory environmental parameters, making it difficult to promptly identify and address equipment malfunctions. Summary of the Invention

[0003] In view of the above-mentioned problems, the present invention provides an oil sample detection method, system, device and medium based on digital twin.

[0004] Therefore, the technical problem solved by this invention is that in the process of multi-instrument collaborative testing of transformer insulating oil samples in the laboratory, the heterogeneous test data output by the moisture analyzer, oil chromatograph, dielectric loss tester and breakdown voltage tester cannot be correlated and fused under a unified framework, and the intermediate process of the test data cannot be restored after multiple steps of processing, resulting in insufficient reliability of the test results, no monitoring of the entire process status and no data traceability.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an oil sample detection method based on digital twins, comprising, Construct a digital twin system architecture consisting of a physical entity layer, a virtual model layer, a twin data layer, and an application service layer; The detection equipment in the physical entity layer is modeled in four dimensions: geometric dimension, physical attribute dimension, rule dimension, and behavior dimension. The oil sample to be tested carrying a unique identifier code in the physical entity layer is modeled in task status according to the detection task set and detection process status. A bidirectional real-time mapping association is established between the physical space of the physical entity layer and the information space of the virtual model layer. The twin data layer stores multi-source heterogeneous data throughout the entire oil sample testing process, and preprocesses the multi-source heterogeneous data to obtain preprocessed data. A data flow chain is constructed using the unique identifier code of the oil sample to be tested as an index, and the preprocessed data is standardized and transformed through a unified data model to obtain standardized data. The standardized data is subjected to data-level fusion processing, the result of the data-level fusion processing is subjected to feature-level fusion processing, and the result of the feature-level fusion processing is subjected to decision-level fusion processing to obtain the oil sample detection fusion result. The application service layer performs visual monitoring of the entire oil sample testing process based on the oil sample testing fusion results, and performs traceability management of the entire oil sample testing process based on the data flow chain.

[0006] As a preferred embodiment of the oil sample detection method based on digital twins described in this invention, the four-dimensional modeling includes: modeling the physical properties of the detection device, and performing structural analysis, acoustic field analysis, electromagnetic field analysis, and multi-physics coupling analysis based on the finite element model of the detection device. Behavioral modeling is performed on the detection equipment, based on the task execution status, external environment, sudden disturbances and internal operating mechanism of the detection equipment in the stages of chromatographic detection, moisture detection, dielectric loss detection and breakdown voltage detection. The task state modeling includes: associating the unique identifier code carried by the radio frequency identification electronic tag with the detection task set of the oil sample to be tested; constructing a virtual model of the oil sample to be tested in the virtual model layer; and synchronizing the state changes of the oil sample to be tested in the stages of chromatographic detection, moisture detection, dielectric loss detection, and breakdown voltage detection to the virtual model of the oil sample to be tested.

[0007] As a preferred embodiment of the oil sample detection method based on digital twins described in this invention, the bidirectional real-time mapping association includes: establishing a communication connection between the physical space and the information space through an industrial Ethernet communication interface; The status data of the physical entities in the physical entity layer is transmitted to the virtual model layer in real time via the industrial Ethernet communication interface. The virtual model layer updates the status of the virtual model within the virtual model layer based on the status data. The virtual model layer performs feature extraction processing on the state data, and transmits the result of the feature extraction processing to the physical entity layer via the industrial Ethernet communication interface. Based on the result of the feature extraction processing, the operating parameters of the detection operation are adjusted.

[0008] As a preferred embodiment of the oil sample detection method based on digital twins described in this invention, the preprocessing includes: using an outlier identification algorithm to identify outliers in the multi-source heterogeneous data, removing the identified outliers from the multi-source heterogeneous data, and obtaining first processed data; Density clustering algorithm is used to identify continuous abnormal segments in the time series data in the first processed data, and the continuous abnormal segments are removed from the first processed data to obtain the second processed data; The missing data caused by sensor offline in the second processed data is filled in using a time-weighted missing data filling algorithm to obtain the preprocessed data; The twin data layer organizes the time-series data collected at a preset sampling frequency in the preprocessed data in a multi-dimensional manner using a spatiotemporal cube model, and stores the multi-dimensionally organized time-series data into the twin data layer through a three-dimensional index.

[0009] As a preferred embodiment of the oil sample testing method based on digital twins described in this invention, the unified data model includes: using a web ontology language to formally describe the entities and relationships between entities in the entire oil sample testing process, and mapping the results of the formal description to the terminology system in the testing field standards and specifications through a terminology mapping table; The data flow chain records the entire process of the preprocessed data being standardized and transformed into standardized data, with each processing step as a unit. The corresponding algorithm parameters are recorded for each processing step in the entire chain.

[0010] As a preferred embodiment of the oil sample detection method based on digital twins described in this invention, the data-level fusion processing includes: aligning similar detection data output by the detection device to a unified time reference using a time synchronization protocol, and performing weighted fusion on the aligned similar detection data based on a dynamic weight calculation model to obtain a data-level fusion result; The feature-level fusion processing includes: using a dimensionality reduction algorithm to reduce the dimensionality of the features in the data-level fusion result to obtain dimensionality-reduced features; using a temporal feature extraction model to extract dynamic features from the temporal detection data in the data-level fusion result to obtain temporal features; and concatenating the dimensionality-reduced features with the temporal features to obtain a feature vector.

[0011] As a preferred embodiment of the oil sample detection method based on digital twins described in this invention, the decision-level fusion processing includes: determining the confidence level of the detection data of the detection instruments in the detection equipment in the feature vector based on a probability allocation model, wherein the confidence level is determined based on the direct measurement results of the detection instruments, the mean of repeated detections, and the comparison results across the detection equipment; The trust levels of the detection instruments in the detection equipment are fused based on the evidence combination rules to obtain the fused detection result of the oil sample.

[0012] This invention provides an oil sample detection system based on digital twins.

[0013] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an oil sample detection system based on digital twin, comprising: a physical entity layer, a virtual model layer, a twin data layer, and an application service layer; The physical entity layer includes detection equipment and oil samples to be tested carrying unique identification codes. The detection equipment completes four-dimensional modeling according to geometric dimension, physical attribute dimension, rule dimension, and behavior dimension. The oil sample to be tested completes task state modeling according to the detection task set and detection process state. A bidirectional real-time mapping relationship is established between the physical space of the physical entity layer and the information space of the virtual model layer. The twin data layer is used to store multi-source heterogeneous data in the entire oil sample testing process, preprocess the multi-source heterogeneous data to obtain preprocessed data, construct a data flow chain with the unique identifier code of the oil sample to be tested as an index, standardize the preprocessed data through a unified data model to obtain standardized data, perform data-level fusion processing on the standardized data, perform feature-level fusion processing on the result of data-level fusion processing, and perform decision-level fusion processing on the result of feature-level fusion processing to obtain the oil sample testing fusion result. The application service layer is used to perform visual monitoring of the entire oil sample testing process based on the oil sample testing fusion results, and to perform traceability management of the entire oil sample testing process based on the data flow chain.

[0014] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the oil sample detection method based on digital twin.

[0015] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the oil sample detection method based on digital twins.

[0016] The beneficial effects of this invention are as follows: By constructing a digital twin system architecture consisting of a physical entity layer, a virtual model layer, a twin data layer, and an application service layer, this invention completes four-dimensional modeling of the detection equipment and establishes a two-way real-time mapping relationship between the physical space and the information space. This allows the flow status of the oil sample to be tested and the operating parameters of the detection equipment in the entire process of chromatographic detection, moisture detection, dielectric loss detection, and breakdown voltage detection to be characterized in real time in the information space. The operating parameters of the detection operation can be dynamically adjusted based on the feature extraction results of the information space.

[0017] After three-step preprocessing and three-level fusion processing, multi-source heterogeneous data is integrated into a unified oil sample testing fusion result. The data flow chain, indexed by a unique identifier code, fully records the algorithm parameters of each processing step in the entire chain. It supports tracing back from the oil sample testing fusion result to each data node of the original multi-source heterogeneous data. The entire oil sample testing process is under a monitorable and traceable control. Attached Figure Description

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

[0019] Figure 1 This is a flowchart illustrating an oil sample detection method based on digital twins, as provided in one embodiment of the present invention.

[0020] Figure 2 This is a preprocessing flowchart of an oil sample detection method based on digital twins, provided as an embodiment of the present invention.

[0021] Figure 3 This is a flowchart illustrating the construction of a unified data model for an oil sample detection method based on digital twins, as provided in one embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram of a four-layer architecture of a digital twin system for an oil sample detection method based on digital twins, provided as an embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram of the interlayer interaction relationship of a digital twin system for an oil sample detection method based on digital twin, provided as an embodiment of the present invention.

[0024] Figure 6 This is a schematic diagram of the entire process of oil sample testing based on digital twin, as provided in one embodiment of the present invention. Detailed Implementation

[0025] To make the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0026] Example 1, referring to Figures 1-6 This is one embodiment of the present invention, which provides an oil sample detection method based on digital twins, comprising: S1 constructs a digital twin system architecture consisting of a physical entity layer, a virtual model layer, a twin data layer, and an application service layer.

[0027] In this embodiment, the specific implementation of the digital twin system architecture in step S1 includes steps S11 to S14, wherein: S11, the physical entity layer, includes a transfer robot, a sorting robot, a pipetting device, a chromatograph sample loading unit, analytical instruments, and transformer insulating oil serving the entire process of transformer insulating oil sample testing, as well as an industrial control computer, sensors, data acquisition cards, and a wireless RFID reader / writer. The transfer robot and sorting robot complete the transportation and sorting of transformer insulating oil samples by receiving control instructions from the industrial control computer. The sensors and data acquisition cards collect real-time data on the operating status during the testing process, and the wireless RFID reader / writer tracks the sample number and testing progress by reading the wireless RFID electronic tags on the oil sample container.

[0028] S12, the virtual model layer is the mirror image of the physical entity layer in the information space, including the geometric model, physical model, rule model and behavior model obtained by modeling the detection equipment and the oil sample to be tested in the physical entity layer; the geometric model describes the shape, position and assembly relationship of the detection equipment, the physical model describes the physical properties and constraint relationship of the detection equipment, the rule model constructs judgment and prediction rules based on the historical operating data of the detection equipment, and the behavior model constructs evaluation and decision-making models based on the external environment, sudden disturbances and internal operating mechanisms.

[0029] S13, the twin data layer carries physical entity data, virtual model data, application service data, and fused derived data throughout the entire oil sample testing process. Physical entity data includes detection data on dissolved gas content, moisture content, and breakdown voltage in the oil sample, as well as equipment operating conditions and equipment status data during the testing process. Virtual model data includes shape, position, and assembly information of the geometric model, attribute and characteristic constraint information of the physical model, sudden disturbances and internal operating mechanisms of the behavioral model, and constraints and correlation information of the rule model. Application service data includes test report management data, equipment health status assessment data, and maintenance decision data. Fusion derived data is multi-source fused data obtained after cleaning, associating, clustering, and fusing the physical entity data, virtual model data, and application service data.

[0030] Understandably, the twin data layer updates and mines physical entity data, virtual model data, application service data, and fused derived data in real time throughout the entire oil sample testing process, ensuring the continuous operation of the data-driven mechanism at each layer of the digital twin system architecture.

[0031] S14, the application service layer integrates the product lifecycle management system, enterprise resource planning system, product data management system and manufacturing execution system. Based on the fusion-derived data output from the twin data layer, the virtual model layer is driven to realize real-time status monitoring, fault warning and optimization decision-making in the oil sample testing process.

[0032] Understandably, the digital twin system architecture carries the physical entities of the entire detection process through the physical entity layer, constructs a mirror model of the physical entities in the information space through the virtual model layer, stores and mines multi-source data in real time through the twin data layer, and presents the oil sample detection fusion results in a visual form through the application service layer. The four layers of the architecture realize the real-time transmission of information through a data-driven mechanism, transforming the isolated operation mode of the physical space and information space in the traditional detection process into a collaborative operation mode of real-time linkage between the physical space and information space.

[0033] Reference Figure 4 In the digital twin system architecture, the physical entity layer consists of physical entities such as flow robots, sorting robots, automated equipment, analytical instruments, sensors, industrial control computers, and RFID readers. The physical entity layer interacts with the detection entity in real time through a two-way monitoring and detection channel. The twin data layer provides data storage, data fusion, data mining, and data transformation functions, carrying multi-source heterogeneous data generated throughout the oil sample testing process. The virtual model layer includes geometric models, physical models, behavioral models, and rule models, constructing a mirror image of the physical entity layer in the information space. The application service layer integrates Product Lifecycle Management (PLM), Enterprise Resource Planning (ERP), Product Data Management (PDM), and Manufacturing Execution System (MES) for real-time monitoring of the oil sample testing process, early warning of faults in the testing process, and optimization decisions in the testing process. The four layers achieve the full-process flow of information between the physical space and the information space through data-driven and interactive iteration mechanisms, forming a complete closed loop of the digital twin system.

[0034] S2 performs four-dimensional modeling of the detection equipment in the physical entity layer according to geometric dimension, physical attribute dimension, rule dimension and behavior dimension. It performs task status modeling of the oil sample to be tested carrying a unique identifier code in the physical entity layer according to the detection task set and detection process status. It establishes a bidirectional real-time mapping relationship between the physical space of the physical entity layer and the information space of the virtual model layer.

[0035] In this embodiment, the specific implementation of four-dimensional modeling in step S2 includes steps S21 to S22, the specific implementation of task state modeling is step S23, and the specific implementation of establishing bidirectional real-time mapping association includes steps S24 to S26, wherein: S21, perform physical property dimension modeling of the testing equipment, and perform structural analysis, acoustic field analysis, electromagnetic field analysis, and multi-physics coupling analysis based on the finite element model of the testing equipment.

[0036] Understandably, the finite element model is constructed based on the actual material parameters, geometric dimensions, and boundary conditions of the testing equipment, forming a digital physical model of the testing equipment in the information space; structural analysis is used to describe the mechanical structural features and constraint relationships of the testing equipment; acoustic field analysis is used to describe the acoustic features generated by the testing equipment during operation; electromagnetic field analysis is used to describe the electromagnetic characteristics of the testing equipment; multiphysics coupling analysis is used to describe the coupling behavior of the testing equipment under the combined action of structural field, acoustic field, and electromagnetic field; and physical property dimension modeling fully describes the physical properties, features, and constraint information of the testing equipment.

[0037] In this embodiment, step S21 uses a formal modeling language to formally define the key elements in the physical entity layer: (1) in, The physical space representing the key elements of the entire oil sample testing process; The notation used in formal modeling languages ​​represents the definition of the left-hand space as being composed of right-hand elements; This represents the set of all detection devices in the physical entity layer; This represents the set of all oil samples to be tested within the physical entity layer; express and The natural connections between them, describing and The natural interaction between them; and All are dynamic sets, and the elements and their states are continuously updated as the detection process progresses.

[0038] Furthermore, in step S21, the following steps are performed: Formal definition: (2) in, This represents the set of all detection devices in the physical entity layer; Represents the first to the second physical entity layer Taiwan testing equipment; This represents the total number of detection devices in the physical entity layer; This represents the index of the testing equipment, with a value range of 1 to... .

[0039] Furthermore, in step S21, the first... Taiwan testing equipment Perform formal modeling: (3) (4) (5) in, Represents the first in the physical entity layer Taiwan testing equipment; Indicates the first Model and technical parameters of the testing equipment; Indicates the first The sensor data set configured on the testing equipment; Indicates the first The operation log information of the testing equipment during the testing process; Indicates the first The total number of sensors configured on the testing equipment; Indicates the first The first of the testing equipment Data objects of a sensor; This represents the sensor index, with a value ranging from 1 to... ; Indicates the first The first of the testing equipment The model number of the sensor; Indicates the first The first of the testing equipment The data set collected in real time by each sensor covers different detection items, including chromatographic datasets, dielectric loss factor datasets, or breakdown voltage datasets.

[0040] It is understandable that the device model parameter of the physical entity data in step S13 comes from... The sensor data comes from The runtime log comes from Through real-time updates Each of and This enables the continuous input of device status data from the physical entity layer to the twin data layer.

[0041] It should be noted that the four-dimensional modeling in step S2 also includes geometric dimension modeling and rule dimension modeling. Geometric dimension modeling constructs a digital model based on the actual layout of the testing equipment, describing the shape, position and assembly relationship of the testing equipment; rule dimension modeling constructs a rule model for judgment, optimization and prediction based on the historical correlation data of the testing equipment.

[0042] S22, perform behavioral dimension modeling of the detection equipment, based on the task execution status, external environment, sudden disturbances and internal operating mechanism of the detection equipment in the stages of chromatographic detection, moisture detection, dielectric loss detection and breakdown voltage detection.

[0043] Understandably, behavioral dimension modeling, at different time scales, integrates the task execution status, external environmental conditions, sudden disturbance events, and internal operating mechanisms of the comprehensive testing equipment during the execution of four types of testing tasks: chromatographic detection, moisture detection, dielectric loss detection, and breakdown voltage detection. Behavioral models are constructed for each of the four testing task stages: chromatographic detection, moisture detection, dielectric loss detection, and breakdown voltage detection. These behavioral models are used to evaluate the operating status of the testing equipment throughout the entire oil sample testing process and output decision instructions for testing operations based on the evaluation results.

[0044] In this embodiment, step S22 uses a formal modeling language to formally define the key elements in the virtual model layer: (6) in, An information space representing the key elements of the entire oil sample testing process; The notation used in formal modeling languages ​​represents the definition of the left-hand space as being composed of right-hand elements; This represents the collection of twin models of all detection devices in the virtual model layer; This represents the set of all twin models of the oil samples to be tested in the virtual model layer; express and The natural connections between them, describing and The natural interaction between them; and All are dynamic sets, and the elements and their states of the sets are updated synchronously with the dynamic operation of the detection process in the physical entity layer.

[0045] Furthermore, in step S22... Formal definition: (7) in, This represents the collection of twin models of all detection devices in the virtual model layer; Represents the first to the second virtual model layer A twin model of the testing equipment; This represents the total number of detection devices in the physical entity layer; This represents the index of the testing equipment, with a value range of 1 to... .

[0046] Furthermore, in step S22, the first... Twin model of Taiwan testing equipment Perform formal modeling: (8) in, Represents the first virtual model layer A twin model of the testing equipment; Indicates the first A digital twin functional model of the testing equipment in the information space; express The set of physical data that can be obtained during oil sample testing; express Predefined rules and behavioral mathematical models.

[0047] Understandable It includes the results of physical attribute dimension modeling and behavioral dimension modeling in step S21; Data collected from the sensors and data acquisition card in step S11 ; Derived from the behavioral model constructed in step S22 and the rule-dimensional modeling results in step S21, This serves as the rule basis for feature extraction processing in step S26.

[0048] S23, associate the unique identification code carried by the radio frequency identification electronic tag with the detection task set of the oil sample to be tested, construct a virtual model of the oil sample to be tested in the virtual model layer, and synchronize the state changes of the oil sample to be tested in the stages of chromatographic detection, moisture detection, dielectric loss detection and breakdown voltage detection to the virtual model of the oil sample to be tested.

[0049] Understandably, the oil sample to be tested consists of the oil sample body and the radio frequency identification (RFID) electronic tag on the container. The RFID electronic tag uniquely identifies the sample number and test batch number of the oil sample through the electronic product code. The electronic product code serves as the unique identifier of the oil sample throughout the entire testing process, ensuring the uniqueness of each sample throughout the entire testing process.

[0050] In this embodiment, step S23 involves... Formal definition: (9) in, This represents the set of all oil samples to be tested within the physical entity layer; Represents the first to the second physical entity layer One oil sample to be tested; This indicates the total number of oil samples to be tested in the physical entity layer; This represents the index of the oil sample to be tested, with a value ranging from 1 to... .

[0051] Furthermore, in step S23, the first... One oil sample to be tested Perform formal modeling: (10) (11) in, Represents the first in the physical entity layer One oil sample to be tested; Indicates the first The electronic product code of the radio frequency identification (RFID) tag attached to each oil sample to be tested; Indicates the first The basic attributes of the oil sample to be tested include information such as the source of the oil sample, the sampling time, and the transformer number to which it belongs; Indicates the first The set of testing tasks required to complete the entire testing process for an oil sample; Indicates the first The detection process status log of each oil sample to be tested; Indicates the first The total number of testing tasks required for each oil sample to be tested; Indicates the first The first oil sample to be tested One detection task; This represents the index of the detection task, with a value range of 1 to... .

[0052] Furthermore, in step S23... Formal definition: (12) in, This represents the set of all twin models of the oil samples to be tested in the virtual model layer; Represents the first to the second virtual model layers. A twin model of the oil sample to be tested; This indicates the total number of oil samples to be tested in the physical entity layer; This represents the index of the oil sample to be tested, with a value ranging from 1 to... .

[0053] Furthermore, in step S23, the first... Twin model of an oil sample to be tested Perform formal modeling: (13) in, Represents the first virtual model layer A twin model of the oil sample to be tested; Indicates the first The electronic product code of the radio frequency identification (RFID) tag attached to each oil sample to be tested; Indicates the first A digital twin functional model of an oil sample to be tested in the information space; express The collection of available detection parameters, detection results, and historical data; express Predefined rules and behavioral mathematical models.

[0054] Understandable exist and The definitions are the same, and they are shared. make sure and A unique correspondence between them; Include and The modeling results; Oil sample detection data components derived from the physical entity data in step S13; drive Follow The state changes at each detection stage are recorded to complete the state evolution.

[0055] Furthermore, automatic identification technology uses radio frequency identification (RFID) readers to automatically read the information on RFID electronic tags. ,Will The detection tasks and The detection process status is associated with In the physical entity layer, the RFID reader reads data in real time. The state changes of the oil sample under test are obtained during four detection stages: chromatographic detection, moisture detection, dielectric loss detection, and breakdown voltage detection. These state changes are then updated to [the relevant data / process]. And synchronize to ,drive Complete the state evolution.

[0056] Understandably, with A uniquely identifiable task state modeling mechanism will In physical space Recorded detection progress and information space Real-time correlation enables accurate characterization of the state changes of the oil sample in the four detection stages—chromatographic detection, moisture detection, dielectric loss detection, and breakdown voltage detection—in the information space, transforming the traditional detection process that relies on manual recording of sample state into a real-time, data-driven approach. Automatic state synchronization method for the driver.

[0057] S24 establishes a communication connection between the physical space and the information space through an industrial Ethernet communication interface.

[0058] Understandably, the industrial Ethernet communication interface adopts a high-speed, low-latency data transmission protocol. The data transmission protocol ensures the real-time performance and stability of data acquisition and transmission between the physical space and the information space, enabling the forward transmission of physical entity state data in step S25 and the reverse transmission of feature extraction results in step S26 to be completed under low-latency conditions.

[0059] In this embodiment, in step S24 and The bidirectional real mapping between them is defined as: (14) in, The physical space representing the key elements of the entire oil sample testing process; An information space representing the key elements of the entire oil sample testing process; Indicates PS and The mapping ratio between them is The bidirectional real mapping relationship, i.e. Each unit and In Maintain synchronous bidirectional real mapping, Each and In Maintain synchronous bidirectional real mapping.

[0060] It is understandable that in formula (14) The physical implementation of bidirectional mapping relies on an industrial Ethernet communication interface. Step S25 realizes the mapping from... to Forward state data transmission in the direction, step S26 realizes the transmission from... to The feature extraction results of the direction are transmitted in reverse. Steps S25 and S26 together constitute the bidirectional mapping mechanism described by formula (14).

[0061] S25, the state data of physical entities in the physical entity layer is transmitted to the virtual model layer in real time via the industrial Ethernet communication interface, and the virtual model layer updates the state of the virtual model within the virtual model layer based on the state data.

[0062] Understandably, the status data includes the operating status data of the detection equipment and the detection process data of the oil sample to be tested. After receiving the status data, the virtual model layer updates accordingly based on the changes in the status data. status and This allows for the real-time mirroring of the operating status of the detection equipment in the physical space and the detection status of the oil sample in the information space.

[0063] S26, the virtual model layer performs feature extraction processing on the state data, and transmits the result of the feature extraction processing to the physical entity layer via the industrial Ethernet communication interface. Based on the result of the feature extraction processing, the operating parameters of the detection operation are adjusted.

[0064] Understandably, the virtual model layer extracts implicit features from the state data through data analysis technology. These implicit features include the abnormal operation features of the detection equipment and the detection state features of the oil sample to be tested. The abnormal operation features of the detection equipment are used to identify the abnormal patterns of the detection equipment in the current operating state, while the detection state features of the oil sample to be tested are used to characterize the state of the oil sample to be tested in the current detection stage.

[0065] Furthermore, after the feature extraction results are transmitted back to the physical entity layer via the industrial Ethernet communication interface, the industrial control computer in the physical entity layer dynamically adjusts the operating parameters of the automated equipment during the oil sample testing process based on the abnormal operation characteristics and detection status characteristics, thereby realizing the reverse control of the physical space detection operation by the information space.

[0066] It is understandable that steps S25 and S26 together realize the bidirectional mapping described by formula (14). The state data of the physical space is transmitted to the information space in the forward direction through step S25. After the information space extracts features from the state data, the feature extraction results are transmitted back to the physical space through step S26. The industrial control computer dynamically adjusts the operating parameters of the detection equipment based on the abnormal operating features identified in the feature extraction results, so that the oil sample detection process can realize real-time optimization of the physical space detection operation based on the analysis results of the information space.

[0067] Reference Figure 5 In the inter-layer interaction relationship of the digital twin system architecture, the physical entity and the virtual model establish a bidirectional synchronous association through mapping iteration. The twin data layer outputs the processing results to the application service layer, physical entity and virtual model respectively through the data-driven mechanism. The application service layer relies on the fusion-derived data output by the twin data layer to drive the visualization presentation and optimization decision of the entire oil sample detection process. The actual operating status of the detection equipment in the physical space and the status of the corresponding twin model in the information space are synchronously displayed in the application service layer. The data-driven and mapping iteration mechanisms between each layer together constitute the engineering implementation basis of the bidirectional mapping mechanism described by formula (14).

[0068] S3 stores multi-source heterogeneous data throughout the entire oil sample testing process using a twin data layer, and preprocesses the multi-source heterogeneous data to obtain preprocessed data.

[0069] Understandably, the multi-source heterogeneous data in the entire oil sample testing process comes from three types of collection channels: The first type is oil sample testing data, which is obtained through data feedback from moisture analyzers, oil chromatographs, dielectric loss testers, and breakdown voltage testers. The raw data is normalized according to the standards and specifications of the testing field to form a complete digital profile of the state of the oil sample to be tested. The second type is the operating status data of the testing equipment, which is constructed by deploying temperature sensors and pressure sensors on the testing equipment at the physical layer to build a high-frequency sensing network and collect the operating status of the testing equipment in all aspects at a preset sampling frequency. The third type is robot operation and maintenance data, which is collected from the timed inspection reports, maintenance reports, and inspection records of the circulation robots and sorting robots in the physical layer to build a multi-source operation and maintenance knowledge base.

[0070] It should be noted that oil sample testing data, testing equipment operation status data, and robot operation and maintenance data have significant differences in data type, sampling frequency, and data format. Directly fusing these data will introduce outlier interference and data loss issues. Therefore, it is necessary to perform preprocessing operations from steps S31 to S34 sequentially on the multi-source heterogeneous data.

[0071] In this embodiment, the specific implementation of the preprocessing in step S3 includes steps S31 to S34, wherein: S31, outlier identification algorithm is used to identify outliers in multi-source heterogeneous data, and the identified outliers are removed from the multi-source heterogeneous data to obtain the first processed data.

[0072] Understandably, outlier identification algorithms are based on The criterion is implemented by calculating the mean of historical datasets of the same type of test data, based on oil sample test data and test equipment operating status data from multi-source heterogeneous data sources. with standard deviation , The statistical threshold is based on Centered on, plus or minus 3 times The confidence interval formed by the boundaries, i.e., the interval Data points whose values ​​fall outside the interval are identified as outliers and removed. The statistical threshold is set based on the assumption that, under a normal distribution, the data falls within... The probability within is Data points falling outside the interval are considered low-probability events; and The data is calculated based on historical datasets of the same type of detection data, with a sample size of no less than 30 sampling points.

[0073] For example, when processing dissolved gas content data output by an oil chromatograph, if the gas content value at a certain sampling time is different from... The deviation is greater than If a statistical threshold is set, the gas content value is identified as an outlier and removed from the multi-source heterogeneous data to obtain the first processed data after outlier removal.

[0074] S32, use density clustering algorithm to identify continuous abnormal segments in the time series data of the first processed data, remove the continuous abnormal segments from the first processed data, and obtain the second processed data.

[0075] It is understandable that density-based clustering algorithms are implemented based on density-based spatial clustering algorithms. The time-series data in the first batch of processed data includes temperature data collected by temperature sensors and pressure data collected by pressure sensors. The density-based clustering algorithm sets a neighborhood radius... The density threshold is determined by the minimum number of samples, MinPts. The density threshold is defined as follows: taking a data point in the time series data as the center, and... Within a neighborhood of a radius of MinPts, if the number of samples contained is greater than or equal to MinPts, the data point is identified as a core point and the area where it is located is identified as a high-density area; if the number of samples in the neighborhood is less than MinPts, the area where the data point is located is identified as a low-density area, and continuous data segments in the low-density area are continuous outlier segments. The setup method is as follows: Based on historical data collection datasets of the same type of time series data, calculate the distance from each data point in the dataset to its MinPts-1 nearest neighbor data point, and take the average of the MinPts-1 nearest neighbor distances of all data points as the mean. The initial value of MinPts is set according to the sampling dimension of the time series data. Set MinPts to , MinPts represents the number of sampling dimensions for time-series data. For one-dimensional time-series data in oil sample testing scenarios, the typical setting for MinPts is an integer between 4 and 6.

[0076] Furthermore, the density-based spatial clustering algorithm determines the boundaries of continuous outlier segments through density reachability, avoiding the misjudgment caused by the inability of a single statistical threshold method to identify the overall characteristics of continuous data segments. After removing the identified continuous outlier segments from the first processed data, the second processed data is obtained.

[0077] For example, when processing temperature data collected by a temperature sensor, if the local density of the temperature data is less than the density threshold within a certain time period, the continuous temperature data collected by the temperature sensor is identified as a continuous abnormal segment and removed to obtain the second processed data.

[0078] S33, the missing data caused by sensor offline in the second processing data is filled in using a time-weighted missing data filling algorithm to obtain preprocessed data.

[0079] It is understandable that the time-series weighted missing data imputation algorithm is based on the improved nearest neighbor algorithm. For missing data caused by sensor offline, it imputs data in the following ways depending on the data type.

[0080] Furthermore, for missing time series data, a preset time interval threshold is set. The setting method is as follows: based on the sensor's rated sampling period. ,Will Set as , This indicates the sensor's rated sampling period; the time interval between the sampling point and the missing point is less than [a certain value]. Historical data, assigned at intervals greater than or equal to The historical data is filled with weights to preserve the temporal continuity of the time series data.

[0081] Furthermore, for spatially missing data, a preset spatial distance threshold is set. The setting method is as follows: based on the nominal spacing between adjacent sensors in the sensor installation layout diagram. ,Will Set as , This indicates the nominal spacing between adjacent sensors in the sensor installation layout diagram; the spatial distance between sensors is less than... Sensor data is assigned to locations with a spatial distance greater than or equal to The sensor data is weighted to fill in missing values, thereby utilizing data from spatially adjacent sensors to assist in filling in missing values.

[0082] Furthermore, when a single parameter is missing, other parameters that are strongly correlated with the missing parameter are used for auxiliary prediction and imputation. Strongly correlated parameters are determined by calculating the Pearson correlation coefficient between the parameters, and the absolute value of the Pearson correlation coefficient is greater than a preset correlation threshold. The parameters are identified as strongly correlated; a preset correlation threshold is used. The setting method is as follows: Calculate the Pearson correlation coefficient between all parameter pairs based on the historical detection dataset, and take the upper quartile of the correlation coefficient distribution as... In this embodiment The typical value range is 0.7 to 0.9; the value of the missing parameter is estimated by the numerical mapping relationship between strongly correlated parameters, and the preprocessed data after filling the missing data is obtained.

[0083] S34, the twin data layer uses a spatiotemporal cube model to organize the time-series data collected at a preset sampling frequency in the preprocessed data in a multi-dimensional manner, and stores the multi-dimensional organized time-series data into the twin data layer through a three-dimensional index.

[0084] It is understood that the time-series data collected at the preset sampling frequency includes vibration data and temperature data of the detection device. The preset sampling frequency is 10Hz in this embodiment. The spatiotemporal cube model organizes the time-series data in multiple dimensions through three dimensions: time dimension, spatial dimension, and attribute dimension. The time dimension records the data collection time, the spatial dimension records the position coordinates of the data acquisition sensor, and the attribute dimension records the physical quantity type and value of the data.

[0085] Furthermore, the three-dimensional index uses time, space, and attribute dimensions as index axes to establish a three-dimensional retrieval structure for time-series data organized in multiple dimensions; the time slice data is a slice of time-series data collected within the current 10-second time window in the spatiotemporal cube model, and the latest time slice data is updated every 10 seconds to ensure that the storage structure of time-series data in the twin data layer matches the query requirements of the real-time monitoring dashboard.

[0086] Understandably, the spatiotemporal cube model transforms time-series data from a linear storage structure to a three-dimensional retrieval structure. In the real-time monitoring scenario of the entire oil sample testing process, the retrieval efficiency of the three-dimensional index is significantly better than that of the traditional linear storage structure, solving the retrieval efficiency problem of high-frequency sampled data in real-time query scenarios.

[0087] S4 uses the unique identifier code of the oil sample to be tested as an index to construct a data flow chain, and performs standardized transformation of the preprocessed data through a unified data model to obtain standardized data.

[0088] It is understood that the preprocessed data includes the preprocessed data output from steps S31 to S33, and the time-series data stored in the twin data layer via a three-dimensional index in step S34; To construct a data flow chain for the index, all test data, equipment operating status data, and robot maintenance data associated with each oil sample in the preprocessed data are... As the retrieval entry point, a data index structure centered on a single oil sample to be tested is formed. Through the data flow chain, all data nodes generated by the oil sample in the four detection stages of chromatographic detection, moisture detection, dielectric loss detection and breakdown voltage detection are connected in series, supporting the reverse tracing from the final detection result to the original collected data.

[0089] In this embodiment, the specific implementation of constructing the unified data model in step S4 includes steps S41 to S42, wherein: S41 uses a network ontology language to formally describe the entities and relationships between entities in the entire oil sample testing process, and maps the results of the formal description to the terminology system in the testing field standards and specifications through a terminology mapping table.

[0090] It is understandable that the Network Ontology Language is a formal language used to describe ontological knowledge structures. This language is used to formally describe entities throughout the entire oil sample testing process. These entities include the QR code and RFID tag information of the oil sample, the sample's attribute information, the model and operating status of the testing equipment, the sensor's installation location and calibration cycle information, robot maintenance fault and repair information provided by the multi-source maintenance knowledge base in step S3, oil sample testing indicators, testing methods, and operator information. Record the QR code and RFID tag information of the oil sample to be tested. Record the volumetric properties of the oil sample to be tested. and Record the model and operating status information of the testing equipment. Record the sensor's installation location and calibration cycle information. Record information about oil sample testing tasks.

[0091] Furthermore, the web ontology language formalizes entities such as the oil sample to be tested, testing equipment, sensors, and robots, as well as the detection, dependency, and operational relationships between entities, into a semantic network by defining entity classes, attributes, and relationships between entities; the description of detection relationships... and Description of the correlation and subordinate relationship between detection actions Belonging to Installation association, operation relationship description of operator and The execution relationships between detection tasks are established; a complete semantic description system for the entire oil sample detection process is formed in the information space.

[0092] Furthermore, the terminology mapping table records the mapping relationship between the terms in the unified data model and the terms in the testing field standards and specifications. The results of the formal description are mapped to the terminology system in the testing field standards and specifications through the terminology mapping table, realizing the semantic alignment between the terms in the unified data model and the terms in the testing field standards and specifications, so that data between different testing systems can interact under a unified semantic framework.

[0093] Understandably, the unified data model built using the web ontology language integrates the scattered entity information in the entire oil sample testing process into a unified semantic framework, and achieves cross-system semantic alignment through a terminology mapping table, eliminating semantic barriers caused by inconsistent terminology between multi-source heterogeneous data.

[0094] S42, the data flow chain records the entire chain of preprocessed data being standardized and transformed into standardized data in units of processing steps. Each processing step in the entire chain records the corresponding algorithm parameters.

[0095] Understandably, the data flow chain records the entire process of raw data from preprocessing and standardization to final standardized data through data association fields. The data association fields use unique identifiers to associate the input and output data of each processing step, forming a traceable data processing record.

[0096] Furthermore, each processing step in the data flow chain stores algorithm parameters through a processing step record field. These algorithm parameters include those used in step S31 for outlier identification. , and Statistical threshold boundary values, density clustering algorithm in step S32 and MinPts, in step S33 , and And the mapping rule parameters in the standardized transformation of the unified data model.

[0097] Furthermore, by combining data association fields with processing step record fields, the data flow chain supports tracing back from standardized data to each data node of preprocessed data. During tracing back, the input data of the previous processing step is located based on the data association fields, and the algorithm parameters used in the processing step are restored based on the processing step record fields, until the original multi-source heterogeneous data collected in step S3 is traced back.

[0098] Understandably, by recording the entire processing steps and algorithm parameters, the data transfer chain enables each data item in the standardized data to be traced back to the original collected data and intermediate processing, solving the problem in traditional detection systems where the intermediate processing process cannot be restored after the detection data has undergone multiple processing steps.

[0099] S5 performs data-level fusion processing on standardized data, then performs feature-level fusion processing on the results of the data-level fusion processing, and finally performs decision-level fusion processing on the results of the feature-level fusion processing to obtain the oil sample detection fusion results.

[0100] It is understandable that there is a sequential dependency between data-level fusion processing, feature-level fusion processing, and decision-level fusion processing. Data-level fusion processing takes standardized data as input and outputs the data-level fusion result as input to feature-level fusion processing. Feature-level fusion processing outputs feature vectors as input to decision-level fusion processing. Decision-level fusion processing outputs the oil sample detection fusion result.

[0101] In this embodiment, the specific implementation of the data-level fusion processing in step S5 is step S51, the specific implementation of the feature-level fusion processing is step S52, and the specific implementation of the decision-level fusion processing includes steps S53 to S54, wherein: S51 uses a time synchronization protocol to align similar detection data output by the detection equipment to a unified time reference, and performs weighted fusion on the aligned similar detection data based on a dynamic weight calculation model to obtain data-level fusion results.

[0102] It is understandable that there is a sampling time difference in the data acquisition of different testing devices. The time synchronization protocol is based on the precise time protocol. The precise time protocol maps the testing data of different testing devices to a unified time axis by adding a timestamp to the same type of testing data output by the testing devices, thereby eliminating the problem of data asynchrony caused by sampling time difference and aligning the same type of testing data to a unified time reference in the time dimension.

[0103] Furthermore, the dynamic weight calculation model assigns dynamic weights to similar test data output by each testing device based on the historical testing accuracy and current operating status of the testing equipment; historical testing accuracy is measured using the root mean square error. Quantification: (15) in, Indicates the first The root mean square error of the testing equipment; Indicates the use of calculation The total number of historical test samples, The value of is not less than 30; This represents the index of historical test samples, with a value ranging from 1 to... ; Indicates the first The testing equipment was at the first The detection values ​​output from the previous historical detection; Indicates the first The standard reference value for the previous historical test.

[0104] Furthermore, preset precision threshold Based on the permissible error limits set for the corresponding testing items in the standards and specifications of the testing field, Set as allowable error limit ;No. Dynamic weights of testing equipment The calculation formula is: (16) in, Indicates the first Dynamic weights of testing equipment; Indicates the first The root mean square error of the testing equipment; This indicates the total number of similar detection devices participating in data-level fusion processing; This represents the summation index of similar testing equipment, with a value range of 1 to... ; Indicates the first The root mean square error of the testing equipment; Indicates the first The operating status coefficient of the testing equipment, when the first When the detection equipment does not trigger the abnormal operation characteristics identified in step S26 When the runtime abnormality characteristic is triggered ; satisfy .

[0105] Understandable Less than and The testing equipment was allocated to more than Greater than or equal to or The dynamic weighting of the detection equipment, the dynamic weighting calculation model will align the similar detection data according to... Weighted fusion is performed to obtain data-level fusion results.

[0106] Furthermore, in the weighted fusion process, the following is adopted: Statistical thresholding and Grubbs' test perform dual filtering on aligned, similar test data; the Grubbs' test calculates the test statistic. Identify the most extreme outliers in a dataset: (17) in, This represents the Grubbs test statistic; This represents the candidate outlier value to be tested, taking the values ​​in the dataset that are equal to or less than the outlier value in the dataset. The data point with the largest deviation; This represents the mean of the aligned detection datasets of the same type. This represents the standard deviation of the aligned detection datasets of the same type.

[0107] when ≥ At that time, the judgment Extreme outliers were identified and removed. Indicates the significance level as The sample size is Critical value of Grubbs' test under the given conditions; This represents the significance level of the Grubbs test, which is set to 0.05 in this embodiment. Indicates the number of samples in the dataset to be tested; Grubbs' test and The statistical threshold dual filtering mechanism performs a second cleaning of abnormal data before weighted fusion, reducing random errors in the data-level fusion results.

[0108] S52, a dimensionality reduction algorithm is used to reduce the dimensionality of the features of the data-level fusion result to obtain dimensionality-reduced features. A time-series feature extraction model is used to extract dynamic features from the time-series detection data in the data-level fusion result to obtain time-series features. The dimensionality-reduced features and time-series features are concatenated to obtain a feature vector.

[0109] Understandably, dimensionality reduction algorithms are based on principal component analysis. The feature matrix representing the data-level fusion result. express The original feature dimension number; Principal component analysis first... The standardized feature matrix is ​​obtained by subtracting the mean of each feature dimension from the standard deviation of that dimension. ; express The covariance matrix is ​​obtained by calculating... ,extract The eigenvalues With the eigenvectors , Indicate the principal component index; select components with a cumulative variance contribution rate greater than or equal to... The former The principal components are used as the dimensionality reduction result. Indicates the number of principal components selected. The cumulative variance contribution rate is greater than or equal to Confirm; Will Project to front The dimensionality reduction features are obtained from the low-dimensional space composed of principal components. , This represents the dimensionality-reduced feature matrix output by principal component analysis. Indicates from the previous The projection matrix formed by the eigenvectors of the principal components.

[0110] Furthermore, the temporal feature extraction model is implemented based on a joint model of attention mechanism and long short-term memory network. The input of the joint model is the temporal detection data sequence in the data-level fusion result. Indicates the length of the time-series detection data sequence. The value is the product of the preset sampling frequency of 10Hz and the duration of the current detection phase. This represents the number of feature dimensions in the time-series detection data. The value of is consistent with the number of sensors involved in the temporal feature extraction in the physical entity data in step S13, and the input format is . The matrix; the Long Short-Term Memory network contains 2 hidden layers. This represents the number of dimensions of each hidden layer in a Long Short-Term Memory (LSTM) network, as shown in this embodiment. The value is set to 128; the Long Short-Term Memory (LSTM) network uses three gating mechanisms—forget gate, input gate, and output gate—to perform sequence modeling of time-series detection data, capturing long-term dependencies within the data. This represents the sequence of hidden states output by the Long Short-Term Memory (LSTM) network. , Indicates the first The hidden state vector at time t. Represents the time index; attention mechanism in Based on this, the attention weights are obtained through the following calculations: (18) (19) (20) in, Indicates the first Attention energy level at any given moment; The learnable weight matrix representing the attention mechanism; The learnable bias representing the attention mechanism; Indicates the first Normalized attention weights at each moment, satisfying ; This represents the time index in the attention weight normalization summation, with a value ranging from 1 to... ; Indicates the first Attention energy level at any given moment; This represents the temporal features of the output after attention weighting, which is... according to The weighted summation result.

[0111] Furthermore, the training dataset for the joint model originates from the historical detection data stored in the twin data layer in step S13, and the training sample format is as follows: The training labels are the category annotations of the oil sample detection status within the corresponding time series data segment; the training objective is to minimize the cross-entropy loss function. : (twenty one) in, This represents the value of the cross-entropy loss function; Indicates the total number of training samples; This represents the training sample index, with a value ranging from 1 to... ; This indicates the total number of categories representing the oil sample's testing status; This represents the detection status category index, with a value range of 1 to... ; Indicates the first The training sample corresponds to the first... The true label of the class detection state, with a value of 0 or 1; The joint model represents the first... The output of the training sample of the th time The predicted probability of the class detection state; training uses the Adam optimizer, with the initial learning rate set to [value missing]. The number of training rounds should be no less than 100, in order to validate the data on the validation set. It will no longer be downgraded to the early shutdown standard; and During training, the parameters of the Long Short-Term Memory network are updated synchronously using the backpropagation algorithm.

[0112] Furthermore, dimensionality reduction features With time series characteristics The concatenation operations are used to create a feature vector. , This represents the concatenated eigenvectors. Reflecting the main characteristics of data-level fusion results in static dimensions, It reflects the dynamic changes of data-level fusion results over time. It includes both static and dynamic features.

[0113] S53, the confidence level of the detection data of the detection instruments in the detection equipment in the feature vector is determined based on the probability allocation model. The confidence level is determined based on the direct measurement results of the detection instruments, the mean of repeated detections, and the comparison results across detection devices.

[0114] It is understandable that the probability allocation model is based on the basic probability allocation theory, and the probability allocation function is used in step S53. Quantifying the credibility of the test evidence: (twenty two) in, The probability assignment value for the empty set is set to 0. A recognition framework representing all possible categories of the detection status of the oil sample to be tested; express any subset of; Indicates the detection evidence pair subset The probability allocation value reflects whether the detection evidence supports the detection result belonging to The reliability of the data is determined based on the following sources of trust for the test data output by the moisture analyzer, oil chromatograph, dielectric loss tester, and breakdown voltage tester. .

[0115] Furthermore, the first source of trust is the direct measurement results from the testing instrument, with a preset deviation threshold. Measurement uncertainty based on the corresponding testing items in the standards and specifications of the testing field Setting, will Set as , This indicates the measurement uncertainty of the corresponding test item; This represents the direct measurement value output by the testing instrument at the current testing stage. This represents the historical baseline value for the corresponding testing item. This represents the absolute value of the deviation between the directly measured value and the historical benchmark value. ;when Less than At that time, the direct measurement results Greater than Greater than or equal to time .

[0116] Furthermore, the second source of trust is the mean of repeated tests, with a preset stability threshold. Based on the repeatability standard deviation limit set for the corresponding test items in the testing field standards and specifications, the following will be implemented. Set at 80% of the repeatability standard deviation limit; In this embodiment, the number of repeated detections is indicated. The value is 3; The standard deviation of repeated test results; Less than When, the mean of repeated tests Greater than Greater than or equal to time .

[0117] Furthermore, the third source of trust is the comparison results across testing devices, with a preset consistency threshold. Based on the maximum permissible deviation settings of two similar testing devices on standard samples, Set to 70% of the maximum permissible deviation. This represents the difference between the test results of different testing instruments for the same oil sample. Less than At the same time, cross-device comparison results Greater than Greater than or equal to time .

[0118] S54, based on the evidence combination rule, fuses the trust levels of the testing instruments in the testing equipment to obtain the oil sample testing fusion result.

[0119] Understandably, the evidence combination rule is based on the Dempster combination rule, which assigns a probability function to any two detection instruments in step S53. and The formula for calculating Dempster's combination rule is: (twenty three) (twenty four) in, express and Subsets after Dempster combination The overall probability allocation value; and Let represent the probability assignment functions of the two detection instruments to be fused; express The probability distribution corresponding to A subset of; express The probability distribution corresponding to A subset of; Indicates all Assigning subsets and Assigning subsets Intersection equals subset pairs Perform summation; The conflict coefficient represents the coefficient of friction, reflecting the overall situation. and The degree of conflict between them The range of values ​​is ; This represents the normalization coefficient, used to eliminate the impact of conflicts on the fusion results.

[0120] The moisture analyzer, oil chromatograph, dielectric loss tester, and breakdown voltage tester were used. The two components are fused one by one according to formula (23). Each fusion will combine two components. merged into a comprehensive Trust baseline The setting method is: take The lower limit of the comprehensive probability allocation value corresponding to the normal detection state is not less than 0.7. , This represents the baseline value of trust level; after multiple rounds of fusion, the category with the highest comprehensive probability allocation value is output as the fusion result of oil sample detection.

[0121] It is understandable that steps S53 and S54, through the complete calculation link of the basic probability allocation and Dempster combination rule constituted by formulas (22) to (24), target the moisture analyzer, oil chromatograph, dielectric loss tester, and breakdown voltage tester in three dimensions: direct measurement, repeated testing, and cross-device comparison. The differences are quantified, and the evidence from multiple instruments is fused into a unified oil sample detection result through the Dempster combination rule. This solves the problem that the output results of multiple heterogeneous detection instruments in traditional detection schemes lack reliable quantification and cannot be effectively fused.

[0122] S6 enables the application service layer to visualize and monitor the entire oil sample testing process based on the fusion results of oil sample testing, and to perform traceability management of the entire oil sample testing process based on the data flow chain.

[0123] Understandably, the application service layer builds a laboratory simulation visualization platform by integrating hardware-in-the-loop simulation technology. Hardware-in-the-loop simulation technology, through... and It directly interfaces with the controller input / output interface of the actual laboratory testing system, enabling the application service layer to simultaneously obtain the twin model state of the information space and the actual testing state of the physical space, and display both synchronously in the visualization platform.

[0124] In this embodiment, the specific implementation of the visual monitoring in step S6 includes steps S61 to S63, the specific implementation of traceability management is step S64, and the specific implementation of the automated oil sample testing process is step S65, wherein: S61, the application service layer performs visual monitoring of the oil sample transfer process, based on... The real-time status of the oil sample is displayed on the visualization platform, showing the current location and testing progress of the oil sample in twelve stages of the process: oil sample warehousing, oil sample transportation, oil sample scanning, oil sample handling, oil sample extraction, oil sample delivery, pressure resistance analysis, dielectric loss analysis, chromatographic analysis, micro-water analysis, oil sample retrieval, and oil sample release.

[0125] Understandably, the visualization platform uses a 3D visualization model to present the real-time flow of the oil sample throughout the entire testing process. The sample management list is the interface within the visualization platform used to display the testing progress of the oil sample. The sample management list records... Corresponding sample number, The voltage level attribute of the transformer in the data The current detection phase recorded in the middle, and The system tracks the progress of the completed testing tasks, and operators can monitor the current status of each oil sample in real time through the sample management list.

[0126] Reference Figure 6 The entire oil sample testing process involves twelve stages: oil sample warehousing, oil sample transportation, oil sample scanning, oil sample handling, oil sample extraction, oil sample delivery, oil sample pressure resistance analysis, oil sample dielectric loss analysis, oil sample chromatographic analysis, oil sample micro-moisture analysis, oil sample retrieval, and oil sample release. During each stage, a radio frequency identification (RFID) reader reads the sample data in real time. Update the current status of the oil sample to be tested to [current status]. And synchronized to and synchronized to and synchronized to Visualization platform basis, visualization platform basis, visualization platform basis The real-time status of the oil sample is displayed in the 3D visualization model, showing the current position and testing progress of the oil sample at each stage, realizing visualized monitoring of the entire oil sample testing process.

[0127] S62, the application service layer performs visual monitoring of the operating status of the testing equipment, based on the oil sample testing fusion results from step S5 and... The system displays the real-time operating status and test task completion of the moisture analyzer, oil chromatograph, dielectric loss tester, and breakdown voltage tester on a visualization platform.

[0128] Understandably, the visualization platform statistically displays the number of tests and pass rates of moisture analyzers, oil chromatographs, dielectric loss testers, and breakdown voltage testers, showing the pass rates for five types of tests: chromatographic detection, trace moisture detection, breakdown voltage detection, dielectric loss detection, and gas content detection. Preset thresholds are set by operators based on the lower limit of the pass rate for the corresponding test items in the testing field's standards and specifications. When the pass rate is lower than the preset threshold, the application service layer triggers a fault warning and outputs optimization decision instructions. The visualization platform displays the laboratory temperature, humidity, and cleanliness in real time. The preset temperature range is set according to the environmental temperature requirements for oil sample testing in the testing field's standards and specifications, while the preset humidity and cleanliness ranges are set according to the laboratory cleanliness level standards. When environmental parameters exceed the upper limit of the preset range or fall below the lower limit, the application service layer triggers an environmental anomaly warning.

[0129] S63, the application service layer performs fault warning and optimization decision-making for the entire oil sample testing process based on the oil sample detection fusion results in step S5.

[0130] It is understandable that the fault warning is based on the abnormal operating characteristics of the detection equipment identified by the feature extraction process in step S26. When the abnormal operating characteristics exceed the normal operating range, the application service layer marks the abnormality in the form of a warning label in the visualization platform. The number and exception type.

[0131] Furthermore, a preset deviation range threshold is defined. The settings are based on the normal fluctuation range of the weighted average of data-level fusion results in historical testing data. Indicates the preset deviation range threshold; This represents the historical baseline value of the data-level fusion results. Based on the steady-state value of the weighted average of data-level fusion results within historical testing cycles; when the comprehensive probability allocation value of the normal testing state in the oil sample testing fusion results is less than... or the weighted average of the data-level fusion results and The absolute value of the difference is greater than At that time, the application service layer according to The decision-making instructions of the behavior model output optimization suggestions, and the optimization parameters are transmitted to the industrial control computer in the physical entity layer through the reverse control mechanism in step S26 to dynamically adjust the operating parameters of the detection equipment.

[0132] S64, Traceability management of the entire oil sample testing process is implemented based on the data flow chain. It can be understood that traceability management is achieved through the data flow chain constructed in step S42, with the data flow chain... As the retrieval entry point, it records all data nodes generated by the oil sample during the four detection stages of chromatographic detection, moisture detection, dielectric loss detection, and breakdown voltage detection, as well as the algorithm parameters saved through the processing step record field for each processing step; Using the data association field in the data flow chain as an index, the input and output data of each processing step are located step by step. The algorithm parameters used in each processing step are restored by the processing step record field. The oil sample detection fusion result in step S5 is traced back to the original multi-source heterogeneous data collected in step S3, realizing the traceability management of the entire detection data chain and meeting the requirements of GB / T27025 for the traceability of detection data.

[0133] Understandably, the data flow chain saves the full-link algorithm parameters through the processing step record field, enabling the complete reproduction of intermediate processing processes that cannot be restored after multiple steps in traditional detection systems, thus solving the problem of untraceable detection data in traditional detection processes.

[0134] S65, the application service layer uses the laboratory information management system to manage the entire process of automated oil sample testing.

[0135] Understandably, the laboratory information management system is a system designed to achieve end-to-end information-based control from sample registration to report generation. During the sample registration stage, the laboratory information management system will... , and Enter the data into the laboratory information management system to receive and update the test data output by the testing instruments in real time at each testing stage. During the report generation phase The test results are summarized to generate a test report. The test data output by the laboratory information management system is used as one of the sources of the first type of oil sample test data in step S3 and is included in the scope of multi-source heterogeneous data collection.

[0136] Furthermore, after receiving the oil sample, laboratory testing personnel complete identity verification at the sample registration workstation and scan the barcode on the oil sample. The oil samples to be tested were placed one by one into the turnover box, and the turnover box was placed in the designated storage location. After the oil samples were put into storage, the sample cabinet door was closed, and the testing process was started at the registration workstation.

[0137] Furthermore, after the oil sample to be tested is stored and the testing task is initiated, the transfer robot delivers the designated oil sample to the automated sorting system. The automated sorting system then sorts the sample according to the information in the laboratory information management system. The recorded testing task information is used to identify each oil sample to be tested. After identification, the oil sample is transferred to the sorting and buffer station for temporary storage, and then sent to each testing system by the automatic transfer system for testing.

[0138] Furthermore, for Includes moisture detection task The robot carries the syringe to the water pipetting module, picks up the injection needle, and extracts the oil sample. The first extracted oil sample is sent to the waste liquid recycling area for recycling. Repeated extractions are performed to ensure sample purity. The injection needle containing the sample is then transferred to the water analyzer for sample loading, injection, and automatic detection. After detection, the water analyzer automatically reads the analytical data and uploads it to the laboratory information management system. The test results are updated to […]. .

[0139] Furthermore, for Includes chromatographic detection tasks The oil sample to be tested is loaded onto the chromatograph. After the sample syringe reaches the loading unit, the syringe cap is removed and the volume is adjusted. After volume adjustment, the syringe cap is replaced and the sample is transferred to the loading unit. The industrial control computer opens the top cover of the chromatograph and automatically loads the sample for detection. After the chromatographic detection is completed, the sample is automatically discharged and transferred to the outgoing buffer. The detection results are updated to... .

[0140] Furthermore, for Includes dielectric loss detection tasks The automatic transfer system places the wide-mouth bottle to be tested into a temporary storage position, and the dielectric loss detection system then... The preceding task information selects the fixture and picks up the oil sample to be tested into the sample pretreatment unit. The sample pretreatment unit automatically mixes the sample. The pipette transfers the oil sample to the instrument's sample cell. The robot rinses the sample cell twice. After rinsing, an appropriate amount of oil sample is taken again and placed into the sample cell. The dielectric loss detection system performs automatic detection twice, and the detection results are updated. After the inspection is completed, the wide-mouth bottle is capped and transferred to the outbound buffer position.

[0141] Furthermore, for Includes breakdown voltage detection task The automatic transfer system places the wide-mouth bottle to be tested into a temporary storage position, and the breakdown voltage detection system... The preceding task information selects the fixture and picks up the oil sample to be tested into the sample pretreatment unit. The sample pretreatment unit automatically mixes the sample. The pipette transfers the oil sample to the instrument's sample cell. The robot rinses the sample cell twice. After rinsing, an appropriate amount of oil sample is taken into the sample cell again. The breakdown voltage detection system automatically detects and records 6 detection data points. The detection results are updated to... After the inspection is completed, the wide-mouth bottle is capped and transferred to the outbound buffer position.

[0142] Furthermore, to complete After all testing tasks are completed, the oil sample container to be tested is automatically transferred by the circulation system. They are transferred to the corresponding automated outbound storage containers for temporary storage. The system records the outbound status and awaits manual pickup, thus automating the entire oil sample testing process.

[0143] Example 2 is an embodiment of the present invention, which provides an oil sample detection system based on digital twins, comprising: Physical entity layer, virtual model layer, twin data layer, and application service layer; The physical entity layer includes the detection equipment and the oil sample to be tested carrying a unique identification code. The detection equipment completes four-dimensional modeling according to geometric dimension, physical attribute dimension, rule dimension, and behavior dimension. The oil sample to be tested completes task state modeling based on the detection task set and detection process state. A two-way real-time mapping relationship is established between the physical space of the physical entity layer and the information space of the virtual model layer.

[0144] Understandably, the testing equipment includes a moisture analyzer, oil chromatograph, dielectric loss tester, breakdown voltage tester, transfer robot, sorting robot, industrial control computer, sensors, data acquisition card, and RFID reader / writer. The RFID electronic tag on the oil sample container uniquely identifies the sample number and batch number of the oil sample through electronic product code. The bidirectional real-time mapping association is established through the industrial Ethernet communication interface. The state data of the physical entities in the physical entity layer is transmitted forward to the virtual model layer through the industrial Ethernet communication interface. After the virtual model layer performs feature extraction processing on the state data, it transmits the results backward to the physical entity layer through the industrial Ethernet communication interface. The industrial control computer dynamically adjusts the operating parameters of the testing equipment based on the feature extraction results.

[0145] The virtual model layer is a mirror image of the physical entity layer in the information space, including the geometric model, physical model, rule model and behavior model corresponding to the detection equipment, as well as the virtual model corresponding to the oil sample to be tested.

[0146] Understandably, the virtual model of the oil sample to be tested uses the electronic product code as an anchor point to record the source information, test task set, current test stage and historical test data of the oil sample to be tested. The radio frequency identification reader reads the electronic product code in real time and synchronizes the state changes of the oil sample to be tested in the stages of chromatographic detection, moisture detection, dielectric loss detection and breakdown voltage detection to the virtual model of the oil sample to be tested.

[0147] The twin data layer is used to store multi-source heterogeneous data in the entire oil sample testing process, preprocess the multi-source heterogeneous data to obtain preprocessed data, construct a data flow chain with the unique identifier code of the oil sample to be tested as an index, standardize the preprocessed data through a unified data model to obtain standardized data, perform data-level fusion processing on the standardized data, perform feature-level fusion processing on the results of data-level fusion processing, and perform decision-level fusion processing on the results of feature-level fusion processing to obtain the oil sample testing fusion result.

[0148] Understandably, preprocessing includes outlier identification, density clustering anomaly identification, and time-weighted missing data imputation. The unified data model is built on ontology, using a network ontology language to formally describe entities and their relationships throughout the oil sample testing process. A terminology mapping table maps terms from the unified data model to the terminology system in testing standards. The data flow chain associates input and output data for each processing step through data association fields, and saves algorithm parameters for each processing step through processing step record fields. Time-series data collected at a preset sampling frequency is organized in multiple dimensions using a spatiotemporal cube model and stored and retrieved using a three-dimensional index. Data-level fusion processing includes a time synchronization module and a dynamic weight calculation module; feature-level fusion processing includes a dimensionality reduction module and a time-series feature extraction module; and decision-level fusion processing includes a probability allocation module and an evidence combination module.

[0149] The application service layer is used to visualize and monitor the entire oil sample testing process based on the fusion results of oil sample testing, and to manage the traceability of the entire oil sample testing process based on the data flow chain.

[0150] Understandably, the application service layer integrates product lifecycle management systems, enterprise resource planning systems, product data management systems, and manufacturing execution systems, and integrates hardware-in-the-loop simulation technology to build a laboratory simulation visualization platform. Hardware-in-the-loop simulation technology directly interfaces the virtual model in the virtual model layer with the actual laboratory testing system. Visual monitoring includes an oil sample flow monitoring module, a testing equipment status monitoring module, and a fault early warning and optimization decision-making module. Traceability management includes a data retrieval module and a traceability restoration module. The data retrieval module uses the electronic product code of the oil sample to be tested as an index to locate the data of each processing step through the data flow chain. The traceability restoration module restores the algorithm parameters of each processing step through the processing step record fields, and traces back from the oil sample test fusion results to the original multi-source heterogeneous data.

[0151] This embodiment also provides an electronic device applicable to a digital twin-based oil sample detection method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the digital twin-based oil sample detection method proposed in the above embodiment.

[0152] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements an oil sample detection method based on digital twins as proposed in the above embodiments.

[0153] The storage medium proposed in this embodiment belongs to the same inventive concept as the oil sample detection method based on digital twin proposed in the above embodiment. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0154] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0155] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A digital-twin-based oil sample detection method, characterized in that, include: Construct a digital twin system architecture consisting of a physical entity layer, a virtual model layer, a twin data layer, and an application service layer; The detection equipment in the physical entity layer is modeled in four dimensions: geometric dimension, physical attribute dimension, rule dimension, and behavior dimension. The oil sample to be tested carrying a unique identifier code in the physical entity layer is modeled in task status according to the detection task set and detection process status. A bidirectional real-time mapping association is established between the physical space of the physical entity layer and the information space of the virtual model layer. The twin data layer stores multi-source heterogeneous data throughout the entire oil sample testing process, and preprocesses the multi-source heterogeneous data to obtain preprocessed data. A data flow chain is constructed using the unique identifier code of the oil sample to be tested as an index, and the preprocessed data is standardized and transformed through a unified data model to obtain standardized data. The standardized data is subjected to data-level fusion processing, the result of the data-level fusion processing is subjected to feature-level fusion processing, and the result of the feature-level fusion processing is subjected to decision-level fusion processing to obtain the oil sample detection fusion result. The application service layer performs visual monitoring of the entire oil sample testing process based on the oil sample testing fusion results, and performs traceability management of the entire oil sample testing process based on the data flow chain.

2. The oil sample detection method based on digital twinning of claim 1, wherein, The four-dimensional modeling includes: modeling the physical properties of the detection device, and performing structural analysis, acoustic field analysis, electromagnetic field analysis, and multi-physics coupling analysis based on the finite element model of the detection device; Behavioral modeling is performed on the detection equipment, based on the task execution status, external environment, sudden disturbances and internal operating mechanism of the detection equipment in the stages of chromatographic detection, moisture detection, dielectric loss detection and breakdown voltage detection. The task state modeling includes: associating the unique identifier code carried by the radio frequency identification electronic tag with the detection task set of the oil sample to be tested; constructing a virtual model of the oil sample to be tested in the virtual model layer; and synchronizing the state changes of the oil sample to be tested in the stages of chromatographic detection, moisture detection, dielectric loss detection, and breakdown voltage detection to the virtual model of the oil sample to be tested.

3. The oil sample detection method based on digital twinning of claim 2, wherein, The bidirectional real-time mapping association includes: establishing a communication connection between the physical space and the information space through an industrial Ethernet communication interface; The status data of the physical entities in the physical entity layer is transmitted to the virtual model layer in real time via the industrial Ethernet communication interface. The virtual model layer updates the status of the virtual model within the virtual model layer based on the status data. The virtual model layer performs feature extraction processing on the state data, and transmits the result of the feature extraction processing to the physical entity layer via the industrial Ethernet communication interface. Based on the result of the feature extraction processing, the operating parameters of the detection operation are adjusted.

4. The oil sample detection method based on digital twinning of claim 3, wherein, The preprocessing includes: using an outlier identification algorithm to identify outliers in the multi-source heterogeneous data, removing the identified outliers from the multi-source heterogeneous data, and obtaining the first processed data; Density clustering algorithm is used to identify continuous abnormal segments in the time series data in the first processed data, and the continuous abnormal segments are removed from the first processed data to obtain the second processed data; The missing data caused by sensor offline in the second processed data is filled in using a time-weighted missing data filling algorithm to obtain the preprocessed data; The twin data layer organizes the time-series data collected at a preset sampling frequency in the preprocessed data in a multi-dimensional manner using a spatiotemporal cube model, and stores the multi-dimensionally organized time-series data into the twin data layer through a three-dimensional index.

5. The oil sample detection method based on digital twinning of claim 4, wherein, The unified data model includes: using a network ontology language to formally describe the entities and relationships between entities in the entire oil sample testing process, and mapping the results of the formal description to the terminology system in the testing field standards and specifications through a terminology mapping table; The data flow chain records the entire process of the preprocessed data being standardized and transformed into standardized data, with each processing step as a unit. The corresponding algorithm parameters are recorded for each processing step in the entire chain.

6. The oil sample detection method based on digital twinning of claim 5, wherein, The data-level fusion processing includes: aligning similar detection data output by the detection device to a unified time reference using a time synchronization protocol, and performing weighted fusion on the aligned similar detection data based on a dynamic weight calculation model to obtain a data-level fusion result; The feature-level fusion processing includes: using a dimensionality reduction algorithm to reduce the dimensionality of the features in the data-level fusion result to obtain dimensionality-reduced features; using a temporal feature extraction model to extract dynamic features from the temporal detection data in the data-level fusion result to obtain temporal features; and concatenating the dimensionality-reduced features with the temporal features to obtain a feature vector.

7. The oil sample detection method based on digital twinning according to claim 6, characterized in that, The decision-level fusion processing includes: determining the trust level of the detection data of the detection instruments in the detection device in the feature vector based on a probability allocation model, wherein the trust level is determined based on the direct measurement results of the detection instruments, the mean of repeated detections, and the comparison results across the detection devices; The trust levels of the detection instruments in the detection equipment are fused based on the evidence combination rules to obtain the fused detection result of the oil sample.

8. A digital twin-based oil sample detection system, employing the digital twin-based oil sample detection method as described in any one of claims 1 to 7, characterized in that, include: Physical entity layer, virtual model layer, twin data layer, and application service layer; The physical entity layer includes detection equipment and oil samples to be tested carrying unique identification codes. The detection equipment completes four-dimensional modeling according to geometric dimension, physical attribute dimension, rule dimension, and behavior dimension. The oil sample to be tested completes task state modeling according to the detection task set and detection process state. A bidirectional real-time mapping relationship is established between the physical space of the physical entity layer and the information space of the virtual model layer. The twin data layer is used to store multi-source heterogeneous data in the entire oil sample testing process, preprocess the multi-source heterogeneous data to obtain preprocessed data, construct a data flow chain with the unique identifier code of the oil sample to be tested as an index, standardize the preprocessed data through a unified data model to obtain standardized data, perform data-level fusion processing on the standardized data, perform feature-level fusion processing on the result of data-level fusion processing, and perform decision-level fusion processing on the result of feature-level fusion processing to obtain the oil sample testing fusion result. The application service layer is used to perform visual monitoring of the entire oil sample testing process based on the oil sample testing fusion results, and to perform traceability management of the entire oil sample testing process based on the data flow chain.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the oil sample detection method based on digital twins as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the oil sample detection method based on digital twins as described in any one of claims 1 to 7.

Images