A Distributed Graph Database Association Analysis Method and System Based on TinkerPop API

Abstract

The application relates to the technical field of graph database analysis, and provides a distributed graph database correlation analysis method and system based on a TinkerPop API, which solves the problems of low efficiency and low accuracy of wide-area infrastructure abnormal event positioning and causal inference. The method comprises the following steps: collecting sound wave sensing signals of a monitoring area and SCADA time series data of key equipment; processing the sound wave signals to obtain strain data, and determining the space-time coordinates of multiple sound source events through sound field reconstruction; generating simulated sound prints based on sound wave segments, forming a structured event stream after comparison and combining with the space-time coordinates; extracting multi-scale feature information of the SCADA data, and generating a causal relationship set after fusing the event stream and the feature information; and constructing a causal correlation network in a distributed graph database by taking the causal relationship as an edge and taking geographical space entities and other elements as vertices. The application improves the efficiency and accuracy of wide-area infrastructure abnormal event positioning and causal inference.

Description

Technical Field

[0001] This application relates to the field of graph database analysis technology, and in particular to a distributed graph database association analysis method and system based on the TinkerPop API. Background Technology

[0002] Distributed graph database association analysis methods have demonstrated significant value in the operation of large-scale enterprises, especially those with extensive infrastructure and complex industrial equipment systems. This method can be used to integrate multi-source data such as Supervisory Control and Data Acquisition (SCADA) systems and IoT sensors to build an association network between equipment status and abnormal events.

[0003] In existing technologies, distributed graph database association analysis methods based on the Graph Computing Framework Application Programming Interface (TinkerPop API) have been applied. For example, by connecting device monitoring data in a time-series database with event records in a business system, the data is first extracted and mapped to vertices and edges in a graph. Then, the Gremlin query language is used to perform relation traversal and pattern discovery to achieve asset topology management or rule-based event association. In business scenarios with well-structured data and clear association rules, this type of method can effectively organize and manage predefined entity relationships.

[0004] However, existing technologies face challenges in adapting to complex scenarios such as wide-area continuous spatial physical monitoring, including the location of acoustic and vibration anomalies in infrastructure such as railway lines and integrated utility tunnels. These scenarios involve the deep fusion of unstructured physical sensor signals and structured equipment SCADA time-series data. The signals themselves are characterized by high dimensionality, continuity, and strong noise. Their relationship with equipment status is not a simple, static correlation that can be described by preset rules, but rather implies dynamic and probabilistic spatiotemporal causal relationships. Therefore, existing technologies suffer from insufficient deep fusion of physical sensor signals and equipment operating status data, making it difficult to effectively uncover dynamic spatiotemporal causal relationships. Summary of the Invention

[0005] This application provides a distributed graph database association analysis method and system based on the TinkerPop API to solve the problems of poor efficiency and low accuracy in the localization of abnormal events and causal inference of wide-area infrastructure in the prior art.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a distributed graph database association analysis method based on the TinkerPop API, comprising:

[0007] Collect acoustic sensor signals and SCADA timing data of key equipment within the monitoring area;

[0008] The acoustic wave sensing signal is processed to obtain strain data, and the strain data is reconstructed into a sound field to obtain multiple sound source events, and the spatiotemporal coordinates of each sound source event are determined.

[0009] The sound wave segment corresponding to each sound source event is input into the Generative Adversarial Network (GAN), which generates a simulated voiceprint. The simulated voiceprint is compared with the sound wave segment, and the structured event stream is obtained by combining the spatiotemporal coordinates.

[0010] Multi-scale feature extraction is performed on the SCADA time series data to obtain feature information, which is used to reflect the operating status of key equipment;

[0011] The structured event stream is aligned and fused with the feature information to obtain fused data. A causal discovery algorithm is then used to analyze the lead-lag relationship and conditional independence test among multiple fused records in the fused data to generate a causal relationship set.

[0012] By calling the operations provided by the TinkerPop API, using the causal relationships in the set of causal relationships as edges and geospatial entities, sensing units, sound source events, and device sources as vertices, a causal association network is constructed in a distributed graph database. The causal association network is used to associate physical disturbances with the operating status of the key equipment.

[0013] Optionally, the step of inputting the sound wave segment corresponding to each sound source event into a generative adversarial network (GAN), generating a simulated voiceprint through the GAN, comparing the simulated voiceprint with the sound wave segment, and combining it with the spatiotemporal coordinates to obtain a structured event stream includes:

[0014] Based on the occurrence time of the sound source event, the sound wave segments within the corresponding time period are extracted from the data matrix;

[0015] The sound wave segment is input into the Generative Adversarial Network (GAN), and the generator of the GAN generates a simulated voiceprint based on the characteristics of the sound wave segment.

[0016] The discriminator of the adversarial generative network compares the differences in frequency domain features and time domain features between the sound wave fragment and the simulated voiceprint to determine the probability that the sound wave fragment belongs to different event categories.

[0017] Based on the probability, an event type label is assigned to the sound wave segment using a softmax classifier;

[0018] The event type labels are associated with the spatiotemporal coordinates to generate structured event records, and all structured event records are combined into a structured event stream in chronological order.

[0019] Optionally, the generator of the adversarial generative network generates simulated voiceprints based on the features of the sound wave segment, including:

[0020] The sound wave fragment is received through the input layer of the generator;

[0021] The sound wave segment is encoded by multiple cascaded neural network layers of the generator to extract the feature vector of the sound wave segment in the frequency domain;

[0022] The feature vector is input into the decoding network of the generator, where the feature vector is weighted and combined with a preset voiceprint feature template.

[0023] The weighted combination result is upsampled through multiple deconvolution layers of the decoding network to generate a simulated voiceprint.

[0024] Optionally, the step of aligning and fusing the structured event stream with the feature information to obtain fused data, and using a causal discovery algorithm to analyze the lead-lag relationship and conditional independence test among multiple fused records in the fused data to generate a causal relationship set, includes:

[0025] Based on the time recorded in the structured event stream, the feature corresponding to the time is found from the feature information;

[0026] The features are merged with the structured event records of the corresponding time in the structured event stream to form a fused record, and multiple fused records are arranged according to time to obtain fused data;

[0027] The fused data was analyzed using a causal discovery algorithm to determine the lead-lag relationships between different variables.

[0028] Based on the aforementioned lead-lag relationship, and combined with the conditional independence test, determine the causal relationship between different variables;

[0029] Calculate the confidence score of the causal relationship, and combine the causal relationship and the confidence score to obtain a set of causal relationships.

[0030] Optionally, the step of reconstructing the sound field from the strain data to obtain multiple sound source events, and determining the spatiotemporal coordinates for each sound source event, includes:

[0031] The strain data are arranged along the spatial and temporal dimensions to form a data matrix;

[0032] Spatial spectrum analysis is performed on the data matrix to calculate the direction of arrival parameters of signals from different spatial azimuth and elevation angles;

[0033] Based on the arrival direction parameters, a beamforming algorithm is used to separate the vibration signal components with concentrated energy in time and space from the data matrix, and each vibration signal component is defined as a sound source event.

[0034] Based on the arrival direction parameter, calculate the distance of each sound source event relative to each sensing unit in the sensing array;

[0035] Based on the distance and the time information corresponding to the strain data in the data matrix, the location coordinates and occurrence time of each sound source event are determined, and the location coordinates and occurrence time constitute spatiotemporal coordinates.

[0036] Secondly, this application provides a distributed graph database association analysis system based on the TinkerPop API, comprising:

[0037] The acquisition module is used to acquire acoustic wave sensor signals and SCADA time-series data of key equipment within the monitoring area.

[0038] The processing module is used to process the acoustic wave sensing signal to obtain strain data, and to reconstruct the sound field of the strain data to obtain multiple sound source events, and to determine the spatiotemporal coordinates of each sound source event.

[0039] The input module is used to input the sound wave segment corresponding to each sound source event into the adversarial generative network, generate a simulated voiceprint through the adversarial generative network, compare the simulated voiceprint with the sound wave segment, and obtain a structured event stream by combining the spatiotemporal coordinates.

[0040] The extraction module is used to perform multi-scale feature extraction on the SCADA time series data to obtain feature information, which is used to reflect the operating status of key equipment;

[0041] The fusion module is used to align and fuse the structured event stream with the feature information to obtain fused data, and to use a causal discovery algorithm to analyze the lead-lag relationship and conditional independence test among multiple fused records in the fused data to generate a causal relationship set.

[0042] The construction module is used to construct a causal relationship network in a distributed graph database by calling the operations provided by the TinkerPop API, using causal relationships in the causal relationship set as edges and geospatial entities, sensing units, sound source events and device sources as vertices. The causal relationship network is used to associate physical disturbances with the operating status of the key equipment.

[0043] Thirdly, this application provides an electronic device, comprising:

[0044] Memory, used to store computer programs;

[0045] A processor, configured to execute the computer program to implement the steps of the distributed graph database association analysis method based on the TinkerPopAPI as described in the first aspect above.

[0046] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the distributed graph database association analysis method based on the TinkerPop API described in the first aspect above.

[0047] The technical solution provided in this application has the following beneficial effects:

[0048] This application provides two heterogeneous but complementary data sources for correlation analysis: physical perception and device status, by collecting acoustic wave and SCADA data. Secondly, it processes acoustic wave signals and reconstructs the sound field, transforming continuous physical disturbances into discrete events with precise spatiotemporal labels, thereby achieving preliminary localization and separation of abnormal physical phenomena. Then, it processes acoustic wave segments through adversarial generative networks and generates simulated acoustic patterns for comparison, combining spatiotemporal coordinates to form a structured event stream. This step elevates unstructured acoustic signals into event descriptions with clear semantics, enhancing the interpretability of the events.

[0049] Meanwhile, multi-scale feature extraction of SCADA data can more comprehensively characterize the operating status of equipment at different time granularities, providing rich state representations for subsequent causal analysis. Then, by aligning and fusing structured event streams with equipment feature information and using causal discovery algorithms for analysis, the potential causal direction and correlation strength between physical events and equipment states can be inferred from time series data, going beyond simple correlation. Finally, by calling the TinkerPop API to build a causal association network in a graph database, unified association modeling and persistence of multi-source heterogeneous data are realized, enabling the dynamic causal relationship between physical disturbances and equipment operating states to be visualized, stored, queried, and inferred.

[0050] Furthermore, this application also achieves high-confidence, refined identification and classification of the original acoustic signal through the generation and discrimination mechanism of adversarial generative networks, and combines it with spatiotemporal information, thereby transforming the original vibration data stream into a semantically clear and spatiotemporally complete structured event sequence, providing high-quality, directly processable input for subsequent causal correlation analysis.

[0051] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart illustrating a distributed graph database association analysis method based on the TinkerPop API, provided in this application embodiment;

[0054] Figure 2 A schematic diagram illustrating a specific implementation of a distributed graph database association analysis method based on the TinkerPop API, provided in this application embodiment;

[0055] Figure 3 This is a schematic diagram of the structure of a distributed graph database association analysis system based on the TinkerPop API, provided in an embodiment of this application. Detailed Implementation

[0056] The root cause of the aforementioned technical problems lies in the fact that existing methods, when processing physical sensing signals and equipment monitoring data, mainly rely on predefined rules or simple data mapping to construct static correlations. When faced with continuous, high-dimensional, and semantically ambiguous physical signals such as acoustic waves, this method struggles to effectively extract "events" with business meaning, and is even less able to deeply explore the potential and dynamic causal relationships between these physical events and the complex operating states of equipment. Therefore, directly applying existing technologies to process wide-area physical monitoring scenarios often results in the correlation analysis between physical disturbances and equipment states remaining superficial, failing to achieve intelligent tracing and reasoning from physical phenomena to the root causes of the system.

[0057] To address the aforementioned issues, this application proposes a distributed graph database association analysis method based on the TinkerPop API. Its core lies in: introducing a generative adversarial network (GAN) to perform refined identification and semantic annotation of sound source events, generating a structured event stream, thereby transforming unstructured physical signals into knowledge units with clearly defined type, time, and location information; simultaneously, multi-scale feature extraction is performed on device SCADA data to form feature information reflecting the comprehensive operating status of the device. Based on this, the structured event stream and device feature information are spatiotemporally aligned and fused, and a causal discovery algorithm is applied to analyze the lead-lag relationship and conditional independence between them, thereby inferring probabilistic causal hypotheses between physical events and device states. Finally, by calling the TinkerPop API, these causal hypotheses are treated as dynamic edges, constructed and continuously updated in the distributed graph database along with vertices such as geospatial data, sensor units, events, and device sources, forming a queryable and reasonable causal association network.

[0058] Therefore, through the aforementioned collaborative working process, this solution achieves deep semantic fusion and dynamic causal mining of physical sensor signals and equipment status data, fundamentally solving the problem of limited accuracy and depth of correlation analysis caused by insufficient data fusion and causal reasoning capabilities in existing technologies, and providing an effective technical means for anomaly localization and root cause analysis of large-scale infrastructure.

[0059] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0060] The core of this application is to provide a distributed graph database association analysis method based on the TinkerPop API, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0061] Step 101: Collect acoustic wave sensor signals and SCADA timing data of key equipment in the monitoring area.

[0062] In step 101, the acoustic wave sensing signal refers to the data collected by the distributed optical fiber acoustic wave sensing system. This system uses communication optical cables laid in the monitoring area, such as along railway lines or integrated pipe corridors, as continuous sensors to convert the changes in optical signals caused by physical disturbances such as vibration or sound waves at various points along the optical cable into electrical signals.

[0063] SCADA time-series data refers to physical measurement data reflecting the operating status of key equipment such as transformers or pumping stations that are continuously acquired at fixed time intervals from the monitoring and data acquisition system of the monitoring area. These data usually include a sequence of multiple parameters such as voltage, current, temperature and pressure changing over time.

[0064] In this embodiment, the original acoustic wave sensing signals reflecting the physical disturbance along the line are first collected in real time through the fiber optic sensing network deployed in the monitoring area. At the same time, the SCADA time-series data of the operating status of each key device in the area are collected synchronously from the monitoring system of each key device in the area, thereby providing two heterogeneous data sources from the physical sensing layer and the device monitoring layer for subsequent correlation analysis.

[0065] Step 102: Process the acoustic wave sensing signal to obtain strain data, and reconstruct the acoustic field of the strain data to obtain multiple acoustic source events, and determine the spatiotemporal coordinates of each acoustic source event.

[0066] Among them, strain data refers to the numerical sequence obtained by demodulating distributed optical fiber sensing signals, which characterizes the intensity of physical disturbances at each point along the optical cable at each sampling time; sound source event refers to a signal component that is identified as originating from an independent physical vibration source and whose energy is relatively concentrated in time and space during the data processing process; spatiotemporal coordinates refer to the combination of geographical location and time point used to uniquely identify the occurrence of a sound source event, including location coordinates and occurrence time.

[0067] In this embodiment, step 102 includes the following process:

[0068] Step 1021: Arrange the strain data along the spatial and temporal dimensions to form a data matrix.

[0069] In step 1021, the data matrix is ​​a mathematical structure obtained by organizing the strain data corresponding to each sensing point arranged along the spatial position of the optical cable and each sampling time arranged along the time axis into a two-dimensional table.

[0070] In this embodiment, the corresponding strain data are first organized according to the spatial position order of each sensing point and the temporal order of each sampling time. Then, a two-dimensional table is filled in according to the rule that the spatial position order corresponds to the row and the temporal order corresponds to the column, thereby constructing a data matrix in which the rows and columns describe the spatial distribution and temporal evolution, respectively.

[0071] Step 1022: Perform spatial spectrum analysis on the data matrix to calculate the arrival direction parameters of signals from different spatial azimuth and elevation angles.

[0072] In step 1022, the direction of arrival parameter refers to a set of angle values ​​that characterize the spatial orientation of the signal source, calculated by this method. This set of angle values ​​includes the azimuth angle and the elevation angle.

[0073] In this embodiment of the application, by performing correlation calculation on signals from different spatial locations in the data matrix, a covariance matrix describing the spatial relationship between signals is formed. Then, the covariance matrix is ​​subjected to eigenvalue decomposition or spectral peak search processing to estimate and extract the arrival direction parameters that can characterize the main spatial source direction of the vibration signal.

[0074] Step 1023: Based on the arrival direction parameters, a beamforming algorithm is used to separate the vibration signal components with concentrated energy in time and space from the data matrix, and each vibration signal component is defined as a sound source event.

[0075] In this embodiment, a digital beamforming filter is constructed for each specific direction based on the direction of arrival parameter. Then, the signal in the data matrix is ​​input into the corresponding filter to enhance the signal components from each direction. Next, the concentration of energy in the time and space axes of each enhanced signal component is analyzed, and those signal components that exhibit concentrated energy in both time and space are extracted. Finally, each such independent component is defined as a sound source event.

[0076] Step 1024: Calculate the distance of each sound source event relative to each sensing unit in the sensing array based on the arrival direction parameter.

[0077] In step 1024, the sensor array refers to a continuous sequence of virtual sensing points distributed along the optical cable for collecting strain data. The spatial distribution of the sensor array determines the spatial organization of the data matrix. Each row of data in the data matrix corresponds to the strain data collected by a specific sensing unit in the sensor array over time. Therefore, the sensor array is the physical basis and data source for constructing the spatial dimension of the data matrix.

[0078] In this embodiment of the application, based on the arrival direction parameters corresponding to each sound source event, combined with the known geographical location information of each sensing unit in the sensing array and the physical direction of the optical cable, the straight-line distance between the location of each sound source event and each sensing unit in the sensing array is determined by geometric trigonometric relationships.

[0079] The explanation of the above geometric triangular relationship can be found in related technologies. This embodiment will not elaborate on the relevant description of the geometric triangular relationship.

[0080] Step 1025: Based on the distance and the time information corresponding to the strain data in the data matrix, determine the location coordinates and occurrence time of each sound source event, wherein the location coordinates and occurrence time constitute spatiotemporal coordinates.

[0081] In this embodiment of the application, the distance information from each sound source event to multiple sensing units is calculated and solved using a multi-point geometric positioning method to determine the precise spatial coordinates of the sound source event. At the same time, based on the starting position of the signal component corresponding to the sound source event on the time axis of the data matrix, the specific time point of its occurrence is determined. Finally, the calculated position coordinates are combined with the determined time point of occurrence to form the spatiotemporal coordinates of the sound source event.

[0082] The explanation of the multi-point geometric positioning method can be found in relevant technologies. This embodiment will not elaborate on the specific calculation process of the method.

[0083] This application uses the above process to resolve the original, continuous acoustic wave sensing signal into a series of discrete, independent physical events with precise geographical locations and times of occurrence, laying a key foundation for the subsequent transformation of physical disturbances into analyzable structured information.

[0084] Step 103: Input the sound wave segment corresponding to each sound source event into the Generative Adversarial Network (GAN), generate a simulated voiceprint through the GAN, compare the simulated voiceprint with the sound wave segment, and combine it with the spatiotemporal coordinates to obtain a structured event stream.

[0085] Among them, the acoustic segment refers to the waveform sequence of the original vibration data extracted from the data matrix within a specific time period corresponding to the acoustic source event separated in step 102;

[0086] Simulated voiceprints refer to waveform data synthesized by a generator in a generative adversarial network based on the characteristics of an input sound wave segment, representing the typical sound characteristics of a certain type of device;

[0087] A structured event stream is a data sequence formed by arranging a series of records with a uniform format, containing event type semantic labels and spatiotemporal coordinates, in chronological order.

[0088] The overall structure of the adversarial generative network can be a deep convolutional generative adversarial network. The generator part adopts an encoder-decoder architecture. The encoder substructure consists of three cascaded one-dimensional convolutional layers. Each layer specifically includes convolution operations, batch normalization processing, and the LeakyReLU activation function. For example, the first layer uses a convolutional kernel with a width of 64 and a stride of 2 to map the input segment from a length of 1000 to a feature map with a length of 500.

[0089] The decoder substructure consists of three cascaded one-dimensional transposed convolutional layers. Each layer includes a transposed convolution operation, batch normalization, and a ReLU activation function. The last layer uses the Tanh activation function to output the waveform. For example, the last layer uses a convolutional kernel with a width of 64 and a stride of 2 to upsample the feature map to an output of length 1000. The discriminator part adopts a classifier architecture, whose substructure consists of four cascaded one-dimensional convolutional layers. Each layer includes a convolution operation, batch normalization, a LeakyReLU activation function, and a dropout layer. Finally, a fully connected layer is connected to output the classification probability.

[0090] The training process of this model first prepares a dataset of sound wave fragments containing labeled event types as real samples, initializes the network parameters of the generator and discriminator, and then fixes the generator parameters and updates the discriminator in each training iteration. That is, the discriminator is fed real samples and fake samples synthesized by the generator. The discriminator is optimized by calculating the binary cross-entropy loss and backpropagation to make it better distinguish between real and fake samples.

[0091] Next, the discriminator parameters are fixed and the generator is updated. By inputting the generator's output into the discriminator with fixed parameters, the loss for classifying generated samples as real is calculated and combined with the feature matching loss with real samples. The generator is then optimized through backpropagation to synthesize more realistic samples. This adversarial training is repeated until the performance of the generator and discriminator reaches a balance and the loss function converges, ultimately resulting in a stable model that can synthesize high-quality simulated voiceprints and accurately distinguish event categories.

[0092] It should be noted that the above structure is exemplary. This application does not impose specific limitations on the internal structure design, parameter design, training process, etc. of the adversarial generative network, and corresponding settings can be made according to the actual situation.

[0093] In this embodiment, step 103 includes the following process, such as... Figure 2 As shown:

[0094] Step 1031: Extract the sound wave segments within the corresponding time period from the data matrix based on the occurrence time of the sound source event.

[0095] In step 1031, the time period refers to a time interval formed by extending a preset duration forward and backward from the timestamp of the sound source event.

[0096] In this embodiment of the application, data of a certain duration is first extracted forward and backward according to the timestamp of the sound source event to determine a corresponding time period. Then, the corresponding column range is located in the time column of the data matrix according to the time period, and the corresponding row is located in the spatial row of the data matrix according to the range of the sensing points corresponding to the sound source event. Finally, all strain values ​​in the row-column intersection area are extracted from the data matrix. These values ​​are arranged in chronological order to constitute the sound wave segment of the sound source event.

[0097] In practical applications, assuming that a sound source event is identified by reconstructing the sound field of distributed fiber optic acoustic wave sensing data of a comprehensive utility tunnel along a railway line, with the timestamp of occurrence being 10:30:00 on the monitoring day, and the preset extraction time is 0.5 seconds before and after, then the determined time period is from 10:29:59.5 to 10:30:00.5 on the monitoring day; from the data matrix constructed by the fiber optic sensing data that has been connected to the enterprise's unified data base IoT platform, data of all virtual sensing channels corresponding to the section from K100+500 to K101+000 of the utility tunnel within this time period is extracted, resulting in a sound wave segment containing the changes in physical vibration intensity along the line within this 1 second. This segment can then be pushed to the intelligent algorithm modeling workbench of the base for further analysis.

[0098] Step 1032: Input the sound wave segment into the Generative Adversarial Network (GAN), and generate a simulated voiceprint based on the characteristics of the sound wave segment through the generator of the GAN.

[0099] Step 1032 may specifically include the following steps:

[0100] A1: The sound wave segment is received through the input layer of the generator.

[0101] In step A1, the generator's input layer receives the acoustic segment data from the previous step. This input layer is typically a fully connected layer or a one-dimensional convolutional layer, and its function is to convert the input data into a tensor format that can be processed internally by the neural network.

[0102] In this embodiment, the acoustic wave segment data is input in the form of a floating-point array, and the input layer contains neuron nodes that match the length of the acoustic wave segment, with each node receiving the value of a sampling point.

[0103] A2: The sound wave segment is encoded through multiple cascaded neural network layers of the generator to extract the feature vector of the sound wave segment in the frequency domain.

[0104] In step A2, the cascaded neural network layer refers to multiple fully connected layers or convolutional layers connected in sequence, used to perform nonlinear transformation and feature abstraction on the input data; the feature vector in the frequency domain refers to a low-dimensional mathematical vector that can characterize the frequency distribution and energy concentration characteristics of the sound wave signal, obtained through neural network learning.

[0105] In this embodiment, after the sound wave segment passes through the input layer, it is fed into multiple cascaded neural network layers of the generator. These layers first extract the local time-frequency features of the signal through convolution operations, then reduce the data dimensionality and enhance the feature robustness through pooling operations, and finally integrate and compress the scattered features into a fixed-length feature vector through a fully connected layer. This feature vector centrally reflects the main characteristics of the original sound wave segment in the frequency domain.

[0106] In practical applications, the encoding part consists of three one-dimensional convolutional layers, each followed by an activation function and a pooling layer. After the sound wave segment is processed by these three layers, the dimension gradually changes from 1000×1 to 250×32, 125×64, and 62×128. Finally, it is flattened and compressed into a feature vector of length 128 through a fully connected layer.

[0107] A3: Input the feature vector into the decoding network of the generator, and perform a weighted combination of the feature vector and a preset voiceprint feature template in the decoding network.

[0108] In step A3, the voiceprint feature template is a mathematical vector that is pre-learned through training and represents the typical acoustic features of different types of devices.

[0109] In this embodiment, the feature vector output from the encoding part is input to the decoding network. The decoding network first maps the feature vector to the same dimension as the preset voiceprint feature template through a fully connected layer. Then, in this layer, the mapped feature vector is weighted and summed with one or more voiceprint feature templates loaded from the pre-trained model library. The weight coefficients are dynamically learned by the network parameters or set to fixed values ​​according to the task, thereby generating a new feature vector that integrates the features of the input segment and the typical device voiceprint characteristics.

[0110] A4: The weighted combination result is upsampled through multiple deconvolution layers of the decoding network to generate a simulated voiceprint.

[0111] In step A4, a deconvolutional layer is a neural network layer used to upsample data to increase its spatial or temporal dimensions; upsampling refers to the process of increasing the length of a data sequence.

[0112] In the embodiments of this application, the fused feature vector obtained after weighted combination is fed into multiple deconvolution layers of the decoding network. Each deconvolution layer performs deconvolution operation on the input data through a specific filter, gradually increasing the length of the data and reducing the number of feature channels. After several layers of such processing, the data is reconstructed into a time-domain waveform sequence with the same length as the original sound wave segment. This waveform sequence is the simulated voiceprint finally output by the generator.

[0113] In practical applications, the decoding network consists of three deconvolutional layers. The first layer upsamples the 256-dimensional fusion vector into data with a length of 500 and 128 channels. The second layer upsamples the data into data with a length of 1000 and 64 channels. The third layer upsamples the data into data with a length of 1000 and 1 channel. This data is the simulated voiceprint.

[0114] Step 1033: By using the discriminator of the adversarial generative network, compare the differences in frequency domain features and time domain features between the sound wave segment and the simulated voiceprint to determine the probability that the sound wave segment belongs to different event categories.

[0115] In step 1033, the discriminator refers to the neural network part of the generative adversarial network that is responsible for distinguishing whether the input data is real data or synthesized data by the generator; the frequency domain features refer to the amplitude and phase distribution characteristics of the signal in the frequency domain after Fourier transform.

[0116] Time-domain characteristics refer to the waveform, amplitude, and statistical properties of a signal on the time axis; the probability of an event category refers to a numerical distribution output by the discriminator that represents the likelihood of an input sound wave segment belonging to each predefined device event type.

[0117] In this embodiment, the discriminator simultaneously receives the original sound wave segment and the generated simulated voiceprint as input. The discriminator internally extracts the spectral features of the two inputs in the frequency domain and the waveform envelope, zero-crossing rate, and other features in the time domain through a multi-layer convolutional neural network. Then, it calculates the degree of difference between the two inputs in each feature dimension. Finally, based on these difference information, it outputs a probability vector through a fully connected layer and an activation function. Each element value in the probability vector represents the confidence level that the input sound wave segment belongs to a specific event category.

[0118] In practical applications, the discriminator extracts two types of features from the acoustic wave segments and simulated acoustic patterns: Mel frequency cepstral coefficients and short-time energy. It calculates the cosine similarity between the two features to obtain a difference score. Then, it compares this score with the classification decision boundary learned from historical railway equipment acoustic event training data loaded from the intelligent algorithm modeling workbench of the enterprise unified data platform. Finally, it outputs a probability distribution that conforms to the event classification system of railway operation and maintenance scenarios, such as 0.05 for train passing, 0.85 for catenary arc discharge, and 0.10 for foreign object intrusion.

[0119] Step 1034: Based on the probability, assign event type labels to the sound wave segment using a softmax classifier.

[0120] In step 1034, the softmax classifier is a mathematical function that converts the raw scores output by the neural network into a normalized probability distribution and selects the class with the highest probability as the final classification result.

[0121] In this embodiment, the probability vector output by the discriminator is input into a softmax classifier. The classifier first normalizes the probability vector to ensure that the sum of all elements is 1, then selects the category index corresponding to the element with the largest value, and finally assigns the corresponding event type label to the sound wave segment according to the preset mapping relationship between the category index and the event type name.

[0122] In practical applications, the original score vector output by the discriminator is 2.1, 5.3, 0.8. After calculation by the softmax function, the normalized probabilities are 0.05, 0.85, 0.10. The second element value of 0.85 is the largest, and its corresponding event type is contact wire arc discharge. Therefore, the event type label assigned to this acoustic segment is contact wire arc discharge.

[0123] Step 1035: Associate the event type label with the spatiotemporal coordinates to generate a structured event record, and combine all structured event records into a structured event stream in chronological order.

[0124] In step 1035, a structured event record refers to a data object containing three core fields: event type label, location coordinates, and occurrence timestamp.

[0125] In this embodiment of the application, the event type label assigned in step 1034 is bound to the spatiotemporal coordinates of the sound source event determined in step 102 to create a data record containing five fields: event type, latitude, longitude, altitude, and timestamp, i.e., a structured event record. After processing all sound source events, all generated structured event records are sorted in ascending order according to their timestamp field, and then these sorted records are connected in sequence to form a continuous, time-progressing structured event stream.

[0126] In practical applications, the spatiotemporal coordinates of a sound source event are located at kilometer marker K100+750 along a certain railway line, corresponding to geographical coordinates of 123 degrees north latitude and 456 degrees east longitude, with a timestamp of 10:30:00 on June 15, 2025. Its event type label is "contact wire arc discharge." The generated structured event record contains fields for event type, kilometer location, geographical coordinates, and timestamp. It should be understood that the aforementioned east longitude, north latitude, and other related values ​​given subsequently are all virtual and hypothetical geographical information, not real geographical information.

[0127] When multiple such records are generated within the same monitoring period, all records are sorted by timestamp and then written into a time-series database or used as a streaming data source through the data platform's data stream processing capabilities. This data is then provided to the subsequent enterprise graph database for building a relational network, thereby forming a structured event stream for analysis.

[0128] In the above process, this application utilizes generative adversarial networks to achieve intelligent recognition and high-confidence classification of unstructured acoustic signals, and combines them with accurate spatiotemporal information, thereby transforming the original physical disturbance data into semantically clear and formatted event sequences, providing high-quality and directly processable input information for subsequent deep fusion and causal analysis with device status data.

[0129] Step 104: Perform multi-scale feature extraction on the SCADA time series data to obtain feature information, which is used to reflect the operating status of key equipment.

[0130] In step 104, feature information refers to a set of quantitative indicators that can more comprehensively characterize the overall operating status of the equipment within a specific time period, obtained by calculating and combining the original SCADA time series data.

[0131] In this embodiment, a SCADA time-series data stream collected from a critical equipment monitoring system is first received. Then, for each physical quantity measurement point of interest in the data stream, parallel calculation and analysis are performed using short time windows and long time windows respectively. Within the short time window, local statistical characteristics that can reflect rapid fluctuations and instantaneous changes in the data are calculated, while within the long time window, global statistical characteristics that can reflect the overall level and trend evolution of the data are calculated.

[0132] Next, the various feature values ​​calculated at different time scales for each measuring point are combined to form a multi-scale feature vector for that measuring point. Finally, the feature vectors of all relevant measuring points are integrated together in a predetermined dimensional order to form a multi-dimensional feature matrix. This feature matrix is ​​the final feature information used to comprehensively reflect the operating status of key equipment.

[0133] Step 105: Align and fuse the structured event stream with the feature information to obtain fused data, and use the causal discovery algorithm to analyze the lead-lag relationship and conditional independence test among multiple fused records in the fused data to generate a causal relationship set.

[0134] Among them, fused data refers to a new dataset generated after alignment and fusion operations, in which each record simultaneously contains event type, spatiotemporal information, and device characteristic information; lead-lag relationship refers to determining the degree to which one variable leads another variable in time by calculating the correlation between two time series at different time offsets;

[0135] Conditional independence test is a statistical test method used to determine whether two variables remain statistically independent given the values ​​of other variables; causal relationship set refers to a data set output by a causal discovery algorithm that contains a set of inferred causal relationships and their confidence scores.

[0136] In this embodiment, step 105 includes the following process:

[0137] Step 1051: Based on the time recorded in the structured event stream, find the feature corresponding to the time from the feature information.

[0138] In step 1051, the time recorded in the structured event stream refers to the occurrence timestamp marked in each structured event record.

[0139] In this embodiment of the application, the timestamp of each record in the structured event stream is read one by one. Then, based on the timestamp, the device feature vector corresponding to the time point that is the same as or closest to the event timestamp is found in the time series corresponding to the feature information generated in step 104, so as to achieve accurate time positioning of the event.

[0140] In practical applications, suppose a structured event record has an occurrence timestamp of 10:30:00 on June 15, 2025. This event originates from acoustic monitoring at a substation along a railway line. The time series of feature information is data collected from the SCADA system of the substation's key equipment and stored in the enterprise's time series database, with one sampling point per minute. Then, the equipment feature vector closest to this timestamp is queried and extracted from the time series database. For example, the multi-dimensional feature data corresponding to 10:30:00 includes transformer oil temperature, load current, etc., and is used as the equipment operating status feature at the same time as the acoustic event.

[0141] Step 1052: Merge the feature with the structured event record of the corresponding time in the structured event stream to form a fused record, and arrange the multiple fused records according to time to obtain fused data.

[0142] In step 1052, a fused record refers to a new data record formed by concatenating the fields of a structured event record with the corresponding device feature vector, thus expanding the information dimensions.

[0143] In this embodiment, the found device feature vector is added as a new field to the corresponding structured event record, thereby generating a fused record containing event attributes, spatial location, timestamp, and multi-dimensional device features. Then, all event records are processed, and feature merging is performed on each record. Finally, all generated fused records are sorted in ascending order according to their timestamp field, and these sorted records are arranged continuously to form a unified multivariate time series data arranged in chronological order, which is the fused data.

[0144] In practical applications, a structured event record contains three fields: event type (e.g., contact network arc discharge), location (e.g., railway mileage K100+750), and timestamp. The corresponding equipment feature vector is extracted from the substation SCADA system and obtained through multi-scale calculation, containing 10 numerical arrays reflecting the equipment's operating status. After merging, a fusion record containing 13 fields is generated. When there are 100 such acoustic event records within the same monitoring period, such as within 10 minutes, the data base's integrated batch processing capability generates 100 fusion records. After being sorted by time, a structured table of 100 rows and 13 columns is formed. This table is the fusion data that can be directly analyzed by subsequent causal discovery algorithms.

[0145] Step 1053: Analyze the fused data using a causal discovery algorithm to determine the lead-lag relationship between different variables.

[0146] In step 1053, the variable refers to each column of data in the fused data, such as event type variables and device characteristic variables. An explanation of the causal discovery algorithm can be found in relevant technical documents, and will not be elaborated upon here.

[0147] In this embodiment, the fused data is input into a causal discovery algorithm for processing. The algorithm first calculates the cross-correlation function or Granger causality test statistic at different time offsets for all variable pairs in the fused data, such as event type variable A and device temperature feature variable B. By searching for the offset that makes the statistic reach its maximum value, the algorithm determines the optimal time delay for variable A to lead variable B, or the optimal time delay for variable B to lead variable A, and finally determines the lead-lag relationship between all variable pairs.

[0148] In practical applications, the algorithm calculates the cross-correlation function between the event type "contact network arc discharge" and the equipment feature "transformer winding temperature". Within the calculation time offset range of -5 seconds to +5 seconds, it is found that the cross-correlation value is the largest, i.e., 0.6, when the event type leads the temperature feature by 3 seconds. Thus, it is determined that the event variable leads the temperature feature variable by 3 seconds.

[0149] Step 1054: Based on the aforementioned lead-lag relationship and combined with the conditional independence test, determine the causal relationship between different variables.

[0150] In step 1054, the conditional independence test is used to eliminate spurious correlations caused by the presence of other common causal variables, thereby more accurately determining direct causal relationships.

[0151] In the embodiments of this application, based on the established lead-lag relationship, the causal discovery algorithm further performs conditional independence tests on all possible variable triples; for example, for variables X and Y with a lead relationship, the algorithm will test whether X and Y are still related given the value of a third variable Z; if X and Y become independent given a certain variable Z, it indicates that the direct causal relationship between X and Y is not valid, and the association may be caused by Z; by systematically performing a large number of such tests, the algorithm can infer the most likely direct causal direction between variables.

[0152] In practical applications, it has been determined that the event variable X leads the feature variable Y. The algorithm introduces a third variable, namely ambient temperature, to conduct a conditional independence test. The calculation shows that when the specific value of ambient temperature Z is given, the partial correlation coefficient between variable X and variable Y is close to 0, that is, the two become independent. Therefore, it is determined that there is no direct causal relationship between event variable X and feature variable Y, and their association is caused by ambient temperature Z as a common cause.

[0153] Step 1055: Calculate the confidence score of the causal relationship, and combine the causal relationship and the confidence score to obtain a set of causal relationships.

[0154] In step 1055, the confidence score is a numerical value used to quantify the reliability of the inferred causal relationship, typically calculated based on the sex ratio of a statistical test or the goodness of fit of a model.

[0155] In this embodiment of the application, a confidence score is calculated for each direct causal relationship determined to be valid in step 1054. This score can be obtained by calculating the p-value of the F-statistic of the Granger causality test or by comparing the merits of models that include and do not include the causal edge using the Bayesian information criterion. Then, the description of each causal relationship, which includes the causal variable, the outcome variable, and the causal direction, is bound to its calculated confidence score to form a causal relationship entry. All inferred causal relationship entries are gathered together to form the final causal relationship set.

[0156] In practical applications, if the event variable "contact wire arc discharge" is determined to be the cause of the increase in the equipment characteristic variable "number of times the protection device operates", and the p-value is 0.01 obtained through Granger causality test, then its confidence score is set to 0.99. The final causal relationship set includes entries such as cause contact wire arc discharge, result number of times the protection device operates, direction cause points to result, and confidence score 0.99.

[0157] This application achieves deep fusion of physical event and equipment status data through the above process, and uses statistical inference methods to mine causal relationships with clear time direction and statistical significance from the fused data, thus providing a direct and reliable data foundation for the subsequent construction of interpretable and quantifiable causal relationship networks.

[0158] Step 106: By calling the operations provided by the TinkerPop API, using the causal relationships in the causal relationship set as edges and geospatial entities, sensing units, sound source events and device sources as vertices, a causal relationship network is constructed in the distributed graph database. The causal relationship network is used to associate physical disturbances with the operating status of the key equipment.

[0159] In this graph data structure, a vertex is the basic unit representing a specific entity or object; an edge is a directed line segment connecting two vertices, representing the relationship between the vertices; the causal relationship network is a graph data model constructed in this application, where the vertices represent entities in the physical world such as geographic locations, fiber optic sensing points, identified acoustic events, and source devices that may cause vibrations, and the directed edges represent causal relationships inferred from the analyzed data, pointing from one entity to another. This network is used to systematically organize and store the dynamic relationship between physical disturbance events and device operating status in a distributed graph database.

[0160] Physical disturbances refer to physical phenomena that cause strain changes in communication optical cables, as monitored by a distributed optical fiber acoustic sensing system. These phenomena include, but are not limited to, vibrations caused by passing trains, abnormal sounds caused by equipment failures, or impacts caused by human activities. They manifest as continuous waveform changes in the sensing signals and are the event sources to be analyzed in this method.

[0161] In this embodiment, step 106 includes the following process:

[0162] Step 1061: In the distributed graph database, for each causal relationship in the set of causal relationships, call the edge addition command provided by the TinkerPop API to add directed edges between the corresponding vertices.

[0163] In step 1061, a directed edge is an edge with a clear direction, used to represent the direction of action of a relationship, such as from cause to effect.

[0164] In this embodiment, each record in the causal relationship set is first read. Based on the cause variable and result variable specified in the record, the vertices representing the two variables are found in the distributed graph database. Then, the specific instructions for adding a directed edge provided in the TinkerPop API are called to create a directed edge from the vertex of the cause variable to the vertex of the result variable. Other information about the causal relationship, such as the confidence score obtained from the causal relationship set, is recorded in the attributes of the directed edge.

[0165] Step 1062: By calling the graph traversal instructions provided by the TinkerPop API, query and analyze the vertices and directed edges to form a causal relationship network.

[0166] In step 1062, the graph traversal instruction refers to the query operation used to access vertices and edges in a graph data structure according to a specific path or rule; the structural form of the causal network can be referred to relevant technologies, and will not be elaborated here.

[0167] In this embodiment, after all directed edges corresponding to causal relationships have been added, the graph database is queried by calling the graph traversal instructions provided by the TinkerPop API. For example, traversal rules are written using the Gremlin query language, starting from a specified vertex and gradually visiting all other vertices and edges connected to it along the direction of the directed edges, thereby exploring the connection structure and paths of the entire graph. The overall graph structure composed of vertices and directed edges revealed by querying and traversing through instructions forms the causal relationship network described in this application.

[0168] In this embodiment, after step 106, the method further includes the following steps:

[0169] B1: Input the vertex data and edge data in the causal association network into the graph neural network, and perform iterative calculations on the vertex data and edge data through the graph neural network to update the feature representation of the vertex data and edge data.

[0170] In this embodiment, the initial feature vectors of all vertices and the attribute vectors of all edges in the causal association network are first input into a pre-trained graph neural network. The network uses its internal message passing mechanism to allow each vertex to receive information from its neighboring vertices and connected edges. Then, through an aggregation function and an update function, the received information and its original features are fused and calculated. After multiple layers of such iterative calculations, each vertex and each edge will generate an updated new feature representation that contains its local graph structure information.

[0171] Explanations regarding the structural design, parameter design, and training process of graph neural networks can be found in relevant technical resources, and will not be elaborated upon here.

[0172] B2: Based on the updated feature representation output by the graph neural network, the attention weight of each associated edge in the causal association network is calculated using an attention mechanism.

[0173] In step B2, the attention weight is a value used to measure the importance of a certain edge in the graph in the current task. This value is obtained by calculating the correlation of the features of the vertices connected to the edge.

[0174] In this embodiment, based on the updated vertex features output by the graph neural network, an attention mechanism is applied to each directed edge in the causal association network. Specifically, for the edge connecting vertex A and vertex B, the updated feature vectors of vertex A and vertex B are first concatenated or multiplied to obtain an initial attention score. Then, the score is input into a single-layer neural network and normalized. Finally, a value between 0 and 1 is calculated, which is the attention weight of the edge. The higher the weight, the more important the edge is in the current context.

[0175] B3: Based on the attention weights, update the weight attributes of the corresponding associated edges in the distributed graph database.

[0176] In this embodiment, the calculated attention weight of each edge is used as a new attribute value for that edge. By calling the update attribute instruction of the TinkerPop API, the value is written into the weight attribute field of the corresponding edge in the distributed graph database, thereby completing the dynamic optimization and update of the edge weights in the graph.

[0177] This application not only dynamically constructs the discovered causal relationships into a visualized graph network through the above process, but also further utilizes graph neural networks and attention mechanisms to intelligently learn and dynamically adjust the strength of relationships in the network, thereby enabling the causal relationship network to more accurately and precisely reflect the complex and dynamically changing dependencies between physical disturbances and device states.

[0178] Figure 3 This application provides a schematic diagram of the structure of a distributed graph database association analysis system based on the TinkerPop API, as shown in the embodiments of this application. Figure 3 As shown, the system includes:

[0179] The acquisition module 31 is used to acquire acoustic wave sensor signals and SCADA timing data of key equipment in the monitoring area.

[0180] The processing module 32 is used to process the acoustic wave sensing signal to obtain strain data, and to reconstruct the sound field of the strain data to obtain multiple sound source events, and to determine the spatiotemporal coordinates of each sound source event.

[0181] The input module 33 is used to input the sound wave segment corresponding to each sound source event into the adversarial generative network, generate a simulated voiceprint through the adversarial generative network, compare the simulated voiceprint with the sound wave segment, and obtain a structured event stream by combining the spatiotemporal coordinates.

[0182] The extraction module 34 is used to perform multi-scale feature extraction on the SCADA time series data to obtain feature information, which is used to reflect the operating status of key equipment.

[0183] The fusion module 35 is used to align and fuse the structured event stream with the feature information to obtain fused data, and to use a causal discovery algorithm to analyze the lead-lag relationship and conditional independence test among multiple fused records in the fused data to generate a causal relationship set.

[0184] The construction module 36 is used to construct a causal relationship network in a distributed graph database by calling the operations provided by the TinkerPop API, using the causal relationships in the causal relationship set as edges and geospatial entities, sensing units, sound source events and device sources as vertices. The causal relationship network is used to associate physical disturbances with the operating status of the key equipment.

[0185] The distributed graph database association analysis system based on the TinkerPop API in this application embodiment is used to implement the aforementioned distributed graph database association analysis method based on the TinkerPop API. Therefore, the specific implementation of the distributed graph database association analysis system based on the TinkerPop API can be found in the embodiment section of the distributed graph database association analysis method based on the TinkerPop API above. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0186] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described distributed graph database association analysis methods based on the TinkerPop API.

[0187] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described distributed graph database association analysis methods based on the TinkerPop API.

[0188] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0189] The embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the distributed graph database association analysis method based on the TinkerPop API.

[0190] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0191] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0192] The foregoing has provided a detailed description of a distributed graph database association analysis method and system based on the TinkerPop API provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A distributed graph database association analysis method based on the TinkerPop API, characterized in that, include: Collect acoustic sensor signals and SCADA time-series data of key equipment within the monitoring area; The acoustic wave sensing signal is processed to obtain strain data, and the strain data is reconstructed into a sound field to obtain multiple sound source events, and the spatiotemporal coordinates of each sound source event are determined. The sound wave segment corresponding to each sound source event is input into the Generative Adversarial Network (GAN), which generates a simulated voiceprint. The simulated voiceprint is compared with the sound wave segment, and the structured event stream is obtained by combining the spatiotemporal coordinates. Multi-scale feature extraction is performed on the SCADA time-series data to obtain feature information, which is used to reflect the operating status of key equipment; The structured event stream is aligned and fused with the feature information to obtain fused data. A causal discovery algorithm is then used to analyze the lead-lag relationship and conditional independence test among multiple fused records in the fused data to generate a causal relationship set. By calling the operations provided by the TinkerPop API, using the causal relationships in the causal relationship set as edges, and the geographic spatial entities corresponding to the monitoring area, the sensing units in the sensing array that collects the acoustic wave sensing signals, the sound source events, and the device sources corresponding to the key equipment as vertices, a causal relationship network is constructed in a distributed graph database. The causal relationship network is used to associate physical disturbances with the operating status of the key equipment. The structured event stream is aligned and fused with the feature information to obtain fused data. A causal discovery algorithm is then used to analyze the lead-lag relationships and conditional independence tests among multiple fused records in the fused data, generating a causal relationship set, including: Based on the time recorded in the structured event stream, the feature corresponding to the time is found from the feature information; The features are merged with the structured event records of the corresponding time in the structured event stream to form a fused record, and multiple fused records are arranged according to time to obtain fused data; The fused data was analyzed using a causal discovery algorithm to determine the lead-lag relationships between different variables. Based on the aforementioned lead-lag relationship, and combined with the conditional independence test, determine the causal relationship between different variables; Calculate the confidence score of the causal relationship, and combine the causal relationship and the confidence score to obtain a set of causal relationships.

2. The distributed graph database association analysis method based on the TinkerPop API according to claim 1, characterized in that, The process of inputting the sound wave segment corresponding to each sound source event into a generative adversarial network (GAN), generating a simulated voiceprint through the GAN, comparing the simulated voiceprint with the sound wave segment, and combining the spatiotemporal coordinates to obtain a structured event stream includes: Based on the occurrence time of the sound source event, the sound wave segments within the corresponding time period are extracted from the data matrix; The sound wave segment is input into the Generative Adversarial Network (GAN), and the generator of the GAN generates a simulated voiceprint based on the characteristics of the sound wave segment. The discriminator of the adversarial generative network compares the differences in frequency domain features and time domain features between the sound wave fragment and the simulated voiceprint to determine the probability that the sound wave fragment belongs to different event categories. Based on the probability, an event type label is assigned to the sound wave segment using a softmax classifier; The event type labels are associated with the spatiotemporal coordinates to generate structured event records, and all structured event records are combined into a structured event stream in chronological order.

3. The distributed graph database association analysis method based on the TinkerPop API according to claim 2, characterized in that, The generator of the adversarial generative network generates simulated voiceprints based on the features of the sound wave fragments, including: The sound wave fragment is received through the input layer of the generator; The sound wave segment is encoded by multiple cascaded neural network layers of the generator to extract the feature vector of the sound wave segment in the frequency domain; The feature vector is input into the decoding network of the generator, where the feature vector is weighted and combined with a preset voiceprint feature template. The weighted combination result is upsampled through multiple deconvolution layers of the decoding network to generate a simulated voiceprint.

4. The distributed graph database association analysis method based on the TinkerPop API according to claim 1, characterized in that, The step of reconstructing the sound field from the strain data to obtain multiple sound source events, and determining the spatiotemporal coordinates for each sound source event, includes: The strain data are arranged along the spatial and temporal dimensions to form a data matrix; Spatial spectrum analysis is performed on the data matrix to calculate the direction of arrival parameters of signals from different spatial azimuth and elevation angles; Based on the arrival direction parameters, a beamforming algorithm is used to separate the vibration signal components with concentrated energy in time and space from the data matrix, and each vibration signal component is defined as a sound source event. Based on the arrival direction parameter, calculate the distance of each sound source event relative to each sensing unit in the sensing array; Based on the distance and the time information corresponding to the strain data in the data matrix, the location coordinates and occurrence time of each sound source event are determined, and the location coordinates and occurrence time constitute spatiotemporal coordinates.

5. The distributed graph database association analysis method based on the TinkerPop API according to claim 1, characterized in that, The process of constructing a causal relationship network in a distributed graph database by calling the operations provided by the TinkerPop API, using causal relationships in the causal relationship set as edges and geospatial entities, sensing units, sound source events, and device sources as vertices, includes: In the distributed graph database, for each causal relationship in the set of causal relationships, the edge addition command provided by TinkerPopAPI is called to add directed edges between the corresponding vertices; By calling the graph traversal instructions provided by the TinkerPop API, the vertices and directed edges are queried and analyzed to form a causal relationship network.

6. The distributed graph database association analysis method based on the TinkerPop API according to claim 1, characterized in that, After constructing the causal relationship network, the following is also included: The vertex data and edge data in the causal association network are input into the graph neural network. The graph neural network iteratively calculates the vertex data and edge data to update the feature representation of the vertex data and edge data. Based on the updated feature representation output by the graph neural network, the attention weight of each associated edge in the causal association network is calculated using an attention mechanism. Based on the attention weights, update the weight attributes of the corresponding associated edges in the distributed graph database.

7. A distributed graph database association analysis system based on the TinkerPop API, characterized in that, include: The acquisition module is used to acquire acoustic wave sensor signals and SCADA time-series data of key equipment within the monitoring area. The processing module is used to process the acoustic wave sensing signal to obtain strain data, and to reconstruct the sound field of the strain data to obtain multiple sound source events, and to determine the spatiotemporal coordinates of each sound source event. The input module is used to input the sound wave segment corresponding to each sound source event into the adversarial generative network, generate a simulated voiceprint through the adversarial generative network, compare the simulated voiceprint with the sound wave segment, and obtain a structured event stream by combining the spatiotemporal coordinates. The extraction module is used to perform multi-scale feature extraction on the SCADA time series data to obtain feature information, which is used to reflect the operating status of key equipment; The fusion module is used to align and fuse the structured event stream with the feature information to obtain fused data, and to use a causal discovery algorithm to analyze the lead-lag relationship and conditional independence test among multiple fused records in the fused data to generate a causal relationship set. The construction module is used to construct a causal relationship network in a distributed graph database by calling the operations provided by the TinkerPop API. The causal relationship network is used to associate physical disturbances with the operating status of the key equipment. The causal relationship network is used to associate physical disturbances with the operating status of the key equipment. The structured event stream is aligned and fused with the feature information to obtain fused data. A causal discovery algorithm is then used to analyze the lead-lag relationships and conditional independence tests among multiple fused records in the fused data, generating a causal relationship set, including: Based on the time recorded in the structured event stream, the feature corresponding to the time is found from the feature information; The features are merged with the structured event records of the corresponding time in the structured event stream to form a fused record, and multiple fused records are arranged according to time to obtain fused data; The fused data was analyzed using a causal discovery algorithm to determine the lead-lag relationships between different variables. Based on the aforementioned lead-lag relationship, and combined with the conditional independence test, determine the causal relationship between different variables; Calculate the confidence score of the causal relationship, and combine the causal relationship and the confidence score to obtain a set of causal relationships.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the distributed graph database association analysis method based on the TinkerPop API as described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the distributed graph database association analysis method based on the TinkerPop API as described in any one of claims 1 to 6.

Images