Multi-source sensor monitoring method and device, electronic equipment and storage medium

Abstract

The invention provides a multi-source sensor monitoring method and device, electronic equipment and a storage medium, and the method comprises the steps: configuring a heterogeneous sensor network according to a monitoring target and environment; a sensor node topological relation graph is established, a unique space-time identifier is distributed, and timestamp synchronous correction is carried out; processing data of different sampling frequencies and then extracting core feature vectors to form a multi-dimensional feature matrix; constructing a dynamic graph structure, learning a space-time dependency relationship between sensors, and generating a node embedding vector; establishing a space-time propagation model to predict a data evolution trend, constructing a normal working model, and calculating a reconstruction error threshold value; when abnormality is detected, triggering a sensor network adaptive reconfiguration mechanism, and adjusting the sampling frequency and sensitivity parameters of the key sensor; and constructing a hierarchical decision tree, fusing the monitoring result, generating confidence evaluation, outputting a differential early warning signal according to the risk level, and generating a monitoring report. According to the invention, the monitoring precision and the abnormal detection and fault diagnosis capability are improved.

Description

Technical Field

[0001] This invention relates to the field of monitoring technology, and specifically to a multi-source sensor monitoring method, device, electronic equipment, and storage medium. Background Technology

[0002] In monitoring activities, data obtained using a single sensor is fragmented, singular, and easily affected by interference, leading to inaccurate detection. Multi-sensor-based monitoring schemes are gaining increasing attention because multiple sensors can acquire more information than a single sensor, providing a better description of the state of the monitored target. However, existing multi-source sensor monitoring schemes have shortcomings that need improvement, such as low monitoring accuracy and weak anomaly detection and fault diagnosis capabilities. Summary of the Invention

[0003] Based on the above-mentioned problems, this invention proposes a multi-source sensor monitoring method, device, electronic device, and storage medium. Through the solution of this invention, the monitoring accuracy, anomaly detection, and fault diagnosis capabilities are improved.

[0004] In view of this, one aspect of the present invention proposes a multi-source sensor monitoring method, comprising: Based on the physical characteristics and environmental parameters of the monitored target, a heterogeneous sensor network, including optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, is dynamically configured. Based on the spatial distribution of sensors, a topological relationship diagram of sensor nodes is established, and a unique spatiotemporal identifier is assigned to each sensor. The raw data collected by each sensor is timestamped and synchronized to establish a unified time reference. The Kalman filter algorithm is used to perform time-domain interpolation and alignment of sensor data with different sampling frequencies; Principal component analysis is used to extract the core feature vectors of each sensor data to form a multidimensional feature matrix. Construct a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights; By combining multidimensional feature matrices, graph attention networks are used to learn the spatiotemporal dependencies between sensors and generate node embedding vectors. Establish a spatiotemporal propagation model to predict the spatiotemporal evolution trend of sensor data; A normal operation model is constructed based on a variational autoencoder, and the reconstruction error threshold is calculated. When an anomaly is detected, the adaptive reconfiguration mechanism of the sensor network is triggered; Based on the spatiotemporal evolution trend, anomaly type, and severity, dynamically adjust the sampling frequency and sensitivity parameters of key sensors; Construct a hierarchical decision tree to weight and fuse monitoring results from sensor level, region level, and system level; A confidence assessment is generated by using a Bayesian network inference mechanism, combining historical data and real-time monitoring results. Differentiated early warning signals are output based on the risk level, and monitoring reports are automatically generated.

[0005] Optionally, the dynamic configuration includes the step of creating a heterogeneous sensor network comprising optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, specifically including: The optimal sensor deployment density is calculated based on the geometric features of the monitoring area corresponding to the monitoring target and the sensing range of the sensor. Using an adaptive coverage optimization algorithm, the influence weight of sensor node m is calculated according to the following formula: .

[0006] Optionally, the multi-dimensional feature fusion in the process of extracting the core feature vectors of each sensor data through principal component analysis to form a multi-dimensional feature matrix adopts a spatiotemporal weighted fusion algorithm: For the feature vector of sensor s at time t, an adaptive weight allocation mechanism is established; The fused feature vector is calculated using the following formula:

[0007] .

[0008] Optionally, the step of constructing a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights includes: A multi-level spatiotemporal graph structure is constructed, and the attention weights between nodes are calculated using the following formula:

[0009] .

[0010] Optionally, the step of constructing a normal operating model based on a variational autoencoder and calculating the reconstruction error threshold includes: A hierarchical variational autoencoder is constructed, and the outlier score is calculated using the following formula:

[0011]

[0012] .

[0013] Optionally, the step of generating a confidence assessment by employing a Bayesian network inference mechanism, combining historical data and real-time monitoring results, includes: System-level decision confidence is calculated using the following formula:

[0014]

[0015] .

[0016] Optionally, the step of constructing a hierarchical decision tree and weightedly fusing the monitoring results at the sensor level, region level, and system level includes: Based on the multidimensional feature matrix and anomaly detection results, local decision results and confidence scores are generated for each sensor node; Based on the sensor’s historical performance and current health status, assign initial weight coefficients to each sensor-level decision. Based on the spatial distribution of sensors and the division of monitoring areas, sensors with geographical proximity are grouped to form monitoring areas; The sensor-level decisions in each region are weighted and averaged to generate regional-level decision results, and the consistency index of decisions within the region is calculated. Collect all regional-level decision-making results and assign system-level weights to each region based on its importance level and decision consistency index. Multiple regional-level decisions are merged into a system-level final decision using a weighted fusion algorithm, and a decision confidence assessment is generated. When there are conflicts in the decision-making results at different levels or in different regions, the conflict resolution mechanism is activated, and the decision-making result with higher confidence is adopted first. Analyze the root causes of conflict decisions, identify the sensors or areas that led to the conflict, and mark them as objects that need further verification. Based on system-level decision-making results, and according to preset risk level thresholds, corresponding early warning levels and handling recommendations are generated. The final decision results are fed back to the adaptive tuning module of the sensor network to optimize subsequent monitoring strategies and parameter configurations.

[0017] A second aspect of the present invention provides a multi-source sensor monitoring device for performing a multi-source sensor monitoring method, comprising: an acquisition module, a processing module, and an execution module; The processing module is configured as follows: Based on the physical characteristics and environmental parameters of the monitored target, a heterogeneous sensor network, including optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, is dynamically configured. Based on the spatial distribution of sensors, a topological relationship diagram of sensor nodes is established, and a unique spatiotemporal identifier is assigned to each sensor. The raw data collected by each sensor is timestamped and synchronized to establish a unified time reference. The Kalman filter algorithm is used to perform time-domain interpolation and alignment of sensor data with different sampling frequencies; Principal component analysis is used to extract the core feature vectors of each sensor data to form a multidimensional feature matrix. Construct a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights; By combining multidimensional feature matrices, graph attention networks are used to learn the spatiotemporal dependencies between sensors and generate node embedding vectors. Establish a spatiotemporal propagation model to predict the spatiotemporal evolution trend of sensor data; A normal operation model is constructed based on a variational autoencoder, and the reconstruction error threshold is calculated. When an anomaly is detected, the adaptive reconfiguration mechanism of the sensor network is triggered; Based on the spatiotemporal evolution trend, anomaly type, and severity, dynamically adjust the sampling frequency and sensitivity parameters of key sensors; Construct a hierarchical decision tree to weight and fuse monitoring results from sensor level, region level, and system level; A confidence assessment is generated by using a Bayesian network inference mechanism, combining historical data and real-time monitoring results. Differentiated early warning signals are output based on the risk level, and monitoring reports are automatically generated.

[0018] A third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of a multi-source sensor monitoring method.

[0019] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a multi-source sensor monitoring method.

[0020] The technical solution of this invention, a multi-source sensor monitoring method, improves monitoring accuracy compared to traditional independent processing methods by modeling the spatiotemporal correlation between sensors through graph neural networks. A multi-dimensional feature fusion mechanism effectively reduces the impact of single sensor failures on overall monitoring performance, while spatiotemporal alignment technology eliminates data latency and inconsistency issues between different sensors. An adaptive reconfiguration mechanism dynamically adjusts sensor parameters according to environmental changes and monitoring needs. Anomaly detection and fault diagnosis capabilities are significantly enhanced, and a redundancy mechanism ensures the system can still operate normally even with partial sensor failures. An intelligent scheduling algorithm reduces the overall energy consumption of the sensor network, a hierarchical decision-making mechanism reduces unnecessary data transmission and computational overhead, and an on-demand activation strategy extends the lifespan of the sensor network. Multi-granularity fusion decision-making provides more accurate and timely early warning information, a Bayesian inference mechanism quantifies the uncertainty of prediction results, and an automatic report generation function enhances the practicality and operability of the monitoring system. Attached Figure Description

[0021] Figure 1 This is a flowchart of a multi-source sensor monitoring method provided in one embodiment of the present invention; Figure 2 This is a schematic block diagram of a multi-source sensor monitoring device provided in one embodiment of the present invention. Detailed Implementation

[0022] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0024] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0025] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0026] The following reference Figures 1 to 2 This invention describes a multi-source sensor monitoring method, apparatus, electronic device, and storage medium provided according to some embodiments of the present invention.

[0027] like Figure 1 As shown, one embodiment of the present invention provides a multi-source sensor monitoring method, including: Based on the physical characteristics and environmental parameters of the monitored target, a heterogeneous sensor network, including optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, is dynamically configured. In this step, the environmental parameters are various environmental-related data of the environment in which the monitoring target is located (e.g., the environment within a preset radius centered on the monitoring target).

[0028] Based on the spatial distribution of sensors, a topological relationship diagram of sensor nodes is established, and a unique spatiotemporal identifier is assigned to each sensor. The raw data collected by each sensor is timestamped and synchronized to establish a unified time reference. The Kalman filter algorithm is used to perform time-domain interpolation and alignment of sensor data with different sampling frequencies; Principal component analysis (PCA) is used to extract the core feature vectors of each sensor's data, forming a multidimensional feature matrix; Construct a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights; By combining multidimensional feature matrices, a graph attention network (GAT) is used to learn the spatiotemporal dependencies between sensors and generate node embedding vectors. Establish a spatiotemporal propagation model to predict the spatiotemporal evolution trend of sensor data; The specific implementation steps of this step are as follows: Based on the node embedding vector of the graph neural network, construct the spatiotemporal state vector of each sensor at time t. This vector integrates the current observation value, historical trend information, and spatial neighborhood characteristics of the sensor; establish a spatiotemporal state transition matrix to describe the propagation path and speed of information and state changes in the sensor network; construct a diffusion propagation model according to the physical characteristics of the monitored object to simulate the propagation law of physical quantities such as temperature, humidity, and gas concentration in space; establish a cascade propagation model to describe the propagation process of abnormal events or state changes in the sensor network, including propagation delay, attenuation coefficient, and impact range; and train a Long Short-Term Memory (LSTM) network using historical spatiotemporal data to learn the sensing... The study analyzes the time-series patterns and periodic variations of sensor data; combines spatial neighborhood information to train a graph convolutional recurrent neural network (GC-RNN) to capture the spatiotemporal correlations and propagation characteristics between sensors; based on the current sensor observation data and spatiotemporal state vector, it predicts the evolution of sensor data at multiple future time steps using a trained spatiotemporal propagation model; calculates the uncertainty interval of the prediction results, provides a confidence assessment for each prediction value, and identifies sensor nodes with high prediction risk; compares and verifies the prediction results with subsequent actual observation data, calculating prediction error and accuracy indicators; based on the verification results, it uses an online learning algorithm to dynamically update the parameters of the spatiotemporal propagation model, improving the model's prediction accuracy and adaptability.

[0029] A normal operation model is constructed based on a variational autoencoder (VAE), and the reconstruction error threshold is calculated. When an anomaly is detected, the adaptive reconfiguration mechanism of the sensor network is triggered; Based on the spatiotemporal evolution trend, anomaly type, and severity, dynamically adjust the sampling frequency and sensitivity parameters of key sensors; Specifically, this step involves: identifying the key sensor nodes most sensitive to and impacted by the current anomaly based on the prediction results of the spatiotemporal propagation model and the severity score of anomaly detection; analyzing the optimal sampling frequency requirement and sensitivity threshold range for each key sensor according to the physical characteristics of the anomaly type (such as temperature anomaly, vibration anomaly, gas leakage, etc.), and establishing a mapping table between anomaly type and sensor parameters; comprehensively considering the rate of change of spatiotemporal evolution trend, anomaly severity level, and current performance status of the sensor, calculating the target sampling frequency and sensitivity parameters for each key sensor; when multiple anomalies occur simultaneously or sensor resources are limited, initiating a parameter conflict resolution mechanism to prioritize the sensor parameter configuration for high-risk areas and key monitoring points; using the sensor network communication protocol, distributing the calculated new parameters to the corresponding key sensor nodes in real time, and recording the timestamp and adjustment range of the parameters; continuously monitoring the sensor performance and monitoring effect after parameter adjustment, and triggering a parameter rollback mechanism or further fine-tuning when the adjustment effect is found to be unsatisfactory to ensure the stability of the overall monitoring quality of the system.

[0030] Construct a hierarchical decision tree to weight and fuse monitoring results from sensor level, region level, and system level; A confidence assessment is generated by using a Bayesian network inference mechanism, combining historical data and real-time monitoring results. Differentiated early warning signals are output based on the risk level, and monitoring reports are automatically generated.

[0031] In this embodiment, the multidimensional feature matrix is ​​the core data structure of the entire technical solution. It can be used as input features for graph neural network nodes to calculate the similarity and correlation between nodes and support the learning of spatiotemporal dependencies; it can also be used as input data for variational autoencoders to build normal working models and support the identification of abnormal patterns; and it can also be used as evidence input for Bayesian network inference to support the generation of multi-level decisions.

[0032] The technical solution adopted in this embodiment improves monitoring accuracy by modeling the spatiotemporal correlation between sensors using graph neural networks, compared to traditional independent processing methods. The multi-dimensional feature fusion mechanism effectively reduces the impact of single sensor failures on overall monitoring performance, while spatiotemporal alignment technology eliminates data latency and inconsistency issues between different sensors. The adaptive reconfiguration mechanism dynamically adjusts sensor parameters according to environmental changes and monitoring needs. Anomaly detection and fault diagnosis capabilities are significantly enhanced, and a redundancy mechanism ensures the system can still operate normally even with partial sensor failures. The intelligent scheduling algorithm reduces the overall energy consumption of the sensor network, the hierarchical decision-making mechanism reduces unnecessary data transmission and computational overhead, and the on-demand activation strategy extends the lifespan of the sensor network. Multi-granularity fusion decision-making provides more accurate and timely early warning information, the Bayesian inference mechanism quantifies the uncertainty of prediction results, and the automatic report generation function enhances the practicality and operability of the monitoring system.

[0033] In some possible embodiments of the present invention, the dynamic configuration includes the step of a heterogeneous sensor network comprising optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, specifically including: The optimal sensor deployment density is calculated based on the geometric features of the monitoring area corresponding to the monitoring target and the sensing range of the sensor. Using an adaptive coverage optimization algorithm, the influence weight of sensor node m is calculated according to the following formula:

[0034] in: The normalized influence weight of sensor node m is in the range [0,1]. This represents the total number of key monitoring points within the monitoring area. Let m be the Euclidean distance (in meters) from sensor node m to monitoring point p. This is the distance decay index, with a value range of [1.5, 3.0], used to adjust the degree of decay of the influence of distance; Let m be the response delay time (in seconds) of sensor m to monitoring point p. This is the time sensitivity coefficient, with a value range of [0.1, 1.0], used to balance the importance of distance and time factors; This represents the total number of sensor nodes in the network. This is the performance correction factor for sensor m, determined based on parameters such as sensor accuracy and stability.

[0035] This embodiment takes into account both spatial distance attenuation and time response characteristics, and achieves accurate quantification of the influence of sensor nodes, thereby improving monitoring coverage efficiency compared to traditional uniform deployment schemes.

[0036] In some possible embodiments of the present invention, the multi-dimensional feature fusion in the multi-dimensional feature matrix formed by extracting the core feature vectors of each sensor data through principal component analysis adopts a spatiotemporal weighted fusion algorithm: For the feature vector of sensor s at time t, an adaptive weight allocation mechanism is established; The fused feature vector is calculated using the following formula:

[0037]

[0038] in: Let be the fused feature vector at time t; The total number of sensors participating in the fusion; Let be the dynamic weighting coefficient of sensor s at time t; Let be the reliability vector of sensor s, which is dynamically updated based on historical performance; This represents the Hadamard product (element-level multiplication). Let be the feature transformation matrix of sensor s; Let be the original feature vector of sensor s at time t; Let be the bias vector of sensor s; The basic weights for the importance of sensor s; The basic weights for the importance of sensor r; Let be the measurement uncertainty of sensor s at time t; Let be the measurement uncertainty of sensor r at time t; Let be the health status factor of sensor s, ranging from [0,1].

[0039] In this embodiment, the weights are dynamically adjusted based on the real-time reliability and measurement uncertainty of the sensors, which effectively suppresses the interference of low-quality data and improves the fusion accuracy compared with the static weighting method.

[0040] In some possible embodiments of the present invention, the step of constructing a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights includes: A multi-level spatiotemporal graph structure is constructed, and the attention weights between nodes are calculated using the following formula:

[0041]

[0042] in: Let be the attention weight of node u to node v in layer l; This is the attention scaling factor, used to control the sharpness of the softmax function; This is a learnable attention parameter vector; Here is the weight matrix for the l-th layer; Let u be the hidden state vector of node u in layer l; Let w be the hidden state vector of node w at level l, where w is the node index variable, representing one of the neighboring nodes of node u. This represents a vector concatenation operation; The enhanced edge feature vector between nodes u and v; Let be the set of neighbors of node u; These are the spatial location vectors of nodes u and v, respectively; This is the decay parameter for spatial correlation; Let be the weight coefficients of the edge features, satisfying ; The Pearson correlation coefficient for the historical time series of nodes u and v; The physical coupling strength between nodes u and v; In the context of a graph attention network: u and v are two specific nodes currently under consideration, and w is the iteration variable when traversing all neighbors of node u. Denotes the set of neighbors of node u; in In the calculation formula, the summation in the denominator This means summing the attention weights of node u over all its neighboring nodes w. Therefore, the complete understanding is: the numerator calculates the original attention score of node u to node v, and the denominator calculates the sum of the attention scores of node u to all its neighboring nodes (including v and other neighboring nodes w). This normalization operation ensures that the sum of the attention weights of node u to all its neighbors is 1.

[0043] In this embodiment, spatial proximity, temporal correlation and physical coupling are considered simultaneously, which improves the ability to capture spatiotemporal dependencies compared to standard graph attention networks.

[0044] In some possible embodiments of the present invention, the step of constructing a normal operating model based on a variational autoencoder and calculating a reconstruction error threshold includes: A hierarchical variational autoencoder is constructed, and the outlier score is calculated using the following formula:

[0045]

[0046]

[0047] in: Let n be the overall anomaly score of sensor n at time t; These are the weighting coefficients for different anomaly detection dimensions; To reconstruct the loss function; Let be the vector of observations of sensor n at time t; This is the reconstructed value vector of the variational autoencoder; This is the L1 regularization coefficient, used to improve the sparsity of the model; For latent variables KL divergence with the prior distribution; The score for continuity anomalies in the time series. The length of the time window; This is the time gradient sensitivity coefficient; This represents the time derivative of the observation sequence.

[0048] The multi-scale anomaly detection model in this embodiment considers reconstruction error, distribution deviation, and temporal continuity simultaneously. Compared with single-dimensional anomaly detection methods, it improves detection accuracy and reduces false alarm rate.

[0049] In some possible embodiments of the present invention, the step of generating a confidence assessment by employing a Bayesian network inference mechanism, combining historical data and real-time monitoring results, includes: System-level decision confidence is calculated using the following formula:

[0050]

[0051]

[0052] in: The system-level decision confidence level, ranging from [0,1]. Total number of regional groups; Let g be the set of decision categories for the g-th region; Evidence for a given region g Make an assumption The posterior probability; The importance weight for decision category c; The influence index of region g; This serves as a correction factor for the overall system reliability. Assumption Observed evidence The likelihood probability; Assumption The prior probability; The set of all possible hypotheses; For decision category c, the time-varying confidence adjustment factor is used. The total number of network layers; This is the set of sensor nodes in layer l; Assess the reliability of sensor node n; This represents the number of sensor nodes in the l-th layer. represents the weight coefficient of the l-th layer.

[0053] The dynamic Bayesian inference model in this embodiment achieves precise quantification of decision confidence through multi-layer evidence fusion and reliability correction. Compared with the traditional voting mechanism, it improves decision accuracy and provides a quantitative assessment of decision uncertainty.

[0054] In some possible embodiments of the present invention, the step of constructing a hierarchical decision tree and weightedly fusing the monitoring results at the sensor level, region level, and system level includes: Based on the multidimensional feature matrix and anomaly detection results, local decision results and confidence scores are generated for each sensor node; Based on the sensor’s historical performance and current health status, assign initial weight coefficients to each sensor-level decision. Based on the spatial distribution of sensors and the division of monitoring areas, sensors with geographical proximity are grouped to form monitoring areas; The sensor-level decisions in each region are weighted and averaged to generate regional-level decision results, and the consistency index of decisions within the region is calculated. Collect all regional-level decision-making results and assign system-level weights to each region based on its importance level and decision consistency index. Multiple regional-level decisions are merged into a system-level final decision using a weighted fusion algorithm, and a decision confidence assessment is generated. When there are conflicts in the decision-making results at different levels or in different regions, the conflict resolution mechanism is activated, and the decision-making result with higher confidence is adopted first. Analyze the root causes of conflict decisions, identify the sensors or areas that led to the conflict, and mark them as objects that need further verification. Based on system-level decision-making results, and according to preset risk level thresholds, corresponding early warning levels and handling recommendations are generated. The final decision results are fed back to the adaptive tuning module of the sensor network to optimize subsequent monitoring strategies and parameter configurations.

[0055] Please see Figure 2 Another embodiment of the present invention provides a multi-source sensor monitoring device for performing a multi-source sensor monitoring method, comprising: an acquisition module, a processing module, and an execution module; The processing module is configured as follows: Based on the physical characteristics and environmental parameters of the monitored target, a heterogeneous sensor network, including optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, is dynamically configured. Based on the spatial distribution of sensors, a topological relationship diagram of sensor nodes is established, and a unique spatiotemporal identifier is assigned to each sensor. The raw data collected by each sensor is timestamped and synchronized to establish a unified time reference. The Kalman filter algorithm is used to perform time-domain interpolation and alignment of sensor data with different sampling frequencies; Principal component analysis is used to extract the core feature vectors of each sensor data to form a multidimensional feature matrix. Construct a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights; By combining multidimensional feature matrices, graph attention networks are used to learn the spatiotemporal dependencies between sensors and generate node embedding vectors. Establish a spatiotemporal propagation model to predict the spatiotemporal evolution trend of sensor data; A normal operation model is constructed based on a variational autoencoder, and the reconstruction error threshold is calculated. When an anomaly is detected, the adaptive reconfiguration mechanism of the sensor network is triggered; Based on the spatiotemporal evolution trend, anomaly type, and severity, dynamically adjust the sampling frequency and sensitivity parameters of key sensors; Construct a hierarchical decision tree to weight and fuse monitoring results from sensor level, region level, and system level; A confidence assessment is generated by using a Bayesian network inference mechanism, combining historical data and real-time monitoring results. Differentiated early warning signals are output based on the risk level, and monitoring reports are automatically generated.

[0056] It should be known that, Figure 2 The block diagram of the multi-source sensor monitoring device shown is for illustrative purposes only, and the number of modules shown does not limit the scope of protection of this invention. The multi-source sensor monitoring device provided in this embodiment can be used to execute various embodiments of the corresponding multi-source sensor monitoring method. For specific implementation details, please refer to the descriptions of the respective method embodiments, which will not be repeated here.

[0057] In another embodiment of the present invention, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of a multi-source sensor monitoring method.

[0058] In another embodiment of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of a multi-source sensor monitoring method.

[0059] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0060] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0061] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0062] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0063] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0064] If the integrated units described above are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0065] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0066] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

[0067] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can easily conceive of variations or substitutions without departing from the spirit and scope of the present invention, and various modifications and alterations can be made, including combinations of the different functions and implementation steps described above, as well as software and hardware implementation methods, all of which are within the protection scope of the present invention.

Claims

1. A multi-source sensor monitoring method, characterized in that, include: Based on the physical characteristics and environmental parameters of the monitored target, a heterogeneous sensor network, including optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, is dynamically configured. Based on the spatial distribution of sensors, a topological relationship diagram of sensor nodes is established, and a unique spatiotemporal identifier is assigned to each sensor. The raw data collected by each sensor is timestamped and synchronized to establish a unified time reference. The Kalman filter algorithm is used to perform time-domain interpolation and alignment of sensor data with different sampling frequencies; Principal component analysis is used to extract the core feature vectors of each sensor data to form a multidimensional feature matrix. Construct a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights; By combining multidimensional feature matrices, graph attention networks are used to learn the spatiotemporal dependencies between sensors and generate node embedding vectors. Establish a spatiotemporal propagation model to predict the spatiotemporal evolution trend of sensor data; A normal operation model is constructed based on a variational autoencoder, and the reconstruction error threshold is calculated. When an anomaly is detected, the adaptive reconfiguration mechanism of the sensor network is triggered; Based on the spatiotemporal evolution trend, anomaly type, and severity, dynamically adjust the sampling frequency and sensitivity parameters of key sensors; Construct a hierarchical decision tree to weight and fuse monitoring results from sensor level, region level, and system level; A confidence assessment is generated by using a Bayesian network inference mechanism, combining historical data and real-time monitoring results. Differentiated early warning signals are output based on the risk level, and monitoring reports are automatically generated.

2. The multi-source sensor monitoring method according to claim 1, characterized in that, The dynamic configuration includes the steps of creating a heterogeneous sensor network comprising optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, specifically including: The optimal sensor deployment density is calculated based on the geometric features of the monitoring area corresponding to the monitoring target and the sensing range of the sensor. Using an adaptive coverage optimization algorithm, the influence weight of sensor node m is calculated according to the following formula: 。 3. The multi-source sensor monitoring method according to claim 2, characterized in that, The process of extracting core feature vectors from each sensor data through principal component analysis to form a multidimensional feature matrix employs a spatiotemporal weighted fusion algorithm for multidimensional feature fusion. For the feature vector of sensor s at time t, an adaptive weight allocation mechanism is established; The fused feature vector is calculated using the following formula: 。 4. The multi-source sensor monitoring method according to claim 3, characterized in that, The steps for constructing a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights include: A multi-level spatiotemporal graph structure is constructed, and the attention weights between nodes are calculated using the following formula: 。 5. The multi-source sensor monitoring method according to claim 4, characterized in that, The steps of constructing a normal working model based on a variational autoencoder and calculating the reconstruction error threshold include: A hierarchical variational autoencoder is constructed, and the outlier score is calculated using the following formula: 。 6. The multi-source sensor monitoring method according to claim 5, characterized in that, The steps for generating confidence assessments using a Bayesian network inference mechanism, combining historical data and real-time monitoring results, include: System-level decision confidence is calculated using the following formula: 。 7. The multi-source sensor monitoring method according to claim 6, characterized in that, The step of constructing a hierarchical decision tree and weightedly fusing monitoring results at the sensor, regional, and system levels includes: Based on the multidimensional feature matrix and anomaly detection results, local decision results and confidence scores are generated for each sensor node; Based on the sensor’s historical performance and current health status, assign initial weight coefficients to each sensor-level decision. Based on the spatial distribution of sensors and the division of monitoring areas, sensors with geographical proximity are grouped to form monitoring areas; The sensor-level decisions in each region are weighted and averaged to generate regional-level decision results, and the consistency index of decisions within the region is calculated. Collect all regional-level decision-making results and assign system-level weights to each region based on its importance level and decision consistency index. Multiple regional-level decisions are merged into a system-level final decision using a weighted fusion algorithm, and a decision confidence assessment is generated. When there are conflicts in the decision-making results at different levels or in different regions, the conflict resolution mechanism is activated, and the decision-making result with higher confidence is adopted first. Analyze the root causes of conflict decisions, identify the sensors or areas that led to the conflict, and mark them as objects that need further verification. Based on system-level decision-making results, and according to preset risk level thresholds, corresponding early warning levels and handling recommendations are generated. The final decision results are fed back to the adaptive tuning module of the sensor network to optimize subsequent monitoring strategies and parameter configurations.

8. A multi-source sensor monitoring device, used to perform the multi-source sensor monitoring method as described in any one of claims 1 to 7, characterized in that, include: The module consists of an acquisition module, a processing module, and an execution module. The processing module is configured as follows: Based on the physical characteristics and environmental parameters of the monitored target, a heterogeneous sensor network, including optical sensors, acoustic sensors, vibration sensors, temperature and humidity sensors, and gas sensors, is dynamically configured. Based on the spatial distribution of sensors, a topological relationship diagram of sensor nodes is established, and a unique spatiotemporal identifier is assigned to each sensor. The raw data collected by each sensor is timestamped and synchronized to establish a unified time reference. The Kalman filter algorithm is used to perform time-domain interpolation and alignment of sensor data with different sampling frequencies; Principal component analysis is used to extract the core feature vectors of each sensor data to form a multidimensional feature matrix. Construct a dynamic graph structure with sensor nodes as vertices and physical correlation as edge weights; By combining multidimensional feature matrices, graph attention networks are used to learn the spatiotemporal dependencies between sensors and generate node embedding vectors. Establish a spatiotemporal propagation model to predict the spatiotemporal evolution trend of sensor data; A normal operation model is constructed based on a variational autoencoder, and the reconstruction error threshold is calculated. When an anomaly is detected, the adaptive reconfiguration mechanism of the sensor network is triggered; Based on the spatiotemporal evolution trend, anomaly type, and severity, dynamically adjust the sampling frequency and sensitivity parameters of key sensors; Construct a hierarchical decision tree to weight and fuse monitoring results from sensor level, region level, and system level; A confidence assessment is generated by using a Bayesian network inference mechanism, combining historical data and real-time monitoring results. Differentiated early warning signals are output based on the risk level, and monitoring reports are automatically generated.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-source sensor monitoring method 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 multi-source sensor monitoring method as described in any one of claims 1 to 7.

Images