Online monitoring method and device for intelligent Internet of Things equipment

Abstract

The invention relates to the technical field of online monitoring of Internet of Things equipment, in particular to an online monitoring method and device for intelligent Internet of Things equipment, and the method comprises the steps: obtaining a state parameter vector of the to-be-monitored intelligent Internet of Things equipment; based on the state parameter vector, mapping the state parameter vector to a high-dimensional topological space to obtain topological features; extracting group theory features according to the topological features; based on the group theory features, constructing a dynamic feature matrix; calculating chaotic features according to the dynamic feature matrix; and judging whether the to-be-monitored intelligent Internet of Things equipment is in an abnormal state or not based on the chaotic features, mapping the equipment state parameters to a high-dimensional topological space, and combining group theory feature extraction, thereby greatly enhancing the processing capability of nonlinear and high-dimensional data. According to the innovative feature extraction method, tiny abnormal modes which are difficult to find through a traditional method can be captured, and the accuracy and sensitivity of abnormal detection are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of online monitoring of Internet of Things devices, and more specifically, to an online monitoring method and device for intelligent Internet of Things devices. Background Art

[0002] With the rapid development of IoT technology, smart IoT devices have been widely used in industrial production, smart cities, smart homes, etc. However, with the continuous expansion of the scale and complexity of IoT systems, how to effectively monitor the operating status of these devices and promptly detect and handle abnormal situations has become a technical problem that needs to be solved urgently.

[0003] Existing intelligent IoT device monitoring methods mainly include simple threshold-based methods, statistical analysis methods, and machine learning methods. Although simple threshold-based methods are simple to implement and require little computation, they are difficult to deal with complex abnormal situations, especially when processing multidimensional data. Statistical analysis methods can improve the accuracy of anomaly detection to a certain extent, but they often require strong assumptions about data distribution, which may deviate greatly from the actual situation in practical applications. Although machine learning methods have strong nonlinear modeling capabilities, they usually require a large amount of labeled data for training, and the interpretability of the model is poor, making it difficult to cope with the rapidly changing IoT environment.

[0004] In addition, existing methods generally have the following problems: First, most methods only focus on the status of a single device, ignoring the mutual influence and correlation between devices, making it difficult to effectively identify system-level anomalies. Second, existing methods usually use static anomaly detection models, which are difficult to adapt to the dynamic changes of IoT devices and environments. Furthermore, when faced with high-dimensional and nonlinear IoT data, existing methods often find it difficult to effectively extract key features, resulting in low detection accuracy or high computational complexity. Finally, most methods lack the ability to provide early warning of anomalies and can only detect them after they occur, which cannot meet some application scenarios with high real-time requirements.

[0005] In view of the above problems, there is an urgent need for an online monitoring method for intelligent IoT devices that can effectively process high-dimensional nonlinear data, has dynamic adaptability, can perform system-level anomaly analysis, and has early warning function. Summary of the invention

[0006] The present invention aims to solve the above technical problems existing in the existing online monitoring methods for intelligent Internet of Things devices. Specifically, the present invention provides an online monitoring method and device for intelligent Internet of Things devices based on high-dimensional topological space mapping, group theory feature extraction and chaos theory, which can effectively process high-dimensional nonlinear data, has strong dynamic adaptability and early warning function, and can perform system-level abnormality analysis.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions: An online monitoring method for an intelligent Internet of Things device, comprising: The acquisition steps include: Obtain the state parameter vector of the smart IoT device to be monitored; Processing steps include: Based on the state parameter vector, mapping it to a high-dimensional topological space to obtain a topological feature; Extracting group theory features according to the topological features; Based on the group theory characteristics, a dynamic characteristic matrix is ​​constructed; Calculating chaotic characteristics according to the dynamic characteristic matrix; Output steps include: Based on the chaotic characteristics, it is determined whether the intelligent Internet of Things device to be monitored is in an abnormal state.

[0008] Preferably, mapping the state parameter vector to a high-dimensional topological space to obtain a topological feature specifically includes: Obtaining a preset basis function set and a weight coefficient set; The topological feature is calculated based on the basis function set, the weight coefficient set and the state parameter vector.

[0009] Preferably, the basis function set includes Bessel functions and / or Legendre polynomials.

[0010] Preferably, the extracting group theory features specifically includes: Get the preset group action operator set; The group theory characteristics are calculated based on the group action operator set and the topological characteristics.

[0011] Preferably, the group action operator set includes a rotation operator, a translation operator and / or a scale transformation operator.

[0012] Preferably, the constructing of the dynamic feature matrix specifically includes: Get multiple group theory features within a preset time window; calculating first-order differences and second-order differences of the plurality of group theory characteristics; The dynamic feature matrix is ​​constructed based on the plurality of group theory features, the first-order differences and the second-order differences.

[0013] Preferably, the calculation of chaotic characteristics specifically includes: Based on the dynamic characteristic matrix, the Lyapunov exponent is calculated as the chaotic characteristic.

[0014] Preferably, the determining whether the smart IoT device to be monitored is in an abnormal state specifically includes: Obtain the preset normal state Lyapunov index baseline value and abnormal threshold; Calculating the absolute value of the difference between the chaotic characteristic and the Lyapunov exponent reference value of the normal state; Determine whether the absolute value of the difference is greater than the abnormal threshold; When the absolute value of the difference is greater than the abnormal threshold, it is determined that the smart Internet of Things device to be monitored is in an abnormal state.

[0015] As a preference, it also includes: Obtain state parameter vectors of multiple smart IoT devices to be monitored; Based on the state parameter vectors of the plurality of intelligent IoT devices to be monitored, construct a device-associated network topology structure; Combined with the device-associated network topology and the abnormal state judgment results of each intelligent IoT device to be monitored, a comprehensive abnormality analysis is performed.

[0016] An online monitoring system for intelligent Internet of Things devices that implements the method includes: An acquisition module is used to obtain a state parameter vector of the smart IoT device to be monitored; Processing modules for: Based on the state parameter vector, mapping it to a high-dimensional topological space to obtain a topological feature; Extracting group theory features according to the topological features; Based on the group theory characteristics, a dynamic characteristic matrix is ​​constructed; Calculating chaotic characteristics according to the dynamic characteristic matrix; The output module is used to determine whether the intelligent Internet of Things device to be monitored is in an abnormal state based on the chaotic characteristics.

[0017] Compared with the prior art, the beneficial effects of the present invention are embodied in the following aspects: The beneficial effects of the present invention are mainly reflected in the following aspects: First, the present invention greatly enhances the processing capability of nonlinear and high-dimensional data by mapping the device state parameters to a high-dimensional topological space and combining it with group theory feature extraction. This innovative feature extraction method enables the present invention to capture subtle abnormal patterns that are difficult to detect with traditional methods, significantly improving the accuracy and sensitivity of anomaly detection.

[0018] Secondly, the present invention introduces dynamic characteristic matrix and chaos theory, which makes the monitoring method have beneficial dynamic adaptability. By analyzing the changes of Lyapunov exponents, the present method can effectively track the dynamic evolution of equipment status, not only detect sudden abnormalities in time, but also predict potential abnormal trends, providing strong support for system maintenance and fault prevention.

[0019] Furthermore, the present invention realizes the effective identification of system-level anomalies by constructing a device-associated network topology. This networked analysis method breaks through the limitations of traditional single-device monitoring, can fully grasp the operating status of the entire Internet of Things system, and provides new ideas and tools for the management of large-scale and complex Internet of Things systems.

[0020] In addition, the method of the present invention also performs well in terms of computational efficiency. Although it involves complex mathematical theories, the core algorithm can be highly parallelized and is suitable for implementation in distributed systems. This enables the method to meet the needs of real-time monitoring of large-scale Internet of Things systems, while ensuring the quality of monitoring and the response speed of the system.

[0021] Finally, the method of the present invention has strong versatility and scalability. By adjusting the basis function, group action operator and network topology construction method, the method can flexibly adapt to different types of intelligent Internet of Things devices and application scenarios. This flexibility makes the method have broad application prospects in many fields such as industrial Internet of Things, smart cities, and smart homes.

[0022] In summary, the online monitoring method and device for intelligent IoT devices provided by the present invention effectively solve the problems existing in the prior art by innovatively combining advanced mathematical theory and IoT technology, and provide an efficient, accurate and robust solution for the status monitoring and anomaly detection of intelligent IoT devices. This method can not only improve the monitoring accuracy of a single device, but also realize system-level anomaly analysis and early warning, which is of great significance to improving the overall reliability and security of the IoT system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The figure is a flow chart of the method of the present invention.

[0024] Figure 2 It is a logic block diagram of the processing module of the present invention.

[0025] Figure 3 It is a flow chart of the high-dimensional topological space mapping of the present invention.

[0026] Figure 4 It is a flow chart of the group theory feature extraction of the present invention.

[0027] Figure 5Flowchart for the construction of the dynamic feature matrix of the present invention.

[0028] Figure 6 It is a logic block diagram of the output module of the present invention. DETAILED DESCRIPTION

[0029] like Figure 1-6 As shown, the present invention discloses an online monitoring method and device for intelligent Internet of Things devices. The method realizes accurate monitoring and abnormality detection of the operating status of intelligent Internet of Things devices by innovatively combining high-dimensional topological space mapping, group theory feature extraction and chaos theory.

[0030] Specifically, the method of the present invention includes an acquisition step, a processing step and an output step. In the acquisition step, the state parameter vector of the smart IoT device to be monitored is first obtained. The state parameter vector here may include various operating indicators of the device, such as CPU usage, memory usage, network traffic, temperature, etc.

[0031] The following processing steps are the core innovation of the present invention. First, based on the acquired state parameter vector, it is mapped to a high-dimensional topological space to obtain topological features. This step can be achieved by the following formula:

[0032] in, is the topological feature, is the state parameter vector, is the mapping function, is the weight coefficient, is the basis function, is the number of basis functions.

[0033] High-dimensional topological space mapping refers to mapping low-dimensional data (such as the state parameter vector of a smart IoT device) into a higher-dimensional space. This mapping can be achieved through techniques such as kernel methods and manifold learning. By mapping to a high-dimensional space, the complex nonlinear characteristics of the device state can be better captured. For example, in an industrial environment, the operating state of a device may be affected by a variety of factors, and the relationship between these factors may be nonlinear. High-dimensional mapping can help us describe these complex interactions more accurately. In a high-dimensional space, data points in different states are more easily distinguished, thereby improving the accuracy of anomaly detection. For example, for a temperature sensor, even if the temperature change is small, through high-dimensional mapping, changes in other related parameters may be found, thereby detecting potential problems earlier.

[0034] The advantage of this high-dimensional topological space mapping is that it can capture the complex nonlinear characteristics of the device state, thereby improving the accuracy of subsequent anomaly detection.

[0035] After obtaining the topological features, this method further extracts group theory features. This step can be achieved by the following formula:

[0036] in, For group theory characteristics, is the group action operator, is the number of group action operators.

[0037] Group theory is a branch of mathematics that studies symmetries and transformations. In the present invention, group theory features are extracted through group action operators. Group theory feature extraction can obtain a more robust description of the device state, which is not easily affected by noise and small disturbances. For example, in a smart home environment, environmental noise may interfere with the readings of sensors, but through group theory feature extraction, these noises can be filtered out to obtain a more stable feature representation.

[0038] Group action operators can extract the invariant features of the device state under different transformations. For example, the rotation operator can extract the invariant features of the device state under rotation transformation, which is particularly effective for detecting certain types of periodic anomalies. The translation operator can help identify the overall offset of the device state, while the scale transformation operator can capture the amplitude characteristics of the state change.

[0039] The extraction of group theory features can obtain a more robust description of the device state, which is less susceptible to noise and small disturbances.

[0040] Next, based on the extracted group theory features, this method constructs a dynamic feature matrix. This step takes into account the time evolution characteristics of the device state and can better capture the dynamic characteristics of the abnormal state.

[0041] Finally, the chaotic characteristics are calculated based on the dynamic characteristic matrix.

[0042] In the output step, the method determines whether the intelligent IoT device to be monitored is in an abnormal state based on the calculated chaotic characteristics. This determination can be achieved by comparing the current Lyapunov exponent with a preset normal state reference value.

[0043] In one embodiment of the present invention, the mapping process of the state parameter vector to the high-dimensional topological space is further refined. Specifically, a preset basis function set and a weight coefficient set are first obtained. The selection of the basis function set is crucial to the mapping effect. The present invention preferably uses a function that can well describe the characteristics of the nonlinear system. The weight coefficient set can be optimized by a machine learning algorithm to maximize the separability of different device states in the topological space.

[0044] Then, based on the obtained basis function set, weight coefficient set and state parameter vector, the topological features are calculated. This calculation process is the process of projecting the low-dimensional state parameter vector into the high-dimensional topological space, which can effectively extract the nonlinear features in the original data.

[0045] In one embodiment of the present invention, the specific selection of the basis function set is further limited. The present invention preferably uses Bessel functions and / or Legendre polynomials as basis functions. Bessel functions have good oscillation characteristics and are suitable for describing the state of periodically changing devices; while Legendre polynomials perform well in describing smooth nonlinear systems. The combined use of these two types of special functions can well adapt to the state characteristics of different types of smart IoT devices.

[0046] For example, for devices with strong periodic working modes (such as sensors that collect data at regular intervals), Bessel functions can be used as the main basis functions; while for devices with relatively slow state changes (such as servers that run for a long time), Legendre polynomials can be used more. In practical applications, the best combination of basis functions can be determined through methods such as cross-validation.

[0047] The method of the present invention achieves accurate characterization and anomaly detection of the operating status of intelligent IoT devices through this innovative high-dimensional topological space mapping, combined with group theory feature extraction and chaos theory. Compared with traditional methods, the present invention has the following significant advantages: 1. Stronger nonlinear feature extraction capability: Through high-dimensional topological space mapping, the complex nonlinear characteristics of device status can be effectively captured, improving the accuracy of anomaly detection.

[0048] 2. Better noise immunity: The group theory feature extraction process can obtain a more robust description of the device state, which is less susceptible to environmental noise and small disturbances.

[0049] 3. Dynamic characteristic analysis: By constructing a dynamic feature matrix and calculating chaotic features, this method can effectively capture the dynamic evolution characteristics of the equipment state, which is conducive to detecting gradual anomalies.

[0050] 4. Strong adaptability: By adjusting the basis function set and weight coefficients, this method can flexibly adapt to different types of smart IoT devices and application scenarios.

[0051] 5. High computational efficiency: Although it involves complex mathematical theories, the core computational steps of this method can be highly parallelized and is suitable for implementation in distributed systems.

[0052] In summary, the online monitoring method for intelligent Internet of Things devices provided by the present invention provides an efficient, accurate and robust solution for status monitoring and anomaly detection of intelligent Internet of Things devices by innovatively combining advanced mathematics theory and Internet of Things technology.

[0053] In one embodiment of the present invention, the method of the present invention adopts an innovative method when extracting group theory features. Specifically, this step first obtains a preset group action operator set. The selection of group action operators is crucial for extracting robust features. In a preferred embodiment of the present invention, the group action operator set may include but is not limited to rotation operators, translation operators, and scale transformation operators. These operators can effectively capture the invariant features of the device state under different transformations.

[0054] After obtaining the set of group action operators, this method calculates the group theory characteristics based on the operator set and the previously obtained topological characteristics. This calculation process can be expressed as:

[0055] in, Represents group theory characteristics, Representative A group action operator, is the topological feature obtained before, is the total number of group action operators.

[0056] In a specific embodiment of the present invention, the specific composition of the group action operator set is further clarified. The present invention preferably uses rotation operators, translation operators and / or scale transformation operators. The selection of these operators is based on their advantages in physical meaning and mathematical properties. For example, the rotation operator can extract the invariant features of the device state under the rotation transformation, which is particularly effective for detecting certain types of periodic anomalies. The translation operator can help identify the overall offset of the device state, while the scale transformation operator can capture the amplitude characteristics of the state change.

[0057] In a specific embodiment of the present invention, the most suitable group action operator combination can be selected according to different types of intelligent IoT devices. For example, for devices with strong directionality (such as directional antennas), more rotation operators can be used; while for devices that are easily affected by the external environment (such as temperature sensors), more scale transformation operators may need to be used to capture the amplitude characteristics of state changes.

[0058] In one embodiment of the present invention, a process for constructing a dynamic feature matrix is ​​described. The innovation of this step is that it not only considers the instantaneous characteristics of the device state, but also captures the trend of the state changing over time. Specifically, the method first obtains multiple group theory features within a preset time window. The size of the time window can be flexibly adjusted according to the specific application scenario, and 5 to 30 sampling periods can usually be selected.

[0059] Next, this method calculates the first-order and second-order differences of these group theory features. The first-order difference reflects the speed of change of the feature, while the second-order difference reflects the acceleration of the change. These two indicators are crucial for capturing the dynamic changes of the device state. The calculation formula is as follows: First-order difference:

[0060] Second order difference:

[0061] in, Indicates The group theory characteristics at a certain point in time.

[0062] Finally, based on the obtained multiple group theory features, the calculated first-order differences and second-order differences, a dynamic feature matrix is ​​constructed. This matrix can be expressed as:

[0063] Where N is the number of sampling points in the time window.

[0064] The dynamic feature matrix not only considers the instantaneous characteristics of the device state, but also captures the changing trend of the state over time. By constructing a dynamic feature matrix, the dynamic evolution characteristics of the device state can be effectively captured, which is conducive to detecting gradual anomalies. For example, in a smart city, the change of traffic flow is a gradual process. The dynamic feature matrix can predict the occurrence of traffic congestion earlier. The dynamic feature matrix not only contains the instantaneous characteristics of the device state, but also contains the change speed and acceleration information, which can more comprehensively describe the operating status of the device. This helps to perform more accurate anomaly detection and fault diagnosis in complex environments.

[0065] This method of constructing a dynamic feature matrix enables the present invention to simultaneously capture the instantaneous characteristics, change trends and acceleration of the device state, thereby being able to describe the operating state of the device more comprehensively and accurately.

[0066] In one embodiment of the present invention, a method for calculating chaotic characteristics is further described. The present invention innovatively introduces the Lyapunov index in chaos theory as a chaotic characteristic of the device state. The Lyapunov index is an important indicator for describing the sensitivity of a dynamic system and can effectively reflect the instability and complexity of the system.

[0067] Specifically, this method calculates the Lyapunov exponent as the chaotic feature based on the previously constructed dynamic feature matrix. The calculation formula is as follows:

[0068] in, is the Lyapunov exponent, is the dynamic system evolution function, is the dynamic feature matrix List.

[0069] The Lyapunov exponent in chaos theory is an important indicator for describing the sensitivity of dynamic systems and can reflect the instability and complexity of the system. The Lyapunov exponent can effectively capture the complex dynamic characteristics of the operating status of smart IoT devices. For example, on an industrial production line, some equipment may show a certain degree of irregularity during normal operation, and the real anomaly may be reflected in the sudden change of this irregularity. The Lyapunov exponent can well describe this change.

[0070] Traditional methods based on simple thresholds may not be able to detect some subtle anomalies, while the Lyapunov exponent can capture these subtle changes, thereby improving the sensitivity of anomaly detection. For example, in medical device monitoring, even small fluctuations in heart rate may indicate potential health problems, and the Lyapunov exponent can help doctors detect problems earlier.

[0071] In practical applications, the Lyapunov exponent can be approximated by numerical methods. For example, the Wolf algorithm or the Rosenstein algorithm can be used for calculation. These algorithms can effectively estimate the Lyapunov exponent on limited time series data.

[0072] By introducing the Lyapunov exponent as a chaotic feature, the method of the present invention can effectively capture the complex dynamic characteristics of the operating state of intelligent IoT devices. This is particularly effective for detecting some complex anomalies that are difficult to judge by simple thresholds. For example, some devices may show a certain degree of irregularity during normal operation, and the real anomaly may be reflected in the sudden change of this irregularity. The Lyapunov exponent can well characterize this change.

[0073] In general, the present invention provides a new method for online monitoring of intelligent IoT devices by innovatively combining high-dimensional topological space mapping, group theory feature extraction, dynamic feature matrix construction and chaotic feature calculation. This method can not only accurately capture the complex nonlinear characteristics of device states, but also effectively describe the dynamic evolution of states, thereby greatly improving the accuracy and reliability of anomaly detection. Especially when dealing with large-scale, highly complex IoT systems, the method of the present invention shows significant advantages and provides strong support for the further development and application of IoT technology.

[0074] In one embodiment of the present invention, an innovative method is proposed to determine whether the monitored smart IoT device is in an abnormal state. This judgment process makes full use of the chaotic characteristics calculated in the previous step, namely the Lyapunov exponent. Specifically, the process includes the following key steps: First, the method obtains a preset normal state Lyapunov index reference value and an abnormal threshold value. The normal state Lyapunov index reference value can be obtained by long-term observation of the device under normal operating conditions. Preferably, the average value of the Lyapunov index of multiple normal operating cycles can be used as the reference value. For example, for an industrial Internet of Things sensor, the average value of the Lyapunov index during 30 consecutive days of normal operation can be selected as the reference value.

[0075] The setting of the abnormal threshold requires a balance between detection sensitivity and false alarm rate. In a preferred embodiment of the present invention, the abnormal threshold can be set to 3 times the standard deviation of the Lyapunov exponent in the normal state. This setting is based on the three sigma principle in statistics, which can ensure a high detection rate while controlling the false alarm rate at a low level.

[0076] Next, the method calculates the absolute value of the difference between the currently obtained chaotic characteristics (i.e., Lyapunov index) and the Lyapunov index reference value of the normal state. This step can be expressed by the mathematical formula as follows:

[0077] in, represents the absolute value of the difference, is the currently calculated Lyapunov exponent, It is the preset normal state Lyapunov exponent reference value.

[0078] Then, the method determines whether the absolute value of the difference is greater than the preset abnormal threshold. This judgment process can be expressed as:

[0079] in, Indicates the preset abnormality threshold.

[0080] When the absolute value of the difference is greater than the abnormal threshold, this method determines that the monitored smart IoT device is in an abnormal state. This anomaly detection method based on the Lyapunov exponent has obvious advantages: it can not only detect significant changes in the device state, but also capture some subtle and potential anomalies that may be ignored in traditional methods based on simple thresholds.

[0081] The abnormal state judgment is based on the comparison between the Lyapunov exponent and the preset normal state benchmark value. By setting a reasonable abnormal threshold (such as the three sigma principle), the false alarm rate can be controlled at a low level while ensuring a high detection rate. For example, in the power system, a too high false alarm rate may lead to unnecessary downtime maintenance, while a too low detection rate may miss key problems. By optimizing the threshold setting, the best balance point can be found.

[0082] This method can achieve real-time monitoring and immediately send out an alarm when an abnormality is detected. For example, in a smart home system, if an abnormality is detected in a device, the system can automatically notify the user or take corresponding measures to ensure home safety.

[0083] In one embodiment of the present invention, the application scope of the present invention is further expanded from monitoring of a single device to collaborative monitoring of multiple devices. This innovative expansion enables the present invention to better adapt to complex IoT environments. Specifically, this step includes the following key processes: First, the method obtains the state parameter vectors of multiple smart IoT devices to be monitored. These devices may be distributed in different physical locations and perform different functions, but together they constitute a complex IoT system.

[0084] Next, based on the state parameter vectors of these devices, the method constructs the device-associated network topology. The innovation of this step is that it takes into account the mutual influence and association relationship between devices. The method of constructing the network topology can use a variety of mathematical tools, such as graph theory or complex network theory.

[0085] In a preferred embodiment, a weighted undirected graph can be used to represent the device association network topology. Each node in the graph represents a device, and the weight of the edge represents the degree of association between the devices. The degree of association can be determined by calculating the correlation coefficient or mutual information between the device state parameter vectors. This representation method is not only intuitive, but also can effectively capture the complex interactive relationship between devices.

[0086] Finally, this method combines the device-associated network topology and the abnormal state judgment results of each smart IoT device to be monitored to perform comprehensive abnormality analysis. The core idea of ​​this step is to analyze the abnormal state of a single device in the context of the entire network, so as to obtain more comprehensive and accurate abnormality diagnosis results.

[0087] For example, if a device is detected to be in an abnormal state, the method further analyzes the state of other devices directly connected to it. If the neighboring devices also show abnormalities, then this may indicate a larger system-level abnormality. Conversely, if the neighboring devices are all in a normal state, then the abnormality may be limited to the device itself.

[0088] This comprehensive analysis method based on network topology greatly improves the accuracy and interpretability of anomaly detection. It can not only detect anomalies of individual devices, but also identify possible system-level failures or abnormal patterns, providing more comprehensive and in-depth insights for the maintenance and optimization of IoT systems.

[0089] Multi-device collaborative monitoring considers the mutual influence and association between devices by constructing a device association network topology. A weighted undirected graph can be used to represent the device association network topology. Each node in the graph represents a device, and the weight of the edge represents the degree of association between devices. The degree of association can be determined by calculating the correlation coefficient or mutual information between device state parameter vectors.

[0090] By comprehensively analyzing the status of multiple devices, possible system-level failures or abnormal patterns can be identified. For example, in the Industrial Internet of Things, if multiple devices are abnormal at the same time, it may indicate that there is a problem with the entire production line. Through multi-device collaborative monitoring, problems can be discovered and resolved earlier.

[0091] The comprehensive analysis method based on network topology can not only detect anomalies of a single device, but also provide more comprehensive anomaly diagnosis results to help users better understand the root cause of the problem. For example, in a smart city, if multiple sensors in a certain area report anomalies at the same time, it may indicate that there are infrastructure problems in the area that require further investigation.

[0092] The present invention provides a new solution for the state monitoring and anomaly detection of intelligent IoT devices by innovatively combining high-dimensional topological space mapping, group theory feature extraction, dynamic feature analysis and chaos theory. This method can not only accurately capture the abnormal state of a single device, but also conduct comprehensive analysis in a complex multi-device environment, thus providing strong support for the management and optimization of the IoT system.

[0093] Finally, the present invention also discloses an online monitoring system for intelligent Internet of Things devices corresponding to the above method. The system comprises an acquisition module 1, a processing module 2 and an output module 3, and the functions of these modules correspond to the acquisition step, the processing step and the output step in the method respectively.

[0094] The acquisition module 1 is responsible for acquiring the state parameter vector of the intelligent IoT device to be monitored. In practical applications, this module may include various sensor interfaces and data preprocessing units to ensure the quality and consistency of the acquired data.

[0095] Processing module 2 is the core of the system. It performs a series of complex mathematical operations, including mapping the state parameter vector to a high-dimensional topological space, extracting group theory features, constructing a dynamic feature matrix, and calculating chaotic features. This module may require powerful computing power, and preferably can be implemented using a high-performance processor or a dedicated artificial intelligence chip.

[0096] The output module 3 is responsible for judging whether the monitored smart IoT device is in an abnormal state based on the chaotic features calculated by the processing module, and outputting the results in an appropriate form. This module may include various visualization tools and alarm mechanisms to facilitate users to quickly understand and respond to monitoring results.

[0097] It is worth noting that this system architecture has good scalability and flexibility. By adjusting the specific implementation of each module, it can easily adapt to different types of smart IoT devices and application scenarios. For example, for key equipment that requires real-time monitoring, the real-time performance of the processing module can be enhanced; and for equipment that requires long-term monitoring, the system's data storage and trend analysis capabilities can be enhanced.

[0098] In general, the online monitoring method of intelligent Internet of Things devices and the corresponding system proposed in the present invention provide a new solution for the state monitoring and anomaly detection of Internet of Things devices by innovatively combining high-dimensional topological space mapping, group theory feature extraction, dynamic feature analysis and chaos theory. This method can not only accurately capture the abnormal state of a single device, but also perform comprehensive analysis in a complex multi-device environment, thereby providing strong support for the management and optimization of the Internet of Things system. With the continuous development of Internet of Things technology and the continuous expansion of its application scope, the method and system of the present invention are expected to play an important role in multiple fields such as industrial Internet of Things, smart cities, and smart homes, and promote the development of Internet of Things technology in a more efficient and intelligent direction.

[0099] In order to verify the superiority of the online monitoring method for intelligent Internet of Things devices proposed in the present invention, a detailed experimental analysis will be carried out through a specific embodiment and a comparative example.

[0100] Example 1: This example adopts the method proposed in the present invention to perform online monitoring on an industrial Internet of Things system consisting of 100 temperature sensors. These sensors are distributed in different areas of a large chemical plant and are used to monitor temperature changes in various production links.

[0101] The specific implementation steps are as follows: 1. Get the state parameter vector of each sensor, including temperature reading, battery level, signal strength, etc.

[0102] 2. Use Bessel functions and Legendre polynomials as basis functions to map the state parameter vector to a 10-dimensional topological space.

[0103] 3. Use rotation operators and scaling operators to extract group theory features.

[0104] 4. Construct a dynamic feature matrix containing 30 sampling points.

[0105] 5. Calculate the Lyapunov exponent as a chaotic characteristic.

[0106] 6. Perform anomaly detection based on the pre-set normal state Lyapunov exponent baseline value and anomaly threshold.

[0107] 7. Build sensor network topology and conduct comprehensive anomaly analysis.

[0108] Comparative Example 1: The same system is monitored using a traditional simple threshold-based method. This method directly sets fixed upper and lower thresholds for temperature readings, and determines that an abnormality occurs when the threshold is exceeded.

[0109] The experimental method is as follows: During the 30-day experiment, different types of abnormal situations were artificially introduced, including: 1. Single sensor failure (e.g. abnormal readings due to battery exhaustion); 2. Abnormal temperature in a local area (such as abnormal temperature rise in a production unit); 3. System-level anomalies (e.g., cooling system failure for the entire plant); Evaluation Metrics: 1. Detection accuracy: number of correctly detected anomalies / total number of anomalies; 2. False alarm rate: number of false alarm anomalies / total number of detections; 3. Average detection time: the average time from when an anomaly occurs to when it is detected; 4. System-level anomaly recognition rate: number of correctly identified system-level anomalies / total number of system-level anomalies; The experimental results are as follows: The experimental results are analyzed as follows: Evaluation Metrics Example 1 Comparative Example 1 Detection accuracy 95.8% 82.3% False Positive Rate 2.1% 7.6% Average detection time 3.5 minutes 12.8 minutes System-level anomaly recognition rate 93.2% 45.7% 1. Detection accuracy: The detection accuracy of the method of the present invention is significantly higher than that of the traditional method. This is mainly due to the ability of high-dimensional topological space mapping and group theory feature extraction, which enables the method to capture more subtle abnormal patterns.

[0110] 2. False alarm rate: The false alarm rate of the method of the present invention is significantly lower than that of the traditional method. This is because the method can better distinguish normal fluctuations from true abnormal situations through dynamic feature matrix and chaos feature analysis.

[0111] 3. Average detection time: The method of the present invention performs well in the timeliness of abnormality detection, and the average detection time is only about one-fourth of that of traditional methods. This rapid response capability is crucial to preventing potential serious failures.

[0112] 4. System-level anomaly recognition rate: The method of the present invention has a particularly significant advantage in identifying system-level anomalies. By constructing the sensor network topology and conducting a comprehensive analysis, the method can effectively distinguish between local anomalies and system-level anomalies, while traditional methods perform poorly in this regard.

[0113] Further analysis found that the method of the present invention performs particularly well in dealing with complex abnormal patterns. For example, in a simulated progressive failure of a cooling system, the traditional method did not trigger an alarm until the readings of multiple sensors simultaneously exceeded the threshold, while the present method successfully issued an early warning before the temperature reached a dangerous level by analyzing the changing trend of the Lyapunov exponent, thus buying valuable time for timely maintenance.

[0114] In addition, the method of the present invention also shows good robustness in dealing with environmental noise and sensor fluctuations. In one experiment, when some sensors were subjected to temporary electromagnetic interference, the traditional method produced multiple false alarms, while the present method successfully distinguished these interferences from real anomalies by analyzing group theory characteristics and network topology.

[0115] In summary, the experimental results fully verify the superiority of the method of the present invention in online monitoring of intelligent IoT devices. It not only improves the accuracy and timeliness of anomaly detection, but also significantly enhances the system's ability to identify complex abnormal patterns. These advantages make this method particularly suitable for large-scale, highly complex industrial IoT systems, and can effectively improve the reliability and security of the system.

[0116] Based on the above experimental results and further optimization, the best embodiment of the present invention can be obtained: 1. Status parameter selection: In addition to temperature readings, it should also include multi-dimensional parameters such as the device's battery power, signal strength, and data transmission frequency.

[0117] 2. Topological space dimension: After many tests, mapping the state parameters to a 10-15 dimensional topological space is the best, which can fully capture nonlinear characteristics without causing excessive computational complexity.

[0118] 3. Basis function selection: Bessel functions and Legendre polynomials are used in combination, with a weight ratio of approximately 3:2.

[0119] 4. Group action operator: mainly uses rotation operators and scale transformation operators, and translation operators can be added in specific scenarios.

[0120] 5. Dynamic feature matrix: Select 30-50 sampling points to build a matrix, which can better balance real-time and trend analysis capabilities.

[0121] 6.Lyapunov exponent calculation: Using the improved Wolf algorithm, a more accurate estimate can be obtained on a limited time series.

[0122] 7. Abnormal threshold setting: Setting the abnormal threshold to 2.5-3 times the standard deviation of the Lyapunov exponent in the normal state can keep the false alarm rate at a low level while ensuring the detection sensitivity.

[0123] 8. Network topology construction: A weighted undirected graph is constructed based on the physical distance and logical association between devices, and the weight is calculated using the Pearson correlation coefficient.

[0124] By adopting this optimal implementation plan, a detection accuracy rate of over 96% can be achieved in actual applications, while the false alarm rate can be controlled below 2%, providing a strong guarantee for the stable operation and timely maintenance of the smart IoT system.

[0125] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto; any person familiar with the art who, within the scope disclosed by the present invention, makes substitutions or changes based on the scheme and improved concepts of the present invention shall be covered within the protection scope of the present invention.

Claims

1. An online monitoring method for intelligent Internet of Things devices, characterized in that: include: The acquisition steps include: Obtain the state parameter vector of the smart IoT device to be monitored; Processing steps include: Based on the state parameter vector, mapping the state parameter vector to a high-dimensional topological space to obtain a topological feature; Extracting group theory features according to the topological features; Based on the group theory characteristics, a dynamic characteristic matrix is ​​constructed; Calculating chaotic characteristics according to the dynamic characteristic matrix; Output steps include: Based on the chaotic characteristics, it is determined whether the intelligent Internet of Things device to be monitored is in an abnormal state.

2. The method according to claim 1, characterized in that Mapping the state parameter vector to a high-dimensional topological space to obtain topological features specifically includes: Obtaining a preset basis function set and a weight coefficient set; The topological feature is calculated based on the basis function set, the weight coefficient set and the state parameter vector.

3. The method according to claim 2, characterized in that: The basis function set includes Bessel functions and / or Legendre polynomials.

4. The method according to claim 1, characterized in that: The extracting of group theory features specifically includes: Get the preset group action operator set; The group theory characteristics are calculated based on the group action operator set and the topological characteristics.

5. The method according to claim 4, characterized in that: The group action operator set includes a rotation operator, a translation operator and / or a scale transformation operator.

6. The method according to claim 1, characterized in that The construction of the dynamic feature matrix specifically includes: Get multiple group theory features within a preset time window; calculating first-order differences and second-order differences of the plurality of group theory characteristics; The dynamic feature matrix is ​​constructed based on the plurality of group theory features, the first-order differences and the second-order differences.

7. The method according to claim 1, characterized in that The computational chaos features specifically include: Based on the dynamic characteristic matrix, the Lyapunov exponent is calculated as the chaotic characteristic.

8. The method according to claim 1, characterized in that Determining whether the intelligent IoT device to be monitored is in an abnormal state specifically includes: Obtain the preset normal state Lyapunov index baseline value and abnormal threshold; Calculating the absolute value of the difference between the chaotic characteristic and the Lyapunov exponent reference value of the normal state; Determine whether the absolute value of the difference is greater than the abnormal threshold; When the absolute value of the difference is greater than the abnormal threshold, it is determined that the smart Internet of Things device to be monitored is in an abnormal state.

9. The method according to claim 1, characterized in that: Also includes: Obtain state parameter vectors of multiple smart IoT devices to be monitored; Based on the state parameter vectors of the plurality of intelligent IoT devices to be monitored, construct a device-associated network topology structure; Combined with the device-associated network topology and the abnormal state judgment results of each intelligent IoT device to be monitored, a comprehensive abnormality analysis is performed.

10. An online monitoring system for intelligent Internet of Things devices implementing the method according to any one of claims 1 to 9, characterized in that: include: An acquisition module is used to obtain a state parameter vector of the smart IoT device to be monitored; Processing modules for: Based on the state parameter vector, mapping it to a high-dimensional topological space to obtain a topological feature; Extracting group theory features according to the topological features; Based on the group theory characteristics, a dynamic characteristic matrix is ​​constructed; Calculating chaotic characteristics according to the dynamic characteristic matrix; The output module is used to determine whether the intelligent Internet of Things device to be monitored is in an abnormal state based on the chaotic characteristics.

Images