Method for monitoring and diagnosing abnormity of Internet of Things equipment

Abstract

The invention discloses an Internet of Things equipment abnormity monitoring and diagnosis method. The method comprises the following steps: S1, collecting at least one monitoring parameter of each monitored equipment in an Internet of Things platform in real time; s2, initializing a control threshold value corresponding to each monitoring parameter; s3, establishing a threshold adjustment model, predicting and evaluating an optimal threshold under different conditions through the model based on the monitoring parameters collected in real time, and performing adaptive adjustment on the control threshold; s4, collecting and analyzing a large amount of historical data, mining relevance and regularity between the equipment state and the monitoring parameters, and determining a state relevance coefficient corresponding to each monitoring parameter; s5, determining health indexes of the monitored devices according to at least one monitoring parameter of each monitored device, the adaptively adjusted control threshold and the state correlation coefficient corresponding to each monitoring parameter; and S6, carrying out equipment monitoring on the Internet of Things platform according to the health index of each monitored equipment and a preset abnormity judgment rule.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things devices, and in particular to a method for monitoring and diagnosing abnormalities of Internet of Things devices. Background Art

[0002] The Internet of Things (IoT) refers to connecting various physical devices and objects to the Internet through various sensors, communication equipment, cloud computing and other technical means to achieve information perception, transmission and processing. With the continuous development and maturity of IoT technology, more and more devices are connected to the network, forming a huge IoT ecosystem. These devices cover various fields, such as smart homes, industrial control, environmental monitoring, etc., which have brought great convenience to people's lives and work. In the IoT ecosystem, the stability and reliability of equipment are crucial. Once an abnormality occurs in a device, it may lead to serious consequences such as system paralysis and data loss. Therefore, it is of great significance to monitor and diagnose abnormalities in IoT devices. By monitoring the operating status of the device in real time and discovering and handling abnormalities in time, the normal operation of the device can be ensured and the stability and reliability of the system can be improved.

[0003] In the prior art, threshold adjustment involves multiple factors, such as device type, working environment, historical data, etc. The adjustment process may be more complicated. Improper threshold adjustment may lead to false alarms or missed alarms, affecting the monitoring effect. In addition, if the relationship between the device status and the monitoring parameters is unclear or changes, the accuracy of the status correlation coefficient may be reduced. Therefore, a method for abnormal monitoring and diagnosis of IoT devices is proposed. Summary of the invention

[0004] The purpose of the present invention is to solve the problems existing in the prior art that threshold adjustment involves multiple factors, such as device type, working environment, historical data, etc., the adjustment process may be more complicated, and improper threshold adjustment may lead to false alarms or missed alarms, affecting the monitoring effect. In addition, if the relationship between the device status and the monitoring parameters is unclear or changes, it may lead to a decrease in the accuracy of the status correlation coefficient. A method for abnormal monitoring and diagnosis of Internet of Things devices is proposed.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for abnormal monitoring and diagnosis of IoT devices, comprising the following steps:

[0007] S1: collect at least one monitoring parameter of each monitored device in the IoT platform in real time;

[0008] S2: Initializing the control threshold corresponding to each of the monitoring parameters;

[0009] S3: Establish a threshold adjustment model, predict and evaluate the optimal threshold under different conditions based on the monitoring parameters collected in real time, and adjust the control threshold adaptively;

[0010] S4: Collect and analyze a large amount of historical data, explore the correlation and regularity between the equipment status and the monitoring parameters, and determine the status correlation coefficient corresponding to each of the monitoring parameters;

[0011] S5: determining the health index of each monitored device according to at least one monitoring parameter of each monitored device, the adaptively adjusted control threshold, and the state correlation coefficient corresponding to each monitoring parameter;

[0012] S6: monitoring the devices on the IoT platform according to the health index of each monitored device and preset abnormality determination rules;

[0013] S7: When an abnormal parameter is identified based on at least one monitoring parameter of each monitored device and the adaptively adjusted control threshold, an alarm message for the monitored device is generated.

[0014] The above further includes:

[0015] Furthermore, in S1, in the Internet of Things platform, each monitored device has its corresponding physical model. Based on the physical model, it is determined which auxiliary sensors are associated and bound with the monitoring device. The auxiliary sensors may include temperature sensors, humidity sensors, and pressure sensors, which are used to monitor different aspects of the device. The Internet of Things platform identifies and connects the auxiliary sensors related to the monitoring device through the definition in the physical model, and collects the monitoring parameters of the monitoring device in real time.

[0016] Further, in S2, the control threshold corresponding to each monitoring parameter is initialized, specifically in the following steps:

[0017] Monitoring over a continuous period of time:

[0018] Obtain multiple sets of monitoring parameter sample sets of the monitored equipment within a continuous period of time;

[0019] Status mark:

[0020] The acquired multiple sets of monitoring parameter sample sets are marked according to the status of the monitored equipment, and the equipment status is divided into two status dimensions: normal and abnormal;

[0021] Basic group and training group:

[0022] Dividing multiple groups of monitoring parameter sample sets into a basic group sample set and a training group sample set according to a certain ratio, wherein the basic group sample set is used to calculate the sample mean and sample standard deviation of the monitoring parameters, and the training group sample set is used for model training and optimization;

[0023] Division ratio:

[0024] The choice of the division ratio needs to be weighed according to the actual situation;

[0025] Sample mean and sample standard deviation:

[0026] For each monitoring parameter in the basic group sample set, calculate its sample mean and sample standard deviation, wherein the sample mean reflects the average level of the monitoring parameter under normal operating conditions, and the sample standard deviation reflects the fluctuation range of the monitoring parameter;

[0027] Determine initial control thresholds:

[0028] According to the sample mean and sample standard deviation of each monitoring parameter, the initial control threshold of the monitoring parameter is determined, and the initial control threshold is set to the sample mean plus or minus a certain multiple of the sample standard deviation.

[0029] Furthermore, in S3, the optimal thresholds under different conditions are predicted and evaluated through the model based on the monitoring parameters collected in real time, and the control thresholds are adaptively adjusted. The specific steps are:

[0030] Data collection and preprocessing:

[0031] Collect historical monitoring data, including various monitoring parameters of the equipment and the corresponding equipment status, and pre-process the collected data;

[0032] Feature selection and extraction:

[0033] Extract features that have a significant impact on the equipment status from the preprocessed data;

[0034] Model training:

[0035] Use the preprocessed historical data as a training set to train the random forest and obtain a threshold adjustment model for predicting device status;

[0036] Real-time data collection:

[0037] Collect various monitoring parameters of the monitored equipment in real time through the Internet of Things platform, and pre-process the collected data to ensure that the monitoring parameters meet the requirements of the threshold adjustment model input;

[0038] Model predictions:

[0039] The preprocessed real-time data is input into the trained threshold adjustment model to obtain the prediction threshold under different conditions. According to the prediction threshold, the state of the equipment is evaluated to determine whether it is in a normal or abnormal state;

[0040] Threshold adjustment strategy:

[0041] Formulate a threshold adjustment strategy based on the predicted results of the threshold adjustment model and the actual status of the device;

[0042] Adaptive Learning:

[0043] The introduction of adaptive threshold adjustment enables the system to automatically adjust the threshold based on historical data and real-time feedback. Through continuous learning and optimization, the threshold adjustment model can adapt to changes in device status.

[0044] Further, in S4, a large amount of historical data is collected and analyzed, the correlation and regularity between the equipment status and the monitoring parameters are mined, and the status correlation coefficient corresponding to each monitoring parameter is determined. The specific steps are:

[0045] Data Collection:

[0046] Obtain multiple sets of monitoring parameter sample sets of the monitored device within a continuous period of time, denoted as {X_i}, where i represents the sample sequence number, X_i represents the i-th set of monitoring parameters, and simultaneously obtain the device status data set corresponding to the monitoring parameter sample set, denoted as {Y_i}, where Y_i represents the device status corresponding to the i-th set of monitoring parameters;

[0047] Data preprocessing:

[0048] Preprocess the monitoring parameter sample set {X_i} and the equipment status data set {Y_i};

[0049] Calculate the mean and standard deviation:

[0050] Calculate the mean E(X) and standard deviation σ(X) of the monitoring parameter sample set {X_i}. The calculation formula of the mean E(X) is E(X) = Σ(X_i) / n, where n is the number of samples. The calculation formula of the standard deviation σ(X) is σ(X) = √[Σ(X_i-E(X))^2 / n]; calculate the mean E(Y) and standard deviation σ(Y) of the equipment status data set {Y_i}. The calculation formula of the mean E(Y) is E(Y) = Σ(Y_i) / n, where n is the number of samples. The calculation formula of the standard deviation σ(Y) is σ(Y) = √[Σ(Y_i-E(Y))^2 / n];

[0051] Compute the covariance:

[0052] According to the mean E(X) of the monitoring parameter sample set and the mean E(Y) of the equipment status data set, the covariance Cov(X,Y) of the monitoring parameter sample set and the equipment status data set is calculated. The calculation formula of the covariance Cov(X,Y) is Cov(X,Y)=Σ[(X_i-E(X))(Y_i-E(Y))] / n;

[0053] Calculate the state correlation coefficient:

[0054] The state correlation coefficient ρ(X,Y) is the ratio of the covariance Cov(X,Y) to the product of the standard deviation σ(X) of the monitoring parameter sample set and the standard deviation σ(Y) of the equipment state data set. It is used to measure the linear correlation between the monitoring parameters and the equipment state. The calculation formula of the state correlation coefficient ρ(X,Y) is ρ(X,Y)=Cov(X,Y) / [σ(X)σ(Y)]; the value range of ρ(X,Y) is [-1,1]. When ρ(X,Y) is close to 1, it indicates that there is a strong positive correlation between the monitoring parameters and the equipment state; when ρ(X,Y) is close to -1, it indicates a strong negative correlation; when ρ(X,Y) is close to 0, it indicates that the correlation is weak or there is no linear correlation.

[0055] Using big data analysis technology:

[0056] Collect and analyze a large amount of historical data, and use classification algorithms to mine the correlation and regularity between equipment status and monitoring parameters.

[0057] Furthermore, in S5, the specific steps are:

[0058] Collecting monitoring parameters: collecting at least one monitoring parameter from the monitored device, wherein the monitoring parameter is a physical quantity reflecting the operating status of the device;

[0059] Comparison of monitoring parameters with control thresholds: Compare the actual value of each monitoring parameter with the corresponding control threshold;

[0060] Determine abnormal status: If the actual value of the monitoring parameter exceeds the control threshold, the monitoring parameter is considered to be in an abnormal state;

[0061] Calculate the abnormal probability: Calculate an abnormal probability value according to the degree of deviation between the actual value of the monitoring parameter and the control threshold. The calculation formula of the abnormal probability value P is expressed as P=f(XT), where X represents the actual value of the monitoring parameter, T represents the control threshold, and f represents a functional relationship;

[0062] Analyze the correlation between monitoring parameters: Use the correlation coefficient to calculate and analyze the correlation between various monitoring parameters;

[0063] Determine the state correlation coefficient: according to the analysis results, assign a state correlation coefficient to each monitoring parameter, which reflects the correlation between the monitoring parameter and other monitoring parameters and the overall state of the equipment. The value range of the state correlation coefficient r is usually [-1, 1], where 1 indicates a complete positive correlation, -1 indicates a complete negative correlation, and 0 indicates no correlation;

[0064] Weighted calculation: multiply the abnormal probability of each monitoring parameter by its corresponding state correlation coefficient to obtain the weighted abnormal probability, sum or average all weighted abnormal probabilities to obtain the health index of the monitored equipment. The calculation formula of the health index HI is expressed as HI=Σ(P_i*r_i) / n, where P_i represents the abnormal probability of the i-th monitoring parameter, r_i represents the state correlation coefficient of the i-th monitoring parameter, and n represents the number of monitoring parameters.

[0065] Furthermore, in S6, the device monitoring of the Internet of Things platform is performed according to the health index of each monitored device and the preset abnormality determination rule, and the specific steps are:

[0066] Define the device cascade relationship: set up a directed graph or tree structure, where nodes represent devices and edges represent connections between devices;

[0067] Configure device cascading relationships: store and manage device cascading relationships through configuration files or databases;

[0068] Analyze device hierarchy: Determine the hierarchical position of each device in the system based on the device cascade relationship;

[0069] Set weight value: set a weight value for each level, which is used to reflect the importance and influence of the device in the system;

[0070] Get health index: Get the health index of each monitored device;

[0071] Weighted calculation: Based on the health index of each device and the corresponding level weight, weighted calculation is performed using the following formula: Among them, i represents the device serial number, Σ represents the sum operation;

[0072] Set abnormal thresholds: Set one or more abnormal thresholds to determine whether the overall health index of the system is in an abnormal state;

[0073] Determine abnormal status: Compare the overall health index of the system with the abnormal threshold. If the overall health index of the system is lower than a certain threshold, the system is considered to be in an abnormal state and the corresponding alarm or maintenance process needs to be triggered.

[0074] Further, in S7, when an abnormal parameter is identified according to at least one monitoring parameter of each monitored device and the adaptively adjusted control threshold, an alarm message for the monitored device is generated, specifically:

[0075] Monitoring parameter collection and threshold comparison:

[0076] Collecting data of at least one monitoring parameter from the monitored device in real time, and comparing the actual value of the monitoring parameter with a preset adaptive control threshold;

[0077] Alarm information generation:

[0078] Once an abnormal parameter is identified, an alarm message is immediately generated, which should include the ID of the monitored device, the ID of the abnormal parameter, the control threshold, the abnormal value, the alarm level, the alarm time and the processing status;

[0079] Alarm level upgrade judgment:

[0080] Key parameter judgment: If the parameter type of the abnormal parameter is a key parameter of the monitored device, the alarm level in the alarm information is upgraded. The upgrade rule is that if r_i>R_th, where r_i represents the state correlation coefficient of the i-th monitoring parameter, and R_th represents the set threshold, the alarm level is upgraded by one level;

[0081] Alarm quantity judgment: If the alarm information generated for the same monitored device within the preset time period exceeds the set number, the alarm level will also be upgraded according to the alarm quantity threshold preset based on the device characteristics and business needs. The upgrade rule is if N>N_th, where N represents the number of alarms generated within the preset time period, and N_th represents the set quantity threshold, the alarm level will be upgraded by one level.

[0082] The present invention has the following beneficial effects:

[0083] 1. In the present invention, a threshold adjustment model is established to perform prediction and evaluation based on factors such as equipment type, working environment, and historical data, which provides a scientific basis for the actual adjustment process and further improves the accuracy of the adjustment. By automating the threshold adjustment, the workload of operation and maintenance personnel is reduced, and the operation and maintenance costs are reduced.

[0084] 2. In the present invention, big data analysis technology is used to explore the correlation and regularity between equipment status and monitoring parameters, thereby determining the status correlation coefficient, thereby improving the accuracy and reliability of the coefficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 This is a step diagram of an abnormal monitoring and diagnosis method for Internet of Things devices proposed by the present invention. DETAILED DESCRIPTION

[0086] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0087] See also Figure 1 As shown, the present invention is a method for abnormal monitoring and diagnosis of IoT devices, comprising the following steps:

[0088] S1: collect at least one monitoring parameter of each monitored device in the IoT platform in real time;

[0089] S2: Initializing the control threshold corresponding to each of the monitoring parameters;

[0090] S3: Establish a threshold adjustment model, predict and evaluate the optimal threshold under different conditions based on the monitoring parameters collected in real time, and adjust the control threshold adaptively;

[0091] S4: Collect and analyze a large amount of historical data, explore the correlation and regularity between the equipment status and the monitoring parameters, and determine the status correlation coefficient corresponding to each of the monitoring parameters;

[0092] S5: determining the health index of each monitored device according to at least one monitoring parameter of each monitored device, the adaptively adjusted control threshold, and the state correlation coefficient corresponding to each monitoring parameter;

[0093] S6: monitoring the devices on the IoT platform according to the health index of each monitored device and preset abnormality determination rules;

[0094] S7: When an abnormal parameter is identified based on at least one monitoring parameter of each monitored device and the adaptively adjusted control threshold, an alarm message for the monitored device is generated.

[0095] In one embodiment, for the above S1, in S1, in the Internet of Things platform, each monitored device has its corresponding physical model. Based on the physical model, it is determined which auxiliary sensors are associated and bound with the monitoring device. The auxiliary sensors can include temperature sensors, humidity sensors, pressure sensors, etc., which are used to monitor different aspects of the device. The Internet of Things platform identifies and connects the auxiliary sensors related to the monitoring device through the definition in the physical model, and collects the monitoring parameters of the monitoring device in real time.

[0096] In one embodiment, for the above S2, in S2, the control threshold corresponding to each monitoring parameter is initialized, specifically in the following steps:

[0097] Monitoring over a continuous period of time:

[0098] To ensure the accuracy and reliability of the control threshold, multiple sets of monitoring parameter samples of the monitored equipment are obtained over a continuous period of time. This period of time should be long enough to cover various operating states of the equipment.

[0099] Status mark:

[0100] Mark the acquired sets of monitoring parameter samples according to the status of the monitored equipment, and divide the equipment status into two status dimensions: normal and abnormal. This requires professional technicians or experts to mark according to the actual situation of the equipment;

[0101] Basic group and training group:

[0102] Divide multiple groups of monitoring parameter sample sets into a basic group sample set and a training group sample set according to a certain ratio, wherein the basic group sample set is used to calculate the sample mean and sample standard deviation of the monitoring parameters, thereby determining the initial control threshold, and the training group sample set is used for model training and optimization;

[0103] Division ratio:

[0104] The selection of the division ratio needs to be weighed according to the actual situation. The ratio of the basic group sample set should be large enough to ensure that the calculated sample mean and sample standard deviation are representative. At the same time, the ratio of the training group sample set should not be too small to ensure the effect of model training.

[0105] Sample mean and sample standard deviation:

[0106] For each monitoring parameter in the basic group sample set, calculate its sample mean and sample standard deviation, wherein the sample mean reflects the average level of the monitoring parameter under normal operating conditions, and the sample standard deviation reflects the fluctuation range of the monitoring parameter;

[0107] Determine initial control thresholds:

[0108] Based on the sample mean and sample standard deviation of each monitoring parameter, the initial control threshold of the monitoring parameter is determined. The initial control threshold is set to the sample mean plus or minus a certain multiple of the sample standard deviation. The selection of this multiple needs to be weighed according to the actual situation to ensure the sensitivity and accuracy of the control threshold.

[0109] In one embodiment, for the above S3, in S3, the optimal threshold under different conditions is predicted and evaluated through a model based on the monitoring parameters collected in real time, and the control threshold is adaptively adjusted, specifically in the following steps:

[0110] Data collection and preprocessing:

[0111] Collect historical monitoring data, including various monitoring parameters of the equipment (such as temperature, pressure, current, etc.) and the corresponding equipment status (normal or abnormal), and pre-process the collected data, including data cleaning (removing outliers, missing values, etc.), data normalization (converting data to the same dimension), etc., to improve data quality and availability;

[0112] Feature selection and extraction:

[0113] Extract features that have a significant impact on the equipment status from the preprocessed data. These features will be used as inputs to the threshold adjustment model. The feature selection process is implemented through correlation analysis, principal component analysis (PCA) and other methods to ensure that the selected features can accurately reflect the operating status of the equipment.

[0114] Model training:

[0115] The preprocessed historical data is used as the training set to train the random forest and obtain the threshold adjustment model for predicting the equipment status. During the training process, the model parameters are optimized through cross-validation, grid search and other methods to improve the prediction performance of the model.

[0116] Real-time data collection:

[0117] Collect various monitoring parameters of the monitored equipment in real time through the Internet of Things platform. Pre-process the collected data to ensure that the monitoring parameters meet the requirements of the threshold adjustment model input;

[0118] Model predictions:

[0119] The preprocessed real-time data is input into the trained threshold adjustment model to obtain the prediction threshold under different conditions. According to the prediction threshold, the state of the equipment is evaluated to determine whether it is in a normal or abnormal state;

[0120] Threshold adjustment strategy:

[0121] Develop a threshold adjustment strategy based on the predicted results of the threshold adjustment model and the actual status of the device. For example, when the device status is abnormal, the threshold can be appropriately lowered to increase the sensitivity of the warning; when the device status is stable, the threshold can be appropriately increased to reduce the false alarm rate.

[0122] Adaptive Learning:

[0123] Adaptive threshold adjustment is introduced to enable the system to automatically adjust the threshold based on historical data and real-time feedback. Through continuous learning and optimization, the threshold adjustment model can adapt to changes in device status and improve the accuracy and efficiency of early warning.

[0124] Model Update and Maintenance:

[0125] Over time, the operating status of the equipment may change, so the threshold adjustment model needs to be updated and maintained regularly. The model can be retrained with new historical data or fine-tuned to adapt to the new equipment status.

[0126] In one embodiment, for the above S4, in S4, the collection and analysis of a large amount of historical data, mining the correlation and regularity between the equipment status and the monitoring parameters, and determining the status correlation coefficient corresponding to each of the monitoring parameters, the specific steps are:

[0127] Data Collection:

[0128] Obtain multiple sets of monitoring parameter sample sets of the monitored device within a continuous period of time, denoted as {X_i}, where i represents the sample number, X_i represents the i-th set of monitoring parameters (which can be a multidimensional vector containing multiple monitoring parameters), and at the same time obtain the device status data set corresponding to the monitoring parameter sample set, denoted as {Y_i}, where Y_i represents the device status corresponding to the i-th set of monitoring parameters (which can be a discrete classification label, such as normal, abnormal, or a continuous numerical value representing a quantitative indicator of the device status);

[0129] Data preprocessing:

[0130] Preprocess the monitoring parameter sample set {X_i} and the equipment status data set {Y_i}, including data cleaning (removing outliers, missing values, etc.), data normalization (converting data to the same dimension), etc.;

[0131] Calculate the mean and standard deviation:

[0132] Calculate the mean E(X) and standard deviation σ(X) of the monitoring parameter sample set {X_i}. The calculation formula of the mean E(X) is E(X) = Σ(X_i) / n, where n is the number of samples. The calculation formula of the standard deviation σ(X) is σ(X) = √[Σ(X_i-E(X))^2 / n]; calculate the mean E(Y) and standard deviation σ(Y) of the equipment status data set {Y_i}. The calculation formula of the mean E(Y) is E(Y) = Σ(Y_i) / n, where n is the number of samples. The calculation formula of the standard deviation σ(Y) is σ(Y) = √[Σ(Y_i-E(Y))^2 / n];

[0133] Compute the covariance:

[0134] According to the mean E(X) of the monitoring parameter sample set and the mean E(Y) of the equipment status data set, the covariance Cov(X,Y) of the monitoring parameter sample set and the equipment status data set is calculated. The calculation formula of the covariance Cov(X,Y) is Cov(X,Y)=Σ[(X_i-E(X))(Y_i-E(Y))] / n;

[0135] Calculate the state correlation coefficient:

[0136] The state correlation coefficient ρ(X,Y) is the ratio of the covariance Cov(X,Y) to the product of the standard deviation σ(X) of the monitoring parameter sample set and the standard deviation σ(Y) of the equipment state data set. It is used to measure the linear correlation between the monitoring parameters and the equipment state. The calculation formula of the state correlation coefficient ρ(X,Y) is ρ(X,Y)=Cov(X,Y) / [σ(X)σ(Y)]; the value range of ρ(X,Y) is [-1,1]. When ρ(X,Y) is close to 1, it indicates that there is a strong positive correlation between the monitoring parameters and the equipment state; when ρ(X,Y) is close to -1, it indicates a strong negative correlation; when ρ(X,Y) is close to 0, it indicates that the correlation is weak or there is no linear correlation.

[0137] Using big data analysis technology:

[0138] Collect and analyze a large amount of historical data, and use classification algorithms to mine the correlation and regularity between equipment status and monitoring parameters to further verify and correct the status correlation coefficient;

[0139] Introducing expert systems:

[0140] Establish an expert system to evaluate and correct the state correlation coefficient by combining the knowledge and experience of domain experts. The expert system can provide additional information and insights to help improve the accuracy and reliability of the state correlation coefficient;

[0141] Adopt multi-source information fusion method:

[0142] Combine information from different sensors and data sources to determine the state correlation coefficient. This can comprehensively consider the influence of multiple factors and improve accuracy and comprehensiveness. For example, the state correlation coefficients of multiple monitoring parameters can be weighted averaged or weighted averaged before fusion processing.

[0143] In one embodiment, for the above S5, in S5, the specific steps are:

[0144] Collect monitoring parameters: collect at least one monitoring parameter from the monitored device, where the monitoring parameter is a physical quantity such as temperature, pressure, current, vibration, etc. that reflects the operating status of the device;

[0145] Comparison of monitoring parameters with control thresholds: Compare the actual value of each monitoring parameter with the corresponding control threshold;

[0146] Determine abnormal status: If the actual value of the monitoring parameter exceeds the control threshold, the monitoring parameter is considered to be in an abnormal state;

[0147] Calculate the abnormal probability: Calculate an abnormal probability value according to the degree of deviation between the actual value of the monitoring parameter and the control threshold. The calculation formula of the abnormal probability value P is expressed as P=f(XT), where X represents the actual value of the monitoring parameter, T represents the control threshold, and f represents a functional relationship. P can be linear (such as P=(XT) / (X_max-T), where X_max represents the possible maximum value of the monitoring parameter) or nonlinear (such as using machine learning algorithms such as logistic regression for modeling);

[0148] Analyze the correlation between monitoring parameters: Use the correlation coefficient to calculate and analyze the correlation between various monitoring parameters;

[0149] Determine the state correlation coefficient: according to the analysis results, assign a state correlation coefficient to each monitoring parameter, which reflects the correlation between the monitoring parameter and other monitoring parameters and the overall state of the equipment. The value range of the state correlation coefficient r is usually [-1, 1], where 1 indicates a complete positive correlation, -1 indicates a complete negative correlation, and 0 indicates no correlation;

[0150] Weighted calculation: multiply the abnormal probability of each monitoring parameter by its corresponding state correlation coefficient to obtain the weighted abnormal probability, sum or average all weighted abnormal probabilities to obtain the health index of the monitored device. The calculation formula of the health index HI is expressed as HI=Σ(P_i*r_i) / n, where P_i represents the abnormal probability of the i-th monitoring parameter, r_i represents the state correlation coefficient of the i-th monitoring parameter, and n represents the number of monitoring parameters;

[0151] Assume that we have a monitored device whose monitoring parameters include temperature T (℃) and vibration V (mm / s), and we know that the adaptive control thresholds of these two monitoring parameters are T_th = 80℃ and V_th = 0.5mm / s. At the same time, we also know that the state correlation coefficients between these two monitoring parameters are r_T = 0.7 and r_V = 0.3;

[0152] Now, we have collected a set of monitoring data: temperature T = 85 ° C, vibration V = 0.4 mm / s. We can calculate the health index of the equipment by following the steps below:

[0153] Calculate the abnormal probability P_T of temperature T:

[0154] P_T = (T-T_th) / (T_max-T_th) = (85-80) / (100-80) = 0.25 (assuming T_max = 100°C is the maximum possible temperature)

[0155] Calculate the abnormal probability P_V of vibration V:

[0156] Since the actual value V=0.4 mm / s of the vibration V does not exceed the control threshold V_th=0.5 mm / s, we can consider that the vibration V is in a normal state, so P_V=0.

[0157] Calculate the health index HI:

[0158] HI=(P_T*r_T+P_V*r_V) / 2=(0.25*0.7+0*0.3) / 2=0.0875

[0159] Since the abnormal probability of temperature T is high, while the vibration V is in a normal state, the health index HI of the equipment is low. This reflects that the equipment has abnormal temperature and needs further inspection or maintenance.

[0160] In one embodiment, for the above S6, in S6, the device monitoring of the Internet of Things platform is performed according to the health index of each monitored device and the preset abnormality determination rule, and the specific steps are:

[0161] Define the device cascade relationship: The device cascade relationship describes the connection and dependency between each device in the monitored device system. Set up a directed graph or tree structure, where nodes represent devices and edges represent connections between devices;

[0162] Configure device cascading relationships: Store and manage device cascading relationships through configuration files or databases. These relationships can be defined based on the physical connection, logical dependency, or business scenario of the device.

[0163] Analyze device hierarchy: Determine the hierarchical position of each device in the system based on the device cascade relationship. Generally, the devices closer to the system entrance or core have higher hierarchical weights.

[0164] Setting weight values: Setting a weight value for each level. The weight value can be fixed or dynamically adjusted according to system characteristics and business requirements. The weight value is used to reflect the importance and influence of the device in the system.

[0165] Get health index: Get the health index of each monitored device;

[0166] Weighted calculation: Perform weighted calculation based on the health index of each device and the corresponding level weight. This is achieved through the following formula: Among them, i represents the device serial number, Σ represents the sum operation;

[0167] Set abnormal thresholds: Set one or more abnormal thresholds to determine whether the overall health index of the system is in an abnormal state;

[0168] Determine abnormal status: Compare the overall health index of the system with the abnormal threshold. If the overall health index of the system is lower than a certain threshold, the system is considered to be in an abnormal state and the corresponding alarm or maintenance process needs to be triggered;

[0169] Assume that we have a simple IoT system with three monitored devices: A (entry device), B (intermediate device) and C (exit device). Their device cascade relationship is as follows:

[0170] A→B→C

[0171] We set the following layer weights: A = 0.5, B = 0.3, C = 0.2 (these weight values ​​can be adjusted according to actual conditions).

[0172] Now, we get the health index of each device: A = 0.8, B = 0.6, C = 0.9.

[0173] According to the weighted calculation formula, we can get the overall health index of the system:

[0174] System overall health index = (0.8*0.5+0.6*0.3+0.9*0.2) / (0.5+0.3+0.2) = 0.74

[0175] Assuming that we set the abnormal threshold to 0.7, the overall health index of the system is higher than the abnormal threshold, so the system is in a normal state. However, if the health index of a device decreases significantly (for example, due to a failure or maintenance), the overall health index of the system will also decrease accordingly and may trigger the abnormal judgment rule.

[0176] In one embodiment, for the above S7, in S7, when an abnormal parameter is identified according to at least one monitoring parameter of each monitored device and the adaptively adjusted control threshold, an alarm message for the monitored device is generated, and the specific steps are:

[0177] Monitoring parameter collection and threshold comparison:

[0178] Collect data of at least one monitoring parameter, such as temperature, pressure, current, etc., from the monitored device in real time, and compare the actual value of the monitoring parameter with a preset adaptive control threshold;

[0179] Alarm information generation:

[0180] Once an abnormal parameter is identified, an alarm message is immediately generated, which should include the ID of the monitored device, the ID of the abnormal parameter, the control threshold, the abnormal value, the alarm level, the alarm time and the processing status;

[0181] Warning information content:

[0182] Monitored device ID: an identifier that uniquely identifies the device;

[0183] Abnormal parameter ID: an identifier that uniquely identifies the abnormal parameter;

[0184] Control threshold: the threshold that triggers the alarm (can be an upper limit or a lower limit);

[0185] Abnormal value: the abnormal value actually monitored;

[0186] Alarm level: a preset level based on the degree of abnormality (such as low, medium, high);

[0187] Alarm time: the timestamp when the alarm is triggered;

[0188] Processing status: the current processing status of the alarm (such as unprocessed, processing, processed);

[0189] Alarm level upgrade judgment:

[0190] Key parameter judgment: If the parameter type of the abnormal parameter is a key parameter of the monitored device (i.e., a monitoring parameter whose state correlation coefficient is greater than a set threshold), the alarm level in the alarm information is upgraded. The state correlation coefficient is a coefficient reflecting the degree of correlation between the monitoring parameter and the overall state of the device. The upgrade rule is that if r_i>R_th, where r_i represents the state correlation coefficient of the i-th monitoring parameter, and R_th represents the set threshold, the alarm level is upgraded by one level;

[0191] Alarm quantity judgment: If the alarm information generated for the same monitored device within a preset time period exceeds the set number, the alarm level is also upgraded. According to the alarm quantity threshold preset according to device characteristics and business needs, the upgrade rule is that if N>N_th, where N represents the number of alarms generated within the preset time period, and N_th represents the set quantity threshold, the alarm level is upgraded by one level;

[0192] Assume that we have a monitored device with ID 001 and two monitoring parameters: temperature T and current I. The control threshold of temperature T is 50℃ to 100℃, and the control threshold of current I is 5A to 10A. The state correlation coefficient r_T=0.8 (for temperature T), r_I=0.5 (for current I), and the set threshold R_th=0.7. The preset time period is within 1 hour, and the set quantity threshold N_th=3;

[0193] Scenario 1: Abnormal temperature

[0194] Monitoring data: temperature T = 105°C, current I = 7A.

[0195] Comparison result: The temperature T exceeds the upper threshold of 100°C, triggering an alarm.

[0196] Warning information:

[0197] Monitored device ID: 001

[0198] Exception parameter ID: T

[0199] Control threshold: 100℃

[0200] Abnormal value: 105℃

[0201] Alarm level: Medium (initial level)

[0202] Alarm time: current timestamp

[0203] Processing status: Unprocessed

[0204] Alarm level upgrade: Since r_T=0.8>R_th=0.7, the alarm level is upgraded to high;

[0205] Scenario 2: Multiple Alarms

[0206] Monitoring data: Within 1 hour, the temperature T of device 001 exceeded the threshold multiple times, triggering 3 alarms;

[0207] Alarm information (the alarm content is similar each time, but the alarm time is different):

[0208] ...(the first two warning messages)

[0209] The third warning information (including the upgraded warning level):

[0210] Monitored device ID: 001

[0211] Exception parameter ID: T

[0212] Control threshold: 100℃

[0213] Abnormal value: (specific value)

[0214] Alarm level: High (upgraded from initial level due to N>N_th=3)

[0215] Alarm time: current timestamp

[0216] Processing status: Unprocessed.

[0217] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for abnormal monitoring and diagnosis of IoT devices, characterized in that: The following steps are involved: S1: collect at least one monitoring parameter of each monitored device in the IoT platform in real time; S2: Initializing the control threshold corresponding to each of the monitoring parameters; S3: Establish a threshold adjustment model, predict and evaluate the optimal threshold under different conditions based on the monitoring parameters collected in real time, and adjust the control threshold adaptively; S4: Collect and analyze a large amount of historical data, explore the correlation and regularity between the equipment status and the monitoring parameters, and determine the status correlation coefficient corresponding to each of the monitoring parameters; S5: determining the health index of each monitored device according to at least one monitoring parameter of each monitored device, the adaptively adjusted control threshold, and the state correlation coefficient corresponding to each monitoring parameter; S6: monitoring the devices on the IoT platform according to the health index of each monitored device and preset abnormality determination rules; S7: When an abnormal parameter is identified based on at least one monitoring parameter of each monitored device and the adaptively adjusted control threshold, an alarm message for the monitored device is generated.

2. The method for abnormal monitoring and diagnosis of IoT devices according to claim 1, characterized in that: In S1, in the Internet of Things platform, each monitored device has its corresponding physical model. Based on the physical model, it is determined which auxiliary sensors are associated and bound with the monitoring device. The auxiliary sensors can include temperature sensors, humidity sensors, and pressure sensors, which are used to monitor different aspects of the device. The Internet of Things platform identifies and connects the auxiliary sensors related to the monitoring device through the definition in the physical model, and collects the monitoring parameters of the monitoring device in real time.

3. The method for abnormal monitoring and diagnosis of IoT devices according to claim 2, characterized in that: In S2, the control threshold corresponding to each monitoring parameter is initialized, specifically in the following steps: Monitoring over a continuous period of time: Obtain multiple sets of monitoring parameter sample sets of the monitored equipment within a continuous period of time; Status mark: The acquired multiple sets of monitoring parameter sample sets are marked according to the status of the monitored equipment, and the equipment status is divided into two status dimensions: normal and abnormal; Basic group and training group: Dividing multiple groups of monitoring parameter sample sets into a basic group sample set and a training group sample set according to a certain ratio, wherein the basic group sample set is used to calculate the sample mean and sample standard deviation of the monitoring parameters, and the training group sample set is used for model training and optimization; Division ratio: The choice of the division ratio needs to be weighed according to the actual situation; Sample mean and sample standard deviation: For each monitoring parameter in the basic group sample set, calculate its sample mean and sample standard deviation, wherein the sample mean reflects the average level of the monitoring parameter under normal operating conditions, and the sample standard deviation reflects the fluctuation range of the monitoring parameter; Determine initial control thresholds: According to the sample mean and sample standard deviation of each monitoring parameter, the initial control threshold of the monitoring parameter is determined, and the initial control threshold is set to the sample mean plus or minus a certain multiple of the sample standard deviation.

4. The method for abnormal monitoring and diagnosis of IoT devices according to claim 1, characterized in that: In S3, the optimal thresholds under different conditions are predicted and evaluated through models based on the monitoring parameters collected in real time, and the control thresholds are adaptively adjusted. The specific steps are as follows: Data collection and preprocessing: Collect historical monitoring data, including various monitoring parameters of the equipment and the corresponding equipment status, and pre-process the collected data; Feature selection and extraction: Extract features that have a significant impact on the equipment status from the preprocessed data; Model training: Use the preprocessed historical data as a training set to train the random forest and obtain a threshold adjustment model for predicting device status; Real-time data collection: Collect various monitoring parameters of the monitored equipment in real time through the Internet of Things platform, and pre-process the collected data to ensure that the monitoring parameters meet the requirements of the threshold adjustment model input; Model predictions: The preprocessed real-time data is input into the trained threshold adjustment model to obtain the prediction threshold under different conditions. According to the prediction threshold, the state of the equipment is evaluated to determine whether it is in a normal or abnormal state; Threshold adjustment strategy: Formulate a threshold adjustment strategy based on the predicted results of the threshold adjustment model and the actual status of the device; Adaptive Learning: The introduction of adaptive threshold adjustment enables the system to automatically adjust the threshold based on historical data and real-time feedback. Through continuous learning and optimization, the threshold adjustment model can adapt to changes in device status.

5. The method for abnormal monitoring and diagnosis of IoT devices according to claim 1, characterized in that: In S4, a large amount of historical data is collected and analyzed, the correlation and regularity between the equipment status and the monitoring parameters are mined, and the status correlation coefficient corresponding to each monitoring parameter is determined. The specific steps are: Data Collection: Obtain multiple sets of monitoring parameter sample sets of the monitored device within a continuous period of time, denoted as {X_i}, where i represents the sample sequence number, X_i represents the i-th set of monitoring parameters, and simultaneously obtain the device status data set corresponding to the monitoring parameter sample set, denoted as {Y_i}, where Y_i represents the device status corresponding to the i-th set of monitoring parameters; Data preprocessing: Preprocess the monitoring parameter sample set {X_i} and the equipment status data set {Y_i}; Calculate the mean and standard deviation: Calculate the mean E(X) and standard deviation σ(X) of the monitoring parameter sample set {X_i}. The calculation formula of the mean E(X) is E(X) = Σ(X_i) / n, where n is the number of samples. The calculation formula of the standard deviation σ(X) is σ(X) = √[Σ(X_i-E(X))^2 / n]; calculate the mean E(Y) and standard deviation σ(Y) of the equipment status data set {Y_i}. The calculation formula of the mean E(Y) is E(Y) = Σ(Y_i) / n, where n is the number of samples. The calculation formula of the standard deviation σ(Y) is σ(Y) = √[Σ(Y_i-E(Y))^2 / n]; Compute the covariance: According to the mean E(X) of the monitoring parameter sample set and the mean E(Y) of the equipment status data set, the covariance Cov(X,Y) of the monitoring parameter sample set and the equipment status data set is calculated. The calculation formula of the covariance Cov(X,Y) is Cov(X,Y)=Σ[(X_i-E(X))(Y_i-E(Y))] / n; Calculate the state correlation coefficient: The state correlation coefficient ρ(X,Y) is the ratio of the covariance Cov(X,Y) to the product of the standard deviation σ(X) of the monitoring parameter sample set and the standard deviation σ(Y) of the equipment state data set. It is used to measure the linear correlation between the monitoring parameters and the equipment state. The calculation formula of the state correlation coefficient ρ(X,Y) is ρ(X,Y)=Cov(X,Y) / [σ(X)σ(Y)]; the value range of ρ(X,Y) is [-1,1]. When ρ(X,Y) is close to 1, it indicates that there is a strong positive correlation between the monitoring parameters and the equipment state; when ρ(X,Y) is close to -1, it indicates a strong negative correlation; when ρ(X,Y) is close to 0, it indicates that the correlation is weak or there is no linear correlation. Using big data analysis technology: Collect and analyze a large amount of historical data, and use classification algorithms to mine the correlation and regularity between equipment status and monitoring parameters.

6. The method for abnormal monitoring and diagnosis of IoT devices according to claim 1, characterized in that: In S5, the specific steps are: Collecting monitoring parameters: collecting at least one monitoring parameter from the monitored device, wherein the monitoring parameter is a physical quantity reflecting the operating status of the device; Comparison of monitoring parameters with control thresholds: Compare the actual value of each monitoring parameter with the corresponding control threshold; Determine abnormal status: If the actual value of the monitoring parameter exceeds the control threshold, the monitoring parameter is considered to be in an abnormal state; Calculate the abnormal probability: Calculate an abnormal probability value according to the degree of deviation between the actual value of the monitoring parameter and the control threshold. The calculation formula of the abnormal probability value P is expressed as P=f(XT), where X represents the actual value of the monitoring parameter, T represents the control threshold, and f represents a functional relationship; Analyze the correlation between monitoring parameters: Use the correlation coefficient to calculate and analyze the correlation between various monitoring parameters; Determine the state correlation coefficient: according to the analysis results, assign a state correlation coefficient to each monitoring parameter, which reflects the correlation between the monitoring parameter and other monitoring parameters and the overall state of the equipment. The value range of the state correlation coefficient r is usually [-1, 1], where 1 indicates a complete positive correlation, -1 indicates a complete negative correlation, and 0 indicates no correlation; Weighted calculation: multiply the abnormal probability of each monitoring parameter by its corresponding state correlation coefficient to obtain the weighted abnormal probability, sum or average all weighted abnormal probabilities to obtain the health index of the monitored equipment. The calculation formula of the health index HI is expressed as HI=Σ(P_i*r_i) / n, where P_i represents the abnormal probability of the i-th monitoring parameter, r_i represents the state correlation coefficient of the i-th monitoring parameter, and n represents the number of monitoring parameters.

7. The method for abnormal monitoring and diagnosis of IoT devices according to claim 1, characterized in that: In S6, the device monitoring of the Internet of Things platform is performed according to the health index of each monitored device and the preset abnormality determination rules, and the specific steps are: Define the device cascade relationship: set up a directed graph or tree structure, where nodes represent devices and edges represent connections between devices; Configure device cascading relationships: store and manage device cascading relationships through configuration files or databases; Analyze device hierarchy: Determine the hierarchical position of each device in the system based on the device cascade relationship; Set weight value: set a weight value for each level, which is used to reflect the importance and influence of the device in the system; Get health index: Get the health index of each monitored device; Weighted calculation: Based on the health index of each device and the corresponding level weight, weighted calculation is performed using the following formula: Among them, i represents the device serial number, Σ represents the sum operation; Set abnormal thresholds: Set one or more abnormal thresholds to determine whether the overall health index of the system is in an abnormal state; Determine abnormal status: Compare the overall health index of the system with the abnormal threshold. If the overall health index of the system is lower than a certain threshold, the system is considered to be in an abnormal state and the corresponding alarm or maintenance process needs to be triggered.

8. The method for abnormal monitoring and diagnosis of IoT devices according to claim 1, characterized in that: In S7, when an abnormal parameter is identified according to at least one monitoring parameter of each monitored device and the adaptively adjusted control threshold, an alarm message for the monitored device is generated. The specific steps are: Monitoring parameter collection and threshold comparison: Collecting data of at least one monitoring parameter from the monitored device in real time, and comparing the actual value of the monitoring parameter with a preset adaptive control threshold; Alarm information generation: Once an abnormal parameter is identified, an alarm message is immediately generated, which should include the ID of the monitored device, the ID of the abnormal parameter, the control threshold, the abnormal value, the alarm level, the alarm time and the processing status; Alarm level upgrade judgment: Key parameter judgment: If the parameter type of the abnormal parameter is a key parameter of the monitored device, the alarm level in the alarm information is upgraded. The upgrade rule is that if r_i>R_th, where r_i represents the state correlation coefficient of the i-th monitoring parameter, and R_th represents the set threshold, the alarm level is upgraded by one level; Alarm quantity judgment: If the alarm information generated for the same monitored device within the preset time period exceeds the set number, the alarm level will also be upgraded according to the alarm quantity threshold preset based on the device characteristics and business needs. The upgrade rule is if N>N_th, where N represents the number of alarms generated within the preset time period, and N_th represents the set quantity threshold, the alarm level will be upgraded by one level.

Images