A method and system for monitoring network faults of Internet of Things

Abstract

The present invention relates to the field of fault monitoring, and specifically to an Internet of Things network fault monitoring method and system. An Internet of Things network fault monitoring system includes: an Internet of Things network information data acquisition module, an Internet of Things network information data prediction construction module, an Internet of Things network information prediction model management module, an Internet of Things network information prediction model real-time training module, a network status deviation value calculation module and an Internet of Things network fault judgment module. The present invention judges the Internet of Things network information data of the current preset time period with the predicted Internet of Things network information data of the current preset time period, and predicts that the Internet of Things network information data can reflect the expected changes of the Internet of Things network information data under the regular Internet of Things network situation, and detects the actual Internet of Things network information data with the expected changes, so as to realize the fault monitoring of the Internet of Things network.

Description

Technical Field

[0001] The present invention relates to the field of fault monitoring, and in particular to a method and system for monitoring faults in an Internet of Things network. Background Art

[0002] With the rapid development of Internet of Things (IoT) technology, more and more devices and sensors are connected through the network, forming a huge IoT ecosystem. These devices cover various fields from smart homes, industrial automation to smart cities and smart healthcare. The popularity of IoT has not only improved the intelligence level of devices and systems, but also brought unprecedented data volume and complexity. For this reason, the reliability and stability of IoT networks have become crucial. However, IoT networks face many potential failure risks due to their wide distribution, diverse devices, and complex network structure. Summary of the invention

[0003] In the present invention, by judging the Internet of Things network information data of the current preset time period with the predicted Internet of Things network information data of the current preset time period, the predicted Internet of Things network information data can reflect the expected changes of the Internet of Things network information data under the regular Internet of Things network situation, and the actual Internet of Things network information data is detected by the expected changes, so as to realize the fault monitoring of the Internet of Things network; when an Internet of Things network failure occurs, the predicted Internet of Things network information data within the current preset time period will be used as the Internet of Things network information data within the current preset time period to predict the predicted Internet of Things network information data for the next preset time period, so as to avoid predicting the faulty Internet of Things network data and improve the prediction accuracy of the Internet of Things network information prediction model; at the same time, the Internet of Things network information prediction model is also trained in real time, so that the Internet of Things network information prediction model can adapt to the changes in the use of the Internet of Things.

[0004] A method for monitoring faults in an Internet of Things network, comprising:

[0005] In the current preset time period, for each analysis time point in the preset time period, perform the following operations: obtain the Internet of Things network information data at the analysis time point, calculate the similarity between the Internet of Things network information data and the predicted Internet of Things network information data corresponding to the analysis time point, and obtain the network status deviation value. The predicted Internet of Things network information data corresponding to the analysis time point is constructed by the Internet of Things network information data corresponding to all analysis time points in the previous preset time period, and then compare the network status deviation value with the deviation threshold. If the network status deviation value is greater than the deviation threshold, output the Internet of Things network fault and send the warning information to the maintenance personnel; if the network status deviation value is not greater than the deviation threshold, no operation is performed; the preset time period includes N analysis time points, and the time intervals between adjacent analysis time points are equal in length;

[0006] At the last analysis time point within the current preset time period, the Internet of Things network information data corresponding to the N analysis time points within the current preset time period are combined in chronological order to form an Internet of Things network information time series data set, and then the Internet of Things network information time series data set is sent to the Internet of Things network information prediction model for processing to construct the predicted Internet of Things network information data corresponding to the N analysis time points within the next preset time period; if an analysis time point of an output Internet of Things network failure occurs within the current preset time period, the predicted Internet of Things network information data corresponding to the N analysis time points within the current preset time period are directly used as the Internet of Things network information data corresponding to the N analysis time points within the current preset time period; and the Internet of Things network information prediction model is trained in real time.

[0007] Preferably, real-time training is performed on the IoT network information prediction model, specifically including the following steps:

[0008] At the last analysis time point in the current preset time period, the IoT network information time series data set in the previous preset time period and the IoT network information time series data set in the current preset time period are combined into a real-time training sample for IoT network information prediction, and the IoT network information prediction model is adjusted through the real-time training sample for IoT network information prediction to realize real-time training of the IoT network information prediction model; if an analysis time point of output IoT network failure occurs in the previous preset time period, the IoT network information time series data set in the current preset time period and the IoT network information time series data set in the M-2 preset time periods before the previous preset time period are combined into a real-time training sample for IoT network information prediction; The IoT network information fitting data set, M is the total number of IoT network information time series data sets in the IoT network information fitting training sample, and the IoT network information fitting data set is sent to the IoT network information fitting model for fitting, and the fitting IoT network information time series data set within the previous preset time period is obtained, and the fitting IoT network information time series data set within the previous preset time period is directly used as the IoT network information time series data set within the previous preset time period; if the analysis time point of the output IoT network failure occurs in both the previous preset time period and the current preset time period, or the analysis time point of the output IoT network failure occurs in the current preset time period, no operation is performed.

[0009] Preferably, the Internet of Things network information prediction model is established based on the LSTM model, including an LSTM layer, a first fully connected layer and a predicted Internet of Things network information data output layer, wherein the LSTM layer is used to receive an Internet of Things network information time series data set as input, analyze the timing characteristics in the Internet of Things network information time series data set, so as to construct an Internet of Things network information time series feature vector; the first fully connected layer is used to fully connect the Internet of Things network information time series feature vector to construct predicted Internet of Things network information data; the predicted Internet of Things network information data output layer is used to output predicted Internet of Things network information data.

[0010] Preferably, the training of the IoT network information prediction model includes the following steps:

[0011] A plurality of Internet of Things network information prediction training samples are obtained, the Internet of Things network information prediction training samples include Internet of Things network information time series data sets of two adjacent preset time periods, all Internet of Things network information prediction training samples are combined into an Internet of Things network information prediction training set, and then the Internet of Things network information prediction training set is sent to the Internet of Things network information prediction model with parameter initialization for training. During the training, the Internet of Things network information time series data set of the previous preset time period in the Internet of Things network information prediction training sample is used as input, and the Internet of Things network information time series data set of the next preset time period in the Internet of Things network information prediction training sample is used as the target output, the Internet of Things network information prediction loss value is calculated, and it is determined whether the Internet of Things network information prediction loss value is within a first preset range. If the Internet of Things network information prediction loss value is within the first preset range, the trained Internet of Things network information prediction model is output; otherwise, the Internet of Things network information prediction model is continuously trained through the Internet of Things network information prediction training set.

[0012] Preferably, the Internet of Things network information fitting model is established based on the Bi-LSTM model, specifically including a Bi-LSTM layer, a second fully connected layer and a fitting Internet of Things network information time series data set output layer, wherein the Bi-LSTM layer is used to receive the Internet of Things network information fitting data set as input, analyze the temporal characteristics of the context between the Internet of Things network information time series data sets within different preset time periods in the Internet of Things network information fitting data set, so as to construct an Internet of Things network information fitting data feature vector; the second fully connected layer is used to fully connect the Internet of Things network information fitting data feature vector to construct a fitting Internet of Things network information time series data set; the fitting Internet of Things network information time series data set output layer is used to output the fitting Internet of Things network information time series data set.

[0013] Preferably, the Internet of Things network information fitting model is trained, specifically comprising the following steps:

[0014] A plurality of Internet of Things network information fitting training samples are obtained, the Internet of Things network information fitting training samples include Internet of Things network information time series data sets of M preset time periods, all Internet of Things network information fitting training samples are combined into an Internet of Things network information fitting training set, and then the Internet of Things network information fitting training set is sent to the parameter-initialized Internet of Things network information fitting model for training. During the training, the Internet of Things network information time series data sets within the first M-2 preset time periods and the last preset time period in the Internet of Things network information fitting training samples are used as input, and the Internet of Things network information time series data set within the second to last preset time period in the Internet of Things network information fitting training samples is used as the target output, the Internet of Things network information fitting loss value is calculated, and it is determined whether the Internet of Things network information fitting loss value is within a second preset range. If the Internet of Things network information fitting loss value is within the second preset range, the trained Internet of Things network information fitting model is output; otherwise, the Internet of Things network information fitting model is continuously trained through the Internet of Things network information fitting training set.

[0015] Preferably, it also includes early prediction of IoT network failures, and the specific steps are as follows:

[0016] In the process of monitoring the Internet of Things network fault at the analysis time point, when the network status deviation value is not greater than the deviation threshold, the network status deviation value corresponding to the analysis time point is recorded, and each time the network status deviation value corresponding to an analysis time point is recorded, it is determined whether the network status deviation value corresponding to the current analysis time point is increasing compared with the network status deviation value corresponding to the previous analysis time point. If the network status deviation value corresponding to the current analysis time point is increasing compared with the network status deviation value corresponding to the previous analysis time point, the network status deviation value corresponding to the current analysis time point is recorded in the linear fitting set; if the network status deviation value corresponding to the current analysis time point is not increasing compared with the network status deviation value corresponding to the previous analysis time point, the elements in the linear fitting set are cleared;

[0017] When the number of elements in the linear fitting set reaches H, linear fitting is performed based on all network status deviation values ​​in the linear fitting set to construct a network status deviation value prediction function, and then the predicted network status deviation value corresponding to the next analysis time point is predicted by the network status deviation value prediction function, and the predicted network status deviation value is compared with the deviation threshold. If the predicted network status deviation value is greater than the deviation threshold, the IoT network prediction fault is output and the warning information is sent to the maintenance personnel; otherwise, the first element in the linear fitting set is deleted.

[0018] An Internet of Things network fault monitoring system, comprising:

[0019] The IoT network information data acquisition module is used to acquire IoT network information data at the analysis time point;

[0020] The predicted IoT network information data construction module is used to send the IoT network information data corresponding to all analysis time points in the previous preset time period into the IoT network information prediction model for processing, and construct the predicted IoT network information data in the current preset time period;

[0021] IoT network information prediction model management module, used for training and storing IoT network information prediction models;

[0022] The IoT network information prediction model real-time training module is used to perform real-time training on the IoT network information prediction model;

[0023] A network status deviation value calculation module is used to calculate the similarity between the IoT network information data and the predicted IoT network information data corresponding to the analysis time point to obtain a network status deviation value;

[0024] The IoT network fault judgment module is used to compare the network status deviation value with the deviation threshold. If the network status deviation value is greater than the deviation threshold, an IoT network fault is output; if the network status deviation value is not greater than the deviation threshold, no operation is performed.

[0025] The present invention has the following advantages:

[0026] The present invention judges the Internet of Things network information data of the current preset time period with the predicted Internet of Things network information data of the current preset time period. The predicted Internet of Things network information data can reflect the expected changes of the Internet of Things network information data under the regular Internet of Things network situation, and detects the actual Internet of Things network information data with the expected changes, so as to realize the fault monitoring of the Internet of Things network; when an Internet of Things network failure occurs, the predicted Internet of Things network information data within the current preset time period will be used as the Internet of Things network information data within the current preset time period to predict the predicted Internet of Things network information data for the next preset time period, so as to avoid predicting the faulty Internet of Things network data and improve the prediction accuracy of the Internet of Things network information prediction model; at the same time, the Internet of Things network information prediction model is also trained in real time, so that the Internet of Things network information prediction model can adapt to the changes in the use of the Internet of Things. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a schematic diagram of the structure of the Internet of Things network fault monitoring system adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0029] Embodiment 1, a method for monitoring Internet of Things network faults, comprising:

[0030] In the current preset time period, for each analysis time point in the preset time period, perform the following operations: obtain the Internet of Things network information data at the analysis time point, perform normalization operations on the Internet of Things network information data during the acquisition process, the Internet of Things network information data includes network traffic, bandwidth utilization, transmission delay and packet loss rate, etc., which can reflect the network status of the Internet of Things at the analysis time point, and perform similarity calculation between the Internet of Things network information data and the predicted Internet of Things network information data corresponding to the analysis time point. The calculation method can use the cosine similarity algorithm to obtain the network status deviation value. The predicted Internet of Things network information data corresponding to the analysis time point is constructed by the Internet of Things network information data corresponding to all analysis time points in the previous preset time period, and then the network status deviation value is compared with the deviation threshold. , the deviation threshold is set by the developer. If the network status deviation value is greater than the deviation threshold, it means that the network status at the current analysis time point does not conform to the change trend, and the IoT network fault is output, and a warning message is sent to the maintenance personnel. The sending method can be a text message to remind the maintenance personnel to inspect the IoT; if the network status deviation value is not greater than the deviation threshold, it means that the network status at the current analysis time point conforms to the change trend, and no operation is required; the preset time period is set by the developer according to the IoT network status change cycle, which is generally 1 day. In the process of IoT executing corresponding performance, since the work tasks executed every day are regular, the changes in the IoT network status are also regular on a daily basis, and the preset time period includes N analysis time points, and the time intervals between adjacent analysis time points are equal in length;

[0031] At the last analysis time point within the current preset time period, the Internet of Things network information data corresponding to the N analysis time points within the current preset time period are combined in chronological order to form an Internet of Things network information time series data set, and then the Internet of Things network information time series data set is sent to the Internet of Things network information prediction model for processing to construct the predicted Internet of Things network information data corresponding to the N analysis time points within the next preset time period; if an analysis time point of an output Internet of Things network failure occurs within the current preset time period, the predicted Internet of Things network information data corresponding to the N analysis time points within the current preset time period are directly used as the Internet of Things network information data corresponding to the N analysis time points within the current preset time period; and the Internet of Things network information prediction model is trained in real time.

[0032] The present application judges the IoT network information data of the current preset time period with the predicted IoT network information data of the current preset time period. The predicted IoT network information data can reflect the expected changes of the IoT network information data under the regular IoT network situation, and the actual IoT network information data is detected by the expected changes to realize fault monitoring of the IoT network; when an IoT network failure occurs, the predicted IoT network information data within the current preset time period will be used as the IoT network information data within the current preset time period to predict the predicted IoT network information data for the next preset time period, so as to avoid predicting the faulty IoT network data and improve the prediction accuracy of the IoT network information prediction model; at the same time, the IoT network information prediction model is trained in real time so that the IoT network information prediction model can adapt to changes in IoT usage.

[0033] The IoT network information prediction model is established based on the LSTM model, including an LSTM layer, a first fully connected layer and a predicted IoT network information data output layer, wherein the LSTM layer is used to receive the IoT network information time series data set as input, analyze the timing characteristics in the IoT network information time series data set to construct the IoT network information time series feature vector; the first fully connected layer is used to fully connect the IoT network information time series feature vector to construct the predicted IoT network information data; the predicted IoT network information data output layer is used to output the predicted IoT network information data.

[0034] Real-time training of the IoT network information prediction model includes the following steps:

[0035] At the last analysis time point within the current preset time period, the IoT network information time series data set within the previous preset time period and the IoT network information time series data set within the current preset time period are combined into a real-time training sample for IoT network information prediction, and the IoT network information prediction model is adjusted through the real-time training sample for IoT network information prediction to achieve real-time training of the IoT network information prediction model; if an analysis time point at which an IoT network failure is output occurs within the previous preset time period, the IoT network information time series data set within the current preset time period and the IoT network information time series data set within the M-2 preset time periods before the previous preset time period are combined into a IoT network information fitting data set Set, M is the total number of IoT network information time series data sets in the IoT network information fitting training samples, and the IoT network information fitting data set is sent to the IoT network information fitting model for fitting, and the fitting IoT network information time series data set within the previous preset time period is obtained, and the fitting IoT network information time series data set within the previous preset time period is directly used as the IoT network information time series data set within the previous preset time period; if the analysis time point of the output IoT network failure occurs in both the previous preset time period and the current preset time period, or the analysis time point of the output IoT network failure occurs in the current preset time period, no operation is performed, that is, the real-time training of the IoT network information prediction model is not performed.

[0036] The training of the IoT network information prediction model includes the following steps:

[0037] A number of IoT network information prediction training samples are obtained. The IoT network information prediction training samples include IoT network information time series data sets of two adjacent preset time periods. It should be noted that the IoT network information time series data sets are constructed by developers based on actual IoT network information data when the IoT is used. All IoT network information prediction training samples form an IoT network information prediction training set, and then the IoT network information prediction training set is sent to the parameter-initialized IoT network information prediction model for training. During the training, the IoT network information time series data set of the previous preset time period in the IoT network information prediction training sample is used as input, and the IoT network information time series data set of the next preset time period in the IoT network information prediction training sample is used as the target output. The IoT network information prediction loss value is calculated to determine whether the IoT network information prediction loss value is within a first preset range. The first preset range is set by the developer. If the IoT network information prediction loss value is within the first preset range, the trained IoT network information prediction model is output; otherwise, the IoT network information prediction model is continuously trained through the IoT network information prediction training set.

[0038] The Internet of Things network information fitting model is established based on the Bi-LSTM model, and specifically includes a Bi-LSTM layer, a second fully connected layer and a fitting Internet of Things network information time series data set output layer, wherein the Bi-LSTM layer is used to receive the Internet of Things network information fitting data set as input, analyze the temporal characteristics of the context between the Internet of Things network information time series data sets within different preset time periods in the Internet of Things network information fitting data set, so as to construct the Internet of Things network information fitting data feature vector; the second fully connected layer is used to fully connect the Internet of Things network information fitting data feature vector to construct the fitting Internet of Things network information time series data set; the fitting Internet of Things network information time series data set output layer is used to output the fitting Internet of Things network information time series data set.

[0039] Training the IoT network information fitting model includes the following steps:

[0040] A plurality of Internet of Things network information fitting training samples are obtained, the Internet of Things network information fitting training samples include Internet of Things network information time series data sets of M preset time periods, all Internet of Things network information fitting training samples are combined into an Internet of Things network information fitting training set, and then the Internet of Things network information fitting training set is sent to the parameter-initialized Internet of Things network information fitting model for training. During the training, the Internet of Things network information time series data sets within the first M-2 preset time periods and the last preset time period in the Internet of Things network information fitting training samples are used as input, and the Internet of Things network information time series data set within the second to last preset time period in the Internet of Things network information fitting training samples is used as the target output, the Internet of Things network information fitting loss value is calculated, and it is determined whether the Internet of Things network information fitting loss value is within a second preset range, and the second preset range is set by the developer. If the Internet of Things network information fitting loss value is within the second preset range, the trained Internet of Things network information fitting model is output; otherwise, the Internet of Things network information fitting model is continuously trained through the Internet of Things network information fitting training set.

[0041] In the actual process of fault monitoring of the IoT network, each time a fault is judged at an analysis time point, the actual IoT network information data has been obtained, that is, the fault has actually occurred. Here is a method for predicting IoT network faults in advance, which specifically includes the following steps:

[0042] In the process of monitoring the Internet of Things network fault at the analysis time point, when the network status deviation value is not greater than the deviation threshold, the network status deviation value corresponding to the analysis time point is recorded, and each network status deviation value corresponding to an analysis time point is recorded to determine whether the network status deviation value corresponding to the current analysis time point is increasing compared to the network status deviation value corresponding to the previous analysis time point. It should be noted that if the network status deviation value corresponding to the previous analysis time point is not recorded, it may be that an Internet of Things network fault has occurred. At this time, no comparison is made, but the next network status deviation value is directly recorded. If the network status deviation value corresponding to the current analysis time point is increasing compared to the network status deviation value corresponding to the previous analysis time point, the network status deviation value corresponding to the current analysis time point is recorded in the linear fitting set; if the network status deviation value corresponding to the current analysis time point is not increasing compared to the network status deviation value corresponding to the previous analysis time point, the elements in the linear fitting set are cleared;

[0043] When the number of elements in the linear fitting set reaches H, the H value is set by the square root personnel, generally 7. Linear fitting is performed based on all network status deviation values ​​in the linear fitting set to construct a network status deviation value prediction function. The predicted network status deviation value corresponding to the next analysis time point is predicted by the network status deviation value prediction function. The predicted network status deviation value is compared with the deviation threshold. If the predicted network status deviation value is greater than the deviation threshold, it means that an IoT network failure will occur at the next analysis time point. The IoT network predicted failure is output and a warning message is sent to the maintenance personnel. Otherwise, the first element in the linear fitting set is deleted. By predicting the network status deviation value in advance, the IoT network failure can be predicted in advance.

[0044] Embodiment 2, a network fault monitoring system of the Internet of Things, such as Figure 1 As shown, including:

[0045] The IoT network information data acquisition module is used to acquire IoT network information data at the analysis time point, and perform normalization operations on the IoT network information data during the acquisition process. The IoT network information data includes network traffic, bandwidth utilization, transmission delay and packet loss rate, etc., which can reflect the network status of the IoT at the analysis time point;

[0046] The predicted IoT network information data construction module is used to send the IoT network information data corresponding to all analysis time points in the previous preset time period into the IoT network information prediction model for processing, and construct the predicted IoT network information data in the current preset time period;

[0047] IoT network information prediction model management module, used for training and storing IoT network information prediction models;

[0048] The IoT network information prediction model real-time training module is used to perform real-time training on the IoT network information prediction model;

[0049] A network status deviation value calculation module is used to calculate the similarity between the IoT network information data and the predicted IoT network information data corresponding to the analysis time point to obtain a network status deviation value;

[0050] The Internet of Things network fault judgment module is used to compare the network status deviation value with the deviation threshold. If the network status deviation value is greater than the deviation threshold, it means that the network status at the current analysis time point does not conform to the change trend, and the Internet of Things network fault is output, and a warning message is sent to the maintenance personnel. The sending method can be SMS, reminding the maintenance personnel to inspect the Internet of Things; if the network status deviation value is not greater than the deviation threshold, it means that the network status at the current analysis time point conforms to the change trend, and no operation is performed.

[0051] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for monitoring Internet of Things network faults, characterized in that: include: In the current preset time period, for each analysis time point in the preset time period, perform the following operations: obtain the Internet of Things network information data at the analysis time point, calculate the similarity between the Internet of Things network information data and the predicted Internet of Things network information data corresponding to the analysis time point, and obtain the network status deviation value. The predicted Internet of Things network information data corresponding to the analysis time point is constructed by the Internet of Things network information data corresponding to all analysis time points in the previous preset time period, and then compare the network status deviation value with the deviation threshold. If the network status deviation value is greater than the deviation threshold, output the Internet of Things network fault and send the warning information to the maintenance personnel; if the network status deviation value is not greater than the deviation threshold, no operation is performed; the preset time period includes N analysis time points, and the time intervals between adjacent analysis time points are equal in length; At the last analysis time point within the current preset time period, the IoT network information data corresponding to the N analysis time points within the current preset time period are combined in chronological order to form an IoT network information time series data set, and then the IoT network information time series data set is sent to the IoT network information prediction model for processing to construct the predicted IoT network information data corresponding to the N analysis time points within the next preset time period; if an analysis time point at which an IoT network failure is output occurs within the current preset time period, the predicted IoT network information data corresponding to the N analysis time points within the current preset time period are directly used as the IoT network information data corresponding to the N analysis time points within the current preset time period; and the IoT network information prediction model is trained in real time; Real-time training of the IoT network information prediction model includes the following steps: At the last analysis time point in the current preset time period, the IoT network information time series data set in the previous preset time period and the IoT network information time series data set in the current preset time period are combined into a real-time training sample for IoT network information prediction, and the IoT network information prediction model is adjusted through the real-time training sample for IoT network information prediction to realize real-time training of the IoT network information prediction model; if an analysis time point of output IoT network failure occurs in the previous preset time period, the IoT network information time series data set in the current preset time period and the IoT network information time series data set in the M-2 preset time periods before the previous preset time period are combined into a real-time training sample for IoT network information prediction; The IoT network information fitting data set, M is the total number of IoT network information time series data sets in the IoT network information fitting training sample, and the IoT network information fitting data set is sent to the IoT network information fitting model for fitting, and the fitting IoT network information time series data set within the previous preset time period is obtained, and the fitting IoT network information time series data set within the previous preset time period is directly used as the IoT network information time series data set within the previous preset time period; if the analysis time point of the output IoT network failure occurs in both the previous preset time period and the current preset time period, or the analysis time point of the output IoT network failure occurs in the current preset time period, no operation is performed.

2. The method for monitoring Internet of Things network faults according to claim 1, characterized in that: The IoT network information prediction model is established based on the LSTM model, including an LSTM layer, a first fully connected layer, and a predicted IoT network information data output layer. The LSTM layer is used to receive the IoT network information time series data set as input, analyze the time series characteristics in the IoT network information time series data set, and construct the IoT network information time series feature vector; The first fully connected layer is used to fully connect the IoT network information time series feature vector to construct predicted IoT network information data; the predicted IoT network information data output layer is used to output predicted IoT network information data.

3. The method for monitoring Internet of Things network faults according to claim 2, characterized in that: The training of the IoT network information prediction model includes the following steps: A plurality of Internet of Things network information prediction training samples are obtained, the Internet of Things network information prediction training samples include Internet of Things network information time series data sets of two adjacent preset time periods, all Internet of Things network information prediction training samples are combined into an Internet of Things network information prediction training set, and then the Internet of Things network information prediction training set is sent to the Internet of Things network information prediction model with parameter initialization for training. During the training, the Internet of Things network information time series data set of the previous preset time period in the Internet of Things network information prediction training sample is used as input, and the Internet of Things network information time series data set of the next preset time period in the Internet of Things network information prediction training sample is used as the target output, the Internet of Things network information prediction loss value is calculated, and it is determined whether the Internet of Things network information prediction loss value is within a first preset range. If the Internet of Things network information prediction loss value is within the first preset range, the trained Internet of Things network information prediction model is output; otherwise, the Internet of Things network information prediction model is continuously trained through the Internet of Things network information prediction training set.

4. The method for monitoring Internet of Things network faults according to claim 3, characterized in that: The IoT network information fitting model is established based on the Bi-LSTM model, which specifically includes a Bi-LSTM layer, a second fully connected layer, and an output layer for fitting IoT network information time series data sets, wherein the Bi-LSTM layer is used to receive the IoT network information fitting data set as input, analyze the temporal characteristics of the context between the IoT network information time series data sets within different preset time periods in the IoT network information fitting data set, and construct a feature vector of IoT network information fitting data; The second fully connected layer is used to fully connect the IoT network information fitting data feature vector to construct a fitting IoT network information time series data set; the fitting IoT network information time series data set output layer is used to output the fitting IoT network information time series data set.

5. The method for monitoring Internet of Things network faults according to claim 4, characterized in that: Training the IoT network information fitting model includes the following steps: A plurality of Internet of Things network information fitting training samples are obtained, the Internet of Things network information fitting training samples include Internet of Things network information time series data sets of M preset time periods, all Internet of Things network information fitting training samples are combined into an Internet of Things network information fitting training set, and then the Internet of Things network information fitting training set is sent to the parameter-initialized Internet of Things network information fitting model for training. During the training, the Internet of Things network information time series data sets within the first M-2 preset time periods and the last preset time period in the Internet of Things network information fitting training samples are used as input, and the Internet of Things network information time series data set within the second to last preset time period in the Internet of Things network information fitting training samples is used as the target output, the Internet of Things network information fitting loss value is calculated, and it is determined whether the Internet of Things network information fitting loss value is within a second preset range. If the Internet of Things network information fitting loss value is within the second preset range, the trained Internet of Things network information fitting model is output; otherwise, the Internet of Things network information fitting model is continuously trained through the Internet of Things network information fitting training set.

6. The method for monitoring Internet of Things network faults according to claim 5, characterized in that: It also includes early prediction of IoT network failures. The specific steps are as follows: In the process of monitoring the Internet of Things network fault at the analysis time point, when the network status deviation value is not greater than the deviation threshold, the network status deviation value corresponding to the analysis time point is recorded, and each time the network status deviation value corresponding to an analysis time point is recorded, it is determined whether the network status deviation value corresponding to the current analysis time point is increasing compared to the network status deviation value corresponding to the previous analysis time point. If the network status deviation value corresponding to the current analysis time point is increasing compared to the network status deviation value corresponding to the previous analysis time point, the network status deviation value corresponding to the current analysis time point is recorded in the linear fitting set; If the network status deviation value corresponding to the current analysis time point does not increase compared with the network status deviation value corresponding to the previous analysis time point, the elements in the linear fitting set are cleared; When the number of elements in the linear fitting set reaches H, linear fitting is performed based on all network status deviation values ​​in the linear fitting set to construct a network status deviation value prediction function, and then the predicted network status deviation value corresponding to the next analysis time point is predicted by the network status deviation value prediction function, and the predicted network status deviation value is compared with the deviation threshold. If the predicted network status deviation value is greater than the deviation threshold, the IoT network prediction fault is output and the warning information is sent to the maintenance personnel; otherwise, the first element in the linear fitting set is deleted.

7. An Internet of Things network fault monitoring system, characterized in that: The system applies an Internet of Things network fault monitoring method as described in any one of claims 1 to 6, including: The IoT network information data acquisition module is used to acquire IoT network information data at the analysis time point; The predicted IoT network information data construction module is used to send the IoT network information data corresponding to all analysis time points in the previous preset time period into the IoT network information prediction model for processing, and construct the predicted IoT network information data in the current preset time period; IoT network information prediction model management module, used for training and storing IoT network information prediction models; The IoT network information prediction model real-time training module is used to perform real-time training on the IoT network information prediction model; A network status deviation value calculation module is used to calculate the similarity between the IoT network information data and the predicted IoT network information data corresponding to the analysis time point to obtain a network status deviation value; The IoT network fault judgment module is used to compare the network status deviation value with the deviation threshold. If the network status deviation value is greater than the deviation threshold, an IoT network fault is output; if the network status deviation value is not greater than the deviation threshold, no operation is performed.

Images