CN117834385A - Fault early warning processing method based on Internet of things system and Internet of things system - Google Patents
Fault early warning processing method based on Internet of things system and Internet of things system Download PDFInfo
- Publication number
- CN117834385A CN117834385A CN202311675441.7A CN202311675441A CN117834385A CN 117834385 A CN117834385 A CN 117834385A CN 202311675441 A CN202311675441 A CN 202311675441A CN 117834385 A CN117834385 A CN 117834385A
- Authority
- CN
- China
- Prior art keywords
- fault
- data
- layer
- neural network
- state
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 25
- 238000003062 neural network model Methods 0.000 claims abstract description 74
- 238000000034 method Methods 0.000 claims abstract description 20
- 238000013528 artificial neural network Methods 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims description 53
- 239000013598 vector Substances 0.000 claims description 39
- 238000000605 extraction Methods 0.000 claims description 29
- 238000011176 pooling Methods 0.000 claims description 23
- 230000007246 mechanism Effects 0.000 claims description 11
- 230000007613 environmental effect Effects 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 7
- 230000010354 integration Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 abstract description 9
- 238000010586 diagram Methods 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/06—Management of faults, events, alarms or notifications
- H04L41/0631—Management of faults, events, alarms or notifications using root cause analysis; using analysis of correlation between notifications, alarms or events based on decision criteria, e.g. hierarchy, tree or time analysis
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0823—Errors, e.g. transmission errors
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Environmental & Geological Engineering (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer And Data Communications (AREA)
Abstract
The embodiment of the application discloses a fault early warning processing method based on an Internet of things system and the Internet of things system. The method specifically comprises the following steps: the industrial Internet of things equipment detects a plurality of state data of the industrial Internet of things equipment and sends at least one target state data which accords with data uploading conditions to an edge server; the edge server adopts a lightweight neural network to perform preliminary fault analysis, and when the edge server analyzes that faults occur, relevant data are uploaded to the cloud server, and a large-scale neural network model is adopted to perform further fault analysis prediction; and early warning is carried out when the cloud server confirms the fault. The scheme can reduce the data uploading quantity and improve the accuracy of fault early warning.
Description
Technical Field
The application relates to the technical fields of the Internet of things and computers, in particular to a fault early warning processing method based on an Internet of things system and the Internet of things system.
Background
The internet of things (Internet of Things, ioT) refers to the concept of connecting everyday physical devices and objects to the internet and enabling data exchange and interconnection through wireless sensors, embedded systems and network communication technologies. Briefly, the internet of things refers to a network of various devices, sensors, and objects that are connected and communicate through the internet.
The core idea of the internet of things is to realize the integration of the physical world and the digital world, so that the intelligent equipment can collect, transmit and analyze data, and realize automatic and intelligent interaction and service based on the data. The internet of things can cover various fields including home automation, smart cities, industrial automation, agriculture, medical care and the like.
In the related application, the device fault early warning can be realized based on the internet of things (IoT) platform, specifically, the internet of things device acquisition data is uploaded to the IoT internet of things platform for fault analysis processing. However, at present, the problem that the fault early warning accuracy of the internet of things platform of the IoT is low due to the fact that the data uploading amount of equipment is large and the fault is analyzed by adopting a simple numerical comparison mode exists in the fault early warning of the internet of things.
Disclosure of Invention
The main purpose of the application is to provide a fault early warning processing method based on an Internet of things system and the Internet of things system, so that the data uploading amount of the industrial Internet of things equipment fault early warning based on the Internet of things system can be reduced, and the early warning accuracy can be improved.
In order to achieve the above purpose, the embodiment of the application provides a fault early warning processing method based on an internet of things system, wherein the internet of things system comprises an edge server, a cloud server and industrial internet of things equipment located in an industrial park; the edge server, the cloud server and the industrial Internet of things device are connected through a network, and the early warning processing method comprises the following steps:
the industrial Internet of things equipment detects a plurality of state data of the industrial Internet of things equipment, judges whether the state data meets a data uploading condition, and sends at least one target state data meeting the condition to the edge server when the state data meets the data uploading condition;
the edge server performs feature extraction processing on target state data by adopting a first neural network model to obtain a state feature set, predicts the fault probability of the industrial Internet of things equipment according to the state feature set, and sends the target state data to the cloud server when the fault probability is greater than a first fault threshold;
the cloud server performs fault prediction processing based on the target state data by adopting a second neural network model to obtain a fault prediction result, wherein the first neural network model is a lightweight neural network model, and the number of model parameters, model complexity and hierarchical structure depth of the second neural network model are larger than those of the first neural network model;
and when the fault prediction result indicates that the industrial Internet of things equipment fails, the cloud server sends early warning information to an early warning terminal.
In an embodiment, determining whether the status data satisfies a data upload condition includes:
identifying a state type of the state data, and acquiring an uploading threshold value corresponding to the state type;
and if the threshold value of the state data is larger than the uploading threshold value, determining that the state data meets the data uploading condition.
In an embodiment, the first neural network model includes at least one convolutional layer, a pooling layer, and a fully-connected layer;
the edge server adopts a first neural network model to conduct feature extraction processing on a plurality of target state data to obtain a state feature set, predicts the fault probability of the industrial Internet of things equipment according to the state feature set, and comprises the following steps:
the edge server adopts a convolution layer to carry out convolution feature extraction processing on the target state data to obtain a state feature set;
carrying out feature pooling treatment on the state feature set by adopting the pooling layer;
and predicting the fault probability of the industrial Internet of things equipment according to the state characteristic set through the full connection layer.
In one embodiment, the first neural network model includes two convolution layers and a pooling layer, the convolution kernel of the convolution layers being 3 by 3, and the size of the pooling layer being 2 by 2.
In an embodiment, the edge server performs convolution feature extraction processing on the target state data by using a convolution layer to obtain a state feature set, including:
and the edge server adopts depth separable convolution to carry out convolution feature extraction processing on the target state data so as to obtain a state feature set.
In an embodiment, the second neural network model comprises: an input encoding layer, an attention layer, a multi-layer encoder, and an output layer; the cloud server performs fault prediction processing based on the target state data by adopting a second neural network model to obtain a fault prediction result, and the method comprises the following steps:
the input coding layer performs feature extraction on a plurality of target state data to obtain a state feature sequence, and performs feature coding processing on the target state feature sequence to obtain a target state feature vector;
the attention layer extracts attention characteristic vectors of key characteristics in the attention state characteristic sequence based on an attention mechanism;
the multi-layer encoder performs multi-layer abstraction and representation on the principal force feature vector to obtain a high-layer feature vector, wherein the high-layer feature vector at least represents relation information and context information among features in the state feature sequence;
and the output layer carries out fault prediction processing on the industrial Internet of things equipment according to the high-level feature vector to obtain a fault prediction result.
In an embodiment, the output layer performs fault prediction processing on the industrial internet of things device according to the high-level feature vector to obtain a fault prediction result, and the method includes:
the output layer performs probability mapping processing on the industrial Internet of things equipment according to the high-level feature vector, and predicts a second fault probability of the industrial Internet of things equipment;
and when the second fault probability is larger than a second fault threshold, determining that the industrial Internet of things equipment breaks down, wherein the second fault threshold is larger than the first fault threshold.
In one embodiment, each encoder is comprised of a multi-headed self-attention mechanism and a feed-forward neural network.
Correspondingly, the embodiment of the application also provides an Internet of things system, wherein the Internet of things system comprises an edge server, a cloud server, an early warning terminal and industrial Internet of things equipment located in an industrial park; the edge server, the cloud server and the industrial Internet of things device are connected through a network, and the early warning processing method comprises the following steps:
the industrial Internet of things equipment is used for detecting a plurality of state data of the industrial Internet of things equipment, judging whether the state data meet data uploading conditions, and sending at least one target state data meeting the conditions to the edge server when the state data meet the data uploading conditions;
the edge server is used for carrying out feature extraction processing on target state data by adopting a first neural network model to obtain a state feature set, predicting the fault probability of the industrial Internet of things equipment according to the state feature set, and sending the target state data to the cloud server when the fault probability is greater than a first fault threshold;
the cloud server is used for performing fault prediction processing based on the target state data by adopting a second neural network model to obtain a fault prediction result, wherein the first neural network model is a lightweight neural network model, and the number of model parameters, the model complexity and the hierarchical structure depth of the second neural network model are greater than those of the first neural network model; and when the fault prediction result indicates that the industrial Internet of things equipment fails, the cloud server sends early warning information to an early warning terminal.
The fault early warning processing method based on the Internet of things system provided by the embodiment of the application specifically comprises the following steps: the industrial Internet of things equipment detects a plurality of state data of the industrial Internet of things equipment, at least one target state data meeting data uploading conditions is sent to an edge server, a lightweight neural network is adopted at the edge server to conduct preliminary fault analysis, when faults occur in the analysis of the edge server, the relevant data are uploaded to a cloud server to conduct further fault analysis prediction through a large-scale neural network model, and when the cloud server confirms the faults, early warning is conducted; according to the scheme, cloud edge cooperation is adopted, early warning is carried out on the edge based on a lightweight model, uploading data is filtered, the cloud data uploading amount and processing amount of an internet of things (IoT) platform are reduced, the data uploading amount and the lifting efficiency can be reduced, and meanwhile the accuracy of fault early warning is improved through cloud edge two-stage fault analysis.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of a scenario of a fault early warning processing method based on an internet of things system in an embodiment of the present application;
fig. 2 is a flowchart of a fault early warning processing method based on an internet of things system in an embodiment of the present application;
fig. 3 is a schematic structural diagram of a first neural network model according to an embodiment of the present application;
FIG. 4 is a schematic structural diagram of a second neural network model according to an embodiment of the present disclosure;
fig. 5 is another flow chart of a fault early warning processing method based on the internet of things system according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.
The embodiment of the application provides a fault early warning processing method based on an internet of things system and the internet of things system, and the fault early warning processing method and the internet of things system are respectively described in detail below.
As shown in fig. 1, the internet of things system realizes fault early warning for industrial internet of things equipment, where the internet of things system may include: the system comprises an edge server, an early warning terminal, a cloud server and industrial Internet of things equipment located in an industrial park; the edge server, the cloud server and the industrial Internet of things device are connected through a network. The cloud server and the edge server can be connected through a cloud network and can be located in the cloud network. In an embodiment, the internet of things may be composed of industrial internet of things equipment, an edge server, an early warning terminal, a cloud server and a cloud network.
The edge server can be a server which is close to the data source position at the edge of the internet of things, and the edge server is used as a computing node deployed at the edge of the internet of things to calculate and process real-time data.
Cloud servers, which may refer to virtual server instances or physical servers deployed on a cloud computing platform. They may be part of a cloud computing infrastructure that divides a physical server into multiple virtual, independent server environments by sharing underlying hardware resources and using virtualization techniques.
Industrial internet of things devices may refer to industrial devices that access the internet of things, including, for example, manufacturing equipment, industrial machinery, industrial controllers, intelligent transportation devices, intelligent instruments, and the like.
The early warning terminal can be electronic equipment for providing early warning reminding, and comprises a notebook computer, a mobile phone, a tablet computer, monitoring center equipment, early warning center equipment and the like.
As shown in fig. 1, the process for implementing fault early warning of industrial internet of things equipment based on the internet of things system comprises the following steps: the industrial Internet of things equipment detects a plurality of state data of the industrial Internet of things equipment, judges whether the state data meets the data uploading condition, and sends at least one target state data meeting the condition to the edge server when the state data meets the data uploading condition;
the edge server performs feature extraction processing on the target state data by adopting a first neural network model to obtain a state feature set, predicts the fault probability of the industrial Internet of things equipment according to the state feature set, and sends the target state data to the cloud server when the fault probability is greater than a first fault threshold;
the cloud server performs fault prediction processing based on the target state by adopting a second neural network model to obtain a fault prediction result, wherein the first neural network is a lightweight neural network model, and the number of model parameters, the model complexity and the hierarchical structure depth of the second neural network model are greater than those of the first neural network model;
when the failure prediction result indicates that the industrial Internet of things equipment fails, the cloud server sends early warning information to the early warning terminal, and monitoring personnel receive the early warning information to take corresponding failure resolution measures.
In some embodiments, as shown in fig. 2, the fault early-warning processing method based on the internet of things system provided in the embodiment of the present application specifically includes:
201. the industrial Internet of things device detects a plurality of state data of the industrial Internet of things device.
Wherein, the status data is data representing the status of the industrial Internet of things equipment, and may include:
sensor data: industrial internet of things devices are typically equipped with a plurality of sensors to monitor various physical quantities, such as temperature, humidity, pressure, vibration, current, voltage, etc. The sensor data records the real-time status of the environment in which the device is located.
Device operating parameters: the operating parameters of the industrial internet of things device describe the state and operating conditions of the device, such as rotational speed, power, load, operating time, operating mode, etc. These parameters may reflect the operating state of the device at a particular point in time.
Error code and log: the error code of the device records the fault or error code that occurs during operation of the device, and the log of the device contains the event and operational records of the device system. Such information may help identify a fault condition of the device.
Operation record: such data includes operation records, maintenance records, repair records, and the like of the device. By analysis of these records, historical operating conditions and fault conditions of the device can be revealed.
Environmental factors (environmental data): the state of industrial internet of things devices is also affected by environmental factors such as ambient temperature, humidity, air pressure, etc. The data for these environmental factors may help understand the operating environment and possible impact of the device.
In practical applications, the status data to be collected or used may be selected according to requirements, for example, in an embodiment, the type of the status data includes at least one of sensor data, device operation parameters, device error log, and environmental data.
202. Whether the status data satisfies the data uploading condition is determined, and when the status data satisfies the data uploading condition, step 203 is executed, and when the status data does not satisfy the data uploading condition, the status data is not uploaded.
In one embodiment, when 10 pieces of status data of the device are collected, if five pieces of status data satisfy the upload condition, the five pieces of data are sent to the edge server. For example, the status data includes temperature, humidity, pressure, vibration, current, voltage, etc., and if the values of temperature and humidity exceed the upload threshold, the temperature and humidity data is sent to the edge server.
The data uploading condition is a condition for judging whether to upload the state data, for example, the uploading threshold corresponding to the state data can be set according to the actual requirement and the type of the state data, for example, when the state data is temperature, humidity, pressure, vibration, current, voltage and the like, the data can be uploaded when the acquired values of the humidity, the pressure, the vibration, the current and the voltage exceed a certain threshold.
In an embodiment, determining whether the status data satisfies the data upload condition may include:
identifying a state type of the state data, and acquiring an uploading threshold value corresponding to the state type;
if the threshold value of the state data is larger than the uploading threshold value, determining that the state data meets the data uploading condition.
Wherein the status types include: sensor data (temperature, humidity, pressure, vibration, current, voltage, etc.), equipment operating parameters, equipment error logs, environmental data, etc., may be specifically defined and set according to actual requirements.
For example, when the state data is identified as the temperature, an uploading threshold corresponding to the temperature is acquired, and when the current temperature value is larger than the uploading threshold, the temperature data is uploaded.
According to the method and the device, the data filtering condition can be set on the side of the Internet of things terminal device such as the industrial Internet of things device, the data uploading amount is reduced, resources can be saved, and the response speed and efficiency of fault early warning can be improved.
203. At least one target state data is sent to the edge server that meets the condition.
The target state data is data satisfying a data uploading condition in the state data, for example, when the state data includes sensor data (temperature, humidity, pressure, vibration, current, voltage, etc.), an operating parameter of the device, an error log of the device, and environmental data, if the sensor data satisfies the uploading condition, the sensor data is the target state data, and the operating parameter of the device satisfies the uploading condition is the target state data.
204. And the edge server performs feature extraction processing on the target state data by adopting a first neural network model to obtain a state feature set, predicts the fault probability of the industrial Internet of things equipment according to the state feature set, and sends the target state data to the cloud server when the fault probability is greater than a first fault threshold.
The first neural network model may be a lightweight neural network model, and the lightweight neural network refers to a neural network model with smaller parameter number and smaller calculation amount complexity. May include: a lightweight convolutional neural network model, a high efficiency neural network model, and so on. The specific network model can be selected according to the actual scene and the requirements.
The state feature set comprises at least one state feature, and the state feature characterizes the state of the industrial Internet of things equipment, and can comprise sensor data features, equipment operation parameter features, equipment error log features, environment data features and the like.
In an embodiment, the first neural network model may be a lightweight convolutional neural network model including at least one convolutional layer, a pooling layer, and a fully-connected layer, and as shown in fig. 3, the first neural network may include 1 to 2 convolutional layers, one pooling layer, and one fully-connected layer.
The step of performing feature extraction processing on the plurality of target state data by the edge server by adopting the first neural network model to obtain a state feature set, and predicting the failure probability of the industrial internet of things device according to the state feature set may include:
the edge server adopts a convolution layer to carry out convolution feature extraction processing on the target state data to obtain a state feature set;
carrying out feature pooling treatment on the state feature set by adopting a pooling layer;
and predicting the fault probability of the industrial Internet of things equipment according to the state characteristic set through the full connection layer.
In an embodiment, to reduce computational complexity and improve failure prediction efficiency, when the first neural network model is a convolutional network structure, fewer convolutional layers and pooling layers are used: the depth of the network is reduced, and only the necessary convolution layers and pooling layers are included. For example, a simple network fabric volume containing only 1 to 2 convolution layers and a maximum pooling layer may be selected.
In an embodiment, to reduce computational complexity and improve failure prediction efficiency, the first neural network model may reduce the number and size of convolution kernels: fewer convolution kernels and smaller sizes are employed to reduce the number of parameters and computational complexity per convolution layer. A 3x3 small size convolution kernel may be used.
In one embodiment, for the size of the pooling window may also be reduced: smaller pooling window sizes, such as 2x2 or 3x3, are used to reduce the size and computational effort of the output feature map.
In one embodiment, to improve feature extraction accuracy and fault prediction accuracy, a convolution operation may be performed using a depth separable convolution (Depthwise Separable Convolution). The step of performing convolution feature extraction processing on the target state data by the edge server by adopting a convolution layer to obtain a state feature set may include: the edge server adopts depth separable convolution to carry out convolution feature extraction processing on the target state data, and a state feature set is obtained.
Depth separable convolution is a lightweight convolution operation that can efficiently extract features. It decomposes the standard convolution operation into two steps, depth convolution and point-by-point convolution. The following is the working principle of depth separable convolution and the feature extraction process:
deep convolution (Depthwise Convolution):
the deep convolution only performs the convolution operation on each channel of the input signature, not all together across the input channel. This means that for each channel of the input profile there is a corresponding convolution kernel.
The depth convolution may capture the spatial correlation in the input signature because the convolution kernel for each channel only focuses on the characteristics of the particular channel.
Point-wise convolution (Pointwise Convolution):
the point-by-point convolution is a 1x1 convolution operation applied after the depth convolution, which convolves the result of the depth convolution on each channel.
Point-by-point convolution combines and organizes the features of each channel extracted by the depth convolution by applying a point-by-point convolution kernel. It can introduce inter-channel interactions and global information.
The feature extraction process of the depth separable convolution is as follows:
the deep convolution performs a convolution operation on each channel of the input feature map to extract the spatial features of each channel.
The point-by-point convolution combines and organizes the depth convolution results of each channel, introducing interaction and global information between channels.
The final output is a feature map calculated by a combination of depth convolution and point-wise convolution. These feature maps may capture spatial relationships and inter-channel associations in the input data.
In an embodiment, various state data of the device and corresponding fault information of the device can be collected in advance as samples, and the neural network model is iteratively trained by adopting a back propagation method, so that a first neural network model for predicting the fault probability based on the state data is finally obtained through training. In addition, in the remaining application scenarios, different training patterns may be employed, which are only examples herein.
205. The cloud server performs fault prediction processing based on target state data by adopting a second neural network model to obtain a fault prediction result, wherein the first neural network model is a lightweight neural network model, and the number of model parameters, the model complexity and the hierarchical structure depth of the second neural network model are greater than those of the first neural network model; and when the failure prediction result indicates that the industrial Internet of things equipment fails, the cloud server sends early warning information to the early warning terminal.
The second neural network model may be a large-scale neural network model, and the number of model parameters, the model complexity and the hierarchical structure depth of the model are larger than those of the lightweight neural network model. For example, the second neural network model may include a recurrent neural network (Recurrent Neural Network, RNN), a Long Short-Term Memory (LSTM), a generate countermeasure network (Generative Adversarial Network, GAN), and so on.
In one embodiment, to improve the accuracy of the fault prediction, referring to fig. 4, the structure of the second neural network model may include: an input encoding layer, an attention layer, a multi-layer encoder, and an output layer; the step of performing, by the cloud server, fault prediction processing based on the target state data by using the second neural network model to obtain a fault prediction result may include:
the input coding layer performs feature extraction on a plurality of target state data to obtain a state feature sequence, and performs feature coding processing on the target state feature sequence to obtain a target state feature vector;
the attention layer extracts attention characteristic of the target state characteristic vector based on an attention mechanism to obtain an attention characteristic vector of key characteristics in the attention state characteristic sequence;
the multi-layer encoder performs multi-layer abstraction and representation on the principal force feature vector to obtain a high-layer feature vector, wherein the high-layer feature vector at least characterizes relation information and context information among features in the state feature sequence;
and the output layer carries out fault prediction processing on the industrial Internet of things equipment according to the high-level feature vector to obtain a fault prediction result.
The multi-layer encoder can be composed of a multi-head self-attention mechanism and a feedforward neural network, and the output layer can be designed according to specific task requirements. For the task of predicting failure probability, the full connection layer may be used to map the output of the encoder to the final probability prediction result.
The method and the device can be combined with the attention mechanism and the feature associated information to predict the equipment faults, and accuracy of fault prediction can be improved. The creation of the second neural network structure introduces a focus mechanism that can accurately capture critical information in the sequence data and make multi-level abstract representations in the multi-layer encoder. The structure can better extract useful characteristics when processing the state data of the industrial Internet of things equipment, thereby improving the prediction performance and accuracy of the fault probability. In addition, the attention mechanism can help the user to know the attention degree of the model to the specific state through the visualized importance weight, so that the interpretability of the model is improved.
In one embodiment, the step of performing fault prediction processing on industrial Internet of things equipment by the output layer according to the high-level feature vector to obtain a fault prediction result "
The output layer carries out probability mapping processing on the industrial Internet of things equipment according to the high-level feature vector, and predicts the second fault probability of the industrial Internet of things equipment;
and when the second fault probability is larger than a second fault threshold, determining that the industrial Internet of things equipment fails, wherein the second fault threshold is larger than the first fault threshold.
The first fault threshold and the second fault threshold can be set according to empirical knowledge or according to actual scene requirements, the second fault threshold is higher than the first fault threshold, and fault prediction is more accurate.
In an embodiment, the industrial internet of things device is determined to be not malfunctioning when the second probability of failure is not greater than a second failure threshold.
In one embodiment, a historical state data set of the device may be collected first, and training is performed on the neural network model through the historical state data set to continuously converge, so as to finally generate a second neural network model for predicting faults.
According to the method and the device, real-time processing can be performed through the small model firstly aiming at industrial Internet of things equipment, and if the judging result is a fault, processing of the large model is further performed on relevant data uploaded to the cloud so as to achieve fine judgment. The filtering conditions of the small model are set so as to improve the data uploading frequency and the uploading data quantity, wherein the requirements of side resources are fewer, the judging threshold is lower, the requirements of large model resources are more, and the judging threshold is higher and more accurate. By the means, early warning precision is improved, and data uploading quantity is reduced.
As shown in fig. 5, in an embodiment, specific status data will be taken as an example to describe a method provided in the embodiment of the present application, specifically:
the status data types include at least: at least one of sensor data, device operating parameters, device error log, and environmental data; the first neural network model comprises at least one convolution layer, a pooling layer and a full connection layer;
the second neural network model includes: an input encoding layer, an attention layer, a multi-layer encoder, and an output layer.
The fault early warning processing method based on the Internet of things system specifically comprises the following steps:
501. the industrial Internet of things device obtains a plurality of state data, wherein the plurality of state data comprises sensor data, device operation parameters, device error logs and environment data.
The status data can be collected by using sensors on the device, such as temperature sensors
502. And the industrial Internet of things equipment respectively judges whether the sensor data, the equipment operation parameters and the environment data are larger than the corresponding uploading thresholds, and if so, the industrial Internet of things equipment sends target state data meeting the thresholds to the edge server.
503. And the edge server performs data integration on the sensor data, the equipment operation parameters, the equipment error log and the environment data to obtain an initial state data sequence.
504. Respectively carrying out convolution characteristic extraction processing on the initial state data sequence by adopting a convolution layer to obtain a state characteristic set; carrying out feature pooling treatment on the state feature set by adopting a pooling layer; and predicting the fault probability of the industrial Internet of things equipment according to the state characteristic set through the full connection layer.
505. And when the fault probability is greater than the first fault threshold, the edge server sends target state data to the cloud server.
506. And the input coding layer in the second neural network model of the cloud server performs feature extraction on the initial state data sequence to obtain a state feature sequence, and performs feature coding processing on the target state feature sequence to obtain a target state feature vector, wherein the state feature sequence comprises sensor data features, environment data features, equipment operation parameter features and equipment error log features.
507. The attention layer of the cloud server extracts attention characteristics of the target state characteristic vector based on an attention mechanism to obtain attention characteristic vectors of key characteristics in the attention state characteristic sequence, wherein the key characteristics at least comprise equipment operation parameter characteristics.
508. The multi-layer encoder of the cloud server performs multi-layer abstraction and representation on the principal force feature vector to obtain a high-layer feature vector, wherein the high-layer feature vector at least represents relation information and context information among features in the state feature sequence; and the output layer carries out fault prediction processing on the industrial Internet of things equipment according to the high-level feature vector to obtain a fault prediction result.
509. And when the failure prediction result indicates that the industrial Internet of things equipment fails, the cloud server sends early warning information to the early warning terminal.
According to the fault early warning processing method based on the Internet of things system, cloud edge cooperation is adopted, early warning is conducted on the basis of a lightweight model at the edge, uploading data and processing amount of cloud data of an internet of things platform are reduced, uploading data amount can be reduced, response efficiency is improved, and meanwhile accuracy of fault early warning is improved through cloud edge two-stage fault analysis.
The foregoing describes in detail a fault early warning processing method based on an internet of things system and the internet of things system provided in the embodiments of the present application, and specific examples are applied herein to illustrate the principles and implementations of the present application, where the descriptions of the foregoing embodiments are only used to help understand the method of the present application and the core idea thereof; meanwhile, those skilled in the art will have variations in the specific embodiments and application scope in light of the ideas of the present application, and the present description should not be construed as limiting the present application in view of the above.
Claims (10)
1. The fault early warning processing method based on the Internet of things system is characterized in that the Internet of things system comprises an edge server, a cloud server and industrial Internet of things equipment; the edge server, the cloud server and the industrial Internet of things device are connected through a network, and the fault early warning processing method comprises the following steps:
the industrial Internet of things equipment detects a plurality of state data of the industrial Internet of things equipment, judges whether the state data meets a data uploading condition, and sends at least one target state data meeting the condition to the edge server when the state data meets the data uploading condition;
the edge server performs feature extraction processing on target state data by adopting a first neural network model to obtain a state feature set, predicts the fault probability of the industrial Internet of things equipment according to the state feature set, and sends the target state data to the cloud server when the fault probability is greater than a first fault threshold;
the cloud server performs fault prediction processing based on the target state data by adopting a second neural network model to obtain a fault prediction result, wherein the first neural network model is a lightweight neural network model, and the number of model parameters, model complexity and hierarchical structure depth of the second neural network model are larger than those of the first neural network model;
and when the fault prediction result indicates that the industrial Internet of things equipment fails, the cloud server sends early warning information to an early warning terminal.
2. The fault pre-warning processing method of claim 1, wherein determining whether the status data satisfies a data upload condition comprises:
identifying a state type of the state data, and acquiring an uploading threshold value corresponding to the state type;
and if the threshold value of the state data is larger than the uploading threshold value, determining that the state data meets the data uploading condition.
3. The fault pre-warning processing method of claim 1, wherein the first neural network model comprises at least one convolutional layer, a pooling layer, and a fully-connected layer;
the edge server adopts a first neural network model to conduct feature extraction processing on a plurality of target state data to obtain a state feature set, predicts the fault probability of the industrial Internet of things equipment according to the state feature set, and comprises the following steps:
the edge server adopts a convolution layer to carry out convolution feature extraction processing on the target state data to obtain a state feature set;
carrying out feature pooling treatment on the state feature set by adopting the pooling layer;
and predicting the fault probability of the industrial Internet of things equipment according to the state characteristic set through the full connection layer.
4. The method of claim 3, wherein the first neural network model comprises two convolutional layers and a pooled layer, the convolutional layers having a convolutional kernel of 3 by 3, and the pooled layer having a size of 2 by 2.
5. The fault prediction method as claimed in claim 3, wherein the edge server performs convolution feature extraction processing on the target state data by using a convolution layer to obtain a state feature set, and the method comprises:
and the edge server adopts depth separable convolution to carry out convolution feature extraction processing on the target state data so as to obtain a state feature set.
6. The fault pre-warning processing method of any one of claims 3 to 5, wherein the second neural network model comprises: an input encoding layer, an attention layer, a multi-layer encoder, and an output layer; the cloud server performs fault prediction processing based on the target state data by adopting a second neural network model to obtain a fault prediction result, and the method comprises the following steps:
the input coding layer performs feature extraction on a plurality of target state data to obtain a state feature sequence, and performs feature coding processing on the target state feature sequence to obtain a target state feature vector;
the attention layer extracts attention characteristic vectors of key characteristics in the attention state characteristic sequence based on an attention mechanism;
the multi-layer encoder performs multi-layer abstraction and representation on the principal force feature vector to obtain a high-layer feature vector, wherein the high-layer feature vector at least represents relation information and context information among features in the state feature sequence;
and the output layer carries out fault prediction processing on the industrial Internet of things equipment according to the high-level feature vector to obtain a fault prediction result.
7. The method for fault early warning processing according to claim 6, wherein the output layer performs fault prediction processing on the industrial internet of things device according to the high-level feature vector to obtain a fault prediction result, and the method comprises the following steps:
the output layer performs probability mapping processing on the industrial Internet of things equipment according to the high-level feature vector, and predicts a second fault probability of the industrial Internet of things equipment;
and when the second fault probability is larger than a second fault threshold, determining that the industrial Internet of things equipment breaks down, wherein the second fault threshold is larger than the first fault threshold.
8. The fault pre-warning method of claim 7, wherein each encoder is comprised of a multi-headed self-attention mechanism and a feed-forward neural network.
9. The fault pre-warning processing method of claim 2, wherein the status data types include at least: at least one of sensor data, device operating parameters, device error log, and environmental data; the first neural network model comprises at least one convolution layer, a pooling layer and a full connection layer;
judging whether the state data meets the data uploading condition or not comprises the following steps: respectively judging whether the sensor data, the equipment operation parameters and the environment data are larger than the corresponding uploading thresholds or not, and if yes, confirming that the data uploading conditions are met;
the edge server adopts a first neural network model to conduct feature extraction processing on a plurality of target state data to obtain a state feature set, predicts the fault probability of the industrial Internet of things equipment according to the state feature set, and comprises the following steps:
the edge server performs data integration on sensor data, equipment operation parameters, equipment error logs and environment data to obtain an initial state data sequence;
respectively carrying out convolution characteristic extraction processing on the initial state data sequence by adopting a convolution layer to obtain a state characteristic set; carrying out feature pooling treatment on the state feature set by adopting the pooling layer; predicting the fault probability of the industrial Internet of things equipment according to the state characteristic set through the full connection layer;
the second neural network model includes: an input encoding layer, an attention layer, a multi-layer encoder, and an output layer; the cloud server performs fault prediction processing based on the target state by adopting a second neural network model to obtain a fault prediction result, and the method comprises the following steps:
the input coding layer performs feature extraction on the initial state data sequence to obtain a state feature sequence, and performs feature coding on the target state feature sequence to obtain a target state feature vector, wherein the state feature sequence comprises sensor data features, environment data features, equipment operation parameter features and equipment error log features;
the attention layer extracts attention characteristics of the target state characteristic vector based on an attention mechanism to obtain attention characteristic vectors of key characteristics in an attention state characteristic sequence, wherein the key characteristics at least comprise equipment operation parameter characteristics;
the multi-layer encoder performs multi-layer abstraction and representation on the principal force feature vector to obtain a high-layer feature vector, wherein the high-layer feature vector at least represents relation information and context information among features in the state feature sequence;
and the output layer carries out fault prediction processing on the industrial Internet of things equipment according to the high-level feature vector to obtain a fault prediction result.
10. The Internet of things system is characterized by comprising an edge server, a cloud server, an early warning terminal and industrial Internet of things equipment located in an industrial park; the edge server, the cloud server and the industrial Internet of things device are connected through a network, and the early warning processing method comprises the following steps:
the industrial Internet of things equipment is used for detecting a plurality of state data of the industrial Internet of things equipment, judging whether the state data meet data uploading conditions, and sending at least one target state data meeting the conditions to the edge server when the state data meet the data uploading conditions;
the edge server is used for carrying out feature extraction processing on target state data by adopting a first neural network model to obtain a state feature set, predicting the fault probability of the industrial Internet of things equipment according to the state feature set, and sending the target state data to the cloud server when the fault probability is greater than a first fault threshold;
the cloud server is used for performing fault prediction processing based on the target state data by adopting a second neural network model to obtain a fault prediction result, wherein the first neural network model is a lightweight neural network model, and the number of model parameters, the model complexity and the hierarchical structure depth of the second neural network model are greater than those of the first neural network model; and when the fault prediction result indicates that the industrial Internet of things equipment fails, the cloud server sends early warning information to an early warning terminal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311675441.7A CN117834385A (en) | 2023-12-07 | 2023-12-07 | Fault early warning processing method based on Internet of things system and Internet of things system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311675441.7A CN117834385A (en) | 2023-12-07 | 2023-12-07 | Fault early warning processing method based on Internet of things system and Internet of things system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117834385A true CN117834385A (en) | 2024-04-05 |
Family
ID=90516406
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311675441.7A Pending CN117834385A (en) | 2023-12-07 | 2023-12-07 | Fault early warning processing method based on Internet of things system and Internet of things system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117834385A (en) |
-
2023
- 2023-12-07 CN CN202311675441.7A patent/CN117834385A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10699195B2 (en) | Training of artificial neural networks using safe mutations based on output gradients | |
| CN111046027B (en) | Missing value filling method and device for time series data | |
| JP2022524244A (en) | Predictive classification of future behavior | |
| CN110768971B (en) | Confrontation sample rapid early warning method and system suitable for artificial intelligence system | |
| CN112202726B (en) | System anomaly detection method based on context sensing | |
| CN115769235A (en) | Method and system for providing an alert related to the accuracy of a training function | |
| CN111798518A (en) | Mechanical arm posture detection method, device and equipment and computer storage medium | |
| CN116842379A (en) | Mechanical bearing residual service life prediction method based on DRSN-CS and BiGRU+MLP models | |
| CN116599857A (en) | Digital twin application system suitable for multiple scenes of Internet of things | |
| CN117595504A (en) | Intelligent monitoring and early warning method for power grid running state | |
| CN118484356A (en) | Method and system for monitoring server state based on RPA | |
| CN113379045A (en) | Data enhancement method and device | |
| JP6846369B2 (en) | Programs, devices and methods for classifying unknown multidimensional vector data groups | |
| CN117951646A (en) | Data fusion method and system based on edge cloud | |
| Chen et al. | NNFacet: Splitting Neural Network for Concurrent Smart Sensors | |
| CN117933531A (en) | Distributed photovoltaic power generation power prediction system and method | |
| Kumar et al. | Building on prior lightweight CNN model combined with LSTM-AM framework to guide fault detection in fixed-wing UAVs | |
| Lavingia et al. | Information retrieval and data analytics in internet of things: current perspective, applications and challenges | |
| Pandiyan et al. | Systematic Review on Fault Diagnosis on Rolling-Element Bearing | |
| CN117834385A (en) | Fault early warning processing method based on Internet of things system and Internet of things system | |
| Gao et al. | Fault prediction in electric power communication network based on improved DenseNet | |
| CN118504847B (en) | Intelligent beam field management method and system based on digital twin technology | |
| CN118506553B (en) | AIoT anomaly identification method, disaster early warning system and road safety system | |
| CN117539674B (en) | Exception handling method, device, equipment and storage medium | |
| CN114997308B (en) | Cloud-edge cooperative self-adaptive industrial data classification method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |