CN115001937B - Smart city Internet of things-oriented fault prediction method and device - Google Patents
Smart city Internet of things-oriented fault prediction method and device Download PDFInfo
- Publication number
- CN115001937B CN115001937B CN202210376991.8A CN202210376991A CN115001937B CN 115001937 B CN115001937 B CN 115001937B CN 202210376991 A CN202210376991 A CN 202210376991A CN 115001937 B CN115001937 B CN 115001937B
- Authority
- CN
- China
- Prior art keywords
- fault prediction
- residual
- result
- feature
- spatial
- 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.)
- Active
Links
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
- G16Y40/00—IoT characterised by the purpose of the information processing
- G16Y40/20—Analytics; Diagnosis
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
- G16Y40/00—IoT characterised by the purpose of the information processing
- G16Y40/40—Maintenance of things
-
- 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
- H04L41/14—Network analysis or design
- H04L41/147—Network analysis or design for predicting network behaviour
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A30/00—Adapting or protecting infrastructure or their operation
- Y02A30/60—Planning or developing urban green infrastructure
Abstract
The invention provides a fault prediction method and device for the Internet of things of a smart city, wherein the method comprises the following steps: acquiring network data to be predicted of intelligent city Internet of things equipment; inputting the network data into a failure prediction model after training to obtain a failure prediction result of the network data; the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the feature data information of the intelligent city Internet of things equipment as a training set and prediction labels corresponding to the training set. The fault prediction model is established based on convolution decomposition operation and improved residual error network structure, so that the fault prediction of the intelligent city Internet of things equipment can be realized through the lightweight model, and the safety and the robustness of the intelligent city Internet of things are effectively ensured.
Description
Technical Field
The invention relates to the technical field of network fault prediction, in particular to a fault prediction method and device for the Internet of things of a smart city.
Background
With the development of intelligent devices and 5G networks, the use of edge intelligent devices has shown explosive growth. In order to provide the unlimited connectivity required for such devices, a basic and smart city internet of things computing architecture of a trusted system is required. In order to realize the prospect of service availability in a trusted system, the sustainability and reliability of the Internet of things in the smart city are very important. But with the complexity of the service and components of the smart city internet of things, faults occur in the smart city internet of things at times, so that the reliability of the smart city internet of things cannot be guaranteed.
The network fault prediction is based on the historical condition of the network node, analyzes in combination with the current node and the network condition, excavates the specific state of each type of fault, further establishes a correlation model between the state and the fault, predicts the type of the fault possibly occurring in the future of the node, and ensures the sustainability and the reliability of the network. The traditional convolutional neural network has strong spatial feature extraction capability and learning capability, and can be applied to fault prediction. However, as network models become deeper, model recognition accuracy decreases, resulting in poorer model results. The residual network is arranged, so that the problem is solved, but the characteristic input of the lower network into the higher network can only be completed through one residual connecting channel due to the existence of the shortcuts of the residual structure, so that the residual block has insufficient weight to learn the multidimensional data characteristic of the nodes under the Internet of things of the smart city, and the problem of insufficient and inaccurate learning of the operation data characteristic of the nodes is caused.
In the internet of things of the smart city, the edge devices are usually small-sized computing devices, such as sensors, mobile terminals and other small-sized intelligent devices, which often have limited computing power, so that a common deep learning model cannot be suitable for the fault prediction of the internet of things of the smart city.
Disclosure of Invention
The invention provides a fault prediction method and device for a smart city Internet of things, which are used for solving the defect that a deep learning model in the prior art cannot be suitable for the smart city Internet of things fault prediction, realizing the fault prediction of smart city Internet of things equipment and effectively guaranteeing the safety and the robustness of an edge side network.
The invention provides a fault prediction method for the Internet of things of a smart city, which comprises the following steps:
acquiring network data to be predicted of intelligent city Internet of things equipment;
inputting the network data into a failure prediction model after training to obtain a failure prediction result of the network data;
the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the feature data information of the intelligent city Internet of things equipment as a training set and prediction labels corresponding to the training set.
According to the fault prediction method for the Internet of things of the smart city, provided by the invention, the training method of the fault prediction model comprises the following steps:
acquiring characteristic data information of the intelligent city Internet of things equipment as a training set of a model;
establishing an initial fault prediction model, inputting the characteristic data information into the initial fault prediction model, obtaining a fault prediction result of a characteristic vector, and evaluating prediction accuracy based on the fault prediction result;
and carrying out iterative updating on the initial fault prediction model, and adjusting model parameters of the initial fault prediction model until the error value of the fault prediction result and the prediction label meets a preset threshold value, and obtaining the trained fault prediction model when the prediction accuracy reaches a target value.
According to the method for predicting the fault of the internet of things oriented to the smart city, which is provided by the invention, an initial fault prediction model is established, the characteristic data information is input into the initial fault prediction model, a fault prediction result of a characteristic vector is obtained, and the prediction accuracy is evaluated based on the fault prediction result, and the method comprises the following steps:
establishing an initial fault prediction model, wherein the initial fault prediction model comprises a convolution layer, a residual error structure layer and a result processing layer;
the convolution layer is used for extracting spatial characteristic information of the input characteristic data information based on the convolution layer, and acquiring a first spatial characteristic by combining an excitation function of the convolution layer;
wherein the convolution layer is established based on a convolution decomposition operation;
the residual structure layer is used for carrying out residual information learning based on the first spatial features output by the convolution layer and outputting second spatial features of the feature data information;
the result processing layer is used for obtaining a fault prediction result of the characteristic data information based on the second spatial characteristic output by the residual error structural layer;
and evaluating the prediction accuracy of the fault prediction model based on the fault prediction result.
According to the fault prediction method for the Internet of things of the smart city, the residual structure layer comprises a first residual connecting channel and a second residual connecting channel;
residual information learning is performed based on the first spatial feature output by the convolution layer, and the residual information learning comprises the following steps:
acquiring a first residual mapping result of the first spatial feature output by the convolution layer through the first residual connecting channel, and acquiring a second residual mapping result of the first spatial feature output by the convolution layer through a second residual connecting channel;
and carrying out multiplication flow residual information learning on the first residual mapping result and the first spatial feature, and carrying out addition flow residual information learning on the second residual mapping result and the first spatial feature.
According to the method for predicting the fault of the internet of things oriented to the smart city, which is provided by the invention, the prediction accuracy of the fault prediction model is evaluated based on the fault prediction result, and the method comprises the following steps:
acquiring the feature number of feature data information, selecting a target feature based on the feature number of the feature data information, and acquiring a data value corresponding to the target feature based on the fault prediction result;
and acquiring a state data value of the intelligent city Internet of things of the target feature at the target moment, and evaluating the prediction accuracy of the fault prediction model by combining the feature number and the data value corresponding to the target feature.
According to the fault prediction method for the internet of things of the smart city provided by the invention, residual information learning is performed based on the first spatial feature output by the convolution layer, and after the second spatial feature of the feature data information is output, the fault prediction method further comprises the following steps:
carrying out batch normalization processing on the second spatial features of the feature data information and inputting the second spatial features to the result processing layer;
the fault prediction result of the feature data information is obtained by the result processing layer based on the second spatial feature output by the residual error structural layer, and the fault prediction result comprises:
and obtaining a fault prediction result of the characteristic data information based on the second spatial characteristics after the batch normalization processing through the result processing layer.
The invention also provides a fault prediction device oriented to the Internet of things of the smart city, which comprises:
the acquisition module is used for acquiring network data to be predicted of the intelligent city Internet of things equipment;
the prediction module is used for inputting the network data into a trained fault prediction model to obtain a fault prediction result of the network data;
the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the feature data information of the intelligent city Internet of things equipment as a training set and prediction labels corresponding to the training set.
The invention also provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the fault prediction method oriented to the intelligent city Internet of things when executing the program.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method of failure prediction for the internet of things of the smart city as described in any of the above.
The invention also provides a computer program product comprising a computer program which, when executed by a processor, implements the method for predicting faults for the internet of things of the smart city as described in any one of the above.
According to the fault prediction method and device for the intelligent city Internet of things, network data to be predicted of intelligent city Internet of things equipment are obtained; the network data is input into the trained fault prediction model to obtain a fault prediction result of the network data, and the fault prediction model is built based on convolution decomposition operation and improved residual network structure, so that the fault prediction of the intelligent city Internet of things equipment can be realized through a lightweight model, and the safety and the robustness of the intelligent city Internet of things are effectively ensured.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a fault prediction method for the Internet of things of a smart city;
FIG. 2 is a second flow chart of the method for predicting the failure of the Internet of things for the smart city according to the present invention;
FIG. 3 is a schematic diagram of the structure of the fault prediction model of the present invention;
FIG. 4 is a schematic diagram of the residual structure layer of the present invention;
fig. 5 is a schematic structural diagram of a fault prediction device for the internet of things of the smart city;
fig. 6 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The method and the device for predicting the faults of the Internet of things for the smart city are described below with reference to fig. 1 to 6.
Referring to fig. 1, the fault prediction method for the internet of things of the smart city provided by the invention comprises the following steps:
Specifically, the smart city internet of things device in this embodiment is a computing device in the smart city internet of things, such as a sensor, a mobile terminal, and other intelligent computing devices. The network data is the data which needs to be subjected to fault prediction in the Internet of things of the smart city. The method comprises the steps of analyzing the historical conditions of network nodes and combining the current node and the network conditions, mining out the specific state of each type of fault, further establishing a correlation model between the state and the fault, and predicting the type of the fault possibly occurring in the future of the node so as to ensure the sustainability and the reliability of the network.
And 120, inputting the network data into a trained fault prediction model to obtain a fault prediction result of the network data.
The fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the feature data information of the intelligent city Internet of things equipment as a training set and prediction labels corresponding to the training set.
Specifically, the step is an application process of the fault prediction model, namely, the trained model is combined into a specific application scene.
And inputting the network data needing to be subjected to fault prediction into a trained model for prediction, thereby obtaining a fault prediction result of the network data.
In the training process of the fault prediction model, an error value is generated between the output result of the model and the prediction label, parameters of the model can be adjusted according to the error value, the model converges when the error value reaches a preset target value, and the fault prediction model at the moment is kept as a model after training is completed so as to conduct fault prediction.
It should be noted that, the fault prediction model in this embodiment is established based on the convolutional decomposition operation and the improved residual network structure, and the convolutional layer is improved based on the concept of the acceptance module, so that the parameter quantity of the convolutional layer is reduced, and the network calculation quantity is reduced. And according to the improved residual network structure, compared with the residual network structure before improvement, under the same parameter quantity, the learning capacity of the network can be improved, and the spatial feature extraction capacity of the network to fault information is improved.
According to the fault prediction method for the intelligent city Internet of things, network data to be predicted of intelligent city Internet of things equipment is obtained; the network data is input into the trained fault prediction model to obtain a fault prediction result of the network data, and the fault prediction model is built based on convolution decomposition operation and improved residual network structure, so that the fault prediction of the intelligent city Internet of things equipment can be realized through a lightweight model, and the safety and the robustness of the intelligent city Internet of things are effectively ensured.
Referring to fig. 2, the training method of the fault prediction model based on the above embodiment includes the following steps:
step 210, obtaining characteristic data information of the intelligent city internet of things equipment as a training set of the model.
And 220, establishing an initial fault prediction model, inputting the characteristic data information into the initial fault prediction model, obtaining a fault prediction result of the characteristic vector, and evaluating the prediction accuracy based on the fault prediction result.
And 230, carrying out iterative updating on the initial fault prediction model, and adjusting model parameters of the initial fault prediction model until the error value of the fault prediction result and the prediction label meets a preset threshold value, and obtaining the trained fault prediction model when the prediction accuracy reaches a target value.
Specifically, the present embodiment is a training process of a failure prediction model, that is, an acquisition process of a model.
Firstly, a large amount of characteristic new data information of intelligent city Internet of things equipment is obtained to be used as a training set of a model, namely a sample. The sequence of fault states for the characteristic new data information may be expressed as: h f ＝(X 1 ,X 2 ,…,X t )。
Wherein X is t The data sequence represented is X t ＝(x t-l ,x t-l+1 ,..,x t-1 )，x t-1 The characteristic vector of the state of the Internet of things of the smart city at the time t-1 is shown.
Then, an initial fault prediction model to be trained is established, new characteristic data information of the intelligent city Internet of things equipment is input into the model, a prediction result is obtained, and then accuracy evaluation is carried out on the prediction result.
The training process of the model is an iterative updating process, namely, the error between the output result of the current model and the expected result is utilized to adjust the model parameters, and the model training is completed under the condition that the error between the current output result and the expected result meets a preset threshold value and the accuracy corresponding to the expected result meets a target value.
Based on the above embodiment, the establishing an initial fault prediction model, inputting the feature data information into the initial fault prediction model, obtaining a fault prediction result of a feature vector, and evaluating prediction accuracy based on the fault prediction result, includes:
establishing an initial fault prediction model, wherein the initial fault prediction model comprises a convolution layer, a residual error structure layer and a result processing layer;
the convolution layer is used for extracting spatial characteristic information of the input characteristic data information based on the convolution layer, and acquiring a first spatial characteristic by combining an excitation function of the convolution layer;
wherein the convolution layer is established based on a convolution decomposition operation;
the residual structure layer is used for carrying out residual information learning based on the first spatial features output by the convolution layer and outputting second spatial features of the feature data information;
the result processing layer is used for obtaining a fault prediction result of the characteristic data information based on the second spatial characteristic output by the residual error structural layer;
and evaluating the prediction accuracy of the fault prediction model based on the fault prediction result.
Specifically, the embodiment is a specific building process of a fault prediction model, including a building process of a plurality of layer structures of the model.
Referring to fig. 3, the fault prediction model in the present implementation includes a convolution layer 310, a residual structure layer 320 (Two-way residual structure), and a result processing layer 330 (softmax).
Firstly, a convolution layer is established, spatial feature information extraction is carried out on Input feature data information (Input), and a first spatial feature is obtained by combining an excitation function of the convolution layer. The convolution decomposition operation in this embodiment is embodied as: the idea improvement convolution layer based on the acceptance module selects two convolution layers of 1×1 (Conv 1×1) and 3×3 (Conv 3×3) to replace 7×7 convolution kernels, namely only three convolution layers in the model, and reduces the parameter quantity of the convolution layers, thereby reducing the network calculation quantity.
It should be noted that, for each convolution layer, an excitation function is followed to enhance the learning ability of the network, and in this embodiment, an improved Relu function is adopted as the excitation function, where the input is negative and the output is 0 in the conventional Relu function, so as to eliminate the output of the negative input and affect the fault information extraction effect. Thus from the newly defined N-Relu function, the formula is as follows:
it should be further noted that the number of active channels of N-Relu in this embodiment is twice that of the Relu before improvement, i.e. N/2 channels are needed to obtain the same N feature maps, so that the number of parameters involved in calculation can be reduced by half.
And then establishing a residual structure layer, inputting the spatial features output by the convolution layer into the residual structure layer for residual information learning, and outputting the learned spatial features.
And further establishing a result processing layer, and inputting the learned spatial characteristics output by the residual structure layer into the result processing layer to obtain a final fault prediction result. And finally, evaluating the accuracy of the fault prediction result.
It should be noted that, in the present embodiment, the result processing layer adopts softmax to perform fault prediction, unlike the SVM commonly used for multi-classification problem, the final output of softmax is probability, so that the fault prediction result can be directly obtained under the internet of things of the smart city, and the final output of SVM is score, which cannot intuitively reflect the prediction situation.
The fault prediction in the embodiment of the smart city internet of things can be abstracted into a multi-classification problem, and softmax regression belongs to a multi-classification model, so that the method and the device can be suitable for the intentional prediction problem in the smart city internet of things scene.
Based on the above embodiments, the residual structure layer includes a first residual connection channel and a second residual connection channel;
residual information learning is performed based on the first spatial feature output by the convolution layer, and the residual information learning comprises the following steps:
acquiring a first residual mapping result of the first spatial feature output by the convolution layer through the first residual connecting channel, and acquiring a second residual mapping result of the first spatial feature output by the convolution layer through a second residual connecting channel;
and carrying out multiplication flow residual information learning on the first residual mapping result and the first spatial feature, and carrying out addition flow residual information learning on the second residual mapping result and the first spatial feature.
Specifically, the residual structure layer in the embodiment is a two-way residual block structure, which is used for solving the problem that the traditional residual structure is insufficient in learning the space characteristics of fault information.
Referring to fig. 4, the residual structure layer of the present embodiment includes two residual connection channels, so that the entire residual block contains both the multiplication information stream and the addition information stream.
The method is characterized in that the two residual error connecting channels perform addition flow residual error information learning according to the spatial characteristics output by the convolution layer, and the first residual error connecting channel can also perform multiplication flow residual error information learning according to the spatial characteristics output by the convolution layer.
It should be noted that the parameter amount of the 1×3 convolution kernel in this embodiment is only 1/3 of that of the conventional 3×3 convolution kernel. The parameter number of the one-dimensional residual block is only 2/3 of that of a traditional residual block. Thus, the residual structure layer, although introducing a multiplication information stream, does not increase the number of parameters compared to the conventional residual structure layer.
In addition, aiming at the intelligent city Internet of things fault prediction scene, the traditional residual structure layer can only learn the residual information of the main information flow, and the improved residual structure layer in the embodiment can learn multiplication factors well while learning the residual information.
The residual block structure of two paths can enable each layer of network to obtain more abundant spatial characteristics of fault information data under the Internet of things of the smart city.
The two-way residual block proposed by this embodiment can be defined as:
y i ＝x i ×F 1 ([x i ,K i ])+(F 2 ([x i ,W i ])+x i ) (2)
wherein x is i Input vector representing residual structure layer, y i Representing the output vector, function F 1 ([x i ,K i ]) And F 2 ([x i ,W i ]) Representing the residual map.
Based on the above embodiment, the evaluating the prediction accuracy of the fault prediction model based on the fault prediction result includes:
acquiring the feature number of feature data information, selecting a target feature based on the feature number of the feature data information, and acquiring a data value corresponding to the target feature based on the fault prediction result;
and acquiring a state data value of the intelligent city Internet of things of the target feature at the target moment, and evaluating the prediction accuracy of the fault prediction model by combining the feature number and the data value corresponding to the target feature.
Specifically, the embodiment provides a specific process of evaluating the accuracy of the failure prediction result. After obtaining the failure prediction result of the model, the accuracy of the model is measured by using Root Mean Square Error (RMSE). As shown in the following formula:
wherein n represents the feature number of data acquisition, namely the feature number of feature data information;
data value representing the ith feature of the model predictive output, i.e. the data value corresponding to the target feature, +.>
And representing the data value of the real ith feature at the t moment, namely the state data value of the intelligent city Internet of things of the target feature at the target moment. From the definition of the above formula, smaller RMSE means higher model accuracy.
Based on the above embodiment, after performing residual information learning based on the first spatial feature output by the convolution layer and outputting the second spatial feature of the feature data information, the method further includes:
carrying out batch normalization processing on the second spatial features of the feature data information and inputting the second spatial features to the result processing layer;
the fault prediction result of the feature data information is obtained by the result processing layer based on the second spatial feature output by the residual error structural layer, and the fault prediction result comprises:
and obtaining a fault prediction result of the characteristic data information based on the second spatial characteristics after the batch normalization processing through the result processing layer.
In particular, the model may be difficult to train because if the neural network is trained too long or too deep during the training process. Therefore, in the embodiment, the second spatial features of the feature data information are subjected to batch normalization processing, and the spatial features after normalization processing are input to the result processing layer, so that a fault prediction result is obtained, and the model training efficiency is improved.
The fault prediction device for the intelligent city internet of things provided by the invention is described below, and the fault prediction device for the intelligent city internet of things described below and the fault prediction method for the intelligent city internet of things described above can be correspondingly referred to each other.
Referring to fig. 5, the present invention further provides a fault prediction device for the internet of things of the smart city, including:
an obtaining module 510, configured to obtain network data to be predicted of the smart city internet of things device;
the prediction module 520 is configured to input the network data into a trained failure prediction model, so as to obtain a failure prediction result of the network data;
the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the feature data information of the intelligent city Internet of things equipment as a training set and prediction labels corresponding to the training set.
Based on the above embodiment, the training process of the failure prediction model includes the following modules:
the training set acquisition module is used for acquiring characteristic data information of the intelligent city Internet of things equipment as a training set of the model;
the model building module is used for building an initial fault prediction model, inputting the characteristic data information into the initial fault prediction model, obtaining a fault prediction result of the characteristic vector and evaluating prediction accuracy based on the fault prediction result;
and the training module is used for carrying out iterative updating on the initial fault prediction model, adjusting the model parameters of the initial fault prediction model until the error value of the fault prediction result and the prediction label meets a preset threshold value, and obtaining the fault prediction model with the prediction accuracy reaching the target value after training.
Based on the above embodiments, the model building module is specifically configured to:
establishing an initial fault prediction model, wherein the initial fault prediction model comprises a convolution layer, a residual error structure layer and a result processing layer;
the convolution layer is used for extracting spatial characteristic information of the input characteristic data information based on the convolution layer, and acquiring a first spatial characteristic by combining an excitation function of the convolution layer;
wherein the convolution layer is established based on a convolution decomposition operation;
the residual structure layer is used for carrying out residual information learning based on the first spatial features output by the convolution layer and outputting second spatial features of the feature data information;
the result processing layer is used for obtaining a fault prediction result of the characteristic data information based on the second spatial characteristic output by the residual error structural layer;
and evaluating the prediction accuracy of the fault prediction model based on the fault prediction result.
Based on the above embodiments, the residual structure layer includes a first residual connection channel and a second residual connection channel;
the residual structure layer is specifically used for:
acquiring a first residual mapping result of the first spatial feature output by the convolution layer through the first residual connecting channel, and acquiring a second residual mapping result of the first spatial feature output by the convolution layer through a second residual connecting channel;
and carrying out multiplication flow residual information learning on the first residual mapping result and the first spatial feature, and carrying out addition flow residual information learning on the second residual mapping result and the first spatial feature.
Based on the above embodiments, the training module is specifically configured to:
acquiring the feature number of feature data information, selecting a target feature based on the feature number of the feature data information, and acquiring a data value corresponding to the target feature based on the fault prediction result;
and acquiring a state data value of the intelligent city Internet of things of the target feature at the target moment, and evaluating the prediction accuracy of the fault prediction model by combining the feature number and the data value corresponding to the target feature.
Based on the above embodiments, the model building module is further configured to:
carrying out batch normalization processing on the second spatial features of the feature data information and inputting the second spatial features to the result processing layer;
and obtaining a fault prediction result of the characteristic data information based on the second spatial characteristics after the batch normalization processing through the result processing layer.
Fig. 6 illustrates a physical schematic diagram of an electronic device, as shown in fig. 6, which may include: processor 610, communication interface (Communications Interface) 620, memory 630, and communication bus 640, wherein processor 610, communication interface 620, and memory 630 communicate with each other via communication bus 640. Processor 610 may invoke logic instructions in memory 630 to perform a smart city internet of things oriented fault prediction method comprising:
acquiring network data to be predicted of intelligent city Internet of things equipment;
inputting the network data into a failure prediction model after training to obtain a failure prediction result of the network data;
the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the feature data information of the intelligent city Internet of things equipment as a training set and prediction labels corresponding to the training set.
Further, the logic instructions in the memory 630 may be implemented in the form of software functional units and stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, where the computer program product includes a computer program, where the computer program can be stored on a non-transitory computer readable storage medium, and when the computer program is executed by a processor, the computer can execute the fault prediction method for the smart city internet of things provided by the above methods, and the method includes:
acquiring network data to be predicted of intelligent city Internet of things equipment;
inputting the network data into a failure prediction model after training to obtain a failure prediction result of the network data;
the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the feature data information of the intelligent city Internet of things equipment as a training set and prediction labels corresponding to the training set.
In still another aspect, the present invention further provides a non-transitory computer readable storage medium having stored thereon a computer program, which when executed by a processor, is implemented to perform the method for predicting a failure of a smart city internet of things provided by the above methods, the method comprising:
acquiring network data to be predicted of intelligent city Internet of things equipment;
inputting the network data into a failure prediction model after training to obtain a failure prediction result of the network data;
the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the feature data information of the intelligent city Internet of things equipment as a training set and prediction labels corresponding to the training set.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (6)
1. The fault prediction method for the Internet of things of the smart city is characterized by comprising the following steps of:
acquiring network data to be predicted of intelligent city Internet of things equipment;
inputting the network data into a failure prediction model after training to obtain a failure prediction result of the network data;
the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the characteristic data information of the intelligent city Internet of things equipment as a training set and a prediction label corresponding to the training set;
the training method of the fault prediction model comprises the following steps:
acquiring characteristic data information of the intelligent city Internet of things equipment as a training set of a model;
establishing an initial fault prediction model, inputting the characteristic data information into the initial fault prediction model, obtaining a fault prediction result of a characteristic vector, and evaluating prediction accuracy based on the fault prediction result;
iteratively updating the initial fault prediction model, and adjusting model parameters of the initial fault prediction model until the error values of the fault prediction result and the prediction label meet a preset threshold value, and the prediction accuracy reaches a target value to obtain a trained fault prediction model;
the establishing an initial fault prediction model, inputting the characteristic data information into the initial fault prediction model, obtaining a fault prediction result of a characteristic vector, and evaluating prediction accuracy based on the fault prediction result, including:
establishing an initial fault prediction model, wherein the initial fault prediction model comprises a convolution layer, a residual error structure layer and a result processing layer;
the convolution layer is used for extracting spatial characteristic information of the input characteristic data information based on the convolution layer, and acquiring a first spatial characteristic by combining an excitation function of the convolution layer;
wherein the convolution layer is established based on a convolution decomposition operation;
the residual structure layer is used for carrying out residual information learning based on the first spatial features output by the convolution layer and outputting second spatial features of the feature data information;
the result processing layer is used for obtaining a fault prediction result of the characteristic data information based on the second spatial characteristic output by the residual error structural layer;
evaluating the prediction accuracy of the fault prediction model based on the fault prediction result;
the residual structure layer comprises a first residual connecting channel and a second residual connecting channel;
residual information learning is performed based on the first spatial feature output by the convolution layer, and the residual information learning comprises the following steps:
acquiring a first residual mapping result of the first spatial feature output by the convolution layer through the first residual connecting channel, and acquiring a second residual mapping result of the first spatial feature output by the convolution layer through a second residual connecting channel;
and carrying out multiplication flow residual information learning on the first residual mapping result and the first spatial feature, and carrying out addition flow residual information learning on the second residual mapping result and the first spatial feature.
2. The smart city internet of things-oriented fault prediction method according to claim 1, wherein the evaluating the prediction accuracy of the fault prediction model based on the fault prediction result comprises:
acquiring the feature number of feature data information, selecting a target feature based on the feature number of the feature data information, and acquiring a data value corresponding to the target feature based on the fault prediction result;
and acquiring a state data value of the intelligent city Internet of things of the target feature at the target moment, and evaluating the prediction accuracy of the fault prediction model by combining the feature number and the data value corresponding to the target feature.
3. The smart city internet of things-oriented fault prediction method of claim 2, further comprising, after performing residual information learning based on the first spatial feature output by the convolution layer and outputting the second spatial feature of the feature data information:
carrying out batch normalization processing on the second spatial features of the feature data information and inputting the second spatial features to the result processing layer;
the fault prediction result of the feature data information is obtained by the result processing layer based on the second spatial feature output by the residual error structural layer, and the fault prediction result comprises:
and obtaining a fault prediction result of the characteristic data information based on the second spatial characteristics after the batch normalization processing through the result processing layer.
4. The utility model provides a trouble prediction device towards intelligent city thing networking which characterized in that includes:
the acquisition module is used for acquiring network data to be predicted of the intelligent city Internet of things equipment;
the prediction module is used for inputting the network data into a trained fault prediction model to obtain a fault prediction result of the network data;
the fault prediction model is built based on convolution decomposition operation and an improved residual error network structure, and is obtained by training the characteristic data information of the intelligent city Internet of things equipment as a training set and a prediction label corresponding to the training set;
the training method of the fault prediction model comprises the following steps:
acquiring characteristic data information of the intelligent city Internet of things equipment as a training set of a model;
establishing an initial fault prediction model, inputting the characteristic data information into the initial fault prediction model, obtaining a fault prediction result of a characteristic vector, and evaluating prediction accuracy based on the fault prediction result;
iteratively updating the initial fault prediction model, and adjusting model parameters of the initial fault prediction model until the error values of the fault prediction result and the prediction label meet a preset threshold value, and the prediction accuracy reaches a target value to obtain a trained fault prediction model;
the establishing an initial fault prediction model, inputting the characteristic data information into the initial fault prediction model, obtaining a fault prediction result of a characteristic vector, and evaluating prediction accuracy based on the fault prediction result, including:
establishing an initial fault prediction model, wherein the initial fault prediction model comprises a convolution layer, a residual error structure layer and a result processing layer;
the convolution layer is used for extracting spatial characteristic information of the input characteristic data information based on the convolution layer, and acquiring a first spatial characteristic by combining an excitation function of the convolution layer;
wherein the convolution layer is established based on a convolution decomposition operation;
the residual structure layer is used for carrying out residual information learning based on the first spatial features output by the convolution layer and outputting second spatial features of the feature data information;
the result processing layer is used for obtaining a fault prediction result of the characteristic data information based on the second spatial characteristic output by the residual error structural layer;
evaluating the prediction accuracy of the fault prediction model based on the fault prediction result;
the residual structure layer comprises a first residual connecting channel and a second residual connecting channel;
residual information learning is performed based on the first spatial feature output by the convolution layer, and the residual information learning comprises the following steps:
acquiring a first residual mapping result of the first spatial feature output by the convolution layer through the first residual connecting channel, and acquiring a second residual mapping result of the first spatial feature output by the convolution layer through a second residual connecting channel;
and carrying out multiplication flow residual information learning on the first residual mapping result and the first spatial feature, and carrying out addition flow residual information learning on the second residual mapping result and the first spatial feature.
5. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the smart city internet of things oriented fault prediction method of any one of claims 1 to 3 when the program is executed.
6. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the smart city internet of things oriented fault prediction method according to any of claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210376991.8A CN115001937B (en) | 2022-04-11 | 2022-04-11 | Smart city Internet of things-oriented fault prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210376991.8A CN115001937B (en) | 2022-04-11 | 2022-04-11 | Smart city Internet of things-oriented fault prediction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115001937A CN115001937A (en) | 2022-09-02 |
| CN115001937B true CN115001937B (en) | 2023-06-16 |
Family
ID=83023890
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210376991.8A Active CN115001937B (en) | 2022-04-11 | 2022-04-11 | Smart city Internet of things-oriented fault prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115001937B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115580635B (en) * | 2022-09-26 | 2023-06-13 | 广州健新科技有限责任公司 | Intelligent fault diagnosis method and system for Internet of things terminal |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3276539A1 (en) * | 2016-07-27 | 2018-01-31 | Samsung Electronics Co., Ltd. | Accelerator in convolutional neural network and method for operating the same |
| WO2018026995A1 (en) * | 2016-08-03 | 2018-02-08 | Schlumberger Technology Corporation | Multi-scale deep network for fault detection |
| JP2018112852A (en) * | 2017-01-11 | 2018-07-19 | 安川情報システム株式会社 | Fault prediction method, fault prediction device and fault prediction program |
| CN110390275A (en) * | 2019-07-04 | 2019-10-29 | 淮阴工学院 | A kind of gesture classification method based on transfer learning |
| WO2019233244A1 (en) * | 2018-06-08 | 2019-12-12 | 腾讯科技（深圳）有限公司 | Image processing method and apparatus, and computer readable medium, and electronic device |
| CN110798417A (en) * | 2019-10-24 | 2020-02-14 | 北京邮电大学 | Signal modulation identification method and device based on cyclic residual error network |
| WO2020215676A1 (en) * | 2019-04-26 | 2020-10-29 | 平安科技（深圳）有限公司 | Residual network-based image identification method, device, apparatus, and storage medium |
| CN112381775A (en) * | 2020-11-06 | 2021-02-19 | 厦门市美亚柏科信息股份有限公司 | Image tampering detection method, terminal device and storage medium |
| CN112910695A (en) * | 2021-01-22 | 2021-06-04 | 湖北工业大学 | Network fault prediction method based on global attention time domain convolutional network |
| CN112953629A (en) * | 2021-03-16 | 2021-06-11 | 北京邮电大学 | Method and system for analyzing uncertainty of optical network fault prediction |
| CN113783717A (en) * | 2021-08-12 | 2021-12-10 | 北京邮电大学 | Smart city network flow prediction method and system |
| CN113923123A (en) * | 2021-09-24 | 2022-01-11 | 天津大学 | Underwater wireless sensor network topology control method based on deep reinforcement learning |
| CN114186500A (en) * | 2022-02-16 | 2022-03-15 | 华中科技大学 | Marine bearing residual life prediction method based on transfer learning and multiple time windows |
-
2022
- 2022-04-11 CN CN202210376991.8A patent/CN115001937B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3276539A1 (en) * | 2016-07-27 | 2018-01-31 | Samsung Electronics Co., Ltd. | Accelerator in convolutional neural network and method for operating the same |
| WO2018026995A1 (en) * | 2016-08-03 | 2018-02-08 | Schlumberger Technology Corporation | Multi-scale deep network for fault detection |
| JP2018112852A (en) * | 2017-01-11 | 2018-07-19 | 安川情報システム株式会社 | Fault prediction method, fault prediction device and fault prediction program |
| WO2019233244A1 (en) * | 2018-06-08 | 2019-12-12 | 腾讯科技（深圳）有限公司 | Image processing method and apparatus, and computer readable medium, and electronic device |
| WO2020215676A1 (en) * | 2019-04-26 | 2020-10-29 | 平安科技（深圳）有限公司 | Residual network-based image identification method, device, apparatus, and storage medium |
| CN110390275A (en) * | 2019-07-04 | 2019-10-29 | 淮阴工学院 | A kind of gesture classification method based on transfer learning |
| CN110798417A (en) * | 2019-10-24 | 2020-02-14 | 北京邮电大学 | Signal modulation identification method and device based on cyclic residual error network |
| CN112381775A (en) * | 2020-11-06 | 2021-02-19 | 厦门市美亚柏科信息股份有限公司 | Image tampering detection method, terminal device and storage medium |
| CN112910695A (en) * | 2021-01-22 | 2021-06-04 | 湖北工业大学 | Network fault prediction method based on global attention time domain convolutional network |
| CN112953629A (en) * | 2021-03-16 | 2021-06-11 | 北京邮电大学 | Method and system for analyzing uncertainty of optical network fault prediction |
| CN113783717A (en) * | 2021-08-12 | 2021-12-10 | 北京邮电大学 | Smart city network flow prediction method and system |
| CN113923123A (en) * | 2021-09-24 | 2022-01-11 | 天津大学 | Underwater wireless sensor network topology control method based on deep reinforcement learning |
| CN114186500A (en) * | 2022-02-16 | 2022-03-15 | 华中科技大学 | Marine bearing residual life prediction method based on transfer learning and multiple time windows |
Non-Patent Citations (5)
| Title |
|---|
| Data clustering-based fault detection in WSNs;Lanlan Rui等;2015 Seventh International Conference on Advanced Computational Intelligence (ICACI);第334-336页 * |
| Human Action Recognition by Learning Spatio-Temporal Features With DNNs;LEI WANG等;Digital Object Identifier 10.1109/ACCESS.;第17913-17917页 * |
| 基于红外图像的变电设备热缺陷自调整残差网络诊断模型;王有元;李后英;梁玄鸿;李昀琪;蔚超;陆云才;;高电压技术;46(第09期);正文第1-2页 * |
| 基于组合模型的短时交通流量预测算法;芮兰兰等;《电 子 与 信 息 学 报》;第38卷(第5期);第1228-1232页 * |
| 支持向量机回归预测在网络故障检测中的应用;孟洛明;北京邮电大学学报;第37卷(第4期);第24-27页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115001937A (en) | 2022-09-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP4198820A1 (en) | Training method for semi-supervised learning model, image processing method, and device | |
| CN112181666A (en) | Method, system, equipment and readable storage medium for equipment evaluation and federal learning importance aggregation based on edge intelligence | |
| CN110926782B (en) | Circuit breaker fault type judgment method and device, electronic equipment and storage medium | |
| CN110991362A (en) | Pedestrian detection model based on attention mechanism | |
| CN110298374B (en) | Driving track energy consumption analysis method and device based on deep learning | |
| CN115001937B (en) | Smart city Internet of things-oriented fault prediction method and device | |
| CN111224905A (en) | Multi-user detection method based on convolution residual error network in large-scale Internet of things | |
| CN113162787B (en) | Method for fault location in a telecommunication network, node classification method and related devices | |
| CN116489038A (en) | Network traffic prediction method, device, equipment and medium | |
| CN116452810A (en) | Multi-level semantic segmentation method and device, electronic equipment and storage medium | |
| CN113688814B (en) | Image recognition method and device | |
| CN113194493B (en) | Wireless network data missing attribute recovery method and device based on graph neural network | |
| WO2024051655A1 (en) | Method and apparatus for processing histopathological whole-slide image, and medium and electronic device | |
| CN116827685B (en) | Dynamic defense strategy method of micro-service system based on deep reinforcement learning | |
| CN113541986B (en) | Fault prediction method and device for 5G slice and computing equipment | |
| CN112396166A (en) | Graph convolution neural network training method and device based on mixed granularity aggregator | |
| CN114445692B (en) | Image recognition model construction method and device, computer equipment and storage medium | |
| CN115426671A (en) | Method, system and equipment for graph neural network training and wireless cell fault prediction | |
| CN115758643A (en) | Network flow prediction method and device based on temporal-spatial feature fusion and storage medium | |
| CN115587616A (en) | Network model training method and device, storage medium and computer equipment | |
| CN115457365A (en) | Model interpretation method and device, electronic equipment and storage medium | |
| CN114819096A (en) | Model training method and device, electronic equipment and storage medium | |
| CN114925720A (en) | Small sample modulation signal identification method based on space-time mixed feature extraction network | |
| CN113743594A (en) | Network flow prediction model establishing method and device, electronic equipment and storage medium | |
| CN110929118A (en) | Network data processing method, equipment, device and medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |