CN116028893A - Intelligent data monitoring method, system and electronic equipment - Google Patents
Intelligent data monitoring method, system and electronic equipment Download PDFInfo
- Publication number
- CN116028893A CN116028893A CN202310307622.8A CN202310307622A CN116028893A CN 116028893 A CN116028893 A CN 116028893A CN 202310307622 A CN202310307622 A CN 202310307622A CN 116028893 A CN116028893 A CN 116028893A
- Authority
- CN
- China
- Prior art keywords
- dimensional
- dimensional signal
- monitoring module
- layer
- data set
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 173
- 238000000034 method Methods 0.000 title claims abstract description 40
- 230000004927 fusion Effects 0.000 claims abstract description 29
- 230000006870 function Effects 0.000 claims description 19
- 230000004913 activation Effects 0.000 claims description 18
- 238000010606 normalization Methods 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 16
- 238000011176 pooling Methods 0.000 claims description 15
- 238000010276 construction Methods 0.000 claims description 13
- 230000009977 dual effect Effects 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000003745 diagnosis Methods 0.000 abstract description 10
- 238000001514 detection method Methods 0.000 abstract description 3
- 239000010410 layer Substances 0.000 description 67
- 238000010586 diagram Methods 0.000 description 5
- 238000000605 extraction Methods 0.000 description 5
- 238000006116 polymerization reaction Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000012938 design process Methods 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 239000013047 polymeric layer Substances 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000002902 bimodal effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000007670 refining Methods 0.000 description 1
Images
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses an intelligent data monitoring method, an intelligent data monitoring system and electronic equipment, and relates to the technical field of data monitoring. According to the intelligent data monitoring method provided by the invention, the one-dimensional signal monitoring module and the two-dimensional signal monitoring module are respectively constructed by acquiring the double-source data set, so that more characteristic information can be provided, and meanwhile, the reliability of data detection and fault diagnosis is improved. And the dual-mode fusion monitoring module is built based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module to monitor the monitoring data in real time, so that the accuracy and reliability of data monitoring can be effectively improved, and efficient and real-time automatic data monitoring and fault diagnosis in a database are realized.
Description
Technical Field
The present invention relates to the field of data monitoring technologies, and in particular, to an intelligent data monitoring method, system and electronic device.
Background
In recent years, with the rapid development of computer technology and rapid expansion of industrial scale, various large and small industrial equipment is popularized and applied in a plurality of fields, and the normal operation of the industrial equipment becomes an important guarantee for safe and reliable operation of various businesses. However, at present, when the industrial equipment fails, the fault detection is still mainly performed manually, and the process is highly dependent on expert knowledge and experience and consumes a great deal of time and labor cost. In addition, a large amount of fault priori knowledge is difficult to effectively manage and utilize, and data resource waste can be caused. Therefore, a highly efficient and feasible set of artificial intelligence methods is urgently needed for automated diagnosis of fault data, thereby optimizing the fault prediction efficiency of the data monitoring system.
Disclosure of Invention
The invention aims to provide an intelligent data monitoring method, an intelligent data monitoring system and electronic equipment, which can realize automatic diagnosis of fault data and improve the fault prediction efficiency and the prediction accuracy.
In order to achieve the above object, the present invention provides the following solutions:
an intelligent data monitoring method, comprising:
acquiring a double-source data set; the dual source data set includes: a one-dimensional time sequence signal identification data set and a two-dimensional signal identification data set;
constructing a one-dimensional signal monitoring module based on the one-dimensional time sequence signal identification data set;
constructing a two-dimensional signal monitoring module based on the two-dimensional signal identification data set;
building a dual-mode fusion monitoring module based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module;
inputting the data to be monitored into the dual-mode fusion monitoring module to output a monitoring result in real time; the monitoring result includes a fault type.
Optionally, before obtaining the dual-source data set, the method further includes:
acquiring monitoring data information with fixed time length in a normal running state as a positive sample, and monitoring data information with fixed time length in different fault types as a negative sample, and labeling the negative sample according to the fault type;
performing first preprocessing on the positive samples and the negative samples to generate a one-dimensional time sequence signal identification data set; the first preprocessing includes: outlier rejection and dimension normalization;
performing second preprocessing on the positive sample and the negative sample, and converting the second preprocessed positive sample and the second preprocessed negative sample into image data to generate a two-dimensional signal identification data set; the second pretreatment includes: wavelet transformation.
Optionally, constructing a one-dimensional signal monitoring module based on the one-dimensional time sequence signal identification data set specifically includes:
constructing a one-dimensional signal identification network; the one-dimensional signal identification network comprises three different types of one-dimensional convolution combinations and two full-connection layers;
and carrying out iterative training and testing on the one-dimensional signal recognition network by adopting the one-dimensional time sequence signal recognition data set until the recognition accuracy of the one-dimensional signal recognition network reaches a first preset value to obtain a trained one-dimensional signal recognition network, and taking the trained one-dimensional signal recognition network as the one-dimensional signal monitoring module.
Optionally, the first one-dimensional convolution combination in the one-dimensional signal identification network comprises: a one-dimensional convolution layer with a convolution kernel of 3 and a step length of 1, a layer of activation functions and a batch of standardization layers;
the second one-dimensional convolution combination in the one-dimensional signal identification network comprises: a one-dimensional convolution layer with a convolution kernel of 5 and a step length of 2, a layer of activation function and a batch of standardization layers;
a third one-dimensional convolution combination in the one-dimensional signal identification network includes: a one-dimensional convolution layer with a convolution kernel of 3 and a step size of 4, a layer activation function, and a batch normalization layer.
Optionally, constructing a two-dimensional signal monitoring module based on the two-dimensional signal identification data set specifically includes:
constructing a two-dimensional signal identification network; the two-dimensional signal recognition network includes: three different types of convolution combinations, a global max pooling layer and two full connection layers;
and carrying out iterative training and testing on the two-dimensional signal recognition network by adopting the two-dimensional signal recognition data set until the recognition accuracy of the two-dimensional signal recognition network reaches a second preset value to obtain a trained two-dimensional signal recognition network, and taking the trained two-dimensional signal recognition network as the two-dimensional signal monitoring module.
Optionally, the first convolution combination of the two-dimensional signal recognition network includes a convolution kernel of 33, a layer of activation functions and a batch of standardization layers;
the second convolution combination of the two-dimensional signal recognition network comprises a convolution kernel of 55, a layer activation function, a batch normalization layer and a pooling kernel of 2 +.>2 a pooling layer;
the third convolution combination of the two-dimensional signal identification network comprises a convolution kernel of 33, a layer activation function, a batch normalization layer and a pooling kernel of 4 +.>4.
Optionally, building a dual-mode fusion monitoring module based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module, which specifically includes:
removing the last full-connection layer of the one-dimensional signal monitoring module and the last full-connection layer of the two-dimensional signal monitoring module, and constructing a tail classification module to obtain the dual-mode fusion monitoring module; the tail classification module comprises: a locally polymerized layer with a core of 4, a one-dimensional convolution layer with a core of 7 and a step of 5, and a fully connected layer.
Optionally, the step size of all convolution layers in the two-dimensional signal identification network is 1.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
according to the intelligent data monitoring method provided by the invention, the one-dimensional signal monitoring module and the two-dimensional signal monitoring module are respectively constructed by acquiring the double-source data set, so that more characteristic information can be provided, and meanwhile, the reliability of data detection and fault diagnosis is improved. And the dual-mode fusion monitoring module is built based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module to monitor the monitoring data in real time, so that the accuracy and reliability of data monitoring can be effectively improved, and efficient and real-time automatic data monitoring and fault diagnosis in a database are realized.
Corresponding to the intelligent data monitoring method provided by the invention, the invention also discloses the following embodiments:
an intelligent data monitoring system is applied to the intelligent data monitoring method; the system comprises:
the data set acquisition unit is used for acquiring a double-source data set; the dual source data set includes: a one-dimensional time sequence signal identification data set and a two-dimensional signal identification data set;
a first module construction unit for constructing a one-dimensional signal monitoring module based on the one-dimensional time sequence signal identification data set;
a second module construction unit for constructing a two-dimensional signal monitoring module based on the two-dimensional signal identification data set;
the third module construction unit is used for constructing a dual-mode fusion monitoring module based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module;
the data monitoring unit is used for inputting the data to be monitored into the dual-mode fusion monitoring module and outputting a monitoring result in real time; the monitoring result includes a fault type.
An electronic device, comprising:
a memory for storing a computer program;
and the processor is connected with the memory and is used for calling and executing the computer program so as to implement the intelligent data monitoring method.
The technical effects achieved by the system and the electronic device provided by the invention are the same as those achieved by the intelligent data monitoring method provided by the invention, so that the detailed description is omitted.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of an intelligent data monitoring method provided by the invention;
FIG. 2 is a schematic diagram of a one-dimensional signal recognition network according to the present invention;
FIG. 3 is a schematic diagram of a two-dimensional signal recognition network according to the present invention;
fig. 4 is a schematic structural diagram of a dual-mode fusion monitoring module provided by the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but 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 invention aims to provide an intelligent data monitoring method, an intelligent data monitoring system and electronic equipment, which can realize automatic diagnosis of fault data and improve the fault prediction efficiency and the prediction accuracy.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
As shown in fig. 1, the intelligent data monitoring method provided by the invention includes:
step 100: a dual source data set is acquired. The dual source data set includes: a one-dimensional time series signal identification data set and a two-dimensional signal identification data set. In the invention, the construction mode of the obtained double-source data set comprises the following steps:
and collecting monitoring data information with fixed time length under normal running state (namely, a state without faults and any abnormal running state) in the database in a large quantity as a positive sample, and collecting monitoring data information with fixed time length under different fault types in the database as a negative sample based on fault priori knowledge. And enabling the number of the fault types to be n, and labeling fault data samples according to the fault types. The data in the database can be monitoring data of equipment such as thermal control equipment, power equipment and the like.
The monitoring data information in the collected positive samples and negative samples is preprocessed (outlier rejection, dimension normalization and the like) and then used as a one-dimensional time sequence signal identification data set. The monitoring data information in the collected positive sample and negative sample is preprocessed (such as wavelet transformation) and then is converted into image data one by one, so that a two-dimensional signal identification data set is generated.
Step 101: and constructing a one-dimensional signal monitoring module based on the one-dimensional time sequence signal identification data set. In the step, in order to realize fault data identification based on One-dimensional signals, a light One-dimensional signal identification network (One-dimensional signal recognition network, OSR-Net) is designed, and characteristic information in One-dimensional time sequence monitoring data is fully extracted with lower calculation cost. As shown in fig. 2, the network structure of the designed OSR-Net comprises a sandwich structure consisting of three different types of convolution combinations and two fully connected layers.
The design process of the network structure of the OSR-Net comprises the following steps: first, a one-dimensional convolution combination F1 is constructed. The structure of the one-dimensional convolution combination F1 comprises a one-dimensional convolution layer with a convolution kernel of 3 and a step size of 1, a layer of activation functions and a batch normalization layer. The one-dimensional convolution combination F1 is mainly used for primary extraction of the characteristic information. Next, a one-dimensional convolution combination S1 is constructed. The structure of the one-dimensional convolution combination S1 comprises a one-dimensional convolution layer with a convolution kernel of 5 and a step size of 2, a layer activation function and a batch normalization layer. The one-dimensional convolution combination S1 is mainly used for screening key characteristic information. Next, a one-dimensional convolution combination T1 is constructed. The structure of the one-dimensional convolution combination T1 comprises a one-dimensional convolution layer with a convolution kernel of 3 and a step length of 4, a layer of activation functions and a batch normalization layer. The one-dimensional convolution combination T1 is mainly used for fine extraction and optimization of key feature information. And finally, constructing two full-connection layers for final fault category classification.
The following is 1024 in one dimension 1, a transmission process of data in an OSR-Net is described by taking one-dimensional time sequence signal data as an example, and the method specifically comprises the following steps:
first, dimension is 10241 is input into a one-dimensional convolution combination F1 of OSR-Net to obtain a one-dimensional feature 1024 +.>1. Next, the one-dimensional feature 1024->1 is input into a one-dimensional convolution combination S1 to obtain one-dimensional characteristics 512 +.>1. Next, one-dimensional feature 512->1 is input into a one-dimensional convolution combination T1 to obtain a one-dimensional feature 128 +.>1. Finally, one-dimensional feature 128->1 are input into two fully connected layers to sequentially obtain one-dimensional characteristics 64 + ->1 and one-dimensional feature n->1。
Further, a one-dimensional signal monitoring module is generated based on the one-dimensional time sequence signal identification data set obtained in the step 100 and the constructed OSR-Net, and the specific steps are as follows:
and performing iterative training and testing on the OSR-Net by adopting a training set and a testing set formed by the one-dimensional time sequence signal identification data set. In the training and testing process, the number of classified categories is n, and the OSR-Net obtained through final training is used as a one-dimensional signal monitoring module until the identification accuracy of data in a test set reaches more than a first preset value (for example, 96%).
Step 102: and constructing a two-dimensional signal monitoring module based on the two-dimensional signal identification data set. In this step, in order to realize image-based fault data recognition, a lightweight Two-dimensional signal recognition network (Two-dimensional signal recognition network, TSR-Net) can be designed to sufficiently extract the feature information of the Two-dimensional signal recognition data set map at a low calculation cost. The network structure of the TSR-Net is shown in FIG. 3, and the two-dimensional signal recognition network includes: three different types of convolution combinations, one global max pooling layer and two fully connected layers.
Based on this, the design process of TSR-Net specifically includes: first, a convolution combination F2 is constructed. The structure of convolution combination F2 includes a convolution kernel of 33, a layer activation function, and a batch normalization layer. The batch normalization layer is used for preliminary extraction of feature information. Next, a convolution combination S2 is constructed. The structure of the convolution combination S2 comprises a convolution kernel of 5 +>5, a layer activation function, a batch normalization layer and a pooling kernel of 2 +.>2. This pooling core is 2->The pooling layer of 2 is mainly used for screening key characteristic information. Next, a convolution combination T2 is constructed. The structure of the convolution combination T2 comprises a convolution kernel of 3 +.>3, a layer activation function, a batch normalization layer and a pooling kernel of 4 +.>4. This pooling core is 4->And 4, the pooling layer is mainly used for the fine extraction and optimization of key feature information. A global maximization layer is then built for feature aggregation and dimension conversion of the two-dimensional signal. Finally, two fully connected layers are constructed for final fault class classification. The step size of all convolution layers in each convolution combination constructed above is 1.
Further, in one dimension 10241024/>1 (i.e., two-dimensional signal diagram), the data transmission process of the above-constructed TSR-Net will be described.
First, dimension is 10241024/>1 into a two-dimensional convolution combination F2 to obtain a feature map 1024 +.>1024. Next, feature map 1024->1024 are input into a two-dimensional convolution combination S2 to obtain a feature map 512 +.>512. Subsequently, the feature map 512->512 is input into a convolution combination T2 to obtain a feature map 128 +.>128. Then, feature map 128->128 are input into a global maximization layer to obtain dimension-converted one-dimensional features 1281. Finally, one-dimensional feature 128->1 are input into two full-connection layers to sequentially obtain one-dimensional characteristics 64 +>1 and n->1。
Further, a two-dimensional signal monitoring module is generated based on the two-dimensional signal identification data set and the TSR-Net, specifically, a training set and a testing set formed by the two-dimensional signal identification data set are adopted to conduct iterative training and testing of the TSR-Net, in the training and testing process, the number of classified categories is n until the identification accuracy of the data of the testing set reaches more than a second preset value (for example, 98%), and the TSR-Net obtained through final training is used as the two-dimensional signal monitoring module.
Step 103: and constructing a dual-mode fusion monitoring module based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module. The structure of the built dual-mode fusion monitoring module is shown in fig. 4, and the dual-mode fusion monitoring module consists of a one-dimensional signal monitoring module, a two-dimensional signal monitoring module and a tail classification module. The input of the dual-mode fusion monitoring module is in two different modes of the same data to be identified, namely a one-dimensional time sequence signal output by the one-dimensional signal monitoring module and a two-dimensional signal diagram output by the two-dimensional signal monitoring module.
Based on the design process of the dual-mode fusion monitoring module is as follows:
removing the last full-connection layer for classification in the one-dimensional signal monitoring module, and outputting the signal with the size of 641, so as to fully extract the characteristic information contained in the one-dimensional signal. Removing the last full-connection layer in the two-dimensional signal monitoring module, and outputting only 64 +.>1 to extract the characteristic information in the two-dimensional signal map. And constructing a tail classification module to fuse the output characteristics of the two large single-mode monitoring modules (namely the one-dimensional signal monitoring module and the two-dimensional signal monitoring module with the last full-connection layer removed), thereby realizing final data classification.
In order to realize accurate extraction of the features, the invention designs a local aggregation layer in the tail classification module. The function of the local polymeric layer is: when the input of the local polymerization layer is one-dimensional feature and the core of the local polymerization layer is set to be m, the average value of every m elements in the one-dimensional feature is sequentially obtained in the local polymerization layer, and the one-dimensional feature composed of a series of average values is output. For example, if the core m of the locally polymerized layer is 4, then its output is 32 in size 1.
Based on the above design, the local polymerization layer has a dimension of 10241 and a dimension of 1024 +.>1024/>1, a data processing process of a tail classification module is described, and the specific steps are as follows:
firstly, the output characteristics of two large single-mode monitoring modules (namely, the output characteristics of a one-dimensional signal monitoring module with the last fully-connected layer removed and the output characteristics of a two-dimensional signal monitoring module with the last fully-connected layer removed) are spliced to obtain a single-mode signal monitoring module with the size of 1281. Subsequently, the size is 128->1 into the local polymeric layer with core 4, the output size is 32 +.>1. Then carrying out feature refining on the one-dimensional convolution layer with the kernel of 7 (i.e. f=7) and the step length of 5 (i.e. s=5) to obtain the one-dimensional convolution layer with the size of 6 +.>1. Finally, the last full-connection layer is based on 6 +.>1, and performing final data classification by the bimodal feature.
In order to further improve the accuracy of data monitoring, after the dual-mode fusion monitoring module is constructed, the invention also carries out distributed training on the dual-mode fusion monitoring module based on the constructed dual-source data set, and the method comprises the following steps:
firstly, fixing internal parameters of a one-dimensional signal monitoring module and a two-dimensional signal monitoring module, respectively inputting a one-dimensional time sequence signal identification data set and a two-dimensional signal identification data set at the input ends of the one-dimensional signal monitoring module and the two-dimensional signal monitoring module, and performing iterative training and testing of a tail classification module, wherein the number of classified categories is n, and canceling parameter fixation until the identification accuracy of the tail classification module on test set data reaches more than 96%. And then, directly carrying out iterative training and testing on the whole dual-mode fusion monitoring module with the parameters cancelled fixed until the identification accuracy of the whole dual-mode fusion monitoring module to the test set data reaches more than 98%, and taking the network model obtained by final training as the dual-mode fusion monitoring module.
Step 104: and inputting the data to be monitored into the dual-mode fusion monitoring module to output a monitoring result in real time. The monitoring result includes the fault type.
Based on the description, the conventional neural network only adopts the single-mode information to carry out data identification, and the characteristic representation completeness of the single-mode information is low, so that the identification accuracy and reliability are poor. According to the invention, the dual-mode fusion monitoring module is adopted to take dual-mode data as a discrimination basis, two different-mode data are converted and fused, and more characteristic information can be provided for final decision of data monitoring, so that the accuracy and reliability of data monitoring are effectively improved, and efficient and real-time data monitoring and fault diagnosis in a database are realized.
Corresponding to the intelligent data monitoring method provided by the invention, the invention also discloses the following embodiments:
an intelligent data monitoring system is applied to the intelligent data monitoring method. The system comprises: the system comprises a data set acquisition unit, a first module construction unit, a second module construction unit, a third module construction unit and a data monitoring unit.
The data set acquisition unit is used for acquiring a double-source data set. The dual source data set includes: a one-dimensional time series signal identification data set and a two-dimensional signal identification data set.
The first module construction unit is used for constructing a one-dimensional signal monitoring module based on the one-dimensional time sequence signal identification data set.
The second module construction unit is used for constructing the two-dimensional signal monitoring module based on the two-dimensional signal identification data set.
The third module construction unit is used for constructing a dual-mode fusion monitoring module based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module.
The data monitoring unit is used for inputting the data to be monitored into the dual-mode fusion monitoring module to output the monitoring result in real time. The monitoring result includes the fault type.
The invention also provides an electronic device, which comprises: memory and a processor.
Wherein the memory is for storing a computer program.
The processor is connected with the memory for retrieving and executing the computer program, so as to implement the intelligent data monitoring method.
Furthermore, the computer program in the above-described memory may be stored in a computer-readable storage medium when it is implemented in the form of a software functional unit and sold or used as a separate 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 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 or 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: various media capable of storing program codes, such as a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.
In summary, the technical scheme provided by the invention can accurately realize automatic data fault type diagnosis in the data monitoring system in real time, greatly improve the efficiency and reliability of data monitoring and fault diagnosis, fully utilize a large amount of fault priori knowledge and avoid the waste of data resources.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (10)
1. An intelligent data monitoring method, comprising:
acquiring a double-source data set; the dual source data set includes: a one-dimensional time sequence signal identification data set and a two-dimensional signal identification data set;
constructing a one-dimensional signal monitoring module based on the one-dimensional time sequence signal identification data set;
constructing a two-dimensional signal monitoring module based on the two-dimensional signal identification data set;
building a dual-mode fusion monitoring module based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module;
inputting the data to be monitored into the dual-mode fusion monitoring module to output a monitoring result in real time; the monitoring result includes a fault type.
2. The intelligent data monitoring method of claim 1, further comprising, prior to obtaining the dual source data set:
acquiring monitoring data information with fixed time length in a normal running state as a positive sample, and monitoring data information with fixed time length in different fault types as a negative sample, and labeling the negative sample according to the fault type;
performing first preprocessing on the positive samples and the negative samples to generate a one-dimensional time sequence signal identification data set; the first preprocessing includes: outlier rejection and dimension normalization;
performing second preprocessing on the positive sample and the negative sample, and converting the second preprocessed positive sample and the second preprocessed negative sample into image data to generate a two-dimensional signal identification data set; the second pretreatment includes: wavelet transformation.
3. The intelligent data monitoring method according to claim 1, wherein constructing a one-dimensional signal monitoring module based on the one-dimensional time sequence signal identification data set specifically comprises:
constructing a one-dimensional signal identification network; the one-dimensional signal identification network comprises three different types of one-dimensional convolution combinations and two full-connection layers;
and carrying out iterative training and testing on the one-dimensional signal recognition network by adopting the one-dimensional time sequence signal recognition data set until the recognition accuracy of the one-dimensional signal recognition network reaches a first preset value to obtain a trained one-dimensional signal recognition network, and taking the trained one-dimensional signal recognition network as the one-dimensional signal monitoring module.
4. The intelligent data monitoring method of claim 3, wherein the first one-dimensional convolution combination in the one-dimensional signal recognition network comprises: a one-dimensional convolution layer with a convolution kernel of 3 and a step length of 1, a layer of activation functions and a batch of standardization layers;
the second one-dimensional convolution combination in the one-dimensional signal identification network comprises: a one-dimensional convolution layer with a convolution kernel of 5 and a step length of 2, a layer of activation function and a batch of standardization layers;
a third one-dimensional convolution combination in the one-dimensional signal identification network includes: a one-dimensional convolution layer with a convolution kernel of 3 and a step size of 4, a layer activation function, and a batch normalization layer.
5. The method for intelligent data monitoring according to claim 4, wherein constructing a two-dimensional signal monitoring module based on the two-dimensional signal identification data set specifically comprises:
constructing a two-dimensional signal identification network; the two-dimensional signal recognition network includes: three different types of convolution combinations, a global max pooling layer and two full connection layers;
and carrying out iterative training and testing on the two-dimensional signal recognition network by adopting the two-dimensional signal recognition data set until the recognition accuracy of the two-dimensional signal recognition network reaches a second preset value to obtain a trained two-dimensional signal recognition network, and taking the trained two-dimensional signal recognition network as the two-dimensional signal monitoring module.
6. The intelligent data monitoring method according to claim 5, wherein the first convolution combination of the two-dimensional signal recognition network comprises a convolution kernel of 33, a layer activation function and a batch normalization layer
The second convolution combination of the two-dimensional signal recognition network comprises a convolution kernel of 55, a layer activation function, a batch normalization layer and a pooling kernel of 2 +.>2 a pooling layer;
7. The intelligent data monitoring method according to claim 6, wherein a dual-mode fusion monitoring module is built based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module, and specifically comprises:
removing the last full-connection layer of the one-dimensional signal monitoring module and the last full-connection layer of the two-dimensional signal monitoring module, and constructing a tail classification module to obtain the dual-mode fusion monitoring module; the tail classification module comprises: a locally polymerized layer with a core of 4, a one-dimensional convolution layer with a core of 7 and a step of 5, and a fully connected layer.
8. The intelligent data monitoring method according to claim 6, wherein the step size of all convolution layers in the two-dimensional signal recognition network is 1.
9. An intelligent data monitoring system, characterized by being applied to the intelligent data monitoring method according to any one of claims 1-8; the system comprises:
the data set acquisition unit is used for acquiring a double-source data set; the dual source data set includes: a one-dimensional time sequence signal identification data set and a two-dimensional signal identification data set;
a first module construction unit for constructing a one-dimensional signal monitoring module based on the one-dimensional time sequence signal identification data set;
a second module construction unit for constructing a two-dimensional signal monitoring module based on the two-dimensional signal identification data set;
the third module construction unit is used for constructing a dual-mode fusion monitoring module based on the one-dimensional signal monitoring module and the two-dimensional signal monitoring module;
the data monitoring unit is used for inputting the data to be monitored into the dual-mode fusion monitoring module and outputting a monitoring result in real time; the monitoring result includes a fault type.
10. An electronic device, comprising:
a memory for storing a computer program;
a processor, coupled to the memory, for retrieving and executing the computer program to implement the intelligent data monitoring method of any of claims 1-8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310307622.8A CN116028893A (en) | 2023-03-28 | 2023-03-28 | Intelligent data monitoring method, system and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310307622.8A CN116028893A (en) | 2023-03-28 | 2023-03-28 | Intelligent data monitoring method, system and electronic equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116028893A true CN116028893A (en) | 2023-04-28 |
Family
ID=86089544
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310307622.8A Pending CN116028893A (en) | 2023-03-28 | 2023-03-28 | Intelligent data monitoring method, system and electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116028893A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117368598A (en) * | 2023-10-09 | 2024-01-09 | 武汉杰创博特自动化科技有限公司 | Electric energy quality monitoring method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111523509A (en) * | 2020-05-08 | 2020-08-11 | 江苏迪赛司自动化工程有限公司 | Equipment fault diagnosis and health monitoring method integrating physical and deep expression characteristics |
CN115204218A (en) * | 2022-06-25 | 2022-10-18 | 江苏大学流体机械温岭研究院 | Intelligent fault identification method for hydraulic plunger pump with integration of vibration characteristics and deep learning |
CN115456088A (en) * | 2022-09-19 | 2022-12-09 | 浙江科技学院 | Motor fault classification method based on self-encoder feature lifting dimension and multi-sensing fusion |
-
2023
- 2023-03-28 CN CN202310307622.8A patent/CN116028893A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111523509A (en) * | 2020-05-08 | 2020-08-11 | 江苏迪赛司自动化工程有限公司 | Equipment fault diagnosis and health monitoring method integrating physical and deep expression characteristics |
CN115204218A (en) * | 2022-06-25 | 2022-10-18 | 江苏大学流体机械温岭研究院 | Intelligent fault identification method for hydraulic plunger pump with integration of vibration characteristics and deep learning |
CN115456088A (en) * | 2022-09-19 | 2022-12-09 | 浙江科技学院 | Motor fault classification method based on self-encoder feature lifting dimension and multi-sensing fusion |
Non-Patent Citations (1)
Title |
---|
陈丹敏等: "基于多源异构信息迁移学习的融合故障诊断方法", 信息工程大学学报, pages 153 - 158 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117368598A (en) * | 2023-10-09 | 2024-01-09 | 武汉杰创博特自动化科技有限公司 | Electric energy quality monitoring method and device |
CN117368598B (en) * | 2023-10-09 | 2024-04-12 | 武汉杰创博特自动化科技有限公司 | Electric energy quality monitoring method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110415215B (en) | Intelligent detection method based on graph neural network | |
CN112767106B (en) | Automatic auditing method, system, computer readable storage medium and auditing equipment | |
CN113608140A (en) | Battery fault diagnosis method and system | |
CN110599459A (en) | Underground pipe network risk assessment cloud system based on deep learning | |
CN116028893A (en) | Intelligent data monitoring method, system and electronic equipment | |
CN112200772A (en) | Pox check out test set | |
CN116823793A (en) | Device defect detection method, device, electronic device and readable storage medium | |
CN116977907A (en) | Image recognition-based power inspection fault diagnosis method and system | |
CN114419005A (en) | Crack automatic detection method based on improved light weight CNN and transfer learning | |
CN114358371A (en) | Photovoltaic short-term power prediction method and device based on deep learning | |
CN115858725B (en) | Text noise screening method and system based on unsupervised graph neural network | |
CN116580232A (en) | Automatic image labeling method and system and electronic equipment | |
CN116129182A (en) | Multi-dimensional medical image classification method based on knowledge distillation and neighbor classification | |
CN115452376A (en) | Bearing fault diagnosis method based on improved lightweight deep convolution neural network | |
CN115880472A (en) | Intelligent diagnosis and analysis system for electric power infrared image data | |
CN114708470A (en) | Illegal picture identification method, medium and computing device | |
CN114077663A (en) | Application log analysis method and device | |
CN111428224A (en) | Computer account login method based on face recognition | |
CN116415137B (en) | Emotion quantification method, device, equipment and storage medium based on multi-modal characteristics | |
CN114638558B (en) | Data set classification method for operation accident analysis of comprehensive energy system | |
CN115641330B (en) | Flexible circuit board defect detection method and system based on image processing | |
CN115116594B (en) | Method and device for detecting effectiveness of medical device | |
CN116484172B (en) | Mechanical fault diagnosis method and system based on robust multi-scale network | |
CN113609996A (en) | Line inspection method, device and storage medium | |
CN116071558A (en) | Processing method and device and electronic equipment |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20230428 |
|
RJ01 | Rejection of invention patent application after publication |