An intelligent operation and maintenance method based on big data and machine learning
Intelligent operation and maintenance methods using big data and machine learning, which utilize convolutional neural networks to process equipment data, solve the real-time problem of equipment operation and maintenance, realize real-time monitoring and prediction of equipment status, and improve the efficiency and accuracy of operation and maintenance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA TOBACCO HENAN IND CO LTD
- Filing Date
- 2022-06-10
- Publication Date
- 2026-05-19
AI Technical Summary
Existing technologies lack real-time capabilities in equipment maintenance, making it impossible to effectively monitor and predict equipment status, resulting in delayed maintenance responses.
The intelligent operation and maintenance method based on big data and machine learning is adopted. The sample data generated by the equipment is processed by convolutional neural network to generate scattered data and perform pixel recognition, filtering and fusion to generate result feature data to judge the equipment status, monitor in real time and determine the operation and maintenance plan.
It enables real-time monitoring and prediction of equipment status, improving the real-time performance and accuracy of operation and maintenance, and enabling timely response to equipment changes.
Smart Images
Figure CN114969050B_ABST
Abstract
Description
Technical Field
[0001] This application seeks protection for big data-based operation and maintenance technology, and particularly relates to an intelligent operation and maintenance method based on big data and machine learning. Background Technology
[0002] As an important data asset, big data is being used more and more widely and deeply. It also has great potential for intelligent operation and maintenance. However, the current operation of equipment is basically in the process of detection, problem discovery and problem solving, which is insufficient in terms of real-time detection of equipment. Summary of the Invention
[0003] To address one or more of the problems mentioned above in the background technology, this application provides an intelligent operation and maintenance method based on big data and machine learning.
[0004] This application provides an intelligent operation and maintenance method based on big data and machine learning, including:
[0005] Acquire sample data, which is the data generated by the device during operation, including set parameters, running data, and data on changes in running data;
[0006] The sample data is divided into two groups and input into the editing end of the convolutional neural network model. The editing end edits the data to generate first scattered data and second scattered data based on the Cartesian coordinate system. The first scattered data and second scattered data are then input into the convolutional layer.
[0007] The convolutional layer performs pixel recognition on the first scattered data and the second scattered data respectively, and filters out the first scattered data and the second scattered data that meet the preset conditions to generate the first filtered scattered data and the second filtered scattered data.
[0008] The first filtered scatter data and the second filtered scatter data are input into the fusion layer of the convolutional neural network, and the first filtered scatter data and the second filtered scatter data are fused to generate fused scatter data.
[0009] The fused scatter data is decoded to generate result feature data.
[0010] Optionally, an operation and maintenance plan is generated based on the feature data, including:
[0011] The operational status is determined based on the aforementioned characteristic data.
[0012] The operation and maintenance plan will be retrieved based on the aforementioned operation and maintenance situation.
[0013] Optionally, the position of the scatter point in the first scatter data and the scatter data in the second scatter data are determined by time and value.
[0014] Optionally, fusing the first filtered scatter data and the second filtered scatter data includes:
[0015] Retain the scatter points with repeated times in the first and second scatter points data, and delete the scatter points with non-repeated times.
[0016] Optionally, the operation and maintenance plan is pre-configured.
[0017] This application also provides an intelligent operation and maintenance device based on big data and machine learning, including:
[0018] The acquisition module is used to acquire sample data, which is the data generated by the device during operation, including set parameters, running data, and data on changes in running data;
[0019] The editing module is used to divide the sample data into two groups and input them into the editing end of the convolutional neural network model. The editing end edits the data to generate first scattered data and second scattered data based on the Cartesian coordinate system, and inputs the first scattered data and second scattered data into the convolutional layer.
[0020] The filtering module is used to perform pixel recognition on the first scattered data and the second scattered data through the convolutional layer, and filter out the first scattered data and the second scattered data that meet the preset conditions to generate the first filtered scattered data and the second filtered scattered data.
[0021] The fusion module is used to input the first filtered scatter data and the second filtered scatter data into the fusion layer of the convolutional neural network, and to fuse the first filtered scatter data and the second filtered scatter data to generate fused scatter data.
[0022] The decoding module 305 is used to decode the fused scatter data and generate result feature data.
[0023] Optionally, an operation and maintenance plan is generated based on the feature data, including:
[0024] The operational status is determined based on the aforementioned characteristic data.
[0025] The operation and maintenance plan will be retrieved based on the aforementioned operation and maintenance situation.
[0026] Optionally, the position of the scatter point in the first scatter data and the scatter data in the second scatter data are determined by time and value.
[0027] Optionally, fusing the first filtered scatter data and the second filtered scatter data includes:
[0028] Retain the scatter points with repeated times in the first and second scatter points data, and delete the scatter points with non-repeated times.
[0029] Optionally, the operation and maintenance plan is pre-configured.
[0030] The advantages of the technical solution in this application compared to the prior art are:
[0031] This application provides an intelligent operation and maintenance method based on big data and machine learning, comprising: acquiring sample data, wherein the sample data is data generated by the device during operation, including set parameters, running data, and data on changes in running data; dividing the sample data into two groups and inputting them respectively into the editing end of a convolutional neural network model, wherein the editing end edits the data to generate first scattered data and second scattered data based on a Cartesian coordinate system, and inputting the first scattered data and second scattered data into a convolutional layer; the convolutional layer performs pixel recognition on the first scattered data and second scattered data respectively, filtering out the first scattered data and second scattered data that meet preset conditions, generating first filtered scattered data and second filtered scattered data; inputting the first filtered scattered data and second filtered scattered data into the fusion layer of the convolutional neural network, and fusing the first filtered scattered data and second filtered scattered data to generate fused scattered data; decoding the fused scattered data to generate result feature data. This application, through intelligent model processing of big data, can monitor the device status in real time and determine the operation and maintenance plan accordingly. Attached Figure Description
[0032] Figure 1 This application describes an intelligent operation and maintenance process based on big data and machine learning.
[0033] Figure 2 This is a flowchart of the data processing in the editing section of this application.
[0034] Figure 3 This is a schematic diagram of the intelligent operation and maintenance device based on big data and machine learning in this application. Detailed Implementation
[0035] The following are examples of specific implementation processes provided to illustrate the technical solutions to be protected in this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can implement this application by different technical means under the guidance of the concept of this application. Therefore, this application is not limited to the specific embodiments below.
[0036] This application provides an intelligent operation and maintenance method based on big data and machine learning, comprising: acquiring sample data, wherein the sample data is data generated by the device during operation, including set parameters, running data, and data on changes in running data; dividing the sample data into two groups and inputting them respectively into the editing end of a convolutional neural network model, wherein the editing end edits the data to generate first scattered data and second scattered data based on a Cartesian coordinate system, and inputting the first scattered data and second scattered data into a convolutional layer; the convolutional layer performs pixel recognition on the first scattered data and second scattered data respectively, filtering out the first scattered data and second scattered data that meet preset conditions, generating first filtered scattered data and second filtered scattered data; inputting the first filtered scattered data and second filtered scattered data into the fusion layer of the convolutional neural network, and fusing the first filtered scattered data and second filtered scattered data to generate fused scattered data; decoding the fused scattered data to generate result feature data. This application, through intelligent model processing of big data, can monitor the device status in real time and determine the operation and maintenance plan accordingly.
[0037] Figure 1 This application describes an intelligent operation and maintenance process based on big data and machine learning.
[0038] Please refer to Figure 1 As shown, S101 acquires sample data, which is the data generated by the device during operation, including set parameters, running data, and data on changes in running data.
[0039] The sample data is collected through sensors installed on the device. The type and specifications of these sensors vary depending on the device, and those skilled in the art can select the appropriate sensor based on the specific circumstances. Data is typically collected at a stable frequency and at preset time intervals. However, in some cases, this frequency can be adjusted. For example, in the operation of power grid equipment, since electricity consumption varies, the data collection frequency can be adjusted according to the amount of electricity consumed.
[0040] The types of data collected include parameters set for the device, data generated during device operation, and data showing changes in that data during device operation. In general, this data can be used to determine the device's operating status.
[0041] The data is collected by sensors, numbered according to the data collection nodes, and then sent to the data computing device for subsequent calculation and analysis.
[0042] In this application, after the data is sent to the computing device, it is summarized according to the data collection time to generate sample data. The sample data includes data collected multiple times, that is, data obtained from multiple rounds of collection within a continuous time range. During collection, if the data is the same as the previously collected data or falls within a preset range of variation, no further data recording is performed. The time period on which the data summarization is based refers to a preset time period.
[0043] On the other hand, the data will be assembled into a sample dataset, packaged and stored separately, which can be a local storage device or a cloud server.
[0044] Please refer to Figure 1 As shown, S102 divides the sample data into two groups and inputs them into the editing end of the convolutional neural network model. The editing end edits the data to generate first scattered data and second scattered data based on the Cartesian coordinate system, and inputs the first scattered data and second scattered data into the convolutional layer.
[0045] The two sets of data that have been segmented are two complete sets of data, including preset parameters, running data, and data showing changes in the running data.
[0046] The specific operation method is as follows: when the sensor collects data, data is collected according to two types of sensors. Then, the data collected by the two sensors at the same time is used as sample data and divided into two groups according to the sensor type as the initial data for data processing.
[0047] The initial data will be sent to the editing end of the convolutional neural network for preliminary processing, that is, the initial data will be edited and filled into a Cartesian coordinate system.
[0048] Figure 2 This is a flowchart of the data processing in the editing section of this application.
[0049] Please refer to Figure 2 As shown, S201 formats the initial data, converting data from different sensors into data of the same format.
[0050] In this stage, the data format is converted according to a pre-set sensor format conversion method. Simultaneously, the units of the data values are standardized.
[0051] Please continue to refer to Figure 2 As shown, S202 fills the initial data of the converted format into the Cartesian coordinate system.
[0052] The horizontal axis of the Cartesian coordinate system represents time, and the vertical axis represents the value of the initial data. In this application, after each type of data is entered into the Cartesian coordinate system, the Cartesian coordinate system is visualized and scaled according to a preset ratio to form a scatter image with a density interval, wherein the first set of data forms the first scatter data, and the second set of data forms the second scatter data.
[0053] Please refer to Figure 1 As shown, the convolutional layer in S103 performs pixel recognition on the first scattered data and the second scattered data respectively, and filters out the first scattered data and the second scattered data that meet the preset conditions to generate the first filtered scattered data and the second filtered scattered data.
[0054] After the images of the first scatter data and the second scatter data are formed, they are input into a convolutional layer for further processing. The convolutional layer has a first convolutional layer, a second convolutional layer, and a third convolutional layer.
[0055] The first, second, and third convolutional layers use different coefficients to perform the same filtering process on the data three times. The steps of this process are as follows:
[0056] Based on a first preset threshold, a second preset threshold, or a third preset threshold with different coefficients, pixel filtering is performed on the images of the first scatter data and the images of the second scatter data. The number of pixels within multiple preset ranges is calculated, and the preset ranges are sorted according to the number of pixels. Then, selection is performed sequentially according to a predetermined selection rule, such as selecting the top 10 preset ranges.
[0057] Then, the pixels within the selected preset range are retained, while the other pixels are deleted.
[0058] Please refer to Figure 1 As shown, in step S104, the first filtered scatter data and the second filtered scatter data are input into the fusion layer of the convolutional neural network, and the first filtered scatter data and the second filtered scatter data are fused to generate fused scatter data.
[0059] Specifically, the fusion refers to fusing the image of the first scatter data and the image of the second scatter data, and the specific steps are as follows:
[0060] Using the same coordinates, the image of the first scatter point and the image of the second scatter point are superimposed, and the pixels in the first and second scatter point images that are in the same preset area are averaged to obtain the image of the fused scatter point data.
[0061] Please refer to Figure 1As shown, S105 decodes the fused scatter data to generate result feature data.
[0062] The result feature data is standard equipment operating status data, and the operation method is as follows:
[0063] The fused scatter data is restored to the initial data, and the operating status of the device is determined based on this data.
[0064] In this application, the operating status of equipment can be predicted by processing big data.
[0065] This application also provides an intelligent operation and maintenance device based on big data and machine learning, including: an acquisition module 301, an editing module 302, a filtering module 303, a fusion module 304, and a decoding module 305.
[0066] Figure 3 This is a schematic diagram of the intelligent operation and maintenance device based on big data and machine learning in this application.
[0067] Please refer to Figure 3 As shown, the acquisition module 301 is used to acquire sample data, which is the data generated by the device when it is working, including the set parameters, the running data, and the data on changes in the running data.
[0068] The sample data is collected through sensors installed on the device. The type and specifications of these sensors vary depending on the device, and those skilled in the art can select the appropriate sensor based on the specific circumstances. Data is typically collected at a stable frequency and at preset time intervals. However, in some cases, this frequency can be adjusted. For example, in the operation of power grid equipment, since electricity consumption varies, the data collection frequency can be adjusted according to the amount of electricity consumed.
[0069] The types of data collected include parameters set for the device, data generated during device operation, and data showing changes in that data during device operation. In general, this data can be used to determine the device's operating status.
[0070] The data is collected by sensors, numbered according to the data collection nodes, and then sent to the data computing device for subsequent calculation and analysis.
[0071] In this application, after the data is sent to the computing device, it is summarized according to the data collection time to generate sample data. The sample data includes data collected multiple times, that is, data obtained from multiple rounds of collection within a continuous time range. During collection, if the data is the same as the previously collected data or falls within a preset range of variation, no further data recording is performed. The time period on which the data summarization is based refers to a preset time period.
[0072] On the other hand, the data will be assembled into a sample dataset, packaged and stored separately, which can be a local storage device or a cloud server.
[0073] Please refer to Figure 3 As shown, the editing module 302 is used to divide the sample data into two groups and input them into the editing end of the convolutional neural network model. The editing end edits the data to generate first scattered data and second scattered data based on the Cartesian coordinate system, and inputs the first scattered data and second scattered data into the convolutional layer.
[0074] The two sets of data that have been segmented are two complete sets of data, including preset parameters, running data, and data showing changes in the running data.
[0075] The specific operation method is as follows: when the sensor collects data, data is collected according to two types of sensors. Then, the data collected by the two sensors at the same time is used as sample data and divided into two groups according to the sensor type as the initial data for data processing.
[0076] The initial data will be sent to the editing end of the convolutional neural network for preliminary processing, that is, the initial data will be edited and filled into a Cartesian coordinate system.
[0077] Figure 2 This is a flowchart of the data processing in the editing section of this application.
[0078] Please refer to Figure 2 As shown, S201 formats the initial data, converting data from different sensors into data of the same format.
[0079] In this stage, the data format is converted according to a pre-set sensor format conversion method. Simultaneously, the units of the data values are standardized.
[0080] Please continue to refer to Figure 2 As shown, S202 fills the initial data of the converted format into the Cartesian coordinate system.
[0081] The horizontal axis of the Cartesian coordinate system represents time, and the vertical axis represents the value of the initial data. In this application, after each type of data is entered into the Cartesian coordinate system, the Cartesian coordinate system is visualized and scaled according to a preset ratio to form a scatter image with a density interval, wherein the first set of data forms the first scatter data, and the second set of data forms the second scatter data.
[0082] Please refer to Figure 3As shown, the filtering module 303 is used to perform pixel recognition on the first scattered data and the second scattered data through the convolutional layer, and filter out the first scattered data and the second scattered data that meet the preset conditions to generate the first filtered scattered data and the second filtered scattered data.
[0083] After the images of the first scatter data and the second scatter data are formed, they are input into a convolutional layer for further processing. The convolutional layer has a first convolutional layer, a second convolutional layer, and a third convolutional layer.
[0084] The first, second, and third convolutional layers use different coefficients to perform the same filtering process on the data three times. The steps of this process are as follows:
[0085] Based on a first preset threshold, a second preset threshold, or a third preset threshold with different coefficients, pixel filtering is performed on the images of the first scatter data and the images of the second scatter data. The number of pixels within multiple preset ranges is calculated, and the preset ranges are sorted according to the number of pixels. Then, selection is performed sequentially according to a predetermined selection rule, such as selecting the top 10 preset ranges.
[0086] Then, the pixels within the selected preset range are retained, while the other pixels are deleted.
[0087] Please refer to Figure 3 As shown, the fusion module 304 is used to input the first filtered scatter data and the second filtered scatter data into the fusion layer of the convolutional neural network, and to fuse the first filtered scatter data and the second filtered scatter data to generate fused scatter data.
[0088] Specifically, the fusion refers to fusing the image of the first scatter data and the image of the second scatter data, and the specific steps are as follows:
[0089] Using the same coordinates, the image of the first scatter point and the image of the second scatter point are superimposed, and the pixels in the first and second scatter point images that are in the same preset area are averaged to obtain the image of the fused scatter point data.
[0090] Please refer to Figure 3 As shown, the decoding module 305 is used to decode the fused scatter data and generate result feature data.
[0091] The result feature data is standard equipment operating status data, and the operation method is as follows:
[0092] The fused scatter data is restored to the initial data, and the operating status of the device is determined based on this data.
[0093] In this application, the operating status of equipment can be predicted by processing big data.
[0094] Although embodiments of the present invention have been shown and described above, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.
Claims
1. An intelligent operation and maintenance method based on big data and machine learning, characterized in that, include: Acquire sample data, which is the data generated by the device during operation, including set parameters, running data, and data on changes in running data; The sample data is divided into two groups and input into the editing end of the convolutional neural network model. The editing end edits the data to generate first scattered data and second scattered data based on the Cartesian coordinate system. The first scattered data and second scattered data are then input into the convolutional layer. The convolutional layer performs pixel recognition on the first scattered data and the second scattered data respectively, and filters out the first scattered data and the second scattered data that meet the preset conditions to generate the first filtered scattered data and the second filtered scattered data. The first filtered scatter data and the second filtered scatter data are input into the fusion layer of the convolutional neural network, and the first filtered scatter data and the second filtered scatter data are fused to generate fused scatter data. The fused scatter data is decoded to generate result feature data; The horizontal axis of the Cartesian coordinate system is time, and the vertical axis is the value of the initial data. After each type of data is filled into the Cartesian coordinate system, the Cartesian coordinate system is visualized and scaled according to a preset ratio to form a scatter image with a density interval. The first set of data forms the first scatter data, and the second set of data forms the second scatter data. After forming the images of the first scatter data and the second scatter data, they are input into a convolutional layer for further processing. The convolutional layer has a first convolutional layer, a second convolutional layer and a third convolutional layer. The first, second, and third convolutional layers use different coefficients to perform the same filtering process on the data three times. The steps of this process are as follows: Based on a first preset threshold, a second preset threshold, or a third preset threshold with different coefficients, pixel filtering is performed on the images of the first scatter data and the images of the second scatter data. The number of pixels within multiple preset ranges is calculated, and the preset ranges are sorted according to the number of pixels. Sequential selection is performed according to a predetermined selection rule. Then, the pixels in the selected preset ranges are retained, and other pixels are deleted. The fusion refers to fusing the image of the first scattered data and the image of the second scattered data. The specific steps are as follows: superimpose the image of the first scattered data and the image of the second scattered data according to the same coordinates, and average the pixels in the first scattered data and the second scattered data that are in the same preset area to obtain the image of the fused scattered data.
2. The intelligent operation and maintenance method based on big data and machine learning according to claim 1, characterized in that, Based on the aforementioned feature data, an operation and maintenance plan is generated, including: The operational status is determined based on the aforementioned characteristic data. The operation and maintenance plan will be retrieved based on the aforementioned operation and maintenance situation.
3. The intelligent operation and maintenance method based on big data and machine learning according to claim 1, characterized in that, The position of the midpoint of the first scatter data and the midpoint of the second scatter data are determined by time and value.
4. The intelligent operation and maintenance method based on big data and machine learning according to claim 1, characterized in that, The step of fusing the first filtered scatter data and the second filtered scatter data includes: Retain the scatter points with repeated times in the first and second scatter points data, and delete the scatter points with non-repeated times.
5. The intelligent operation and maintenance method based on big data and machine learning according to claim 2, characterized in that, The operation and maintenance plan is pre-configured.
6. An intelligent operation and maintenance device based on big data and machine learning, characterized in that, include: The acquisition module is used to acquire sample data, which is the data generated by the device during operation, including set parameters, running data, and data on changes in running data; The editing module is used to divide the sample data into two groups and input them into the editing end of the convolutional neural network model. The editing end edits the data to generate first scattered data and second scattered data based on the Cartesian coordinate system, and inputs the first scattered data and second scattered data into the convolutional layer. The filtering module is used to perform pixel recognition on the first scattered data and the second scattered data through the convolutional layer, and filter out the first scattered data and the second scattered data that meet the preset conditions to generate the first filtered scattered data and the second filtered scattered data. The fusion module is used to input the first filtered scatter data and the second filtered scatter data into the fusion layer of the convolutional neural network, and to fuse the first filtered scatter data and the second filtered scatter data to generate fused scatter data. The decoding module is used to decode the fused scatter data and generate result feature data; The horizontal axis of the Cartesian coordinate system is time, and the vertical axis is the value of the initial data. After each type of data is filled into the Cartesian coordinate system, the Cartesian coordinate system is visualized and scaled according to a preset ratio to form a scatter image with a density interval. The first set of data forms the first scatter data, and the second set of data forms the second scatter data. After forming the images of the first scatter data and the second scatter data, they are input into a convolutional layer for further processing. The convolutional layer has a first convolutional layer, a second convolutional layer and a third convolutional layer. The first, second, and third convolutional layers use different coefficients to perform the same filtering process on the data three times. The steps of this process are as follows: Based on a first preset threshold, a second preset threshold, or a third preset threshold with different coefficients, pixel filtering is performed on the images of the first scatter data and the images of the second scatter data. The number of pixels within multiple preset ranges is calculated, and the preset ranges are sorted according to the number of pixels. Sequential selection is performed according to a predetermined selection rule. Then, the pixels in the selected preset ranges are retained, and other pixels are deleted. The fusion refers to fusing the image of the first scattered data and the image of the second scattered data. The specific steps are as follows: superimpose the image of the first scattered data and the image of the second scattered data according to the same coordinates, and average the pixels in the first scattered data and the second scattered data that are in the same preset area to obtain the image of the fused scattered data.
7. The intelligent operation and maintenance device based on big data and machine learning according to claim 6, characterized in that, Based on the aforementioned feature data, an operation and maintenance plan is generated, including: The operational status is determined based on the aforementioned characteristic data. The operation and maintenance plan will be retrieved based on the aforementioned operation and maintenance situation.
8. The intelligent operation and maintenance device based on big data and machine learning according to claim 6, characterized in that, The position of the midpoint of the first scatter data and the midpoint of the second scatter data are determined by time and value.
9. The intelligent operation and maintenance device based on big data and machine learning according to claim 6, characterized in that, The step of fusing the first filtered scatter data and the second filtered scatter data includes: Retain the scatter points with repeated times in the first and second scatter points data, and delete the scatter points with non-repeated times.
10. The intelligent operation and maintenance device based on big data and machine learning according to claim 7, characterized in that, The operation and maintenance plan is pre-configured.