CN113139518B - Section bar cutting state monitoring method based on industrial internet - Google Patents
Section bar cutting state monitoring method based on industrial internet Download PDFInfo
- Publication number
- CN113139518B CN113139518B CN202110529552.1A CN202110529552A CN113139518B CN 113139518 B CN113139518 B CN 113139518B CN 202110529552 A CN202110529552 A CN 202110529552A CN 113139518 B CN113139518 B CN 113139518B
- Authority
- CN
- China
- Prior art keywords
- classification
- feature map
- softmax
- differential
- feature
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a profile cutting state monitoring method based on an industrial internet, which uses a computer vision method based on deep learning to perform feature extraction and classification on vibration of a profile in a cutting process so as to perform intelligent monitoring on vibration exceeding a preset limit in the profile cutting process. Like this, prevent that the lathe from influencing the cutting quality because of excessive vibration at the in-process to the section bar cutting, damage the driving motor of cutting knife even.
Description
Technical Field
The present invention relates to the field of industrial internet, and more particularly, to a profile cutting state monitoring method based on industrial internet, a profile cutting state monitoring system based on industrial internet, and an electronic device.
Background
As a product of the deep integration of new generation information technology and manufacturing industry, the industrial internet is changing the innovation mode, production mode, management mode and service mode of the traditional industry, promoting the generation of many new technologies, new modes, new models and new industries, and increasingly becoming an important support for the revolution of new industry and an important foundation for the "internet + advanced manufacturing industry". Through the connection of intelligent machines, the industrial internet finally realizes man-machine interconnection, combines software and big data analysis, remodels global industry, arouses productivity, makes the world become better, faster, safer, cleaner, more economic.
In many industrial plants, it is necessary to cut profiles, for example during the machining of profiles using machine tools. In general, the cutting blade of a machine tool directly acts on the profile to cut, and the profile inevitably generates vibration during cutting, which not only affects the cutting quality, but also may damage the driving motor of the cutting blade. If can carry out effectual control and judgement through industry internet technique to the shake of equipment cutting in-process to in time make the adjustment, will greatly improve the quality of section bar, reduce the loss of equipment.
In this way, the computer vision method based on deep learning can extract and classify the vibration of the profile in the cutting process, so as to convert the monitoring of the vibration exceeding the predetermined limit in the profile cutting process into the classification problem based on the high-dimensional image features extracted by the deep neural network model, so how to monitor the specific jitter condition of the cutting equipment based on the deep neural network through the industrial internet becomes the key for solving the problem.
It is therefore desirable to provide a solution for monitoring vibrations exceeding a predetermined limit during the profile cutting process.
Disclosure of Invention
The present application is proposed to solve the above-mentioned technical problems. Embodiments of the present application provide a profile cutting state monitoring method, system and electronic device based on the industrial internet, which use a computer vision method based on deep learning to perform feature extraction and classification on the vibration of a profile in the cutting process, so as to perform intelligent monitoring on the vibration exceeding a predetermined limit in the profile cutting process. Like this, prevent that the lathe from influencing the cutting quality because of excessive vibration at the in-process to the section bar cutting, damage the driving motor of cutting knife even.
According to an aspect of the present application, there is provided an industrial internet-based section bar cutting state monitoring method, which includes the steps of:
obtaining a plurality of frame images of a profile in time sequence in a cutting process by a camera disposed at an end side of an industrial device through an industrial internet, the industrial device being communicably connected to the industrial internet;
respectively passing the plurality of frame images through a convolutional neural network to obtain a plurality of feature maps in a time sequence;
calculating the difference between every two adjacent feature maps in the plurality of feature maps to obtain a plurality of difference feature maps;
Calculating a weighting coefficient for each of the plurality of difference profiles to obtain a plurality of weighting coefficients, comprising: calculating a classification probability value of Softmax-like of each differential feature map, wherein the classification probability value of Softmax-like is the sum of power functions with natural constants of the feature values of the ith position in the differential feature map divided by the sum of power functions with natural constants of the feature values of the 1 st to ith positions in the differential feature map; and calculating the information entropy of the class Softmax-like classification function value to obtain a weighting coefficient corresponding to the difference feature map, wherein the information entropy of the class Softmax-like classification function value is the negative value of the logarithm value of the class Softmax-like classification function value plus a preset compensation item;
calculating a weighted sum of the plurality of differential feature maps with the plurality of weighting coefficients to obtain a classification feature map; and
and passing the classification characteristic diagram through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the vibration in the profile cutting process exceeds a preset limit or not.
In the method for monitoring the cutting state of the profile based on the industrial internet, the calculating a classification probability value of the class Softmax of each differential feature map includes: calculating a classification probability value of class Softmax of each differential feature map according to the following formula: pi ═ Σ Fi exp(xi)/∑ Fd～Fi exp (xi), wherein the numerator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the ith differential feature map, and the denominator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the 1 st to ith differential feature maps.
In the above method for monitoring a cutting state of a profile based on the industrial internet, the calculating an information entropy of the classification function value of the class Softmax to obtain a weighting coefficient corresponding to the differential feature map includes: calculating information entropy of the Softmax-like classification function value to obtain a weighting coefficient corresponding to the differential feature map, wherein gi ═ log (pi) + α, pi denotes the Softmax-like classification function value of the ith differential feature map, and a denotes a compensation term.
In the above method for monitoring the cutting state of the profile based on the industrial internet, the compensation term causes the weighting coefficient of the differential characteristic diagram not to be zero.
In the above method for monitoring a cutting status of a profile based on the industrial internet, the step of passing the classification feature map through a classifier to obtain a classification result includes: passing the classified feature map through one or more fully-connected layers to encode the classified feature map through the one or more fully-connected layers to obtain a classified feature vector; and inputting the classification feature vector into a Softmax classification function to obtain the classification result.
In the section bar cutting state monitoring method based on the industrial internet, the convolutional neural network is a deep residual error network.
According to another aspect of the present application, there is provided an industrial internet-based profile cutting status monitoring system, including:
an image acquisition unit for obtaining a plurality of frame images in time series of a profile in a cutting process, the profile being acquired by a camera disposed on an end side of an industrial apparatus, the industrial apparatus being communicably connected to an industrial internet, through the industrial internet;
a feature map generation unit configured to pass the plurality of frame images obtained by the image acquisition unit through a convolutional neural network, respectively, to obtain a plurality of feature maps in a time sequence;
a difference feature map generation unit configured to calculate a difference between every two adjacent feature maps in the plurality of feature maps obtained by the feature map generation unit to obtain a plurality of difference feature maps;
a weighting coefficient generation unit configured to calculate a weighting coefficient for each of the plurality of differential feature maps obtained by the differential feature map generation unit to obtain a plurality of weighting coefficients, including:
a classification probability value operator unit, configured to calculate a classification probability value of a Softmax-like of each differential feature map generation unit, where the classification probability value of the Softmax-like is a sum of power functions with natural constants as bases of feature values of respective positions in an ith differential feature map divided by a sum of power functions with natural constants as bases of feature values of respective positions in the 1 st to ith differential feature maps; and
An information entropy calculation subunit operable to calculate an information entropy of the Softmax-like classification function value to obtain a weighting coefficient corresponding to the differential feature map obtained by the differential feature map generation unit, the information entropy of the Softmax-like classification function value obtained by the classification probability value calculation subunit being a negative value of a logarithm value of the Softmax-like classification function value plus a predetermined compensation term;
a classification feature map generation unit configured to calculate a weighted sum of the plurality of difference feature maps with the plurality of weighting coefficients obtained by the weighting coefficient generation unit to obtain a classification feature map; and
and the classification result generating unit is used for enabling the classification characteristic diagram obtained by the classification characteristic diagram generating unit to pass through a classifier so as to obtain a classification result, wherein the classification result is used for indicating whether the vibration in the section bar cutting process exceeds a preset limit or not.
According to still another aspect of the present application, there is provided an electronic apparatus including: a processor; and a memory in which computer program instructions are stored, which, when executed by the processor, cause the processor to perform the industrial internet-based profile cutting status monitoring method as described above.
Compared with the prior art, the embodiment of the application provides the profile cutting state monitoring method, the profile cutting state monitoring system and the electronic equipment based on the industrial internet, which use a computer vision method based on deep learning to perform feature extraction and classification on the vibration of the profile in the cutting process so as to perform intelligent monitoring on the vibration exceeding the preset limit in the cutting process of the profile. Like this, prevent that the lathe from influencing the cutting quality because of excessive vibration at the in-process to the section bar cutting, damage the driving motor of cutting knife even.
Drawings
The above and other objects, features and advantages of the present application will become more apparent by describing in more detail embodiments of the present application with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings, like reference numbers generally indicate like parts or steps.
Fig. 1 is an application scene diagram of a profile cutting state monitoring method based on the industrial internet according to an embodiment of the application.
Fig. 2 is a flowchart of the method for monitoring the cutting state of the profile based on the industrial internet according to the embodiment of the application.
Fig. 3 is a schematic configuration diagram of the industrial internet-based profile cutting status monitoring method according to an embodiment of the present application.
Fig. 4 is a flowchart illustrating that, in the method for monitoring a cutting status of a profile based on the industrial internet according to an embodiment of the present application, a weighting coefficient of each of the plurality of differential feature maps is calculated to obtain a plurality of weighting coefficients.
Fig. 5 is a flowchart of obtaining classification results by passing the classification feature map through a classifier in the method for monitoring the cutting status of the profile based on the industrial internet according to the embodiment of the application.
Fig. 6 is a block diagram of an industrial internet-based profile cutting status monitoring system according to an embodiment of the present application.
Fig. 7 is a block diagram of a weighting coefficient generation unit in the industrial internet-based profile cutting status monitoring system according to an embodiment of the present application.
Fig. 8 is a block diagram of a classification result generation unit in the industrial internet-based profile cutting status monitoring system according to an embodiment of the present application.
Fig. 9 is a block diagram of an electronic device according to an embodiment of the application.
Detailed Description
Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be understood that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and that the present application is not limited by the example embodiments described herein.
Overview of a scene
As mentioned above, in many industrial plants, it is necessary to cut the section bar, for example during the machining of the section bar using a machine tool. In general, the profile cutting knife of a machine tool directly acts on the profile to cut, and the profile inevitably generates vibration during cutting, which not only affects the cutting quality, but also may damage the driving motor of the cutting knife, therefore, it is desirable to provide a scheme for monitoring the vibration exceeding a predetermined limit during the profile cutting process.
The applicant of the present application uses a computer vision method based on deep learning to perform feature extraction and classification on vibrations of a profile during a cutting process to convert monitoring of vibrations exceeding a predetermined limit during the cutting process of the profile into a classification problem based on high-dimensional image features extracted by a deep neural network model.
Specifically, in the technical scheme of the application, a series of frame images in time sequence of the section bar obtained by a camera at the end side of the industrial equipment in the cutting process are obtained through the industrial internet, the frame images are respectively passed through a convolutional neural network to obtain a plurality of feature maps in time sequence, and every two adjacent feature maps in the feature maps are subjected to differential operation to obtain a differential feature map. Thus, the obtained differential feature map essentially extracts the high-dimensional features of the image variations of the profile during the cutting process.
However, the contributions of these high-dimensional features in image classification are different, and it is obvious from the physical point of view that the higher-dimensional features closer in time contribute more in classification, and therefore, appropriate weights should be set in chronological order for the obtained plurality of differential feature maps. In consideration of the fact that most classifiers are classified by using a Softmax classification function, in the technical scheme of the application, the weights of the differential feature maps are calculated by combining a classification probability form of class Softmax and an idea that small probability events like information entropy contain larger information.
In particular, for a plurality of differential feature maps, for example, denoted as differential feature maps Fd, …, Fm, for each differential feature map a classification probability value of the Softmax class is first calculated, i.e. pi ═ Σ ∑ Σ Fi exp(xi)/∑ Fd～Fi exp (xi), where the numerator is the sum of power functions with a natural constant as the bottom of the feature value of each position in the i-th differential feature map, and the denominator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the 1 st to i-th differential feature maps. Thus, it can be seen that the value of pi decreases from 1 to m, and p1 equals 1, whereas p2 to pm are less than one. Therefore, the information entropy of pi is further evaluated as a weighting factor, i.e., gi ═ log (pi) + α, where α is used as a compensation term in order to make the weighting factor of the differential feature map Fd nonzero.
Then, the calculated weighting coefficients are used for weighting and summing the differential characteristic graphs, so that a classification characteristic graph can be obtained, and the classification characteristic graph is passed through a classifier to obtain a classification result, wherein the classification result indicates whether the vibration in the section bar cutting process exceeds a preset limit or not.
Based on this, the application provides a section bar cutting state monitoring method based on industrial internet, which includes: obtaining a plurality of frame images of a profile in time sequence in a cutting process by a camera disposed at an end side of an industrial device through an industrial internet, the industrial device being communicably connected to the industrial internet; respectively passing the plurality of frame images through a convolutional neural network to obtain a plurality of feature maps in a time sequence; calculating the difference between every two adjacent feature maps in the plurality of feature maps to obtain a plurality of difference feature maps; calculating a weighting coefficient for each of the plurality of difference profiles to obtain a plurality of weighting coefficients, comprising: calculating a classification probability value of Softmax-like of each differential feature map, wherein the classification probability value of Softmax-like is the sum of power functions with natural constants of the feature values of the ith position in the differential feature map divided by the sum of power functions with natural constants of the feature values of the 1 st to ith positions in the differential feature map; and calculating the information entropy of the class Softmax-like classification function value to obtain a weighting coefficient corresponding to the difference feature map, wherein the information entropy of the class Softmax-like classification function value is the negative value of the logarithm value of the class Softmax-like classification function value plus a preset compensation item; calculating a weighted sum of the plurality of differential feature maps with the plurality of weighting coefficients to obtain a classification feature map; and passing the classification characteristic diagram through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the vibration in the profile cutting process exceeds a preset limit or not.
Fig. 1 is an application scenario diagram of a profile cutting state monitoring method based on the industrial internet according to an embodiment of the application. As shown in fig. 1, in this application scenario, first, an image of the profile during cutting on the machine tool is acquired by a camera (e.g., C as illustrated in fig. 1); then, the obtained image is input into a server (for example, S as illustrated in fig. 1) deployed with an industrial internet-based profile cutting state monitoring algorithm, wherein the server can process the image of the profile during the cutting process based on the industrial internet-based profile cutting state monitoring algorithm to obtain the classification result indicating whether the vibration during the profile cutting process exceeds a predetermined limit.
Having described the general principles of the present application, various non-limiting embodiments of the present application will now be described with reference to the accompanying drawings.
Exemplary method
Fig. 2 is a flowchart of the method for monitoring the cutting state of the profile based on the industrial internet according to the embodiment of the application. As shown in fig. 2, the method for monitoring the cutting state of the profile based on the industrial internet according to the embodiment of the present application includes the steps of: s110, obtaining a plurality of frame images of the sectional material in the cutting process, wherein the frame images are obtained by a camera arranged at the end side of an industrial device in a time sequence, and the industrial device is connected with the industrial internet in a communication mode; s120, respectively passing the plurality of frame images through a convolutional neural network to obtain a plurality of feature maps in a time sequence; s130, calculating the difference between every two adjacent feature maps in the plurality of feature maps to obtain a plurality of difference feature maps; s140, calculating a weighting coefficient of each of the plurality of differential feature maps to obtain a plurality of weighting coefficients, including: calculating a classification probability value of Softmax-like of each differential feature map, wherein the classification probability value of Softmax-like is the sum of power functions with natural constants of the feature values of the ith position in the differential feature map divided by the sum of power functions with natural constants of the feature values of the 1 st to ith positions in the differential feature map; and calculating the information entropy of the class Softmax-like classification function value to obtain a weighting coefficient corresponding to the difference feature map, wherein the information entropy of the class Softmax-like classification function value is the negative value of the logarithm value of the class Softmax-like classification function value plus a preset compensation item; s150, calculating the weighted sum of the differential feature maps by the weighting coefficients to obtain a classification feature map, and S160, passing the classification feature map through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the vibration in the profile cutting process exceeds a preset limit or not.
Fig. 3 illustrates an architecture diagram of the industrial internet-based profile cutting status monitoring method according to an embodiment of the present application. As shown IN fig. 3, IN this network architecture, first, a plurality of frame images (for example, IN as illustrated IN fig. 3) IN time series during a cutting process of a profile acquired by a camera disposed at an end side of an industrial apparatus are obtained through an industrial internet; then, passing the plurality of frame images through a convolutional neural network (e.g., CNN as illustrated in fig. 3) respectively to obtain a plurality of feature maps (e.g., Ft1 to Ftn as illustrated in fig. 3) in time sequence; calculating a difference between each two adjacent feature maps of the plurality of feature maps to obtain a plurality of difference feature maps (e.g., Fd 1-Fdn as illustrated in FIG. 3); calculating a weighting coefficient of each of the plurality of differential feature maps to obtain a plurality of weighting coefficients (e.g., C1-Cn as illustrated in fig. 3), wherein the calculating of the weighting coefficients comprises: firstly, calculating a classification probability value of class Softmax of each differential feature map; and aggregating the information entropy of the classification function value of the class Softmax to obtain a weighting coefficient corresponding to the differential feature map; calculating a weighted sum of the plurality of differential feature maps with the plurality of weighting coefficients to obtain a classification feature map (e.g., F as illustrated in fig. 3); finally, the classification feature map is passed through a classifier to obtain a classification result (e.g., K as illustrated in fig. 3).
In step S110, a plurality of frame images in chronological order of the profile obtained by a camera disposed on an end side of an industrial apparatus in the middle of a cutting process are obtained through an industrial internet to which the industrial apparatus is communicably connected. That is, a time-series frame image of the profile acquired by the camera on the end side of the industrial equipment during the cutting process is first obtained through the industrial internet. In a specific implementation, a monitoring video of the profile in the cutting process can be collected by the camera, and a plurality of frame images are extracted from the monitoring video at preset time intervals.
In step S120, the plurality of frame images are respectively passed through a convolutional neural network to obtain a plurality of feature maps in time order. That is, the frame images are respectively passed through a convolutional neural network to obtain a plurality of feature maps in time sequence, and the feature maps comprise high-dimensional implicit features of the section bar cutting at each time point.
Those skilled in the art will appreciate that the deep convolutional neural network has excellent performance in extracting local spatial features of an image. In one particular example of the present application, the deep convolutional neural network is implemented as a deep residual network, e.g., ResNet 150. It should be known to those skilled in the art that, compared to the conventional convolutional neural network, the deep residual network is an optimized network structure proposed on the basis of the conventional convolutional neural network, which mainly solves the problem of gradient disappearance during the training process. The depth residual error network introduces a residual error network structure, the network layer can be made deeper through the residual error network structure, and the problem of gradient disappearance can not occur. The residual error network uses the cross-layer link thought of a high-speed network for reference, breaks through the convention that the traditional neural network only can provide N layers as input from the input layer of the N-1 layer, enables the output of a certain layer to directly cross several layers as the input of the later layer, and has the significance of providing a new direction for the difficult problem that the error rate of the whole learning model is not reduced and inversely increased by superposing multiple layers of networks.
In step S130, a difference between every two adjacent feature maps in the plurality of feature maps is calculated to obtain a plurality of difference feature maps. That is, every two adjacent feature maps in the plurality of feature maps are subjected to a difference operation to obtain a difference feature map. Here, the obtained differential feature map essentially extracts the high-dimensional features of the image variations of the profile during the cutting process. However, the contributions of these high-dimensional features in image classification are different, and it is obvious from the physical point of view that the higher-dimensional features closer in time contribute more in classification, and therefore, appropriate weights should be set in chronological order for the obtained plurality of differential feature maps.
In addition, since most classifiers are classified by using a Softmax classification function, in the technical solution of the present application, the weights of the respective differential feature maps are calculated by combining a classification probability form of class Softmax and an idea that information included in a small probability event like information entropy is larger.
In step S140, a weighting coefficient of each of the plurality of differential feature maps is calculated to obtain a plurality of weighting coefficients. In an embodiment of the present application, a process of calculating a weighting coefficient of each of the plurality of differential feature maps to obtain a plurality of weighting coefficients includes: firstly, calculating Softma-like of each differential feature map x classification probability value. Specifically, the classification probability value of the class Softmax of each differential feature map is calculated according to the following formula: pi ═ Σ Fi exp(xi)/∑ F1～Fi exp (xi), wherein the numerator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the ith differential feature map, and the denominator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the 1 st to ith differential feature maps.
Then, the information entropy of the Softmax-like classification function value is calculated to obtain a weighting coefficient corresponding to the differential feature map. Specifically, in the application embodiment, the information entropy of the Softmax-like classification function value is calculated by the following formula to obtain the weighting coefficient corresponding to the differential feature map, where gi is-log (pi) + α, pi represents the Softmax-like classification function value of the ith differential feature map, and a represents the compensation term. It can be seen that the value of pi decreases from 1 to m, and p1 equals 1, whereas p2 to pm are less than one. Therefore, the information entropy of pi is further evaluated as a weighting coefficient, i.e., gi ═ log (pi) + α, where α is used as a compensation term so that the weighting coefficient of the differential feature map F1 is not zero.
Fig. 4 illustrates a flowchart of calculating a weighting coefficient of each of the plurality of differential feature maps to obtain a plurality of weighting coefficients in the industrial internet-based profile cutting state monitoring method according to an embodiment of the present application. As shown in fig. 4, calculating a weighting factor of each of the plurality of difference feature maps to obtain a plurality of weighting factors includes: s141, calculating a classification probability value of Softmax-like of each differential feature map, wherein the classification probability value of Softmax-like is a sum of power functions with natural constants as bases of feature values of the ith differential feature map and a sum of power functions with natural constants as bases of feature values of the 1 st to ith differential feature maps; and S142, calculating the information entropy of the class function value of the Softmax-like to obtain the weighting coefficient corresponding to the difference feature map, wherein the information entropy of the class function value of the Softmax-like is the negative value of the logarithm value of the class function value of the Softmax-like added with a preset compensation item.
In step S150, a weighted sum of the differential feature maps is calculated with the weighting coefficients to obtain a classification feature map. That is, a position-weighted sum of the plurality of differential feature maps is calculated with the plurality of weighting coefficients to obtain the classification feature map.
In step S160, the classification feature map is passed through a classifier to obtain a classification result, wherein the classification result is used to indicate whether the vibration in the profile cutting process exceeds a predetermined limit. That is, the classification feature map is passed through a classifier to obtain a classification result indicating whether the vibration during the profile cutting process exceeds a predetermined limit.
In an embodiment of the present application, a process of passing the classification feature map through a classifier to obtain a classification result includes: s151, passing the classification feature map through one or more fully-connected layers to encode the classification feature map through the one or more fully-connected layers to obtain a classification feature vector; and S152, inputting the classification feature vector into a Softmax classification function to obtain the classification result, as shown in FIG. 5.
In summary, a profile cutting state monitoring method based on the industrial internet is clarified based on the embodiments of the present application, which extracts and classifies the features of the profile vibration during the cutting process based on the deep learning computer vision method to convert the monitoring of the vibration exceeding the predetermined limit during the profile cutting process into the classification problem based on the high-dimensional image features extracted by the deep neural network model.
Exemplary System
Fig. 6 illustrates a block diagram of an industrial internet-based profile cutting status monitoring system according to an embodiment of the present application. As shown in fig. 6, the industrial internet-based section bar cutting state monitoring system 600 according to the embodiment of the present application includes: an image acquisition unit 610 for obtaining a plurality of frame images in time series of the profile in the middle of the cutting process, acquired by a camera disposed at an end side of an industrial apparatus communicably connected to the industrial internet, through the industrial internet; a feature map generation unit 620, configured to pass the plurality of frame images obtained by the image acquisition unit 610 through a convolutional neural network respectively to obtain a plurality of feature maps in a time sequence; a difference feature map generating unit 630, configured to calculate a difference between every two adjacent feature maps in the plurality of feature maps obtained by the feature map generating unit 620 to obtain a plurality of difference feature maps; the weighting coefficient generating unit 640 is configured to calculate a weighting coefficient of each of the plurality of differential feature maps obtained by the differential feature map generating unit 630 to obtain a plurality of weighting coefficients, and includes: a classification probability value calculation operator unit 641 configured to calculate a classification probability value of a Softmax class of each of the differential feature maps generated by the differential feature map generation unit, where the classification probability value of the Softmax class is a sum of power functions with a natural constant as a base of feature values of respective positions in the ith differential feature map divided by a sum of power functions with a natural constant as a base of feature values of respective positions in the 1 st to ith differential feature maps; and a weighting coefficient composition subunit 642 for calculating an information entropy of the Softmax-like classification function value to obtain a weighting coefficient corresponding to the differential feature map obtained by the differential feature map generating unit 630, the information entropy of the Softmax-like classification function value obtained by the classification probability value calculating subunit being a negative value of a logarithm value of the Softmax-like classification function value plus a predetermined compensation term; a classification feature map generation unit 660 configured to calculate a weighted sum of the plurality of differential feature maps with the plurality of weighting coefficients obtained by the weighting coefficient generation unit 640 to obtain a classification feature map; and a classification result generating unit 660 for passing the classification feature map obtained by the classification feature map generating unit 650 through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the vibration in the profile cutting process exceeds a predetermined limit or not
In one example, in the industrial internet-based profile cutting status monitoring system 600, as shown in fig. 7, the weighting factor generating unit 640 includes: a classification probability value calculation operator unit 641 configured to calculate a classification probability value of a Softmax class of each of the differential feature maps generated by the differential feature map generation unit, where the classification probability value of the Softmax class is a sum of power functions with a natural constant as a base of feature values of respective positions in the ith differential feature map divided by a sum of power functions with a natural constant as a base of feature values of respective positions in the 1 st to ith differential feature maps; and an information entropy calculating subunit 642 configured to calculate an information entropy of the Softmax-like classification function value to obtain a weighting coefficient corresponding to the differential feature map obtained by the differential feature map generating unit, wherein the information entropy of the Softmax-like classification function value obtained by the classification probability value calculating subunit is a negative value of a logarithm value of the Softmax-like classification function value plus a predetermined compensation term.
In one example, in the industrial internet-based profile cutting status monitoring system 600, the classification probability value sub-unit 641 is further configured to: calculating a classification probability value of class Softmax of each differential feature map according to the following formula: pi ═ Σ Fi exp(xi)/∑ Fd～Fi exp (xi), wherein the numerator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the ith differential feature map, and the denominator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the 1 st to ith differential feature maps.
In one example, in the industrial internet-based profile cutting status monitoring system 600, the information entropy calculating subunit 642 is further configured to: calculating information entropy of the Softmax-like classification function value to obtain a weighting coefficient corresponding to the differential feature map, wherein gi ═ log (pi) + α, pi denotes the Softmax-like classification function value of the ith differential feature map, and a denotes a compensation term.
In one example, in the industrial internet-based profile cutting status monitoring system 600, the compensation term causes the weighting coefficient of the differential feature map to be non-zero.
In one example, in the industrial internet-based profile cutting status monitoring system 600, as shown in fig. 8, the classification result generating unit 660 includes: a classified feature vector generating sub-unit 661, configured to pass the classified feature map through one or more fully-connected layers to encode the classified feature map through the one or more fully-connected layers to obtain a classified feature vector; and a classification subunit 662, configured to input the classification feature vector into a Softmax classification function to obtain the classification result.
In one example, in the industrial internet-based profile cutting status monitoring system 600, the convolutional neural network is a deep residual network.
Here, it can be understood by those skilled in the art that the detailed functions and operations of the respective units and modules in the above-described industrial internet-based profile cutting status monitoring system 600 have been described in detail in the above description of the industrial internet-based profile cutting status monitoring method with reference to fig. 1 to 5, and thus, a repetitive description thereof will be omitted.
As described above, the industrial internet-based profile cutting status monitoring system 600 according to the embodiment of the present application may be implemented in various terminal devices, for example, a server applied to an industrial internet-based profile cutting status monitoring algorithm, and the like. In one example, the industrial internet-based profile cutting status monitoring system 600 according to the embodiment of the present application may be integrated into a terminal device as one software module and/or hardware module. For example, the industrial internet-based profile cutting status monitoring system 600 may be a software module in the operating system of the terminal device, or may be an application developed for the terminal device; of course, the industrial internet-based profile cutting status monitoring system 600 can also be one of the hardware modules of the terminal device.
Alternatively, in another example, the industrial internet-based profile cutting status monitoring system 600 and the terminal device may also be separate devices, and the industrial internet-based profile cutting status monitoring system 600 may be connected to the terminal device through a wired and/or wireless network and transmit the interactive information according to an agreed data format.
Exemplary electronic device
Next, an electronic apparatus according to an embodiment of the present application is described with reference to fig. 9.
As shown in fig. 9, the electronic device 10 includes at least one processor 11 and at least one memory 12.
The processor 11 may be a Central Processing Unit (CPU) or other form of processing unit having data processing capabilities and/or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.
The memory 12 may include at least one computer program product that may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. The volatile memory may include, for example, Random Access Memory (RAM), cache memory (cache), and/or the like. The non-volatile memory may include, for example, Read Only Memory (ROM), hard disk, flash memory, etc. At least one computer program instruction may be stored on the computer readable storage medium and executed by the processor 11 to implement the industrial internet-based profile cutting status monitoring method of the various embodiments of the present application described above and/or other desired functions. Various contents such as a weighting coefficient, a weighted sum, etc. may also be stored in the computer-readable storage medium.
In one example, the electronic device 10 may further include: an input device 13 and an output device 14, which are interconnected by a bus system and/or other form of connection mechanism (not shown).
The input device 13 may include, for example, a keyboard, a mouse, and the like.
The output device 14 can output various information including the classification result to the outside. The output devices 14 may include, for example, a display, speakers, a printer, and a communication network and its connected remote output devices, among others.
Of course, for simplicity, only some of the components of the electronic device 10 relevant to the present application are shown in fig. 9, and components such as buses, input/output interfaces, and the like are omitted. In addition, the electronic device 10 may include any other suitable components depending on the particular application.
Claims (10)
1. A section bar cutting state monitoring method based on industrial Internet is characterized by comprising the following steps:
obtaining a plurality of frame images of a profile in time sequence in a cutting process by a camera disposed at an end side of an industrial device through an industrial internet, the industrial device being communicably connected to the industrial internet;
respectively passing the plurality of frame images through a convolutional neural network to obtain a plurality of feature maps in a time sequence;
Calculating the difference between every two adjacent feature maps in the plurality of feature maps to obtain a plurality of difference feature maps;
calculating a weighting coefficient for each of the plurality of difference profiles to obtain a plurality of weighting coefficients, comprising:
calculating a classification probability value of Softmax-like of each differential feature map, wherein the classification probability value of Softmax-like is the sum of power functions with natural constants of the feature values of the ith position in the differential feature map divided by the sum of power functions with natural constants of the feature values of the 1 st to ith positions in the differential feature map; and
calculating the information entropy of the class function value of Softmax to obtain a weighting coefficient corresponding to the difference feature map, wherein the information entropy of the class function value of Softmax is the negative value of the logarithm value of the class function value of Softmax plus a preset compensation item;
calculating a weighted sum of the plurality of differential feature maps with the plurality of weighting coefficients to obtain a classification feature map; and
and passing the classification characteristic diagram through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the vibration in the profile cutting process exceeds a preset limit or not.
2. The industrial internet-based profile cutting status monitoring method according to claim 1, wherein calculating a classification probability value of class Softmax of each differential feature map comprises:
calculating a classification probability value of class Softmax of each differential feature map according to the following formula: pi ═ Σ Fi exp(xi)/∑ Fd～Fi exp (xi), wherein the numerator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the ith differential feature map, and the denominator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the 1 st to ith differential feature maps.
3. The industrial internet-based profile cutting status monitoring method according to claim 2, wherein calculating the information entropy of the Softmax-like classification function value to obtain the weighting coefficient corresponding to the differential feature map comprises:
calculating information entropy of the Softmax-like classification function value to obtain a weighting coefficient corresponding to the differential feature map, wherein gi ═ log (pi) + α, pi denotes the Softmax-like classification function value of the ith differential feature map, and a denotes a compensation term.
4. The industrial internet-based profile cutting status monitoring method according to claim 3, wherein the compensation term makes the weighting coefficient of the differential feature map non-zero.
5. The industrial internet-based section bar cutting state monitoring method according to claim 1, wherein the passing the classification feature map through a classifier to obtain a classification result comprises:
passing the classified feature map through one or more fully-connected layers to encode the classified feature map through the one or more fully-connected layers to obtain a classified feature vector; and
inputting the classification feature vector into a Softmax classification function to obtain the classification result.
6. The industrial internet-based profile cutting status monitoring method according to claim 1, wherein the convolutional neural network is a deep residual network.
7. The utility model provides a section bar cutting state monitoring system based on industry internet which characterized in that includes:
an image acquisition unit for obtaining a plurality of frame images in time series of a profile in a cutting process, the profile being acquired by a camera disposed on an end side of an industrial apparatus, the industrial apparatus being communicably connected to an industrial internet, through the industrial internet;
a feature map generation unit configured to pass the plurality of frame images obtained by the image acquisition unit through a convolutional neural network, respectively, to obtain a plurality of feature maps in a time sequence;
A difference feature map generation unit configured to calculate a difference between every two adjacent feature maps in the plurality of feature maps obtained by the feature map generation unit to obtain a plurality of difference feature maps;
a weighting coefficient generation unit configured to calculate a weighting coefficient for each of the plurality of differential feature maps obtained by the differential feature map generation unit to obtain a plurality of weighting coefficients, including:
a classification probability value operator unit, configured to calculate a classification probability value of a Softmax-like of each of the differential feature maps generated by the differential feature map generation unit, where the classification probability value of the Softmax-like is a sum of power functions with a natural constant as a base of feature values of respective positions in the ith differential feature map divided by a sum of power functions with a natural constant as a base of feature values of respective positions in the 1 st to ith differential feature maps; and
an information entropy calculation subunit operable to calculate an information entropy of the Softmax-like classification function value to obtain a weighting coefficient corresponding to the differential feature map obtained by the differential feature map generation unit, the information entropy of the Softmax-like classification function value obtained by the classification probability value calculation subunit being a negative value of a logarithm value of the Softmax-like classification function value plus a predetermined compensation term;
A classification feature map generation unit configured to calculate a weighted sum of the plurality of difference feature maps with the plurality of weighting coefficients obtained by the weighting coefficient generation unit to obtain a classification feature map; and
and the classification result generating unit is used for enabling the classification characteristic diagram obtained by the classification characteristic diagram generating unit to pass through a classifier so as to obtain a classification result, wherein the classification result is used for indicating whether the vibration in the section bar cutting process exceeds a preset limit or not.
8. The industrial internet-based profile cutting status monitoring system according to claim 7, wherein the classification result generation unit comprises:
a classification feature vector generation subunit, configured to pass the classification feature map through one or more fully-connected layers to encode the classification feature map through the one or more fully-connected layers to obtain a classification feature vector; and
and the classification subunit is used for inputting the classification feature vector into a Softmax classification function so as to obtain the classification result.
9. The industrial internet-based profile cutting status monitoring system according to claim 7, wherein the classification probability value obtaining unit is configured to calculate a classification probability value of a Softmax-like class of each of the differential feature maps according to the following formula: pi ═ Σ Fi exp(xi)/∑ Fd～Fi exp (xi), wherein the numerator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the ith differential feature map, and the denominator is the sum of power functions with a natural constant as the bottom of the feature values of each position in the 1 st to ith differential feature maps.
10. An electronic device, comprising:
a processor; and
a memory in which computer program instructions are stored, which, when executed by the processor, cause the processor to perform the industrial internet-based profile cutting status monitoring method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110529552.1A CN113139518B (en) | 2021-05-14 | 2021-05-14 | Section bar cutting state monitoring method based on industrial internet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110529552.1A CN113139518B (en) | 2021-05-14 | 2021-05-14 | Section bar cutting state monitoring method based on industrial internet |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113139518A CN113139518A (en) | 2021-07-20 |
| CN113139518B true CN113139518B (en) | 2022-07-29 |
Family
ID=76817545
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110529552.1A Active CN113139518B (en) | 2021-05-14 | 2021-05-14 | Section bar cutting state monitoring method based on industrial internet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113139518B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102014251A (en) * | 2009-09-04 | 2011-04-13 | 卡西欧计算机株式会社 | Image processing apparatus and image processing method |
| CN103208012A (en) * | 2013-05-08 | 2013-07-17 | 重庆邮电大学 | Light face recognition method |
| CN105158762A (en) * | 2014-03-31 | 2015-12-16 | 霍尼韦尔国际公司 | Identifying and tracking convective weather cells |
| CN105787458A (en) * | 2016-03-11 | 2016-07-20 | 重庆邮电大学 | Infrared behavior identification method based on adaptive fusion of artificial design feature and depth learning feature |
| CN108986100A (en) * | 2017-06-02 | 2018-12-11 | 宏达国际电子股份有限公司 | image cutting method and device |
| CN109844815A (en) * | 2016-08-22 | 2019-06-04 | 皇家飞利浦有限公司 | The image procossing based on feature is carried out using from the characteristic image of different iterative extractions |
| CN109977826A (en) * | 2019-03-15 | 2019-07-05 | 百度在线网络技术（北京）有限公司 | The classification recognition methods of object and device |
| CN111291614A (en) * | 2020-01-12 | 2020-06-16 | 杭州电子科技大学 | Child epilepsy syndrome classification method based on transfer learning multi-model decision fusion |
| CN112182797A (en) * | 2020-09-11 | 2021-01-05 | 新兴铸管股份有限公司 | Continuous casting cooling bed billet tracking method |
| CN112733750A (en) * | 2021-01-14 | 2021-04-30 | 成都缘能缘彩网络科技有限公司 | Dynamic water flow image-based sewage treatment detection neural network training method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10788836B2 (en) * | 2016-02-29 | 2020-09-29 | AI Incorporated | Obstacle recognition method for autonomous robots |
-
2021
- 2021-05-14 CN CN202110529552.1A patent/CN113139518B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102014251A (en) * | 2009-09-04 | 2011-04-13 | 卡西欧计算机株式会社 | Image processing apparatus and image processing method |
| CN103208012A (en) * | 2013-05-08 | 2013-07-17 | 重庆邮电大学 | Light face recognition method |
| CN105158762A (en) * | 2014-03-31 | 2015-12-16 | 霍尼韦尔国际公司 | Identifying and tracking convective weather cells |
| CN105787458A (en) * | 2016-03-11 | 2016-07-20 | 重庆邮电大学 | Infrared behavior identification method based on adaptive fusion of artificial design feature and depth learning feature |
| CN109844815A (en) * | 2016-08-22 | 2019-06-04 | 皇家飞利浦有限公司 | The image procossing based on feature is carried out using from the characteristic image of different iterative extractions |
| CN108986100A (en) * | 2017-06-02 | 2018-12-11 | 宏达国际电子股份有限公司 | image cutting method and device |
| CN109977826A (en) * | 2019-03-15 | 2019-07-05 | 百度在线网络技术（北京）有限公司 | The classification recognition methods of object and device |
| CN111291614A (en) * | 2020-01-12 | 2020-06-16 | 杭州电子科技大学 | Child epilepsy syndrome classification method based on transfer learning multi-model decision fusion |
| CN112182797A (en) * | 2020-09-11 | 2021-01-05 | 新兴铸管股份有限公司 | Continuous casting cooling bed billet tracking method |
| CN112733750A (en) * | 2021-01-14 | 2021-04-30 | 成都缘能缘彩网络科技有限公司 | Dynamic water flow image-based sewage treatment detection neural network training method |
Non-Patent Citations (1)
| Title |
|---|
| 李智勇 等.基于蛙眼检测模型的双阈值判定目标跟踪方法.《电子测量与仪器学报》.2019,第29卷(第2期),1-9. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113139518A (en) | 2021-07-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108762228B (en) | Distributed PCA-based multi-working-condition fault monitoring method | |
| Chien et al. | A system for online detection and classification of wafer bin map defect patterns for manufacturing intelligence | |
| US8781982B1 (en) | System and method for estimating remaining useful life | |
| Bordoloi et al. | Optimum multi-fault classification of gears with integration of evolutionary and SVM algorithms | |
| CN112613646A (en) | Equipment state prediction method and system based on multi-dimensional data fusion | |
| CN106228270B (en) | Energy consumption prediction method and system for big data driven extrusion equipment | |
| Tran et al. | Algorithm for estimating online bearing fault upon the ability to extract meaningful information from big data of intelligent structures | |
| CN110737976A (en) | mechanical equipment health assessment method based on multi-dimensional information fusion | |
| DE102021207269A1 (en) | METHOD AND SYSTEM FOR LEARNING PERTURBATION QUANTITIES IN MACHINE LEARNING | |
| CN116563293B (en) | Photovoltaic carrier production quality detection method and system based on machine vision | |
| CN111325410A (en) | General fault early warning system based on sample distribution and early warning method thereof | |
| CN116652823A (en) | Automatic monitoring system and method for grinding machine | |
| CN116872089A (en) | Grinding self-adaptive control system and method of numerically controlled grinder | |
| CN113139518B (en) | Section bar cutting state monitoring method based on industrial internet | |
| US11035666B2 (en) | Inspection-guided critical site selection for critical dimension measurement | |
| Sbarbaro et al. | Self-tuning control of non-linear systems using Gaussian process prior models | |
| CN116542956A (en) | Automatic detection method and system for fabric components and readable storage medium | |
| CN113159164B (en) | Industrial Internet equipment collaborative operation method based on distribution type | |
| CN113191306A (en) | Equipment abnormal state prediction method based on edge calculation | |
| Koyuncu et al. | Designing robust modified R control charts for asymmetric distributions under ranked set and median ranked set sampling | |
| Zhang | Deep learning driven tool wear identification and remaining useful life prediction | |
| CN109523195B (en) | Sample multi-modal classification-based fault detection method for hierarchical non-Gaussian algorithm | |
| Pham et al. | Application of the bees algorithm to the selection features for manufacturing data | |
| US20200143283A1 (en) | Time-series feature extraction apparatus, time-series feature extraction method and recording medium | |
| JP2020187616A (en) | Plant monitoring model creation device, plant monitoring model creation method, and plant monitoring model creation program |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20220705 Address after: 226001 4th and 5th floors of building 10B, Zilang science and Technology City, No. 60, Chongzhou Avenue, Nantong City, Jiangsu Province Applicant after: Jiangsu Zhongtian Internet Technology Co.,Ltd. Address before: 310000 No.48, Jiuhuan Road, Jianggan District, Hangzhou City, Zhejiang Province Applicant before: Hangzhou Xuyan Technology Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |