Device and method for automatically acquiring raw material information on cold rolling raw material pickling line

Abstract

The invention belongs to the field of intelligent manufacturing of metallurgical enterprises, and discloses a device and a method for automatically acquiring raw material information on a cold rolling raw material pickling line, and the specific technical scheme is that the device comprises a camera, a proximity switch, an IO input and output acquisition module, an image analysis and recognition computer, a loudspeaker, a network switch, a PCS server and a production execution system server; the method specifically comprises the steps that a raw material steel coil is hoisted to a raw material pickling line inlet saddle by a crown block, a proximity switch senses a metal object, a camera is triggered to capture a steel coil appearance image, an image analysis and recognition host analyzes the captured image to obtain a steel coil number, the steel coil number is compared with a plan of an MES, and if the plan is hit, the steel coil number is identified. The method comprises the following steps: sending a steel coil number including information such as a corresponding steel type and specification acquired from an MES system to a PCS system, thereby realizing automation of information acquisition of the PCS system, and if a plan is not hit, informing a production line worker through a loudspeaker that the on-line steel coil is not in the plan and needs to be manually verified and post-processed.

Description

Technical Field

[0001] The invention belongs to the field of intelligent manufacturing of metallurgical enterprises, and specifically discloses a device and method for automatically acquiring raw material information by a process control system of a cold-rolled raw material pickling line. Background Art

[0002] The cold rolling raw material pickling line is the first process of cold rolling production. The steel coils are connected by head-to-tail welding. The entire production line often has 3 to 5 steel coils processed continuously. Since the material number sprayed on it will disappear under the high temperature in the production line furnace, the process control system (PCS) at the entrance of the production line obtains accurate steel coil information, which is the basis for production and tracking. Traditionally, the method of using the material loader to load materials on demand, the production line recorder to manually copy and enter the coil number and information in the PCS system. With the advancement of technology, some raw material warehouses before the pickling line have become intelligent management, so that each coil on the line can be accurately located and marked, and the information can be transmitted to the PCS, but the investment in such intelligent warehouses ranges from millions to tens of millions, so 90% of such production lines in China still use manual methods. Due to the large number of manual operations in the manual method, it is inevitable that the PCS will get the wrong steel coil number, resulting in mixed steel coil numbers, leading to production quality problems. Therefore, how to use a lower cost to realize the automatic acquisition of steel coil information by the PCS system has become an urgent problem to be solved. Summary of the invention

[0003] In order to solve the technical problem that the cold-rolled raw material pickling line process control system mainly relies on manual verification to obtain steel coils and information, which is prone to errors and low efficiency, the present invention provides a device and method for automatically obtaining raw material information by the cold-rolled raw material pickling line process control system, which is used to reduce production line mixing errors and provide high work efficiency. Due to its high cost-effectiveness, it can be widely used in the automation system of cold-rolled raw material pickling lines.

[0004] To achieve the above-mentioned purpose, the technical solution adopted by the present invention is: a device for automatically acquiring raw material information of a cold-rolled raw material pickling line, including an MES server and a PCS server. The MES server and the image recognition and analysis host are connected by Ethernet. The purpose of the Ethernet connection is for the image recognition and analysis host to retrieve steel coil information according to the steel coil number. The image recognition and analysis host and the PCS server communicate by using an RS232 serial interface, which is used to transmit the retrieved steel coil information to the PCS server to guide production.

[0005] The image recognition analysis host is connected to the camera through a network switch to receive pictures captured by the camera.

[0006] The proximity switch is connected to the IO input node of the camera to trigger the camera to capture images. The image recognition and analysis host is connected to the loudspeaker through its audio output interface. When an abnormality is identified, it reminds on-site manual intervention.

[0007] The image recognition and analysis host is the core of the system. The steel coil number recognition model built into the image recognition and analysis host recognizes the steel coil number of the cold-rolled raw material. As the data source of the AI ​​development platform, after training, the steel coil number recognition model accurately recognizes the steel coil number of the cold-rolled raw material. After recognizing the steel coil number, the image recognition and analysis host retrieves the steel coil information from the database DB2 in the MES server through Ethernet, and sends the steel coil information to the PCS server through the RS232 serial port.

[0008] After obtaining the coil number of the analyzed steel coil, the image recognition analysis host obtains the production plan details from the MES server to further verify the coil number. In the event of incorrect coil recognition or failure to lift the steel coil as planned, an alarm is sounded on site through a loudspeaker to prompt manual intervention.

[0009] The image recognition analysis host receives the original image captured by the camera through Ethernet, rotates the original photo clockwise 9 times at a unit of 10 degrees to obtain 9 photos, and rotates the original photo counterclockwise 9 times at a unit of 10 degrees to obtain 9 photos. After the original photo is rotated, 19 photos at different angles are obtained, which solves the problem that the printed steel coil number is not in the center of the steel coil due to the deflection of the steel coil during the lifting process, affecting the model recognition.

[0010] The images captured by the camera are preprocessed to improve the image quality. When the maximum grayscale average value of the image is detected to be greater than 105, the system will enhance the image. Image enhancement is to transform a weak-contrast image into a strong-contrast image. Transforming the image grayscale can increase the image dynamic range and expand the contrast, making the image clear and obvious, that is, the coil number of the steel coil is clearer. Image enhancement uses nonlinear transformation to transform the grayscale of the image, including exponential transformation and logarithmic transformation;

[0011] The exponential transformation is used to expand the high grayscale area of ​​the image and is used for overly bright images. The exponential transformation formula is as follows:

[0012] g(x,y)=b c[f(x，y)－a] -1

[0013] Where a, b, and c are all adjustable parameters used to adjust the shape and position of the curve. Parameter a determines the starting position of the exponential function curve, b is the base, parameter c determines the degree of the transformation curve, and f(x, y) is the function that needs to be transformed.

[0014] The logarithmic transformation is used to expand the low grayscale area of ​​the image and is used for too dark images. The logarithmic transformation formula is as follows:

[0015] g(x, y)=a1+{ln[f(x, y)+1]} / (b1 lnc1)

[0016] In the formula, a1, b1, and c1 are all adjustable parameters used to adjust the shape and position of the curve. Parameter a1 is the starting position of the transformation curve, and parameters b1 and c1 are used to determine the change rate of the transformation curve.

[0017] The image recognition analysis host obtains the production plan details from the MES server to further verify the accuracy of the steel coil number. If the steel coil is identified incorrectly or the steel coil is not lifted as planned, an on-site alarm will be sounded through the loudspeaker to prompt manual intervention.

[0018] The method for automatically obtaining raw material information of cold-rolled raw material pickling line, the specific steps are as follows:

[0019] Step 1: Use the overhead crane to hoist the steel coil to the entrance saddle of the pickling line production line. The proximity switch senses that the material has arrived at the saddle, and the camera is triggered to capture the picture through the camera IO input node. The picture captured by the camera is uploaded to the image recognition analysis host through the network switch;

[0020] Step 2: Image recognition analysis The host analyzes the captured image through the built-in roll number recognition model to obtain the steel roll number of the entrance saddle. The roll number recognition model is obtained by capturing nearly 2,000 photos of steel coils on site, all of which are trained by the AI ​​development platform.

[0021] Step 3: The image analysis and recognition host obtains the production plan details from the MES server. If the identified steel coil number is not in the production plan details, it is determined that an abnormal situation has occurred. The image recognition and analysis host triggers a loudspeaker alarm reminder for the above abnormal situation and recommends manual intervention.

[0022] Step 4: If the identified steel coil number is within the production plan, the image recognition analysis host retrieves the steel coil information through the MES server, and sends it to the PCS server serial port through the identification computer serial port, and the PCS server data receiving program.

[0023] In step three, there are two types of abnormal situations. One is that the hoisting worker does not organize the steel coils according to the production plan and hoists the unplanned steel coils onto the saddle. The other is that the coil number recognition model makes an error in recognizing the coil number.

[0024] The present invention realizes the automation of the acquisition of information sources (steel coil number, steel grade, specification, processing requirements) of the PCS system of the cold-rolled raw material pickling line, avoids the errors caused by manual copying and data entry, and improves work efficiency. In particular, for this enterprise that mainly produces stainless steel and special steel, the steel grade changes frequently during production, and there are two feeding areas, A and B, with cross-production. Before implementation, it was time-consuming and labor-intensive to manually check the steel coil number, and often only the last few digits of the steel coil number were checked. At the same time, the information system imported the plan in batches, and the coils were manually loaded one by one, which easily led to errors. After implementation, automatic identification devices were installed in front of the uncoiler in the two areas to realize automatic loading one by one, effectively solving the problem of mixed steel coil numbers caused by manual work. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is the overall structure diagram of the present invention.

[0026] In the figure, 1 is the MES server, 2 is the PCS server, 3 is the image recognition analysis host, 4 is the network switch, 5 is the camera, 6 is the proximity switch, and 7 is the loudspeaker. DETAILED DESCRIPTION

[0027] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0028] like Figure 1 As shown, a device for automatically acquiring raw material information of a cold-rolled raw material pickling line includes an MES server 1 and a PCS server 2. The MES server 1 and the image recognition and analysis host 3 are connected by Ethernet. The purpose of the Ethernet connection is for the image recognition and analysis host 3 to retrieve the steel coil information according to the steel coil number. The image recognition and analysis host 3 and the PCS server 2 communicate by using an RS232 serial interface, which is used to transmit the retrieved steel coil information to the PCS server 2 to guide production.

[0029] The image recognition and analysis host 3 is connected to the camera 5 via a network switch 4 and is used to receive pictures captured by the camera 5 .

[0030] The proximity switch 6 is connected to the IO input node of the camera 5 to trigger the camera 5 to capture an image. The image recognition and analysis host 3 is connected to the loudspeaker 7 through its audio output interface. When an abnormality is recognized, it reminds on-site manual intervention.

[0031] The image recognition and analysis host 3 is the core of the system. Through the software equipped by it, the ORC (optical character recognition) technology is used to extract text information from the picture captured by the camera 5, and automatically identify the material number (material identity information) printed on the surface of the steel coil, instead of manually reading the material number. After successfully reading the steel coil number, the raw steel type, specification and other information can be obtained from the MES system according to the steel coil number.

[0032] The steel coil number recognition model built into the image recognition analysis host 3 recognizes the cold-rolled raw steel coil number. As the data source of the AI ​​development platform, after training, the steel coil number recognition model accurately recognizes the cold-rolled raw steel coil number. After recognizing the steel coil number, the image recognition analysis host 3 retrieves the steel coil information from the database DB2 in the MES server 1 through Ethernet, and the image recognition analysis host 3 sends the steel coil information to the PCS server 2 through the RS232 serial port. The PCS server 2 guides production based on the information obtained. Through the above-mentioned automatic recognition of steel coil numbers, automatic retrieval and transmission of steel coil information, work efficiency is improved and errors caused by manual operation are reduced.

[0033] After obtaining the coil number of the analyzed steel coil, the image recognition analysis host 3 obtains the production plan details from the MES server 1 to further verify the coil number. In the event of steel coil recognition errors or steel coils not being lifted as planned, an alarm is issued on site through the loudspeaker 7 to prompt manual intervention.

[0034] Create an on-site steel coil dataset and upload real pictures, define the data identification field roll number as the annotation label, select the corresponding Key / Value content area in the picture, automatically identify the content of the selected area to complete the transcription, and the annotation personnel can complete the annotation by checking and correcting the identification result roll number.

[0035] During the model training phase, the existing labeled data can be used to select an algorithm for training based on the determined preliminary model type.

[0036] After confirming that the model effect is available, the model can be deployed to the image recognition analysis host 3, so that the image recognition analysis host 3 has the ability to recognize the coil number of the on-site steel coil image.

[0037] When the coil number of the on-site steel coil is printed in the hot rolling process, the printing position is in the center of the end face of the steel coil. Before the steel coil is put on the cold rolling raw material pickling line, it is hoisted by multiple overhead cranes, and the steel coil is deflected. The steel coil number is no longer in the center of the steel coil. It is possible that a large deflection will occur, which has a great impact on the accuracy of the model's recognition of the steel coil number. In order to solve the problem of steel coil deflection, the present invention adopts a method of software rotation of the captured photo, that is, the original photo is rotated clockwise (10, 20...90 degrees) and counterclockwise (10, 20...90 degrees) according to 10 degrees as a unit. After a photo is rotated, 19 photos of different angles will be obtained for model analysis, and a steel coil number can always be collected in the center of the image, providing basic data for the accuracy of model analysis.

[0038] In order to verify the accuracy of the model in identifying the coil number, after the coil number is obtained in the image recognition analysis host 3, it will be compared with the production plan of the cold-rolled raw material pickling line and the inventory coil details in the MES system. If it is not in the above detailed list, it means that the coil is incorrectly identified or the coil is not lifted as planned, and an alarm will be issued on site through the loudspeaker 7 to remind manual intervention.

[0039] The identification host Windows application and PCS server 2 use RS232 serial communication. The purpose of using this communication standard is to better solve the network isolation problem between the image recognition analysis host 3 and the PCS server 2. Compared with the widely used Ethernet communication, which is susceptible to network security (such as viruses), this communication method is safe and reliable. Although the speed is slow, it is sufficient to meet the transmission of steel coil information (coil number, specification, steel type, production requirements). Specifically, in the present invention, a direct communication line is welded between the serial port of the identification host and the serial port of the PCS server 2, and data sending and receiving programs are deployed respectively to realize information transmission.

[0040] The image recognition analysis host 3 receives the original image captured by the camera 5 through Ethernet, rotates the original photo clockwise 9 times at a unit of 10 degrees to obtain 9 photos, and rotates the original photo counterclockwise 9 times at a unit of 10 degrees to obtain 9 photos. After the original photo is rotated, 19 photos at different angles are obtained, which solves the problem that the printed steel coil number is not in the center of the steel coil due to the deflection of the steel coil during the lifting process, affecting the model recognition.

[0041] The image captured by camera 5 is preprocessed to improve the image quality. When the maximum grayscale average value of the image is detected to be greater than 105, the system enhances the image. Image enhancement is to transform a weak-contrast image into a strong-contrast image. Transforming the image grayscale can increase the image dynamic range and expand the contrast, making the image clear and obvious, that is, the coil number of the steel coil is clearer. Image enhancement uses nonlinear transformation to transform the grayscale of the image, including exponential transformation and logarithmic transformation;

[0042] The exponential transformation is used to expand the high grayscale area of ​​the image and is used for overly bright images. The exponential transformation formula is as follows:

[0043] g(x,y)=b c[f(x，y)－a] -1

[0044] In the formula, a, b, and c are all adjustable parameters used to adjust the shape and position of the curve. Parameter a determines the starting position of the exponential function curve, b is the base, parameter c determines the degree of the transformation curve, and f(x, y) is the function to be transformed;

[0045] The logarithmic transformation is used to expand the low grayscale area of ​​the image and is used for too dark images. The logarithmic transformation formula is as follows:

[0046] g(x, y)=a1+{ln[f(x, y)+1]} / (b1 lnc1)

[0047] In the formula, a1, b1, and c1 are all adjustable parameters used to adjust the shape and position of the curve. Parameter a1 is the starting position of the transformation curve, and parameters b1 and c1 are used to determine the change rate of the transformation curve.

[0048] The image recognition analysis host 3 obtains the production plan details from the MES server 1 to further verify the accuracy of the coil number. If an error is found in the coil identification or the coil is not lifted as planned, an on-site alarm is sounded through the loudspeaker 7 to remind manual intervention.

[0049] The method for automatically obtaining raw material information of cold-rolled raw material pickling line, the specific steps are as follows:

[0050] Step 1: Use an overhead crane to hoist the steel coil to the entrance saddle of the pickling line production line. The proximity switch 6 installed at this position senses the metal object and triggers the camera 5 to capture the appearance image of the steel coil. The camera 5 is triggered to capture the image through the camera 5IO input node. The image captured by the camera 5 is uploaded to the image recognition analysis host 3 through the network switch 4. The image should be consistent with the real business scenario as much as possible and cover all kinds of situations as much as possible.

[0051] Step 2: Image recognition and analysis: Host 3 analyzes the captured image through the built-in roll number recognition model to obtain the steel roll number of the entrance saddle. The roll number recognition model is obtained by capturing nearly two thousand photos of steel coils on site, all of which are trained through the AI ​​development platform.

[0052] Step 3: The image analysis and recognition host obtains the production plan details from the MES server 1. If the identified steel coil number is not in the production plan details, it is determined that an abnormal situation has occurred. The image recognition and analysis host 3 triggers the loudspeaker 7 to alarm and recommend manual intervention.

[0053] There are two types of abnormal situations. One is that the hoisting workers fail to organize the lifting of steel coils according to the production plan and lift the unplanned steel coils onto the saddle. The other is that the coil number recognition model makes an error in identifying the coil number.

[0054] Step 4: Compare the steel coil number with the MES plan. If the identified steel coil number is in the production plan, the image recognition analysis host 3 retrieves the steel coil information through the MES server 1, and sends it to the serial port of the PCS server 2 through the identification computer serial port, and the data receiving program of the PCS server 2.

[0055] The present invention realizes the automation of obtaining information sources (steel coil number, steel type, specification, processing requirements) of the PCS system of the cold-rolled raw material pickling line, avoids errors caused by manual copying and inputting of data, and improves work efficiency.

[0056] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims

1. A device for automatically acquiring raw material information for a cold-rolled raw material pickling line, characterized in that: It includes an MES server and a PCS server. The MES server and the image recognition and analysis host are connected by Ethernet. The image recognition and analysis host retrieves the steel coil information according to the steel coil number. The image recognition and analysis host and the PCS server communicate by using an RS232 serial interface to transmit the retrieved steel coil information to the PCS server. The image recognition analysis host is connected to the camera through a network switch to receive the pictures captured by the camera; The proximity switch is connected to the IO input node of the camera to trigger the camera to capture the image. The image recognition and analysis host is connected to the loudspeaker through its audio output interface. The steel coil number recognition model built into the image recognition analysis host recognizes the steel coil number of the cold-rolled raw material. After recognizing the steel coil number, the image recognition analysis host retrieves the steel coil information from the database in the MES server through Ethernet, and sends the steel coil information to the PCS server through the RS232 serial port; After obtaining the coil number of the analyzed steel coil, the image recognition analysis host obtains the production plan details from the MES server to further verify the coil number. In the event of incorrect coil recognition or failure to lift the steel coil as planned, an alarm is sounded on site through a loudspeaker to prompt manual intervention.

2. The device for automatically acquiring raw material information of a cold-rolled raw material pickling line according to claim 1 is characterized in that: The image recognition analysis host receives the original image captured by the camera through Ethernet, rotates the original photo clockwise 9 times at a unit of 10 degrees to obtain 9 photos, and rotates the original photo counterclockwise 9 times at a unit of 10 degrees to obtain 9 photos. After rotating an original photo, 19 photos at different angles are obtained.

3. The device for automatically acquiring raw material information of a cold-rolled raw material pickling line according to claim 1 is characterized in that: Pre-process the pictures captured by the camera. When the maximum grayscale average value of the picture is detected to be greater than 105, the system will enhance the picture and transform the image grayscale. Image enhancement uses nonlinear transformation to transform the grayscale of the picture, including exponential transformation and logarithmic transformation. The exponential transformation is used to expand the high grayscale area of ​​the image. The exponential transformation formula is as follows: g(x,y)=b c[f(x，y)－a] -1 In the formula, a, b, and c are all adjustable parameters used to adjust the shape and position of the curve. Parameter a determines the starting position of the exponential function curve, b is the base, parameter c determines the degree of the transformation curve, and f(x, y) is the function to be transformed; Logarithmic transformation is suitable for processing overly bright images. Logarithmic transformation is used to expand the low grayscale area of ​​the image. The logarithmic transformation formula is as follows: g(x, y)=a1+{ln[f(x, y)+1]} / (b1 lnc1) In the formula, a1, b1, and c1 are all adjustable parameters used to adjust the shape and position of the curve. Parameter a1 is the starting position of the transformation curve, and parameters b1 and c1 are used to determine the change rate of the transformation curve. The image recognition analysis host obtains the production plan details from the MES server. If it is identified that the steel coil is incorrectly identified or the steel coil is not lifted as planned, an on-site alarm will be sounded through the loudspeaker to prompt manual intervention.

4. A method for automatically obtaining raw material information for a cold-rolled raw material pickling line, characterized in that: The specific steps are as follows: Step 1: Use the overhead crane to hoist the steel coil to the entrance saddle of the pickling line production line. The proximity switch senses that the material has arrived at the saddle, and the camera is triggered to capture the picture through the camera IO input node. The picture captured by the camera is uploaded to the image recognition analysis host through the network switch; Step 2: Image recognition and analysis The host analyzes the captured image through the built-in roll number recognition model to obtain the steel roll number of the entrance saddle; Step 3: The image analysis and recognition host obtains the production plan details from the MES server. If the identified steel coil number is not in the production plan details, it is determined that an abnormal situation has occurred. The image recognition and analysis host triggers a loudspeaker alarm reminder for the above abnormal situation and recommends manual intervention; Step 4: If the identified coil number is within the production plan, the image recognition analysis host retrieves the coil information through the MES server and sends it to the PCS server serial port through the identification computer serial port, the PCS server data receiving program.

5. The method for automatically acquiring raw material information of a cold-rolled raw material pickling line according to claim 4 is characterized in that: In step 2, the built-in coil number recognition model is obtained by capturing photos of steel coils on site and training them through the AI ​​development platform.

6. The method for automatically acquiring raw material information of a cold-rolled raw material pickling line according to claim 4 is characterized in that: In step three, there are two types of abnormal situations. One is that the hoisting worker fails to organize the hoisting of the steel coils according to the production plan and hoists the unplanned steel coils onto the saddle. The other is that the coil number recognition model makes an error in recognizing the coil number.

Images