Heat-resistant Steel Bar Impurity Identification System Using a Feedforward Neural Network Model

Abstract

The present invention relates to a heat-resistant steel bar impurity identification system applying a feedforward neural network model, comprising: a bar cutting mechanism for performing cutting on a heat-resistant steel bar blank after the production of a set length is completed to obtain a heat-resistant steel bar finished product of a single set length; and a third analysis device for intelligently identifying the on-site impurity presence level of the heat-resistant steel bar finished product by using a feedforward neural network model that has completed multiple learning actions based on various visually targeted information screened on the surface of the heat-resistant steel bar finished product of a single set length obtained by cutting. Through the present invention, various visually targeted information screened on the surface of the heat-resistant steel bar finished product of a single set length can be extracted, and then a feedforward neural network model that has completed multiple learning actions can be used to obtain the on-site impurity presence level of the heat-resistant steel bar finished product output therefrom, thereby completing the intelligent determination of the on-site impurity distribution state of the heat-resistant steel bar finished product.

Description

Technical Field

[0001] The present invention relates to the field of neural networks, and in particular, to a heat-resistant steel bar impurity identification system applying a feedforward neural network model. Background Art

[0002] The production of stainless steel bars emerged with the development of stainless steel. Due to the increasingly wide range of applications of stainless steel bars, such as the foundations of high-rise buildings in cold regions, isolation nets beside highways, household daily necessities, etc., the hot rolling production of stainless steel bars has been greatly developed. With the rapid development of cutting-edge technologies such as petroleum, chemical industry, energy, atomic energy, aerospace, and ocean development, higher comprehensive performance requirements are put forward for stainless steel, not only requiring good corrosion resistance, but also requiring high strength, high temperature and high pressure resistance, radiation protection, low temperature resistance, etc., which further expands the variety types of stainless steel. Heat-resistant steel bars are an important type branch of stainless steel bars, divided into two categories: steel for pressure processing and steel for cutting processing, and divided into four categories: austenitic, ferritic, martensitic, and precipitation hardening types according to organizational characteristics.

[0003] Obviously, the on-site impurity distribution state of heat-resistant steel bar products is of great significance to the various heat-resistant properties of heat-resistant steel bar products. However, in the prior art, there is a lack of a high-precision analysis mechanism for the on-site impurity distribution state of heat-resistant steel bar products, resulting in heat-resistant steel bar products of uneven quality flowing into the market, posing potential safety hazards to the use of related manufactured products. Summary of the Invention

[0004] In order to solve the technical problems in the prior art, the present invention provides a heat-resistant steel bar impurity identification system applying a feedforward neural network model, which can extract the respective Y component values, respective U component values, and respective V component values of each pixel point occupied by the heat-resistant steel bar product in the received multi-level optimized image in the YUV space, and obtain the upper limit value of the Y component, the lower limit value of the Y component, the upper limit value of the U component, the lower limit value of the U component, the upper limit value of the V component, and the lower limit value of the V component corresponding to the heat-resistant steel bar product, so as to provide reliable information for the intelligent identification of the on-site impurity existence level of the heat-resistant steel bar product, and input the respective Y component values, respective U component values, and respective V component values of each pixel point occupied by the heat-resistant steel bar product in the received multi-level optimized image in the YUV space, as well as the upper limit value of the Y component, the lower limit value of the Y component, the upper limit value of the U component, the lower limit value of the U component, the upper limit value of the V component, and the lower limit value of the V component corresponding to the heat-resistant steel bar product into the feedforward neural network model that has completed multiple learning actions, and execute the feedforward neural network model that has completed multiple learning actions to obtain the on-site impurity existence level of the heat-resistant steel bar product output by it, thereby completing the intelligent determination of the on-site impurity distribution state of the heat-resistant steel bar product.

[0005] According to the present invention, there is provided a heat-resistant steel bar impurity identification system applying a feedforward neural network model, and the system includes:

[0006] A bar cutting mechanism, configured to perform cutting on a heat-resistant steel bar blank after the production of a set length is completed, so as to obtain a finished heat-resistant steel bar of a single set length;

[0007] A surface capture mechanism, connected to the bar cutting mechanism, and configured to perform a screen capture action on the surface of the finished heat-resistant steel bar of a single set length obtained by cutting after detecting that the bar cutting mechanism has completed the cutting operation, so as to obtain and output a corresponding finished product environment picture;

[0008] A multi-level optimization mechanism, connected to the surface capture mechanism and including a distortion correction device, a box filter device, and a data sharpening device, configured to continuously perform distortion correction processing, box filter processing, and spatial domain differential method sharpening processing on the received finished product environment picture, so as to obtain and output a corresponding multi-level optimized picture;

[0009] A first analysis device, configured to perform multiple learning on a feedforward neural network to obtain a feedforward neural network model that has completed multiple learning actions, and the number of learning times corresponding to the multiple learning is proportional to the number of pixel points of the received multi-level optimized picture;

[0010] A second analysis device, connected to the multi-level optimization mechanism, and configured to extract the respective Y component values, respective U component values, and respective V component values of each pixel point occupied by the finished heat-resistant steel bar in the received multi-level optimized picture in the YUV space, and obtain the upper limit value of the Y component, the lower limit value of the Y component, the upper limit value of the U component, the lower limit value of the U component, the upper limit value of the V component, and the lower limit value of the V component corresponding to the finished heat-resistant steel bar;

[0011] A third analysis device, respectively connected to the first analysis device and the second analysis device, and configured to input the respective Y component values, respective U component values, and respective V component values of each pixel point occupied by the finished heat-resistant steel bar in the received multi-level optimized picture in the YUV space, and the upper limit value of the Y component, the lower limit value of the Y component, the upper limit value of the U component, the lower limit value of the U component, the upper limit value of the V component, and the lower limit value of the V component corresponding to the finished heat-resistant steel bar into the feedforward neural network model that has completed multiple learning actions, and execute the feedforward neural network model that has completed multiple learning actions, so as to obtain the on-site impurity presence level of the finished heat-resistant steel bar output by it;

[0012] Wherein, the higher the on-site impurity presence level of the finished heat-resistant steel bar output by the feedforward neural network model that has completed multiple learning actions, the more surface impurities the finished heat-resistant steel bar has.

[0013] It can be seen that the present invention has at least the following three main inventive concepts:

[0014] Inventive concept 1: Perform multiple learning on a feedforward neural network to obtain a feedforward neural network model that has completed multiple learning actions, where the number of learning times corresponding to the multiple learning is proportional to the number of pixels in the received multi-level optimization screen;

[0015] Inventive concept 2: Extract the respective Y component values, respective U component values, and respective V component values of each pixel occupied by the finished heat-resistant steel bar in the received multi-level optimization screen in the YUV space, and obtain the upper limit value of the Y component, the lower limit value of the Y component, the upper limit value of the U component, the lower limit value of the U component, the upper limit value of the V component, and the lower limit value of the V component corresponding to the finished heat-resistant steel bar, so as to provide reliable information for the intelligent identification of the on-site impurity existence level of the finished heat-resistant steel bar;

[0016] Inventive concept 3: Input the respective Y component values, respective U component values, and respective V component values of each pixel occupied by the finished heat-resistant steel bar in the received multi-level optimization screen in the YUV space, as well as the upper limit value of the Y component, the lower limit value of the Y component, the upper limit value of the U component, the lower limit value of the U component, the upper limit value of the V component, and the lower limit value of the V component corresponding to the finished heat-resistant steel bar into the feedforward neural network model that has completed multiple learning actions, and execute the feedforward neural network model that has completed multiple learning actions to obtain the on-site impurity existence level of the finished heat-resistant steel bar output by it, thereby completing the intelligent determination of the on-site impurity distribution state of the finished heat-resistant steel bar. Brief Description of the Drawings

[0017] The following will describe the embodiments of the present invention with reference to the drawings, where:

[0018] Figure 1 FIG. is a structural block diagram of a heat-resistant steel bar impurity identification system applying a feedforward neural network model according to the primary embodiment of the present invention.

[0019] Figure 2 FIG. is a structural block diagram of a heat-resistant steel bar impurity identification system applying a feedforward neural network model according to the secondary embodiment of the present invention.

[0020] Figure 3 FIG. is a structural block diagram of a heat-resistant steel bar impurity identification system applying a feedforward neural network model according to the tertiary embodiment of the present invention. Detailed Description of the Invention

[0021] The following will refer to the drawings to describe in detail the embodiments of the heat-resistant steel bar impurity identification system applying a feedforward neural network model of the present invention.

[0022] Figure 1A structural block diagram of a heat-resistant steel bar impurity identification system applying a feedforward neural network model according to the primary embodiment of the present invention. The system includes:

[0023] A bar cutting mechanism, configured to perform cutting on a heat-resistant steel bar blank after the production of a set length is completed to obtain a finished heat-resistant steel bar of a single set length;

[0024] Specifically, the bar cutting mechanism, configured to perform cutting on a heat-resistant steel bar blank after the production of a set length is completed to obtain a finished heat-resistant steel bar of a single set length, includes: a length measurement unit, a cutting execution tool, and a microcontroller built in the bar cutting mechanism;

[0025] A surface capture mechanism, connected to the bar cutting mechanism, configured to perform a screen capture action on the surface of the finished heat-resistant steel bar of a single set length obtained by cutting after detecting that the bar cutting mechanism has completed the cutting operation, so as to obtain and output a corresponding finished product environment screen;

[0026] A multi-level optimization mechanism, connected to the surface capture mechanism and including a distortion correction device, a box filter device, and a data sharpening device, configured to continuously perform distortion correction processing, box filter processing, and spatial domain differential method sharpening processing on the received finished product environment screen, so as to obtain and output a corresponding multi-level optimization screen;

[0027] A first analysis device, configured to perform multiple learning on a feedforward neural network to obtain a feedforward neural network model that has completed multiple learning actions, where the number of learning times corresponding to the multiple learning is proportional to the number of pixel points of the received multi-level optimization screen;

[0028] A second analysis device, connected to the multi-level optimization mechanism, configured to extract the respective Y component values, respective U component values, and respective V component values of each pixel point occupied by the finished heat-resistant steel bar in the received multi-level optimization screen in the YUV space, and obtain the upper limit value of the Y component, the lower limit value of the Y component, the upper limit value of the U component, the lower limit value of the U component, the upper limit value of the V component, and the lower limit value of the V component corresponding to the finished heat-resistant steel bar;

[0029] A third analysis device, which is respectively connected to the first analysis device and the second analysis device, is configured to input the Y - component values, U - component values, and V - component values of each pixel point occupied by the finished heat - resistant steel rod in the received multi - level optimized image in the YUV space, as well as the upper limit value of the Y - component, lower limit value of the Y - component, upper limit value of the U - component, lower limit value of the U - component, upper limit value of the V - component, and lower limit value of the V - component corresponding to the finished heat - resistant steel rod into the feed - forward neural network model that has completed multiple learning actions, and execute the feed - forward neural network model that has completed multiple learning actions to obtain the on - site impurity presence level of the finished heat - resistant steel rod output by it;

[0030] Among them, the higher the on - site impurity presence level of the finished heat - resistant steel rod output by the feed - forward neural network model that has completed multiple learning actions, the more surface impurities the finished heat - resistant steel rod has;

[0031] Among them, the multi - level optimization mechanism, which is connected to the surface capture mechanism and includes a distortion correction device, a box - type filtering device, and a data sharpening device, is configured to continuously perform distortion correction processing, box - type filtering processing, and spatial domain differential method sharpening processing on the received finished product environment image to obtain and output the corresponding multi - level optimized image, including: the distortion correction device, the box - type filtering device, and the data sharpening device respectively perform distortion correction processing, box - type filtering processing, and spatial domain differential method sharpening processing on the received image signal.

[0032] Figure 2 It is a structural block diagram of a heat - resistant steel rod impurity identification system applying a feed - forward neural network model shown according to a secondary embodiment of the present invention.

[0033] Compared with Figure 1 , Figure 2 the heat - resistant steel rod impurity identification system applying a feed - forward neural network model in

[0034] The area analysis mechanism is respectively connected to the first analysis device, the second analysis device, the third analysis device, and the multi - level optimization mechanism, and is configured to respectively measure the current real - time area values of the first analysis device, the second analysis device, the third analysis device, and the multi - level optimization mechanism;

[0035] Among them, the area analysis mechanism is respectively connected to the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism, and is used to measure the current real-time area values of the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism respectively. The area analysis mechanism includes multiple area measurement units, which are used to be respectively connected to the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism to complete the separate measurement of the current real-time area values of the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism respectively;

[0036] Among them, the area analysis mechanism includes multiple area measurement units, which are used to be respectively connected to the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism to complete the separate measurement of the current real-time area values of the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism respectively. The multiple area measurement units are multiple visual detectors, which are used to be respectively connected to the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism to complete the separate measurement of the current real-time area values of the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism respectively;

[0037] Among them, the multiple area measurement units are multiple visual detectors, which are used to be respectively connected to the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism to complete the separate measurement of the current real-time area values of the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism respectively. The structures of the multiple visual detectors are the same;

[0038] Among them, the multiple area measurement units are multiple visual detectors, which are used to be respectively connected to the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism to complete the separate measurement of the current real-time area values of the first analysis device, the second analysis device, the third analysis device, and the multi-level optimization mechanism respectively. Each of the multiple visual detectors includes a planar camera.

[0039] Figure 3 It is a structural block diagram of a heat-resistant steel bar impurity identification system applying a feedforward neural network model shown in a secondary embodiment of the present invention.

[0040] Compared with Figure 1 , Figure 3The impurity identification system for heat-resistant steel bars using a feedforward neural network model may further include:

[0041] An information storage device, respectively connected to the first parsing device, the second parsing device, the third parsing device, and multiple area measurement units of the multi-level optimization mechanism, for performing immediate storage of the real-time state parameters of the first parsing device, the second parsing device, the third parsing device, and the multi-level optimization mechanism;

[0042] Among them, the information storage device, respectively connected to the first parsing device, the second parsing device, the third parsing device, and multiple area measurement units of the multi-level optimization mechanism, for performing immediate storage of the real-time state parameters of the first parsing device, the second parsing device, the third parsing device, and the multi-level optimization mechanism includes: the information storage device is a dynamic storage device;

[0043] Among them, the information storage device, respectively connected to the first parsing device, the second parsing device, the third parsing device, and multiple area measurement units of the multi-level optimization mechanism, for performing immediate storage of the real-time state parameters of the first parsing device, the second parsing device, the third parsing device, and the multi-level optimization mechanism includes: the information storage device is a TF memory card.

[0044] In addition, in the impurity identification system for heat-resistant steel bars using a feedforward neural network model of the present invention, the multi-level optimization mechanism, connected to the surface capture mechanism and including a distortion correction device, a box filter device, and a data sharpening device, for continuously performing distortion correction processing, box filter processing, and spatial differential method sharpening processing on the received finished product environment picture to obtain and output a corresponding multi-level optimized picture further includes: the multi-level optimization mechanism further includes a parallel data bus, respectively connected to the distortion correction device, the box filter device, and the data sharpening device, for realizing parallel data communication between any two of the distortion correction device, the box filter device, and the data sharpening device.

[0045] By using the impurity identification system for heat-resistant steel bars using a feedforward neural network model of the present invention, aiming at the technical problem that there is a lack of a targeted and high-precision identification mechanism for the on-site impurity existence level of heat-resistant steel bars in the prior art, by parallelly inputting the visual information obtained by specifically screening the surface of the finished heat-resistant steel bars of a single set length into the feedforward neural network model that has completed multiple learning actions to obtain the on-site impurity existence level of the finished heat-resistant steel bars output by it, thereby completing the intelligent determination of the on-site impurity distribution state of the finished heat-resistant steel bars.

[0046] It is understood that although the present invention has been disclosed above in preferred embodiments, the above embodiments are not intended to limit the present invention. For any person skilled in the art, without departing from the scope of the technical solution of the present invention, many possible changes and modifications can be made to the technical solution of the present invention by using the technical content disclosed above, or it can be modified into equivalent embodiments with equivalent changes. Therefore, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A heat-resistant steel bar impurity identification system using a feedforward neural network model, characterized in that: The system comprises: A rod cutting mechanism, used for cutting the heat-resistant steel rod blank after the heat-resistant steel rod blank has been produced to a set length to obtain a single heat-resistant steel rod finished product of the set length; A surface capture mechanism is connected to the rod cutting mechanism and is used to perform a picture capture action on the surface of a single heat-resistant steel rod finished product of a set length obtained by cutting after detecting that the rod cutting mechanism has completed the cutting operation, so as to obtain and output a corresponding finished product environment picture; A multi-level optimization mechanism, connected to the surface capture mechanism and comprising a distortion correction device, a box filter device and a data sharpening device, for continuously performing distortion correction processing, box filter processing and spatial differential method sharpening processing on the received finished environment picture to obtain and output a corresponding multi-level optimized picture; A first analysis device is used to perform multiple learning on the feedforward neural network to obtain a feedforward neural network model that completes multiple learning actions, wherein the number of learning times corresponding to the multiple learning actions is proportional to the number of pixels of the received multi-level optimized picture; A second analysis device is connected to the multi-level optimization mechanism, and is used to extract each Y component value, each U component value, and each V component value in the YUV space of each pixel point occupied by the heat-resistant steel bar finished product in the received multi-level optimization picture, and obtain the Y component upper limit value, Y component lower limit value, U component upper limit value, U component lower limit value, V component upper limit value, and V component lower limit value corresponding to the heat-resistant steel bar finished product; A third analysis device is connected to the first analysis device and the second analysis device respectively, and is used to input each Y component value, each U component value and each V component value in the YUV space of each pixel point occupied by the heat-resistant steel bar finished product in the received multi-level optimization picture, and the Y component upper limit value, Y component lower limit value, U component upper limit value, U component lower limit value, V component upper limit value and V component lower limit value corresponding to the heat-resistant steel bar finished product into the feedforward neural network model that has completed multiple learning actions, and execute the feedforward neural network model that has completed multiple learning actions to obtain the on-site impurity existence level of the heat-resistant steel bar finished product outputted by it; Among them, the higher the on-site impurity presence level of the heat-resistant steel bar finished product output by the feedforward neural network model that has completed multiple learning actions, the more the number of surface impurities of the heat-resistant steel bar finished product; A multi-level optimization mechanism, connected to the surface capture mechanism and comprising a distortion correction device, a box filter device and a data sharpening device, for continuously performing distortion correction processing, box filter processing and spatial differential method sharpening processing on the received finished environment picture to obtain and output a corresponding multi-level optimized picture, comprising: the distortion correction device, the box filter device and the data sharpening device respectively performing distortion correction processing, box filter processing and spatial differential method sharpening processing on the received picture signal; The multi-stage optimization mechanism also includes a parallel data bus, which is respectively connected to the distortion correction device, the box filter device and the data sharpening device, and is used to realize parallel data communication between the distortion correction device, the box filter device and the data sharpening device.

2. The heat-resistant steel bar impurity identification system using a feedforward neural network model as claimed in claim 1, characterized in that: The system further comprises: an area analysis mechanism, connected to the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism, respectively, and used to measure the current real-time area values ​​of the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism respectively; Among them, the area analysis mechanism is respectively connected to the first analysis device, the second analysis device, the third analysis device and the multi-level optimization mechanism, and is used to respectively measure the current real-time area values ​​of the first analysis device, the second analysis device, the third analysis device and the multi-level optimization mechanism. The area analysis mechanism includes multiple area measurement units, which are respectively connected to the first analysis device, the second analysis device, the third analysis device and the multi-level optimization mechanism to complete the respective measurements of the current real-time area values ​​of the first analysis device, the second analysis device, the third analysis device and the multi-level optimization mechanism.

3. The heat-resistant steel bar impurity identification system using a feedforward neural network model as claimed in claim 2, characterized in that: The area analysis mechanism includes a plurality of area measurement units, which are used to be connected to the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism respectively, so as to complete the respective measurement of the current real-time area values ​​of the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism respectively, including: the plurality of area measurement units are a plurality of visual detectors, which are used to be connected to the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism respectively, so as to complete the respective measurement of the current real-time area values ​​of the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism respectively.

4. The heat-resistant steel bar impurity identification system using a feedforward neural network model as claimed in claim 3, characterized in that: The multiple area measurement units are multiple visual detectors, which are used to be connected to the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism respectively, so as to complete the respective measurements of the current real-time area values ​​of the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism, including: the structures of the multiple visual detectors are the same.

5. The heat-resistant steel bar impurity identification system using a feedforward neural network model as claimed in claim 4, characterized in that: The multiple area measurement units are multiple visual detectors, which are used to be connected to the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism respectively to complete the respective measurements of the current real-time area values ​​of the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism. It also includes: each of the multiple visual detectors includes a planar camera.

6. The heat-resistant steel bar impurity identification system using a feedforward neural network model as claimed in any one of claims 2 to 5, characterized in that: The system further comprises: An information storage device is respectively connected to the first analysis device, the second analysis device, the third analysis device and multiple area measurement units of the multi-stage optimization mechanism, and is used to perform instant storage of real-time status parameters of the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism.

7. The heat-resistant steel bar impurity identification system using a feedforward neural network model as claimed in claim 6, characterized in that: An information storage device is respectively connected to the first analysis device, the second analysis device, the third analysis device and multiple area measurement units of the multi-stage optimization mechanism, and is used to perform instant storage of real-time status parameters of the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism, including: the information storage device is a dynamic storage device.

8. The heat-resistant steel bar impurity identification system using a feedforward neural network model as claimed in claim 6, characterized in that: An information storage device is respectively connected to the first analysis device, the second analysis device, the third analysis device and multiple area measurement units of the multi-stage optimization mechanism, and is used to perform instant storage of real-time status parameters of the first analysis device, the second analysis device, the third analysis device and the multi-stage optimization mechanism, including: the information storage device is a TF memory card.

Images