Two-parameter optimized intelligent rating method for material pearlite spheroidization

Abstract

The invention relates to the technical field of pearlite spheroidizing rating, in particular to a two-parameter optimized intelligent rating method for material pearlite spheroidizing, which is characterized in that a convolutional neural network is used for feature extraction, key parts of pearlite spheroidizing images, which have great influence on rating results, can be highlighted in a targeted manner through the setting of convolution kernels, and the accuracy of the rating results is improved. And the noise part of the pearlite spheroidized image generating negative interference on the rating result is weakened or eliminated, so that the de-noising degree of the image is improved. And then training the XGBoost algorithm through the feature image, and obtaining the number of decision trees which have great influence on the precision of the XGBoost algorithm and the optimal value of the depth of the decision trees, thereby obtaining an optimal classification model. According to the method, two weak classifiers, namely the convolutional neural network and the XGBoost algorithm, are combined to form a strong classifier, and the under-fitting problem of the convolutional neural network is eliminated through the XGBoost algorithm, so that the rating accuracy is effectively improved.

Description

Technical Field

[0001] The invention relates to the technical field of pearlite spheroidization rating, in particular to a double-parameter optimized material pearlite spheroidization intelligent rating method. Background Art

[0002] The pearlite spheroidization phenomenon of steel is one of the common structural changes of steel after long-term service at high temperature. The thermal strength and mechanical properties of steel will gradually decrease with the increase of pearlite spheroidization degree. Spheroidization grade is an important evaluation index of equipment material spheroidization and is of great significance for evaluating equipment safety.

[0003] The work of pearlite spheroidization rating is generally done by experienced operators, who refer to relevant rating standards for manual rating. However, the manual rating of pearlite spheroidization requires extremely high knowledge and experience of the operator, and the rating process is not fully quantified. There are subjective cognitive errors in the operator's rating process, which will cause incorrect labeling of some data samples. Computer vision technology based on deep learning can realize automatic rating of pearlite spheroidization in pearlite spheroidization rating, adapt to the needs of multiple locations and multiple tasks at the same time, so computer vision technology is widely used in pearlite spheroidization rating. Patent CN114581719A discloses a specific application of computer vision technology in pearlite spheroidization rating, which is a pearlite spheroidization intelligent rating method based on noisy learning. This method can effectively reduce the negative impact of label noise and improve the efficiency of pearlite spheroidization rating. However, in the process of using the above-mentioned current technology, only a single deep learning model is used to process the denoised image data. Due to the influence of the deep learning model's own system error, underfitting will occur, resulting in a reduction in the accuracy of the rating, which cannot meet the requirements of the actual rating work. Therefore, a high-accuracy rating method is urgently needed. Summary of the invention

[0004] In order to avoid and overcome the technical problems existing in the prior art, the present invention provides a dual-parameter optimized material pearlite spheroidization intelligent rating method, which can effectively improve the accuracy of pearlite spheroidization rating.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A dual-parameter optimized material pearlite spheroidization intelligent rating method, the method comprising the following steps:

[0007] S1. Obtain pearlite spheroidization images of the material, and annotate the spheroidization grades of each pearlite spheroidization image according to the set standard, and store the annotated pearlite spheroidization images in the original data set;

[0008] S2, using a neural network to extract features from the pearlite spheroidization image in the original data set to obtain a corresponding feature image;

[0009] S3, input the obtained feature image into the XGBoost algorithm to perform spheroidization level classification; optimize the number and depth of decision trees in the XGBoost algorithm according to the accuracy of the classification results to obtain an optimal classification model that meets the classification accuracy;

[0010] S4. Inputting the pearlite spheroidization image without the spheroidization grade marked into the neural network for feature extraction, and inputting the feature image obtained after the feature extraction into the optimal classification model for classification to obtain the spheroidization grade corresponding to the pearlite spheroidization image data.

[0011] As a further solution of the present invention: the optimization process of the number of decision trees and the depth of decision trees is as follows:

[0012] S31, combining the number of decision trees N_estiamators and the depth of the decision tree Max_depth to form a combination parameter in the form of a binary, and randomly generating M combination parameters;

[0013] S32, start iterative optimization, substitute each combination parameter into the XGBoost algorithm in turn, and calculate the average classification accuracy of each combination parameter under all feature images;

[0014] Obtain the combination parameter with the highest current average classification accuracy, and use the combination parameter as the initial value of the optimal combination parameter in the iterative optimization process;

[0015] Obtain the average value of all current combination parameters, and use the average value as the initial value of the average value of the combination parameters in the iterative optimization process;

[0016] S33, calculate the search result factor A during the t-th iteration optimization t , search result factor A t The calculation formula is as follows:

[0017]

[0018] In the formula, A0 represents the initial value of the search result factor, which is randomly selected in each iterative optimization; e represents a natural constant; T max It represents the maximum number of iterations in the iterative optimization process; t represents the current number of iterations;

[0019] S34. According to A t The update direction of the combined parameters is determined based on the size of ; the specific judgment process is as follows:

[0020] When|A t When |≥1, the combined parameters are updated as follows:

[0021]

[0022] Where X(t) represents the combination parameter of the current iterative optimization, that is, the combination parameter of the t-th iterative optimization; X(t+1) represents the combination parameter of the t+1-th iterative optimization; X rand (t) represents the combination parameter randomly generated during the tth iteration optimization; X best (t) represents the combination parameter with the highest average classification accuracy among all the combination parameters obtained at the tth iteration optimization; r1, r2, r3, r4 and q are all random numbers; X m (t) represents the average value of the combined parameters of all the combined parameters obtained at the tth iterative optimization;

[0023] When 0.5≤|A t When |≤1, the combined parameters are updated as follows:

[0024] X(t+1)=ΔX(t)-A t |JX best (t)-X(t)|;

[0025] ΔX(t)=X best (txt);

[0026] Where ΔX(t) represents X best The difference between (t) and X(t); J represents a random number;

[0027] When|A t When |≤0.5, the combined parameters are updated as follows:

[0028] X(t+1)=X best (t)-A t |ΔX(t)|;

[0029] S35, when the number of iterative optimization reaches the maximum number of iterations T max When , the current combination parameters are output, and the XGBoost algorithm using the combination parameters is the optimal classification model.

[0030] As a further solution of the present invention: the specific process of obtaining the combined parameters in step S31 is as follows:

[0031] S311, since the combined parameter is a two-dimensional parameter, a search range is set in the corresponding two-dimensional space;

[0032] S312, randomly selecting N groups of combination parameter groups within the search range, each group of combination parameter groups includes K combination parameters, and the product of N and K is M;

[0033] S313, performing Latin hypercube sampling on each group of combined parameter groups, and obtaining a corresponding sample matrix;

[0034] S314, initializing each combination parameter using an initialization formula;

[0035] The initialization formula is as follows:

[0036] X n =L n +(U n -L n )×l n ;

[0037] Where, X n represents the nth group of combined parameters after initialization, n∈[1,N]; L n Indicates the lower limit of the value of the nth group of combined parameters before initialization; U n Indicates the upper limit of the value of the nth group of combined parameter groups before initialization; l n Represents the sample matrix formed by Latin hypercube sampling of the nth group of combined parameters before initialization.

[0038] As a further solution of the present invention: a pearlite spheroidized sample of the material is obtained, and a metallographic photograph of the pearlite spheroidized sample is taken, and the metallographic photograph is the pearlite spheroidized image.

[0039] As a further solution of the present invention: performing image data amplification operation on the pearlite spheroidization images in the original data set to increase the number of pearlite spheroidization images in the original data set.

[0040] As a further solution of the present invention: the image data augmentation operation includes image horizontal flipping, image vertical flipping, image cropping, image rotation, image translation, image blurring, image sharpening, and adding noise.

[0041] As a further solution of the present invention: the neural network used is a convolutional neural network, and the convolutional neural network is used to extract features of the pearlite spheroidization image to obtain a corresponding feature image.

[0042] As a further solution of the present invention: the value range of the search result factor initial value A0 is [-1, 1]; the value ranges of r1, r2, r3, r4 and q are all [0, 1]; the value range of J is [0, 2].

[0043] An electronic device includes a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are connected in sequence, the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the above-mentioned rating method.

[0044] A readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the above-mentioned rating method.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The present invention uses a convolutional neural network for feature extraction. Through the setting of the convolution kernel, it can specifically highlight the key parts of the pearlite spheroidization image that have a greater impact on the rating results, weaken or eliminate the noise parts of the pearlite spheroidization image that have a negative impact on the rating results, thereby improving the denoising degree of the image. Then, the XGBoost algorithm is trained through the feature image, and the optimal values ​​of the number of decision trees and the depth of the decision trees that have a greater impact on the accuracy of the XGBoost algorithm are obtained, thereby obtaining the optimal classification model. The present invention combines two "weak classifiers" of the convolutional neural network and the XGBoost algorithm to form a "strong classifier", and eliminates the underfitting problem of the convolutional neural network through the XGBoost algorithm, thereby effectively improving the accuracy of the rating. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is the main flow chart of the rating method in the present invention.

[0048] Figure 2 This is a metallographic diagram of pearlite spheroidization in an embodiment of the present invention.

[0049] Figure 3 It is a convolutional neural network for extracting features from metallographic images in the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] See also Figure 1 to Figure 3 In the embodiment of the present invention, a dual-parameter optimized material pearlite spheroidization intelligent rating method mainly includes the following contents:

[0052] 1. Get the original data.

[0053] A pearlite spheroidized sample of steel is obtained, and a metallographic photograph of the pearlite spheroidized sample is taken. The metallographic photograph constitutes a pearlite spheroidized image, and the spheroidization grade of each pearlite spheroidized image is annotated according to the set standards such as DL / T 773-2016 Spheroidization Rating Standard for 12Cr1MoV Steel for Thermal Power Plants and DLT674-1999 Pearlite Spheroidization Rating Standard for No. 20 Steel for Thermal Power Plants. The annotated pearlite spheroidized images are stored in the original data set.

[0054] The image data augmentation operation is performed on the pearlite spheroidization images in the original data set to increase the number of pearlite spheroidization images in the original data set. The image data augmentation operation includes image horizontal flipping, image vertical flipping, image cropping, image rotation, image translation, image blurring, image sharpening, and adding noise.

[0055] 2. Feature extraction.

[0056] The convolutional neural network in the neural network is used to extract features of the pearlite spheroidization image in the original data set. Through the setting of the convolution kernel, the key parts of the pearlite spheroidization image that have a greater impact on the rating results can be highlighted in a targeted manner, and the noise parts of the pearlite spheroidization image that have a negative impact on the rating results can be weakened or eliminated, thereby improving the denoising degree of the image to obtain the corresponding feature image.

[0057] 3. Establish the optimal classification model.

[0058] First, combine the number of decision trees N_estiamators and the depth of the decision tree Max_depth to form a combination parameter in the form of a binary. In this example, the initial value range of the number of decision trees N_estiamators is [0,300], and the initial value range of the depth of the decision tree Max_depth is [0,20]. Since the initial value of the combination parameter is a two-dimensional parameter, a search range is set in its corresponding two-dimensional space. Randomly select N groups of combination parameter groups within the search range, each group of combination parameter groups contains K combination parameters, and the product of N and K is M, that is, there are M combination parameters in total.

[0059] Then, Latin hypercube sampling is performed on each group of combination parameters, and the corresponding sample matrix is ​​obtained, and each combination parameter is initialized using the initialization formula.

[0060] Then, iterative optimization begins, and each initialized combination parameter is substituted into the XGBoost algorithm in turn, and the average classification accuracy of each combination parameter under all feature images is calculated. The combination parameter with the highest current average classification accuracy is obtained, and this combination parameter is used as the initial value of the optimal combination parameter in the iterative optimization process, and this combination parameter is used as the initial value of the combination parameter in the iterative optimization process. The average value of all current combination parameters is obtained, and this average value is used as the initial value of the average value of the combination parameter in the iterative optimization process.

[0061] Then calculate the search result factor A during the t-th iteration optimization t , and according to A t The size of is used to determine the update direction of the combined parameters.

[0062] The specific judgment process is as follows:

[0063] When|A t When |≥1, the combined parameters are updated as follows:

[0064]

[0065] When 0.5≤|A t When |≤1, the combined parameters are updated as follows:

[0066] X(t+1)=ΔX(t)-A t |JX best (t)-X(t)|;

[0067] ΔX(t)=X best (txt).

[0068] When|A t When |≤0.5, the combined parameters are updated as follows:

[0069] X(t+1)=X best (t)-A t |ΔX(t|.

[0070] When the number of iterative optimization reaches the maximum number of iterations T max When , the current combination parameters are output, and the XGBoost algorithm using the combination parameters is the optimal classification model.

[0071] 4. Rating.

[0072] The pearlite spheroidization image without the spheroidization grade marked is input into the neural network for feature extraction, and the feature image obtained after feature extraction is input into the optimal classification model for classification to obtain the spheroidization grade corresponding to the pearlite spheroidization image data. The accuracy rate of the intelligent spheroidization rating of the pearlite material of the present invention can reach 83%.

[0073] 5. Electronic equipment.

[0074] The electronic device can be the mobile device itself, or a stand-alone device independent of the mobile device, which can communicate with the mobile device to receive the collected input signals from the mobile device and send the selected target decision behavior to the mobile device. The electronic device includes one or more processors and a memory.

[0075] The processor may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0076] The memory may include one or more computer program products, and the computer program product 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) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may run the program instructions to implement a material pearlite spheroidization intelligent grading method according to various embodiments of the present invention described above.

[0077] In one example, the electronic device may further include: an input device and an output device, and these components are interconnected through a bus system and / or other forms of connection mechanisms. For example, the input device may include various devices such as an on-board diagnostic system (OBD), a camera, an industrial camera, etc. The input device may also include, for example, a keyboard, a mouse, etc. The output device may include, for example, a display, a speaker, a printer, and a communication network and a remote output device connected thereto, etc.

[0078] In addition, the electronic device may further include any other appropriate components according to specific application scenarios.

[0079] The present invention also includes a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the phased array nonlinear imaging crack length prediction method according to various embodiments of the present invention described in the above-mentioned pearlite spheroidization intelligent rating method section of this specification.

[0080] The computer program product may be written in any combination of one or more programming languages ​​to write program code for performing the operations of the embodiments of the present invention, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0081] 6. Storage media.

[0082] In addition, an embodiment of the present invention may also be a computer-readable storage medium having computer program instructions stored thereon, and when the computer program instructions are executed by a processor, the processor executes a material pearlite spheroidization intelligent grading method in this specification.

[0083] The computer readable storage medium may adopt any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may, for example, include but is not limited to an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0084] The basic principle of the present invention is described above in conjunction with specific embodiments. However, it should be pointed out that the advantages, strengths, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. must be possessed by each embodiment of the present invention. In addition, the specific details disclosed above are only for the purpose of illustration and facilitation of understanding, rather than limitation, and the above details do not limit the present invention to be implemented by adopting the above specific details.

[0085] The block diagrams of the devices, apparatuses, equipment, and systems involved in the present invention are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The word "such as" used here refers to the phrase "such as but not limited to", and can be used interchangeably with it.

[0086] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A dual-parameter optimized material pearlite spheroidization intelligent rating method, It is characterized in that The method comprises the following steps: S1, obtaining pearlite spheroidization images of the material, and marking the spheroidization grade of each pearlite spheroidization image according to the set standard, and storing the marked pearlite spheroidization images in the original data set; S2, using a neural network to extract features from the pearlite spheroidization image in the original data set to obtain a corresponding feature image; S3, input the obtained feature image into the XGBoost algorithm to perform spheroidization level classification; optimize the number and depth of decision trees in the XGBoost algorithm according to the accuracy of the classification results to obtain an optimal classification model that meets the classification accuracy; S4, inputting the pearlite spheroidization image without the spheroidization grade marked into the neural network for feature extraction, and inputting the feature image obtained after the feature extraction into the optimal classification model obtained in step S3 for classification, so as to obtain the spheroidization grade corresponding to the pearlite spheroidization image data.

2. According to the dual-parameter optimized material pearlite spheroidization intelligent rating method of claim 1, It is characterized in that The optimization process of the number of decision trees and the depth of decision trees is as follows: S31, combining the number of decision trees N_estiamators and the depth of the decision tree Max_depth to form a combination parameter in the form of a binary, and randomly generating M combination parameters; S32, start iterative optimization, substitute each combination parameter into the XGBoost algorithm in turn, and calculate the average classification accuracy of each combination parameter under all feature images; Obtain the combination parameter with the highest current average classification accuracy, and use the combination parameter as the initial value of the optimal combination parameter in the iterative optimization process; Obtain the average value of all current combination parameters, and use the average value as the initial value of the average value of the combination parameters in the iterative optimization process; S33, calculate the search result factor A during the t-th iteration optimization t , search result factor A t The calculation formula is as follows: In the formula, A 0 represents the initial value of the search result factor, which is randomly selected in each iterative optimization; e represents a natural constant; T max Indicates the maximum number of iterations in the iterative optimization process; t represents the current iteration number; S34. According to A t The size of is used to determine the update direction of the combined parameters. The specific determination process is as follows: When|A t When |≥1, the combined parameters are updated as follows: Where X(t) represents the combination parameter of the current iterative optimization, that is, the combination parameter of the t-th iterative optimization; X(t+1) represents the combination parameter of the t+1-th iterative optimization; X rand (t) represents the combination parameter randomly generated during the tth iteration optimization; X best (t) represents the combination parameter with the highest average classification accuracy among all the combination parameters obtained at the tth iteration optimization; r 1 、r 2 、r 3 、r 4 and q are both random numbers; X m (t) represents the average value of the combined parameters of all the combined parameters obtained at the tth iterative optimization; When 0.5≤|A t When |≤1, the combined parameters are updated as follows: X(t+1)=ΔX(t)-A t |JX best (t)-X(t)|; ΔX(t)=X best (t)-X(t); Where ΔX(t) is used for the beam X best The difference between (t) and X(t); J represents a random number; When|A t When |≤0.5, the combined parameters are updated as follows: X(t+1)=X best (t)-A t |ΔX(t)|; S35, when the number of iterative optimization reaches the maximum number of iterations T max When , the current combination parameters are output, and the XGBoost algorithm using the combination parameters is the optimal classification model.

3. A dual-parameter optimized material pearlite spheroidization intelligent grading method according to claim 2, It is characterized in that The specific process of obtaining the combined parameters in step S31 is as follows: S311, since the combined parameter is a two-dimensional parameter, a search range is set in the corresponding two-dimensional space; S312, randomly selecting N groups of combination parameter groups within the search range, each group of combination parameter groups includes K combination parameters, and the product of N and K is M; S313, performing Latin hypercube sampling on each group of combined parameter groups, and obtaining a corresponding sample matrix; S314, initializing each combination parameter using an initialization formula; The initialization formula is as follows: X n =L n +(U n -L n )×l n ; Where, X n represents the nth group of combined parameters after initialization, n∈[1,N]; L n Indicates the lower limit of the value of the nth group of combined parameters before initialization; U n Indicates the upper limit of the value of the nth group of combined parameter groups before initialization; l n Represents the sample matrix formed by Latin hypercube sampling of the nth group of combined parameters before initialization.

4. A dual-parameter optimized material pearlite spheroidization intelligent grading method according to claim 3, It is characterized in that A pearlite spheroidized sample of the material is obtained, and a metallographic photograph of the pearlite spheroidized sample is taken. The metallographic photograph is the pearlite spheroidized image.

5. A dual-parameter optimized material pearlite spheroidization intelligent grading method according to claim 4, It is characterized in that An image data amplification operation is performed on the pearlite spheroidization images in the original data set to increase the number of pearlite spheroidization images in the original data set.

6. A dual-parameter optimized material pearlite spheroidization intelligent grading method according to claim 5, It is characterized in that Image data augmentation operations include image horizontal flipping, image vertical flipping, image cropping, image rotation, image translation, image blurring, image sharpening, and adding noise.

7. A dual-parameter optimized material pearlite spheroidization intelligent grading method according to claim 6, It is characterized in that The neural network used is a convolutional neural network, which is used to extract features of the pearlite spheroidization image to obtain the corresponding feature image.

8. A dual-parameter optimized material pearlite spheroidization intelligent grading method according to claim 7, It is characterized in that Search result factor initial value A 0 The value range of r is [-1,1]; 1 、r 2 、r 3 、r 4 The value ranges of and q are both [0,1]; the value range of J is [0,2].

9. An electronic device, It is characterized in that The method comprises a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are connected in sequence, the memory is used to store a computer program, the computer program comprises program instructions, and the processor is configured to call the program instructions to execute a material pearlite spheroidization intelligent grading method as described in any one of claims 1 to 8.

10. A readable storage medium, It is characterized in that The storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor, the processor executes the material pearlite spheroidization intelligent grading method as described in any one of claims 1 to 8.

Images