CN115081536A

CN115081536A - Heat treatment device and heat treatment method for hardness hardening and tempering of hardware product workpiece

Info

Publication number: CN115081536A
Application number: CN202210800455.6A
Authority: CN
Inventors: 高巧英; 高星; 高玮鸿
Original assignee: Xuzhou Rur Tools Making Co ltd
Current assignee: Xuzhou Rur Tools Making Co ltd
Priority date: 2022-07-08
Filing date: 2022-07-08
Publication date: 2022-09-20

Abstract

The application relates to the field of intelligent manufacturing, and specifically discloses a hardware product workpiece hardness tempering heat treatment device and a heat treatment method thereof, wherein local high-dimensional correlation characteristic distribution of a thermal infrared image of a hardware product workpiece is extracted through a convolutional neural network model, implicit correlation characteristic information of temperature measurement values of all positions of the workpiece is further fused with characteristic information of the local high-dimensional correlation characteristic distribution and the implicit correlation characteristic information of the temperature measurement values of all positions of the workpiece, and the characteristic information of the local high-dimensional correlation characteristic distribution and the implicit correlation characteristic information of all positions of the workpiece are further fused to judge the distribution rationality of the temperature of the hardware product workpiece to be detected, and when the characteristics are fused, the temperature measurement characteristic matrix is further modified to introduce robustness through the minimum loss around the temperature local characteristic information, so that the characteristic local equivalent clustering performance of the whole characteristics is promoted, and further the joint characterization capability of the temperature measurement characteristic matrix on the global temperature characteristics is promoted. Thus, the classification accuracy can be improved, and the hardness tempering performance can be ensured.

Description

Heat treatment device and heat treatment method for hardness hardening and tempering of hardware product workpiece

Technical Field

The application relates to the field of intelligent manufacturing, in particular to a hardware product workpiece hardness tempering heat treatment device and a heat treatment method thereof.

Background

The hardware is a tool obtained by processing and casting metals such as gold, silver, copper, iron, tin and the like, and has the functions of fixing objects, processing objects, decorating and the like. In the process, hardness tempering is one of the key processing procedures, and the hardness tempering is a double heat treatment method of quenching and high-temperature tempering, so that the hardness of the hardware product workpiece can meet the due requirement, and the processed product workpiece has good comprehensive mechanical properties.

In the heat treatment process of hardening and tempering hardware products, the key of the hardening and tempering process is to keep the workpiece to be processed heated as uniformly as possible. During current temperature regulation and control scheme, detect the temperature that the work piece heated through temperature sensor, when the temperature is inhomogeneous, come to carry hot-blast to corresponding position through the hot-blast main. However, this way of regulation ignores a fundamental fact: if the workpiece itself as a whole has temperature deviations among its respective portions, heat flows among the respective portions of the workpiece itself, that is, the respective portions of the workpiece cannot be regarded as the portions that are not related to each other when temperature regulation is performed. However, in the conventional method, hot air is passed through a position of a workpiece having a relatively low temperature, and the hot air passed through the position is often a high-temperature portion, which results in a decrease in hardness hardening and tempering performance.

Therefore, a heat treatment device for hardness tempering of hardware products is expected to ensure uniformity of heated temperature distribution during hardware product processing, and further ensure hardness tempering performance of the hardware products.

Disclosure of Invention

The present application is proposed to solve the above-mentioned technical problems. The embodiment of the application provides a hardware product workpiece hardness tempering's heat treatment device and heat treatment method thereof, it extracts the local high dimension correlation characteristic distribution of the hot infrared image of hardware product workpiece through convolution neural network model, and the implicit correlation characteristic information of the temperature measurement value of each position of work piece, and then fuse these characteristic information between them and come right the distribution rationality of hardware product workpiece temperature is treated to detect judges, and when the characteristic is fused, it is further right temperature measurement characteristic matrix revises to introduce the robustness through the minimizing loss around temperature local characteristic information, promote the characteristic local and equivalent to the holistic clustering performance of characteristic, and then promoted temperature measurement characteristic matrix is to the joint representation ability of global temperature characteristic. Therefore, the classification accuracy can be improved, and the hardness tempering performance can be ensured.

According to one aspect of the application, a hardware product workpiece hardness quenching and tempering heat treatment device is provided, which comprises: a training module comprising: the hardware product detection device comprises a first training data unit, a second training data unit and a third training data unit, wherein the first training data unit is used for acquiring temperature data of a plurality of positions of a hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; the temperature characteristic extraction unit is used for constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors and then obtaining a temperature measurement characteristic matrix through a first convolution neural network; the temperature characteristic correction unit is used for correcting the characteristic value of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix; the second training data unit is used for acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by the infrared camera; the thermal infrared coding unit is used for enabling the thermal infrared image to pass through a second convolutional neural network so as to obtain a thermal infrared temperature distribution characteristic matrix; the characteristic matrix fusion unit is used for fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; the loss calculation unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification loss function value; a training unit for training the first convolutional neural network, the second convolutional neural network, and the classifier based on the classification loss function value; and an inference module comprising: the temperature data acquisition unit is used for acquiring temperature data of a plurality of positions of the hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; the temperature coding unit is used for constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors and then obtaining a temperature measurement characteristic matrix through the first convolution neural network trained and completed by the training module; the thermal infrared image acquisition unit is used for acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; the thermal infrared image coding unit is used for enabling the thermal infrared image to pass through the second convolutional neural network which is trained in the training stage so as to obtain a thermal infrared temperature distribution characteristic matrix; the characteristic fusion unit is used for fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; and the preprocessing evaluation result generating unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, and the classification result is used for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

According to another aspect of the application, a heat treatment method of a heat treatment device for quenching and tempering the hardness of a hardware product workpiece is provided, and comprises the following steps: a training phase comprising: acquiring temperature data of a plurality of positions of a hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors, and then obtaining a temperature measurement characteristic matrix through a first convolution neural network; correcting the characteristic value of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix; acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; enabling the thermal infrared image to pass through a second convolution neural network to obtain a thermal infrared temperature distribution characteristic matrix; fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; passing the classification characteristic matrix through a classifier to obtain a classification loss function value; training the first convolutional neural network, the second convolutional neural network, and the classifier based on the classification loss function values; and an inference phase comprising: acquiring temperature data of a plurality of positions of a hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors, and then obtaining a temperature measurement characteristic matrix through the first convolution neural network trained and completed by a training module; acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; passing the thermal infrared image through the second convolutional neural network trained in the training stage to obtain a thermal infrared temperature distribution characteristic matrix; fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; and enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

Compared with the prior art, the utility model provides a hardware product work piece hardness quenching and tempering heat treatment device and heat treatment method thereof, it extracts the local high dimension correlation characteristic distribution of the hot infrared image of hardware product work piece through convolution neural network model, and the implicit correlation characteristic information of the temperature measurement value of each position of work piece, and then fuse these characteristic information between them and come right the distribution rationality of waiting to detect hardware product work piece temperature judges, and when the characteristic fuses, it is further right temperature measurement characteristic matrix revises to introduce the robustness through the minimizing loss around temperature local characteristic information, promote the characteristic local clustering performance equivalent to the characteristic is holistic, and then promoted temperature measurement characteristic matrix is to the joint representation ability of global temperature characteristic. Therefore, the classification accuracy can be improved, and the hardness tempering performance can be ensured.

Drawings

The above and other objects, features and advantages of the present application will become more apparent by describing in more detail embodiments of the present application with reference to the attached drawings. The accompanying drawings are included to provide a further understanding of the embodiments of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings, like reference numbers generally represent like parts or steps.

FIG. 1 is a schematic view of a heat treatment device for hardness tempering of a hardware product workpiece according to an embodiment of the application.

FIG. 2 is a block diagram of a heat treatment device for hardness tempering of a hardware product workpiece according to an embodiment of the present application.

FIG. 3A is a flowchart of a training phase in the heat treatment method of the heat treatment device for hardness tempering of the hardware product workpiece according to the embodiment of the application.

FIG. 3B is a flow chart of the inference stage in the heat treatment method of the heat treatment device for hardness tempering of the hardware product workpiece according to the embodiment of the application.

FIG. 4 is a schematic diagram of the architecture of the training phase in the heat treatment method of the heat treatment apparatus for hardness tempering of hardware products according to the embodiment of the present application.

Fig. 5 is a schematic configuration diagram of an inference stage in a heat treatment method of a heat treatment apparatus for hardness tempering of a hardware product workpiece according to an embodiment of the present application.

Detailed Description

Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be understood that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and that the present application is not limited by the example embodiments described herein.

Scene overview: as mentioned above, in the heat treatment process of hardness tempering of hardware workpiece, it is the key of the hardness tempering process to keep the workpiece to be processed heated as uniformly as possible. During current temperature regulation and control scheme, detect the temperature that the work piece heated through temperature sensor, when the temperature is inhomogeneous, come to carry hot-blast to corresponding position through the hot-blast main. However, this way of regulation ignores a fundamental fact: if the workpiece itself as a whole has temperature deviations among its respective portions, heat flows among the respective portions of the workpiece itself, that is, the respective portions of the workpiece cannot be regarded as the portions that are not related to each other when temperature regulation is performed. However, in the conventional method, hot air is passed through a position of a workpiece having a relatively low temperature, and the hot air passed through the position is often a high-temperature portion, which results in a decrease in hardness hardening and tempering performance.

Based on this, in the technical scheme of this application, the temperature sensor who disposes in each position of waiting to detect the work piece comes the representation the heating temperature value of waiting to detect work piece each position to through the thermal infrared image of waiting to detect the work piece comes the heat flow that the representation work piece takes place, in order to assist to detect the homogeneity of being heated of work piece accurately judges, can make the hardness quenching and tempering effect of hardware products work piece better like this.

Specifically, in the technical scheme of the application, the temperature data of a plurality of positions of the hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological style, are firstly acquired. It should be understood that, considering that the convolutional neural network has excellent performance in extracting the implicit associated features of the data, the convolutional neural network can be used to extract the high-dimensional implicit associated features of the temperature data of the plurality of positions of the workpiece to be detected.

Specifically, the temperature data of the positions of the hardware product workpiece to be detected are constructed into a temperature matrix according to the preset topological patterns of the temperature sensors. In the embodiment of the present application, a matrix may be constructed based on the preset topological pattern of the plurality of temperature sensors, and it should be understood that the plurality of sensors are not necessarily disposed in a strict matrix form, and therefore, when the matrix is constructed based on the preset topological pattern, a canonical matrix is required to be constructed, for example, M times N, that is, M rows, each row having N positions (ideally, corresponding to N sensors). Then, filling each temperature data into the corresponding matrix position to obtain a temperature initial matrix. However, it is contemplated that there may be some locations in the temperature initiation matrix where there is no temperature dataAnd also considering that the temperature data is gradual and continuous, therefore, the neighborhood-based eigenvalue padding is further performed on the positions of the temperature initial matrix not filled with temperature data. For example, M in the matrix ₁₁ Is T1, M ₁₃ At T2, then M is vacant ₁₁ The complement is (T1+ T2)/2.

In this way, the constructed temperature matrix is processed through the first convolutional neural network to extract local implicit topological correlation characteristic information of the temperature data of the plurality of positions, so that a temperature measurement characteristic matrix is obtained.

And then, acquiring a thermal infrared image of the hardware product workpiece to be detected through an infrared camera, and performing feature extraction on the thermal infrared image through a second convolutional neural network to extract local high-dimensional implicit feature information of the thermal infrared image of the hardware product workpiece to be detected, so as to obtain a thermal infrared temperature distribution feature matrix. In this way, the temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix are subjected to matrix fusion and then classified, so that a classification result for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not can be obtained.

However, considering that the thermal infrared temperature distribution feature matrix obtained from the thermal infrared image performs feature extraction of the workpiece as an organic thermal whole, and the temperature measurement feature matrix is obtained by feature extraction of a temperature matrix in which the temperatures of a plurality of sensors are arranged, which is associated with the local temperatures obtained by each sensor, it is desirable that the temperature measurement feature matrix can also represent a global temperature expression to some extent before being fused with the thermal infrared temperature distribution feature matrix.

Based on this, the temperature measurement feature matrix is corrected:

in this way, robustness can be introduced by the minimized loss of the temperature local feature information to improve the clustering performance of the feature local equivalent to the feature whole, so that in the iterative training process through back propagation, the dependency divergence on the global expected feature caused by parameter adjustment is reduced, and the joint characterization capability of the temperature measurement feature matrix on the global temperature feature is improved. Furthermore, the accuracy of classification is improved, so that the hardness tempering performance is ensured.

Based on this, this application has proposed a hardware product work piece hardness quenching and tempering's heat treatment device, and it includes training module and inference module. Wherein, the training module includes: the hardware product detection device comprises a first training data unit, a second training data unit and a third training data unit, wherein the first training data unit is used for acquiring temperature data of a plurality of positions of a hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; the temperature characteristic extraction unit is used for constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors and then obtaining a temperature measurement characteristic matrix through a first convolution neural network; the temperature characteristic correction unit is used for correcting the characteristic value of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix; the second training data unit is used for acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by the infrared camera; the thermal infrared coding unit is used for enabling the thermal infrared image to pass through a second convolutional neural network so as to obtain a thermal infrared temperature distribution characteristic matrix; the characteristic matrix fusion unit is used for fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; the loss calculation unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification loss function value; a training unit to train the first convolutional neural network, the second convolutional neural network, and the classifier based on the classification loss function value. Wherein, the inference module comprises: the temperature data acquisition unit is used for acquiring temperature data of a plurality of positions of the hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; the temperature coding unit is used for constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors and then obtaining a temperature measurement characteristic matrix through the first convolution neural network trained and completed by the training module; the thermal infrared image acquisition unit is used for acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; the thermal infrared image coding unit is used for enabling the thermal infrared image to pass through the second convolutional neural network which is trained in the training stage so as to obtain a thermal infrared temperature distribution characteristic matrix; the characteristic fusion unit is used for fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; and the preprocessing evaluation result generating unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, and the classification result is used for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

FIG. 1 is a schematic view illustrating a scene of a heat treatment device for hardness tempering of a hardware product workpiece according to an embodiment of the present application. As shown in fig. 1, in the training phase of the application scenario, first, temperature data of a plurality of positions of a hardware product workpiece to be detected (e.g., P as illustrated in fig. 1) is acquired by a plurality of temperature sensors (e.g., T as illustrated in fig. 1) deployed in a preset topological pattern, and a thermal infrared image of the hardware product workpiece to be detected is acquired by an infrared camera (e.g., C as illustrated in fig. 1). Then, the obtained temperature data of the plurality of positions and the thermal infrared image of the hardware product workpiece to be detected are input into a server (for example, S as illustrated in fig. 1) deployed with a hardware product workpiece hardness-tempering heat treatment algorithm, wherein the server can train the first convolutional neural network, the second convolutional neural network and the classifier of the hardware product workpiece hardness-tempering heat treatment device based on the hardware product workpiece hardness-tempering heat treatment algorithm and the temperature data of the plurality of positions and the thermal infrared image of the hardware product workpiece to be detected.

After the training is completed, in the inference phase, first, temperature data of a plurality of positions of a hardware product workpiece to be detected (for example, P as illustrated in fig. 1) is acquired by a plurality of temperature sensors (for example, T as illustrated in fig. 1) deployed in a preset topological pattern, and a thermal infrared image of the hardware product workpiece to be detected is acquired by an infrared camera (for example, C as illustrated in fig. 1). Then, the temperature data of the multiple positions and the thermal infrared image of the hardware product workpiece to be detected are input into a server (for example, S as illustrated in fig. 1) deployed with a hardware product workpiece hardness tempering heat treatment algorithm, wherein the server can process the temperature data of the multiple positions and the thermal infrared image of the hardware product workpiece to be detected by the hardware product workpiece hardness tempering heat treatment algorithm to generate a classification result for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

Having described the general principles of the present application, various non-limiting embodiments of the present application will now be described with reference to the accompanying drawings.

An exemplary system: FIG. 2 illustrates a block diagram of a heat treatment apparatus for hardness tempering of a hardware product workpiece according to an embodiment of the present application. As shown in fig. 2, the heat treatment apparatus 200 for hardening and tempering the hardness of the hardware product workpiece according to the embodiment of the present application includes: a training module 210 and an inference module 220. Wherein, the training module 210 includes: the first training data unit 2101 is used for acquiring temperature data of a plurality of positions of a hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological style; the temperature characteristic extraction unit 2102 is used for constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors and then obtaining a temperature measurement characteristic matrix through a first convolution neural network; a temperature characteristic correction unit 2103, configured to correct a characteristic value of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix; the second training data unit 2104 is used for acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by the infrared camera; the thermal infrared encoding unit 2105 is used for enabling the thermal infrared image to pass through a second convolutional neural network so as to obtain a thermal infrared temperature distribution characteristic matrix; a feature matrix fusion unit 2106, configured to fuse the corrected temperature measurement feature matrix and the thermal infrared temperature distribution feature matrix to obtain a classification feature matrix; a loss calculating unit 2107, configured to pass the classification feature matrix through a classifier to obtain a classification loss function value; a training unit 2108 for training the first convolutional neural network, the second convolutional neural network and the classifier based on the classification loss function value. The inference module 220 includes: the temperature data acquisition unit 221 is used for acquiring temperature data of a plurality of positions of the hardware product workpiece to be detected, acquired by a plurality of temperature sensors deployed in a preset topological pattern; the temperature coding unit 222 is configured to construct temperature data of multiple positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the multiple temperature sensors, and then obtain a temperature measurement characteristic matrix through the first convolutional neural network trained by the training module; the thermal infrared image acquisition unit 223 is used for correcting the characteristic values of all positions in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix; the thermal infrared image acquisition unit 224 is used for acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; a thermal infrared image encoding unit 225, configured to pass the thermal infrared image through the second convolutional neural network trained in the training stage to obtain a thermal infrared temperature distribution feature matrix; a feature fusion unit 226, configured to fuse the corrected temperature measurement feature matrix and the thermal infrared temperature distribution feature matrix to obtain a classification feature matrix; and a preprocessing evaluation result generating unit 227, configured to pass the classification feature matrix through a classifier to obtain a classification result, where the classification result is used to indicate whether the temperature distribution of the hardware product workpiece to be detected is reasonable.

Specifically, in the embodiment of the present application, in the training module 210, the first training data unit 2101 and the temperature feature extraction unit 2102 are configured to acquire temperature data of a plurality of positions of a hardware product workpiece to be detected, which is acquired by a plurality of temperature sensors deployed in a preset topological pattern, construct the temperature data of the plurality of positions of the hardware product workpiece to be detected as a temperature matrix according to the preset topological pattern of the plurality of temperature sensors, and then obtain a temperature measurement feature matrix through a first convolutional neural network. As mentioned above, in the heat treatment process of hardness tempering of hardware workpiece, it is the key of the hardness tempering process to keep the workpiece to be processed heated as uniformly as possible. Therefore, in the technical scheme of the application, in order to better obtain the global temperature information of the workpiece to be machined, the temperature sensors arranged at the positions of the workpiece to be machined are selected to represent the heating temperature values of the positions of the workpiece to be machined, and the thermal flow information generated by the workpiece is represented by the thermal infrared image of the workpiece to be machined, so that the heating uniformity of the workpiece to be machined can be accurately judged, and the hardness tempering effect of the hardware product workpiece can be better.

Specifically, in the technical scheme of the application, the temperature data of a plurality of positions of the hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern, are firstly acquired. It should be understood that, considering that the convolutional neural network has an excellent performance in extracting the implicit associated features of the data, the convolutional neural network may be used to extract the high-dimensional implicit associated features of the temperature data of the plurality of positions of the workpiece to be detected. Specifically, the temperature data of the positions of the hardware product workpiece to be detected are constructed into a temperature matrix according to the preset topological patterns of the temperature sensors. And then, processing the temperature matrix in a first convolution neural network to extract high-dimensional implicit topological correlation characteristic information of the temperature data of the plurality of positions, so as to obtain a temperature measurement characteristic matrix.

More specifically, in an embodiment of the present application, the temperature characteristic extraction unit includes: and the matrix constructing subunit is used for constructing a matrix based on the preset topological patterns of the plurality of temperature sensors. It should be understood that the plurality of sensors need not be deployed in a strict matrix, and therefore, when matrix construction is based on a predetermined topological pattern,a canonical matrix is constructed, for example, M by N, i.e., M rows, each having N positions (ideally corresponding to N sensors). And the temperature value filling subunit is used for respectively filling the temperature data of the positions of the hardware product workpiece to be detected into the corresponding positions of the matrix to obtain a temperature initial matrix. And the temperature value supplementing subunit is used for performing neighborhood-based characteristic value supplementing on the position which is not filled with the temperature data in the temperature initial matrix to obtain the temperature matrix. It should be appreciated that, given that there may be some locations in the temperature initiation matrix that may not have temperature data, and also given that the temperature data is gradual and continuous, the neighborhood-based eigenvalue padding is further performed for locations in the temperature initiation matrix that are not populated with temperature data. For example, in one specific example, M in the temperature initiation matrix ₁₁ Is T1, M ₁₃ At T2, then M is vacant ₁₁ The complement is (T1+ T2)/2.

In this way, the constructed temperature matrix is processed through the first convolutional neural network to extract high-dimensional implicit topological correlation characteristic information of the temperature data of the plurality of positions, so that a temperature measurement characteristic matrix is obtained. Accordingly, in one particular example, the layers of the first convolutional neural network each perform in a layer forward pass: performing convolution processing on input data to obtain a convolution characteristic diagram; performing mean pooling on the convolution feature map based on local channel dimensions to obtain a pooled feature map; and performing nonlinear activation on the pooled feature map to obtain an activated feature map; wherein the output of the last layer of the first convolutional neural network is the temperature measurement characteristic matrix, and the input of the first layer of the first convolutional neural network is the temperature matrix.

Specifically, in the embodiment of the present application, in the training module 210, the temperature characteristic correction unit 2103 is configured to correct the characteristic value of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix. It should be understood that, in the technical solution of the present application, it is desirable that the temperature measurement feature matrix is capable of representing a global temperature expression to some extent before being fused with the thermal infrared temperature distribution feature matrix, considering that the thermal infrared temperature distribution feature matrix obtained by the thermal infrared image subsequently performs feature extraction on the workpiece as an organic thermal whole, and the temperature measurement feature matrix is obtained by feature extraction from a temperature matrix after temperature arrangement of a plurality of sensors, which is associated feature extraction of local temperatures obtained by each of the sensors. Therefore, in the technical solution of the present application, the temperature measurement feature matrix needs to be further corrected.

More specifically, in an embodiment of the present application, the temperature characteristic correction unit includes: and the local temperature characteristic representation subunit is used for calculating a logarithmic function value of a sum value of the eigenvalue and one of each position in the temperature measurement characteristic matrix as the local temperature characteristic representation of each temperature in the temperature measurement characteristic matrix. And the global temperature characteristic representation subunit is used for calculating a logarithmic function value of a summation value of the summation values of the characteristic values of all the positions in the temperature measurement characteristic matrix and a summation value of one as the global temperature characteristic representation of the temperature measurement characteristic matrix. And the temperature characteristic representation adjusting subunit is used for dividing the local temperature characteristic representation of each position in the temperature measurement characteristic matrix by the global temperature characteristic representation of the temperature measurement characteristic matrix to obtain a corrected characteristic diagram of each position in the temperature measurement characteristic matrix so as to obtain the corrected temperature measurement characteristic matrix. That is, in one specific example, the correction formula is expressed as:

wherein,

a matrix of characteristics representing said temperature measurements is represented,

representing the corrected temperature measurement feature matrix. Should be takenIt can be understood that robustness can be introduced by the minimized loss of the temperature local feature information to improve the clustering performance of the feature local equivalent to the feature whole, so that in the iterative training process through back propagation, the dependency divergence on the global expected feature caused by parameter adjustment is reduced, and the joint characterization capability of the temperature measurement feature matrix on the global temperature feature is improved.

Specifically, in this embodiment of the application, in the training module 210, the second training data unit 2104, the thermal infrared encoding unit 2105 and the feature matrix fusion unit 2106 are configured to acquire a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera, and pass the thermal infrared image through a second convolutional neural network to obtain a thermal infrared temperature distribution feature matrix, and then fuse the corrected temperature measurement feature matrix and the thermal infrared temperature distribution feature matrix to obtain a classification feature matrix. That is, in the technical scheme of this application, further still need gather through infrared camera wait to detect the hot infrared image of hardware products work piece, and will hot infrared image carries out the feature extraction through the second convolution neural network in order to extract detect the local high-dimensional implicit characteristic information of the hot infrared image of hardware products work piece to obtain the hot infrared temperature distribution characteristic matrix. Accordingly, in one particular example, the layers of the second convolutional neural network each perform in a forward pass of the layers: performing convolution processing on input data to obtain a convolution characteristic diagram; performing mean pooling based on local channel dimensions on the convolution feature map to obtain a pooled feature map; and performing nonlinear activation on the pooled feature map to obtain an activated feature map; and the output of the last layer of the second convolutional neural network is the thermal infrared temperature distribution characteristic matrix, and the input of the first layer of the second convolutional neural network is the thermal infrared image. Then, the corrected temperature measurement feature matrix and the thermal infrared temperature distribution feature matrix can be fused to obtain a classification feature matrix with global temperature feature information.

More particularly, in the present applicationIn an embodiment, the feature matrix fusion unit is further configured to: fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix according to the following formula to obtain the classification characteristic matrix; wherein the formula is:

wherein,

for the purpose of the classification feature matrix,

for the corrected temperature measurement feature matrix,

is a characteristic matrix of the thermal infrared temperature distribution "

"represents the addition of elements at the corresponding positions of the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix,

is a weighting parameter for controlling a balance between the corrected temperature measurement feature matrix and the thermal infrared temperature distribution feature matrix in the classification feature matrix.

Specifically, in the embodiment of the present application, in the training module 210, the loss calculating unit 2107 and the training unit 2108 are configured to pass the classification feature matrix through a classifier to obtain a classification loss function value, and train the first convolutional neural network, the second convolutional neural network, and the classifier based on the classification loss function value. That is, in the technical solution of the present application, further, the classification feature matrix with the global temperature feature information is passed through a classifier to obtain a classification loss function value for training. Then, the first convolutional neural network, the second convolutional neural network and the classifier are trained in a back propagation iterative mode based on the classification loss function values, so that dependency divergence on global expected features caused by parameter adjustment is reduced, and the joint characterization capability of the global temperature features is improved. Furthermore, the classification accuracy is improved, and the hardness tempering performance is ensured.

More specifically, in this embodiment of the application, the loss calculating unit is further configured to: the loss calculation unit is further configured to: the classifier processes the classification feature matrix to generate a classification result according to the following formula, wherein the formula is as follows:

wherein

Representing the projection of the classification feature matrix as a vector,

to

Is a weight matrix of the fully connected layers of each layer,

to is that

A bias matrix representing the layers of the fully-connected layer; and calculating a cross entropy value between the classification result and a real value as the classification loss function value.

Specifically, in the embodiment of the present application, in the inference module 220, first, temperature data of a plurality of positions of the hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern, are acquired. And then, constructing the temperature data of the positions of the hardware product workpiece to be detected into a temperature matrix according to the preset topological style of the temperature sensors, and then obtaining a temperature measurement characteristic matrix through the first convolution neural network trained and completed by a training module. And then, the thermal infrared image acquisition unit is used for acquiring the thermal infrared image of the hardware product workpiece to be detected, which is acquired by the infrared camera. Then, the thermal infrared image passes through the second convolutional neural network which is trained in the training stage to obtain a thermal infrared temperature distribution characteristic matrix. And then, fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix. And finally, passing the classification characteristic matrix through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

In summary, the heat treatment device 200 for hardware product workpiece hardness tempering based on the embodiment of the present application is set forth, and extracts the local high-dimensional correlation feature distribution of the thermal infrared image of the hardware product workpiece and the implicit correlation feature information of the temperature measurement value of each position of the workpiece through a convolutional neural network model, so as to fuse the feature information of the local high-dimensional correlation feature distribution and the implicit correlation feature information of the temperature measurement value of each position of the workpiece, judge the distribution rationality of the temperature of the hardware product workpiece to be detected, and further correct the temperature measurement feature matrix during feature fusion, so as to introduce robustness through the minimized loss around the temperature local feature information, improve the clustering performance of the feature local equivalent to the feature whole, and further improve the joint characterization capability of the temperature measurement feature matrix to the global temperature feature. Therefore, the classification accuracy can be improved, and the hardness tempering performance can be ensured.

As described above, the heat treatment apparatus 200 for hardness-tempering of hardware products according to the embodiment of the present application can be implemented in various terminal devices, such as a server for heat treatment of hardness-tempering of hardware products. In one example, the heat treatment apparatus 200 for hardness refining of hardware product workpiece according to the embodiment of the present application may be integrated into a terminal device as a software module and/or a hardware module. For example, the heat treatment apparatus 200 for hardening and tempering the hardware product workpiece may be a software module in the operating system of the terminal device, or may be an application program developed for the terminal device; of course, the heat treatment apparatus 200 for hardening and tempering the hardware workpiece may also be one of many hardware modules of the terminal equipment.

Alternatively, in another example, the hardware product workpiece hardness-adjusted heat treatment apparatus 200 and the terminal device may be separate devices, and the hardware product workpiece hardness-adjusted heat treatment apparatus 200 may be connected to the terminal device through a wired and/or wireless network and transmit the interaction information according to the agreed data format.

An exemplary method: FIG. 3A illustrates a flow chart of a training phase in a heat treatment method of a heat treatment apparatus for hardness tempering a hardware product workpiece according to an embodiment of the present application. As shown in fig. 3A, the heat treatment method of the heat treatment device for hardness tempering of the hardware product workpiece according to the embodiment of the application comprises the following steps: a training phase comprising the steps of: s110, acquiring temperature data of a plurality of positions of the hardware product workpiece to be detected, wherein the temperature data are acquired by a plurality of temperature sensors deployed in a preset topological pattern; s120, constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to the preset topological style of the plurality of temperature sensors, and then obtaining a temperature measurement characteristic matrix through a first convolution neural network; s130, correcting the characteristic value of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix; s140, acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; s150, enabling the thermal infrared image to pass through a second convolutional neural network to obtain a thermal infrared temperature distribution characteristic matrix; s160, fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; s170, enabling the classification characteristic matrix to pass through a classifier to obtain a classification loss function value; s180, training the first convolutional neural network, the second convolutional neural network and the classifier based on the classification loss function value.

FIG. 3B illustrates a flow chart of an inference stage in a heat treatment method of a heat treatment apparatus for hardness tempering of a hardware product workpiece according to an embodiment of the present application. FIG. 3B shows a heat treatment method of the heat treatment device for hardness tempering of the hardware product workpiece according to the embodiment of the application, which comprises the following steps: an inference phase comprising the steps of: s210, acquiring temperature data of a plurality of positions of the hardware product workpiece to be detected, wherein the temperature data are acquired by a plurality of temperature sensors deployed in a preset topological pattern; s220, constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors, and then obtaining a temperature measurement characteristic matrix through the first convolution neural network trained by a training module; s230, acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; s240, passing the thermal infrared image through the second convolutional neural network trained in the training stage to obtain a thermal infrared temperature distribution characteristic matrix; s250, fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; and S260, enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

FIG. 4 is a schematic diagram illustrating an architecture of a training phase in a heat treatment method of a heat treatment device for hardness tempering of a hardware product workpiece according to an embodiment of the application. As shown in fig. 4, in the training phase, in the network architecture, firstly, the obtained temperature data (e.g., P1 as illustrated in fig. 4) of a plurality of positions of the hardware product workpiece to be detected is constructed into a temperature matrix (e.g., M as illustrated in fig. 4) according to the preset topological pattern of the plurality of temperature sensors, and then the temperature measurement feature matrix (e.g., MF1 as illustrated in fig. 4) is obtained through a first convolutional neural network (e.g., CNN1 as illustrated in fig. 4); then, correcting the eigenvalues of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix (for example, MF2 as illustrated in fig. 4); then, passing the obtained thermal infrared image (e.g., P2 as illustrated in fig. 4) through a second convolutional neural network (e.g., CNN2 as illustrated in fig. 4) to obtain a thermal infrared temperature distribution feature matrix (e.g., MF3 as illustrated in fig. 4); then, fusing the corrected temperature measurement feature matrix and the thermal infrared temperature distribution feature matrix to obtain a classification feature matrix (e.g., MF as illustrated in fig. 4); then, passing the classification feature matrix through a classifier (e.g., a classifier as illustrated in fig. 4) to obtain a classification loss function value (e.g., a CLV as illustrated in fig. 4); finally, the first convolutional neural network, the second convolutional neural network, and the classifier are trained based on the classification loss function values.

Fig. 5 illustrates a schematic architecture diagram of an inference stage in a heat treatment method of a heat treatment device for hardness tempering of a hardware product workpiece according to an embodiment of the present application. As shown in fig. 5, in the inference stage, in the network architecture, firstly, the obtained temperature data (e.g., P1 as illustrated in fig. 5) of the multiple positions of the hardware product workpiece to be detected is constructed as a temperature matrix (e.g., M as illustrated in fig. 5) according to the preset topological pattern of the multiple temperature sensors, and then the first convolutional neural network (e.g., CN1 as illustrated in fig. 5) trained by the training module is used to obtain a temperature measurement feature matrix (e.g., MF1 as illustrated in fig. 5); then, passing the obtained thermal infrared image (e.g., P2 as illustrated in fig. 5) through the second convolutional neural network (e.g., CN2 as illustrated in fig. 5) completed by training phase to obtain a thermal infrared temperature distribution feature matrix (e.g., MF2 as illustrated in fig. 5); then, fusing the corrected temperature measurement feature matrix and the thermal infrared temperature distribution feature matrix to obtain a classification feature matrix (e.g., MF as illustrated in fig. 5); and finally, passing the classification feature matrix through a classifier (for example, a classifier as illustrated in fig. 5) to obtain a classification result, wherein the classification result is used for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

In summary, the heat treatment method of the heat treatment device for hardware product workpiece hardness tempering based on the embodiment of the application is clarified, the local high-dimensional correlation feature distribution of the thermal infrared image of the hardware product workpiece and the implicit correlation feature information of the temperature measurement value of each position of the workpiece are extracted through a convolutional neural network model, the feature information of the local high-dimensional correlation feature distribution and the implicit correlation feature information of the temperature measurement value of each position of the workpiece are further fused to judge the distribution rationality of the temperature of the hardware product workpiece to be detected, and in the feature fusion process, the temperature measurement feature matrix is further corrected to introduce robustness through the minimized loss surrounding the temperature local feature information, so that the clustering performance of the feature local equivalent to the feature whole is improved, and the joint characterization capability of the temperature measurement feature matrix to the global temperature feature is further improved. Therefore, the classification accuracy can be improved, and the hardness tempering performance can be ensured.

The basic principles of the present application have been described above with reference to specific embodiments, but it should be noted that advantages, effects, etc. mentioned in the present application are only examples and are not limiting, and the advantages, effects, etc. must not be considered to be possessed by various embodiments of the present application. Furthermore, the foregoing disclosure of specific details is for the purpose of illustration and description and is not intended to be limiting, since the foregoing disclosure is not intended to be exhaustive or to limit the disclosure to the precise details disclosed.

The block diagrams of devices, apparatuses, systems referred to in this application are only given as illustrative examples and are not intended to require or imply that the connections, arrangements, configurations, etc. must be made in the manner shown in the block diagrams. These devices, apparatuses, devices, systems may be connected, arranged, configured in any manner, as will be appreciated by one skilled in the art. Words such as "including," "comprising," "having," and the like are open-ended words that mean "including, but not limited to," and are used interchangeably herein. The words "or" and "as used herein mean, and are used interchangeably with, the word" and/or, "unless the context clearly dictates otherwise. The word "such as" is used herein to mean, and is used interchangeably with, the phrase "such as but not limited to".

It should also be noted that in the devices, apparatuses, and methods of the present application, the components or steps may be decomposed and/or recombined. These decompositions and/or recombinations are to be considered as equivalents of the present application.

The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. The utility model provides a heat treatment device of hardware products work piece hardness quenching and tempering which characterized in that includes: a training module comprising: the hardware product detection device comprises a first training data unit, a second training data unit and a third training data unit, wherein the first training data unit is used for acquiring temperature data of a plurality of positions of a hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; the temperature characteristic extraction unit is used for constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors and then obtaining a temperature measurement characteristic matrix through a first convolution neural network; the temperature characteristic correction unit is used for correcting the characteristic value of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix; the second training data unit is used for acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by the infrared camera; the thermal infrared coding unit is used for enabling the thermal infrared image to pass through a second convolutional neural network so as to obtain a thermal infrared temperature distribution characteristic matrix; the characteristic matrix fusion unit is used for fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; the loss calculation unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification loss function value; a training unit for training the first convolutional neural network, the second convolutional neural network, and the classifier based on the classification loss function value; and an inference module comprising: the temperature data acquisition unit is used for acquiring temperature data of a plurality of positions of the hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; the temperature coding unit is used for constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors and then obtaining a temperature measurement characteristic matrix through the first convolution neural network trained and completed by the training module; the thermal infrared image acquisition unit is used for acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; the thermal infrared image coding unit is used for enabling the thermal infrared image to pass through the second convolutional neural network which is trained in the training stage so as to obtain a thermal infrared temperature distribution characteristic matrix; the characteristic fusion unit is used for fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; and the preprocessing evaluation result generation unit is used for enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, and the classification result is used for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

2. The hardware product workpiece hardness tempering heat treatment apparatus according to claim 1, wherein said temperature characteristic extraction unit comprises: the matrix construction subunit is used for constructing a matrix based on preset topological patterns of the plurality of temperature sensors; the temperature value filling subunit is used for respectively filling the temperature data of the positions of the hardware product workpiece to be detected into the corresponding positions of the matrix to obtain a temperature initial matrix; and the temperature value supplementing subunit is used for performing neighborhood-based characteristic value supplementing on the position which is not filled with the temperature data in the temperature initial matrix to obtain the temperature matrix.

3. The hardware product work piece hardness tempering heat treatment apparatus of claim 2, wherein said temperature characteristic correction unit comprises: the local temperature characteristic representation subunit is used for calculating a logarithmic function value of a sum value of the eigenvalue and one of each position in the temperature measurement characteristic matrix as the local temperature characteristic representation of each temperature in the temperature measurement characteristic matrix; a global temperature feature representation subunit, configured to calculate a logarithmic function value of a sum of one and a sum of feature values of all positions in the temperature measurement feature matrix as a global temperature feature representation of the temperature measurement feature matrix; and the temperature characteristic representation adjusting subunit is used for dividing the local temperature characteristic representation of each position in the temperature measurement characteristic matrix by the global temperature characteristic representation of the temperature measurement characteristic matrix to obtain a corrected characteristic diagram of each position in the temperature measurement characteristic matrix so as to obtain the corrected temperature measurement characteristic matrix.

4. The apparatus for heat treatment of hardness tempering of a hardware product workpiece of claim 3, wherein each layer of said first convolutional neural network is separately in forward pass of layer: performing convolution processing on input data to obtain a convolution characteristic diagram; performing mean pooling based on local channel dimensions on the convolution feature map to obtain a pooled feature map; and carrying out nonlinear activation on the pooling feature map to obtain an activation feature map; wherein the output of the last layer of the first convolutional neural network is the temperature measurement characteristic matrix, and the input of the first layer of the first convolutional neural network is the temperature matrix.

5. The hardware product workpiece hardness-conditioned thermal processing apparatus of claim 4, wherein each layer of the second convolutional neural network performs in a forward pass of the layer respectively: performing convolution processing on input data to obtain a convolution characteristic diagram; performing mean pooling based on local channel dimensions on the convolution feature map to obtain a pooled feature map; and carrying out nonlinear activation on the pooling feature map to obtain an activation feature map; and the output of the last layer of the second convolutional neural network is the thermal infrared temperature distribution characteristic matrix, and the input of the first layer of the second convolutional neural network is the thermal infrared image.

6. The hardware product workpiece hardness tempering heat treatment apparatus according to claim 5, wherein said feature matrix fusion unit is further configured to: fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix according to the following formula to obtain the classification characteristic matrix; wherein the formula is:

wherein,

for the purpose of the classification feature matrix,

for the corrected temperature measurement feature matrix,

is a characteristic matrix of the thermal infrared temperature distribution "

7. The hardware product work piece hardness quenched and tempered heat treatment apparatus of claim 6, wherein the loss calculation unit is further configured to: the classifier processes the classification feature matrix to generate a classification result according to the following formula, wherein the formula is as follows:

wherein

Representing the projection of the classification feature matrix as a vector,

to

Is a weight matrix of the fully connected layers of each layer,

to

8. A heat treatment method of a heat treatment device for hardware product workpiece hardness tempering is characterized by comprising the following steps: a training phase comprising: acquiring temperature data of a plurality of positions of a hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors, and then obtaining a temperature measurement characteristic matrix through a first convolution neural network; correcting the characteristic value of each position in the temperature measurement characteristic matrix to obtain a corrected temperature measurement characteristic matrix; acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; enabling the thermal infrared image to pass through a second convolution neural network to obtain a thermal infrared temperature distribution characteristic matrix; fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; passing the classification characteristic matrix through a classifier to obtain a classification loss function value; training the first convolutional neural network, the second convolutional neural network, and the classifier based on the classification loss function values; and an inference phase comprising: acquiring temperature data of a plurality of positions of a hardware product workpiece to be detected, which are acquired by a plurality of temperature sensors deployed in a preset topological pattern; constructing temperature data of a plurality of positions of the hardware product workpiece to be detected into a temperature matrix according to preset topological patterns of the plurality of temperature sensors, and then obtaining a temperature measurement characteristic matrix through the first convolution neural network trained and completed by a training module; acquiring a thermal infrared image of the hardware product workpiece to be detected, which is acquired by an infrared camera; enabling the thermal infrared image to pass through the second convolutional neural network trained in the training stage to obtain a thermal infrared temperature distribution characteristic matrix; fusing the corrected temperature measurement characteristic matrix and the thermal infrared temperature distribution characteristic matrix to obtain a classification characteristic matrix; and enabling the classification characteristic matrix to pass through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the temperature distribution of the hardware product workpiece to be detected is reasonable or not.

9. The heat treatment method of the hardware product workpiece hardness tempering heat treatment device according to claim 8, wherein constructing the temperature data of the plurality of positions of the hardware product workpiece to be detected as a temperature matrix according to the preset topological pattern of the plurality of temperature sensors and then obtaining a temperature measurement characteristic matrix through a first convolutional neural network comprises: constructing a matrix based on preset topological patterns of the plurality of temperature sensors; respectively filling the temperature data of the positions of the hardware product workpiece to be detected into the corresponding positions of the matrix to obtain a temperature initial matrix; and performing neighborhood-based eigenvalue completion on the position which is not filled with temperature data in the temperature initial matrix to obtain the temperature matrix.

10. The heat treatment method of the heat treatment device for hardness tempering of hardware product workpieces according to claim 9, wherein correcting the eigenvalues of each position in the temperature measurement eigen matrix to obtain a corrected temperature measurement eigen matrix comprises: calculating a logarithmic function value of a summation value of the eigenvalue and one of each position in the temperature measurement characteristic matrix as a local temperature characteristic representation of each temperature in the temperature measurement characteristic matrix; calculating a logarithm function value of a summation value of all the eigenvalues of all the positions in the temperature measurement characteristic matrix and a summation value of one as a global temperature characteristic representation of the temperature measurement characteristic matrix; and dividing the local temperature characteristic representation of each position in the temperature measurement characteristic matrix by the global temperature characteristic representation of the temperature measurement characteristic matrix to obtain a corrected characteristic graph of each position in the temperature measurement characteristic matrix to obtain the corrected temperature measurement characteristic matrix.