Heat treatment process design method based on large model principle

Abstract

The invention discloses a heat treatment process design method based on a large model principle, and particularly relates to the field of heat treatment processes, which comprises the following steps: step S100, collecting a large batch of heat treatment process parameter data and material performance parameter data corresponding to the heat treatment process parameter data, preprocessing the data, and establishing a big data set; step S200, establishing a heat treatment process design model by using the established big data set; step S300, training the established heat treatment process design model; and S400, after the performance of the target material is determined, the target material is input into the established heat treatment process design model. According to the method, through the established heat treatment process design model, after the performance of the target material is determined, the corresponding process design parameters can be rapidly and automatically output, the process design time is greatly shortened, accurate prediction can be carried out based on a big data set, the number of unnecessary tests is reduced, and the trial and error cost is reduced.

Description

Technical Field

[0001] The present invention relates to the field of heat treatment technology, and more specifically, to a heat treatment technology design method based on a large model principle. Background Art

[0002] Heat treatment is a process that changes the structure of metal materials or alloys by heating, keeping them warm, and cooling them to obtain the desired properties.

[0003] In the field of traditional heat treatment process design, the determination of process parameters mainly depends on the long-term experience accumulated by process personnel. This approach has many disadvantages. For example, it is extremely inefficient because process personnel need to spend a lot of time trying and adjusting various parameter combinations. Moreover, it is extremely difficult to train an experienced heat treatment process personnel, who need to gradually master the complex relationship between various material properties, process conditions and final performance in long-term practice. This not only consumes a lot of manpower, material resources and time costs, but also due to the limitations of individual experience, it is difficult to ensure the optimality and consistency of process design.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a heat treatment process design method based on the large model principle to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: A heat treatment process design method based on the large model principle comprises the following steps: Step S100, collecting a large amount of heat treatment process parameter data and corresponding material performance parameter data, pre-processing these data, and establishing a large data set; Step S200, establishing a heat treatment process design model using the established big data set; Step S300, training the established heat treatment process design model; Step S400, after determining the target material properties, input it into the established heat treatment process design model, and automatically output the corresponding heat treatment process design parameters.

[0007] In a preferred embodiment, step S100 specifically includes the following contents: Heat treatment process parameters include heating parameters, insulation parameters and cooling parameters; Material performance parameters include mechanical performance parameters, physical performance parameters and chemical performance parameters; After cleaning and transforming the data, a large data set is created.

[0008] In a preferred embodiment, step S200 specifically includes the following contents: The heat treatment process design model includes an input layer, multiple hidden layers and an output layer; The number of nodes in the input layer corresponds to heating parameters, insulation parameters and cooling parameters; the number of nodes in the output layer corresponds to mechanical property parameters, physical property parameters and chemical property parameters; The number of hidden layers and nodes is adjusted and optimized according to the data complexity and model performance, and the best configuration is determined by experimenting with different structures.

[0009] In a preferred embodiment, step S300 specifically includes the following contents: Convert data into a format suitable for model input; For the input layer, the heating parameters, insulation parameters and cooling parameters are organized into vector form. Each sample corresponds to a vector, and the length of the vector is equal to the number of nodes in the input layer. For the output layer, the mechanical property parameters, physical property parameters and chemical property parameters are organized into corresponding vectors, whose length is equal to the number of nodes in the output layer; Using the loss function, the average of the squares of the differences between the predicted values ​​and the true values ​​is calculated using the formula: ;in, represents the loss function, represents the number of samples, represents the true value, Represents the predicted value.

[0011] In a preferred embodiment, step S300 specifically further includes the following contents: During the training process, observe the changes in the loss function value in real time; The sample is input into the model, starting from the input layer, through the calculation of the hidden layer, and finally obtaining the prediction result of the output layer for forward model training; According to the predicted results and actual results obtained by forward model training, the value of the loss function is calculated. Then, the reverse model training is adopted. Starting from the output layer, the gradient of the loss function is back-propagated to the input layer, and the gradient of each parameter is calculated. According to the calculated gradient, the parameters of the model are updated. Repeat forward model training and reverse model training, and continuously adjust model parameters after multiple training cycles.

[0012] In a preferred embodiment, step S400 specifically includes the following contents: Clarify the specific indicators and values ​​of the target material performance required, and convert the target performance data into a format that matches the model input requirements; The converted target performance data is input into the trained heat treatment process design model. The data is gradually transferred from the input layer to the output layer through the hidden layer, and finally the prediction result of the heat treatment process design parameters corresponding to the target performance is obtained at the output layer.

[0013] In a preferred embodiment, step S400 specifically further includes the following contents: Determine the sensitivity of different process parameters to the final material properties, and observe the changes in the material properties output by the model by changing the value of a single process parameter while keeping other parameters unchanged; Based on the process parameters predicted by the model, a set of optimal process parameter combinations is found so that multiple performance indicators can achieve a relatively optimal balance while satisfying certain constraints.

[0014] In a preferred embodiment, step S400 specifically further includes the following contents: Collect the actual performance data of the material after heat treatment according to the process parameters output by the model and the detailed records of the corresponding process parameters, feed these new data back to the model, and use the incremental learning method to update the model; Set a prediction error threshold and issue a warning signal in time when the prediction error exceeds the prediction error threshold.

[0015] The technical effects and advantages of the heat treatment process design method based on the large model principle of the present invention are as follows: 1. In the traditional method, the process personnel need to spend a lot of time trying and adjusting various parameter combinations to determine the appropriate heat treatment process. However, the present invention can quickly and automatically output the corresponding process design parameters after determining the target material properties by establishing a heat treatment process design model, which greatly shortens the process design time; 2. Traditional process design relies on experience and often requires multiple experiments to find the right combination of parameters, which not only wastes time but also increases the cost of raw materials, energy and equipment. The model of the present invention can make accurate predictions based on large data sets, reducing the number of unnecessary experiments and reducing the cost of trial and error. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention is a schematic flow chart of a heat treatment process design method based on the large model principle. DETAILED DESCRIPTION

[0017] 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. Example

[0018] Figure 1 The present invention provides a heat treatment process design method based on the large model principle, which specifically includes the following steps: Step S100, collecting a large amount of heat treatment process parameter data and corresponding material performance parameter data, pre-processing these data, and establishing a large data set; Step S200, establishing a heat treatment process design model using the established big data set; Step S300, training the established heat treatment process design model; Step S400, after determining the target material properties, input it into the established heat treatment process design model, and automatically output the corresponding heat treatment process design parameters.

[0019] Step S100 specifically includes the following contents: Heat treatment process parameters include heating parameters, insulation parameters and cooling parameters; Material performance parameters include mechanical performance parameters, physical performance parameters and chemical performance parameters; After cleaning and transforming the data, a large data set is created.

[0020] It should be added that a large data set containing a variety of heat treatment process parameters (heating, insulation, cooling) and material performance parameters (mechanical, physical, chemical) can more comprehensively reflect the changes in materials during heat treatment. The model can learn from these rich data the complex interactions between different parameters, such as how heating temperature and cooling rate jointly affect the mechanical properties of materials.

[0021] Sufficient data samples can make the model better generalize to different situations. Whether it is data from different materials, different process requirements or different production environments, it helps the model learn more general rules, so that when facing new heat treatment process design tasks, it can more accurately output reasonable process parameters and performance predictions.

[0022] Step S200 specifically includes the following contents: The heat treatment process design model includes an input layer, multiple hidden layers and an output layer; The number of nodes in the input layer corresponds to heating parameters, insulation parameters and cooling parameters; the number of nodes in the output layer corresponds to mechanical property parameters, physical property parameters and chemical property parameters; The number of hidden layers and nodes is adjusted and optimized according to the data complexity and model performance, and the best configuration is determined by experimenting with different structures.

[0023] It should be added that the input layer and the output layer correspond to the heat treatment process parameters and the material performance parameters respectively. This structure can directly establish the mapping relationship between process parameters and performance. The model can learn how the heating parameters, insulation parameters and cooling parameters specifically affect the mechanical properties, physical properties and chemical properties, so as to accurately predict the material properties based on the input process parameters, or conversely, output the appropriate process parameters according to the expected performance.

[0024] The presence of multiple hidden layers enables the model to handle complex nonlinear relationships. The relationship between heat treatment process parameters and material properties is often not a simple linear relationship, but a complex result of the interaction of multiple factors. The hidden layer can extract high-level features from the data through layer-by-layer abstraction and combination, and learn these complex interactive relationships. For example, the combined effect of heating temperature and cooling rate on material strength and toughness may be a complex nonlinear relationship, and the hidden layer can effectively capture and model this relationship.

[0025] By experimenting with different hidden layer structures and node configurations, the model structure that best suits specific heat treatment process data can be found. This experimental exploration method can fully tap the potential of the model and continuously improve the accuracy and generalization ability of the model. For example, in the process of modeling a new alloy steel heat treatment process data, by trying different combinations of the number of hidden layers (from 2 to 5 layers) and the number of nodes (from dozens to hundreds), the structure that makes the model perform best on the validation set and test set is found, thereby obtaining more reliable process parameter output and performance prediction.

[0026] Step S300 specifically includes the following contents: Convert data into a format suitable for model input; For the input layer, the heating parameters, insulation parameters and cooling parameters are organized into vector form. Each sample corresponds to a vector, and the length of the vector is equal to the number of nodes in the input layer. For the output layer, the mechanical property parameters, physical property parameters and chemical property parameters are organized into corresponding vectors, whose length is equal to the number of nodes in the output layer; Using the loss function, the average of the squares of the differences between the predicted values ​​and the true values ​​is calculated using the formula: ; in, represents the loss function, represents the number of samples, represents the true value, Represents the predicted value.

[0028] It should be added that data organized in vector form can be directly used as input to the model without complex data conversion and preprocessing. This reduces the data preparation time before model training and improves the training efficiency of the model. Especially for large-scale data sets, this efficient data processing method can significantly shorten the model training cycle and speed up the design of heat treatment processes.

[0029] MSE calculates the average of the squares of the differences between the predicted values ​​and the true values, which can intuitively reflect the size of the error between the model's predicted values ​​and the actual values. The square of the difference amplifies larger errors, causing the model to pay more attention to those predictions that deviate greatly from the true values. By minimizing MSE, the model can continuously adjust parameters to reduce prediction errors and improve prediction accuracy.

[0030] Step S300 specifically includes the following contents: During the training process, observe the changes in the loss function value in real time; The sample is input into the model, starting from the input layer, through the calculation of the hidden layer, and finally obtaining the prediction result of the output layer for forward model training; According to the predicted results and actual results obtained by forward model training, the value of the loss function is calculated. Then, the reverse model training is adopted. Starting from the output layer, the gradient of the loss function is back-propagated to the input layer, and the gradient of each parameter is calculated. According to the calculated gradient, the parameters of the model are updated. Repeat forward model training and reverse model training, and continuously adjust model parameters after multiple training cycles.

[0031] It should be added that by observing the changes in the loss function value during the training process in real time, you can intuitively understand the training progress of the model. By observing the downward trend of the loss function value with the training cycle, you can judge whether the model is constantly learning and improving. If the loss function value continues to decrease, it means that the model is moving in a better direction and the training effect is good; if the loss function value stagnates or fluctuates, you may need to adjust the training strategy, such as adjusting the learning rate, increasing the training data, etc.

[0032] After multiple training cycles, the model parameters are adjusted continuously, so that the model gradually converges to a better state. Each training cycle is an update of the model parameters. By repeating this process, the model can continuously learn and improve, improving its ability to fit the data and prediction accuracy.

[0033] As the training cycle increases, the model can gradually adapt to different data samples and situations, reducing the risk of overfitting to specific data. At the same time, repeated training can also allow the model to better capture the potential laws and trends in the data, improving the stability and reliability of the model.

[0034] Step S400 specifically includes the following contents: Clarify the specific indicators and values ​​of the target material performance required, and convert the target performance data into a format that matches the model input requirements; The converted target performance data is input into the trained heat treatment process design model. The data is gradually transferred from the input layer to the output layer through the hidden layer, and finally the prediction result of the heat treatment process design parameters corresponding to the target performance is obtained at the output layer.

[0035] Step S400 specifically includes the following contents: Determine the sensitivity of different process parameters to the final material properties, and observe the changes in the material properties output by the model by changing the value of a single process parameter while keeping other parameters unchanged; Based on the process parameters predicted by the model, a set of optimal process parameter combinations is found so that multiple performance indicators can achieve a relatively optimal balance while satisfying certain constraints.

[0036] It should be added that in practical applications, materials often need to meet multiple performance indicators at the same time, such as high strength, high toughness, good corrosion resistance, etc. Multi-objective optimization can find a set of optimal process parameter combinations based on the process parameters predicted by the model, so that multiple performance indicators can achieve a relatively optimal balance under certain constraints. This can better meet complex engineering needs and improve the comprehensive performance of materials.

[0037] For example, in the aerospace field, materials need to have sufficient strength to withstand high loads, good toughness to prevent brittle fracture, and certain corrosion resistance to adapt to harsh environments. Through multi-objective optimization, a combination of process parameters that can meet the strength and toughness requirements and ensure a certain degree of corrosion resistance can be found, providing more ideal materials for the manufacture of aerospace components.

[0038] Multi-objective optimization provides a more flexible process design method. Traditional single-objective optimization can usually only optimize one performance indicator, but in actual production, different performance indicators often have a mutually restrictive relationship. Multi-objective optimization can consider multiple performance indicators at the same time, providing process personnel with more choices and possibilities.

[0039] For example, increasing the hardness of a material may lead to a decrease in toughness. Through multi-objective optimization, a suitable balance between hardness and toughness can be found, and the optimal combination of process parameters can be determined according to specific application requirements. This flexibility can better adapt to different production requirements and market demands.

[0040] Step S400 specifically includes the following contents: Collect the actual performance data of the material after heat treatment according to the process parameters output by the model and the detailed records of the corresponding process parameters, feed these new data back to the model, and use the incremental learning method to update the model; Set a prediction error threshold and issue a warning signal in time when the prediction error exceeds the prediction error threshold.

[0041] It should be added that as actual production proceeds, the actual performance data of the material after heat treatment according to the process parameters output by the model and the detailed records of the corresponding process parameters are collected and fed back to the model for incremental learning, so that the model can continuously adapt to new data and changes in actual production conditions. This helps to improve the accuracy and generalization ability of the model, enabling it to better predict the heat treatment results under different materials and process conditions.

[0042] For example, over time, the properties of raw materials may change, or the performance of production equipment may fluctuate. With incremental learning, the model can learn these new changes and adjust its predictions accordingly, maintaining a high level of accuracy.

[0043] By setting a prediction error threshold and issuing a warning signal in time when the prediction error exceeds the threshold, process personnel can promptly detect the deviation between the model prediction results and the actual situation. This enables process personnel to take measures at an early stage to prevent the problem from further expanding, thereby reducing production losses and quality risks.

[0044] For example, if the error between the material properties predicted by the model and the actual properties exceeds a certain level, it may mean that an abnormal situation has occurred in the production process, such as raw material quality problems, equipment failure, or improper process parameter settings. Timely early warning signals can allow process personnel to quickly troubleshoot problems and take appropriate corrective measures.

[0045] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0046] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0047] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0048] In the several embodiments provided in the present application, it should be understood that the disclosed methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0049] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0050] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0051] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0052] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0053] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A heat treatment process design method based on the large model principle, characterized in that: The steps include: Step S100, collecting a large amount of heat treatment process parameter data and corresponding material performance parameter data, pre-processing these data, and establishing a large data set; Step S200, establishing a heat treatment process design model using the established big data set; Step S300, training the established heat treatment process design model; Step S400, after determining the target material properties, input them into the established heat treatment process design model, and automatically output the corresponding heat treatment process design parameters.

2. The heat treatment process design method based on the large model principle according to claim 1 is characterized in that: Step S100 specifically includes the following contents: Heat treatment process parameters include heating parameters, insulation parameters and cooling parameters; Material performance parameters include mechanical performance parameters, physical performance parameters and chemical performance parameters; After cleaning and transforming the data, a large data set is created.

3. The heat treatment process design method based on the large model principle according to claim 2 is characterized in that: Step S200 specifically includes the following contents: The heat treatment process design model includes an input layer, multiple hidden layers and an output layer; The number of nodes in the input layer corresponds to heating parameters, insulation parameters and cooling parameters; the number of nodes in the output layer corresponds to mechanical property parameters, physical property parameters and chemical property parameters; The number of hidden layers and nodes is adjusted and optimized according to the data complexity and model performance, and the best configuration is determined by experimenting with different structures.

4. The heat treatment process design method based on the large model principle according to claim 3 is characterized in that: Step S300 specifically includes the following contents: Convert data into a format suitable for model input; For the input layer, the heating parameters, insulation parameters and cooling parameters are organized into vector form. Each sample corresponds to a vector, and the length of the vector is equal to the number of nodes in the input layer. For the output layer, the mechanical property parameters, physical property parameters and chemical property parameters are organized into corresponding vectors, whose length is equal to the number of nodes in the output layer; Using the loss function, the average of the squares of the differences between the predicted values ​​and the true values ​​is calculated using the formula: Among them, MSE represents the loss function, n represents the number of samples, and y i represents the true value, Represents the predicted value.

5. The heat treatment process design method based on the large model principle according to claim 4 is characterized in that: Step S300 specifically includes the following contents: During the training process, observe the changes in the loss function value in real time; The sample is input into the model, starting from the input layer, through the calculation of the hidden layer, and finally obtaining the prediction result of the output layer for forward model training; According to the predicted results and actual results obtained by forward model training, the value of the loss function is calculated. Then, the reverse model training is adopted. Starting from the output layer, the gradient of the loss function is back-propagated to the input layer, and the gradient of each parameter is calculated. According to the calculated gradient, the parameters of the model are updated. Repeat forward model training and reverse model training, and continuously adjust model parameters after multiple training cycles.

6. The heat treatment process design method based on the large model principle according to claim 5 is characterized in that: Step S400 specifically includes the following contents: Clarify the specific indicators and values ​​of the target material performance required, and convert the target performance data into a format that matches the model input requirements; The converted target performance data is input into the trained heat treatment process design model. The data is gradually transferred from the input layer to the output layer through the hidden layer, and finally the prediction result of the heat treatment process design parameters corresponding to the target performance is obtained at the output layer.

7. The heat treatment process design method based on the large model principle according to claim 6 is characterized in that: Step S400 specifically includes the following contents: Determine the sensitivity of different process parameters to the final material properties, and observe the changes in the material properties output by the model by changing the value of a single process parameter while keeping other parameters unchanged; Based on the process parameters predicted by the model, a set of optimal process parameter combinations is found so that multiple performance indicators can achieve a relatively optimal balance while satisfying certain constraints.

8. The heat treatment process design method based on the large model principle according to claim 7 is characterized in that: Step S400 specifically includes the following contents: Collect the actual performance data of the material after heat treatment according to the process parameters output by the model and the detailed records of the corresponding process parameters, feed these new data back to the model, and use the incremental learning method to update the model; Set a prediction error threshold and issue a warning signal in time when the prediction error exceeds the prediction error threshold.

Images