CN115985424A - BP neural network-based part heat treatment performance prediction method and system - Google Patents
BP neural network-based part heat treatment performance prediction method and system Download PDFInfo
- Publication number
- CN115985424A CN115985424A CN202211660049.0A CN202211660049A CN115985424A CN 115985424 A CN115985424 A CN 115985424A CN 202211660049 A CN202211660049 A CN 202211660049A CN 115985424 A CN115985424 A CN 115985424A
- Authority
- CN
- China
- Prior art keywords
- predicted
- performance
- average cooling
- training
- end quenching
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses a method and a system for predicting heat treatment performance of a part based on a BP neural network, which relate to the technical field of material performance prediction, and the method comprises the following steps: obtaining the average cooling speed of the part to be predicted at a plurality of sampling points to be predicted; inputting the average cooling speed of all sampling points to be predicted into a performance prediction model to obtain the performance of the part to be predicted; the performance prediction model is constructed based on a BP neural network; the performance includes: hardness, yield strength, tensile strength, elongation, reduction of area, fatigue strength, fracture toughness, impact toughness and abrasion resistance. The invention improves the prediction precision of the heat treatment performance of the part.
Description
Technical Field
The invention relates to the technical field of material performance prediction, in particular to a method and a system for predicting heat treatment performance of a part based on a BP neural network.
Background
In the manufacturing process of the metal parts, a heat treatment process is often needed to improve the microstructure and improve the mechanical property. The temperature of each position of the part in the heat treatment process has certain difference, so that the mechanical properties of different positions of the part are different. Particularly, after the large-scale component is quenched, the temperature difference in the cooling process at different positions is large, and the mechanical property distribution is extremely uneven. With the continuous development of the manufacturing industry level, the requirement on the heat treatment performance of the parts is continuously improved, and the design of a heat treatment process and a tool is urgently needed by predicting the heat treatment performance distribution, so that the performance distribution of the parts is improved, and the service life is prolonged.
The heat treatment process is very complex in its influence on the properties, with the temperature history being the most critical factor affecting the material properties. The material can obtain different metallographic structures, precipitated phases, dislocation densities and the like through different processes of heating, heat preservation, cooling and the like, and further the mechanical properties of the material are influenced. The current prediction of mechanical properties is mainly based on the establishment of a mathematical model between temperature-structure-properties. In the document, test verification of a hardness prediction model (Sondongli et al. Mechanical engineering materials 2008,32 (3): 29-31, 34), a Maynier and Carsic hardness calculation model is adopted to predict hardness distribution after end quenching. The model calculates the hardness distribution through the material composition and the tissue content, so the accuracy of the model is easily influenced by the calculation accuracy of the tissue content. The research result shows that the hardness prediction model proposed by Carsic can not accurately predict the hardness value of all the materials after cooling. A mechanical property calculation model in the aluminum alloy aging process is summarized in a document of aluminum alloy aging precipitation kinetics and strengthening model (Wang Xiaona, and the like, chinese non-ferrous metals school report, 2013 (10): 2754-2768). Researches show that the existing prediction model for aluminum alloy aging strengthening mainly calculates the microstructure evolution and the precipitated phase morphology, and then calculates the mechanical properties through different strengthening mechanism models. At present, a quantitative calculation model between temperature, structure and performance in a heat treatment process is complex, parameters are difficult to obtain, the difficulty in calculating the performance of a part is high, the accuracy is low, and engineering application is difficult to realize.
Disclosure of Invention
The invention aims to provide a method and a system for predicting the heat treatment performance of a part based on a BP neural network, which improve the prediction precision of the heat treatment performance of the part.
In order to achieve the purpose, the invention provides the following scheme:
a part heat treatment performance prediction method based on a BP neural network comprises the following steps:
acquiring the average cooling speed of the part to be predicted at a plurality of sampling points to be predicted;
inputting the average cooling speed of all sampling points to be predicted into a performance prediction model to obtain the performance of the part to be predicted; the performance prediction model is constructed based on a BP neural network; the properties include: hardness, yield strength, tensile strength, elongation, reduction of area, fatigue strength, fracture toughness, impact toughness, and abrasion resistance.
Optionally, the obtaining the average cooling rate of the part to be predicted at the plurality of sampling points to be predicted specifically includes:
carrying out end quenching treatment on the part to be predicted to obtain a temperature-time curve of the part to be predicted at a plurality of sampling points to be predicted in the end quenching treatment process;
and determining the average cooling rate of each sampling point to be predicted according to the temperature-time curve of each sampling point to be predicted.
Optionally, the training method of the performance prediction model includes:
determining an end quenching sample; the end quenching sample and the part to be predicted are made of the same material;
carrying out end quenching treatment on the end quenching sample to obtain temperature-time curves and performance numerical values of the end quenching sample at a plurality of training sampling points in the end quenching treatment process;
determining the average cooling rate of each training sampling point according to the temperature-time curve of each training sampling point;
and training the BP neural network by taking the average cooling rate of all the training sampling points as input and the numerical values of the performance of all the training sampling points as output to obtain the performance prediction model.
A BP neural network-based part thermal processing performance prediction system, the system comprising:
the first average cooling speed acquisition module is used for acquiring the average cooling speed of the part to be predicted at a plurality of sampling points to be predicted;
the prediction module is used for inputting the average cooling speed of all sampling points to be predicted into the performance prediction model to obtain the performance of the part to be predicted; the performance prediction model is constructed based on a BP neural network; the properties include: hardness, yield strength, tensile strength, elongation, reduction of area, fatigue strength, fracture toughness, impact toughness, and abrasion resistance.
Optionally, the first average cooling rate obtaining module specifically includes:
the first end quenching processing unit is used for carrying out end quenching processing on the part to be predicted and acquiring temperature-time curves of the part to be predicted at a plurality of sampling points to be predicted in the process of end quenching processing;
and the first average cooling rate determining unit is used for determining the average cooling rate of each sampling point to be predicted according to the temperature-time curve of each sampling point to be predicted.
Optionally, the prediction module comprises: a performance prediction model training sub-module, the performance prediction model training sub-module comprising:
the end quenching sample determining unit is used for determining an end quenching sample; the end quenching sample and the part to be predicted are made of the same material;
the second end quenching processing unit is used for carrying out end quenching processing on the end quenching sample to obtain temperature-time curves and performance values of the end quenching sample at a plurality of training sampling points in the end quenching processing process;
the second average cooling rate determining unit is used for determining the average cooling rate of each training sampling point according to the temperature-time curve of each training sampling point;
and the training unit is used for training the BP neural network by taking the average cooling rate of all the training sampling points as input and the numerical values of the performance of all the training sampling points as output to obtain the performance prediction model.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention discloses a method and a system for predicting heat treatment performance of a part based on a BP neural network, wherein the method comprises the following steps: acquiring the average cooling speed of the part to be predicted at a plurality of sampling points to be predicted; inputting the average cooling speed of all sampling points to be predicted into a performance prediction model to obtain the performance of the part to be predicted; the performance prediction model is constructed based on a BP neural network; the performance includes: hardness, yield strength, tensile strength, elongation, reduction of area, fatigue strength, fracture toughness, impact toughness, and abrasion resistance. Compared with a quantitative calculation model aiming at the temperature-structure-performance in the heat treatment process, the method can realize the prediction of the performance of the part by only acquiring the average cooling speed of the part to be predicted at a plurality of sampling points to be predicted as input quantity, the average cooling speed can be acquired only by end quenching treatment, and the prediction precision of the heat treatment performance of the part is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a schematic flow chart of a method for predicting heat treatment performance of a part based on a BP neural network according to embodiment 1 of the present invention;
FIG. 2 is a schematic structural diagram of a block-shaped part made of a 42CrMo material;
FIG. 3 is a linear regression plot of the training results;
FIG. 4 is a graphical representation of the predicted hardness for a specific example;
fig. 5 is a schematic structural diagram of a system for predicting heat treatment performance of a part based on a BP neural network according to embodiment 2 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method and a system for predicting the heat treatment performance of a part based on a BP neural network, and aims to improve the prediction precision of the heat treatment performance of the part.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example 1
Fig. 1 is a schematic flow chart of a method for predicting heat treatment performance of a part based on a BP neural network according to embodiment 1 of the present invention. As shown in fig. 1, the method for predicting heat treatment performance of a part based on a BP neural network in this embodiment includes:
step 101: and obtaining the average cooling speed of the part to be predicted at a plurality of sampling points to be predicted.
Step 102: inputting the average cooling speed of all sampling points to be predicted into a performance prediction model to obtain the performance of the part to be predicted; the performance prediction model is constructed based on a BP neural network; the performance includes: hardness, yield strength, tensile strength, elongation, reduction of area, fatigue strength, fracture toughness, impact toughness, and abrasion resistance.
As an optional implementation manner, step 101 specifically includes:
and performing end quenching treatment on the part to be predicted to obtain the temperature-time curves of the part to be predicted at a plurality of sampling points to be predicted in the end quenching treatment process.
And determining the average cooling rate of each sampling point to be predicted according to the temperature-time curve of each sampling point to be predicted.
As an alternative embodiment, the method for training the performance prediction model includes:
determining an end quenching sample; the end quenching sample and the part to be predicted are made of the same material.
And carrying out end quenching treatment on the end quenching sample to obtain temperature-time curves and performance numerical values of the end quenching sample at a plurality of training sampling points in the end quenching treatment process.
And determining the average cooling rate of each training sampling point according to the temperature-time curve of each training sampling point.
And training the BP neural network by taking the average cooling rate of all the training sampling points as input and the numerical values of the performance of all the training sampling points as output to obtain a performance prediction model.
In practical operation, the training process of the performance prediction model includes:
(1) And processing the material consistent with the part to be predicted into an end quenching sample.
Specifically, the chemical components, initial structure and grain size of the end-quenched sample material are consistent with the state of the part to be predicted before heat treatment.
(2) End quenching treatment: heating the end quenching sample to austenitizing temperature by adopting a heating process consistent with the thermal treatment of the performance of the part to be predicted, carrying out heat preservation for the same time, discharging the sample out of the furnace, transferring the sample to an end quenching platform, starting a water spraying device, carrying out end face quenching treatment, and taking down the sample after the sample is completely cooled.
Specifically, if the actual performance heat treatment includes tempering treatment, the cooled end-quenched sample needs to be subjected to tempering treatment of the same process.
(3) Sampling the end quenching sample at regular intervals along the axial direction (i.e. the axial direction of the end quenching sample, such as the length direction in fig. 2), and performing mechanical property detection and analysis to obtain a performance value.
When the predicted property is hardness, a parallel plane was ground in the direction along the axis of the end-quenched sample to a depth of about 1.5mm. The grinding wheel grinding method is adopted, flowing water cooling is kept in the grinding process, and the situation that the structure is changed due to overhigh temperature of a sample is prevented. And selecting a position with a certain distance from the quenching end face on the grinding plane along the direction away from the quenching end face to measure the hardness of the sample.
When the predicted performance is yield strength, tensile strength, elongation and reduction of area, a position at a certain distance from the quenching end face is selected to prepare a tensile sample with the thickness of 1mm along the cross section, and the sample is subjected to a room temperature tensile experiment to obtain the yield strength, tensile strength, elongation and reduction of area at different positions of the end quenching sample.
When the predicted performance is fatigue strength, fracture toughness, impact toughness and abrasion resistance, a cross section of a position at a certain distance from a quenching end face is taken as a research object, a corresponding sample is designed for detection, and the thickness of the sample is not more than 3mm.
(4) Finite element numerical simulation is carried out on the end quenching treatment process, a temperature-time curve of a sampling position (training sampling point) in the end quenching process is calculated, and the average cooling speed of each training sampling point is obtained by averaging the cooling speeds at certain intervals in the temperature-time curve within the temperature range below Ac3 (Ac 3 represents the temperature at which all ferrite is transformed into austenite when the hypoeutectoid steel is heated).
Specifically, a finite element geometric model of the end quenching sample is established, and the model is subjected to meshing with a mesh size smaller than 1 mm. The method comprises the steps of setting data of material density, heat conductivity and specific heat capacity of a model, setting initial temperature of the model as heat treatment heating temperature, setting heat exchange coefficient of a water quenching end as heat exchange coefficient of water, and setting other boundaries as air cooling heat exchange. And (3) performing transient temperature field calculation on the model to obtain temperature history data at nodes (the distance corresponding to each node is the same, and is shown as a point taking position in fig. 2) at a certain distance from the quenching end face, and obtaining a temperature-time curve.
(5) Designing a BP neural network, taking average cooling rate data at different positions as an input layer and corresponding mechanical property detection data (performance numerical values) as an output layer, and selecting the number of hidden layers, a training method and training parameters of a reasonable Back Propagation (BP) neural network to train the neural network so as to obtain a mapping relation between the cooling rate and the mechanical property.
Specifically, the BP neural network has a three-layer structure of an input layer, a hidden layer and an output layer. And (5) the input layer is the average cooling speed extracted in the step (4). And (4) outputting the corresponding performance value in the step (3). The number of hidden layer neuron nodes satisfies the following conditions:
2 u >v(1)
wherein u is the number of hidden layer neuron nodes; v is the number of input layer nodes; e is the number of output layer nodes; r is an integer between 0 and 10. In the invention, the number of nodes of an input layer is determined by a cooling rate array, the number of nodes of an output layer is certain performance data (the cooling rate array is that a cooling curve is divided into a section (for example, one section at every 10 ℃), then the average cooling rate of the section is taken, and the whole cooling process has a plurality of average cooling rates, namely the cooling rate array, so that e =1.
The hidden layer activation function is a sigmoid function, and the training function is Bayesian regularization. The data proportion of the training set, validation set and test set in the input data is approximately 8:1:1.
compared with the prior art, the invention has the beneficial effects that:
the invention fully utilizes the more accurate temperature calculation result in the numerical simulation, adopts the BP neural network algorithm to establish the nonlinear relation between the temperature history and the heat treatment performance, avoids establishing a temperature-tissue-performance model to predict the heat treatment performance distribution, further reduces the complexity of the model and simultaneously improves the calculation precision.
The method can be applied to the prediction of the heat treatment performance of most metals such as steel materials, aluminum alloys, titanium alloys, zirconium alloys and the like. The invention is also applicable to more complex performance heat treatment processes, such as multi-pass quenching or multi-pass tempering.
The experimental and simulation technology of the invention has the advantages of lower difficulty, simple parameter setting, short research and development period, controllable cost and better engineering application value.
The specific embodiment is as follows: a block part (100 mm multiplied by 50 mm) prepared by 42CrMo material is quenched, the quenching process is that the block part is taken out of a furnace and quenched by water after being kept at 850 ℃ for 1h, and the hardness distribution after quenching needs to be predicted for example.
The method comprises the following steps: the material consistent with the part was processed into end quenched samples.
In this example, the material was 42CrMo, the same material was processed into an end-quenched sample, and the sample size is shown in FIG. 2. The chemical composition, initial structure and grain size of the material are consistent with those of the massive parts before quenching.
Step two: heating the end quenching sample to austenitizing temperature by adopting a heating process consistent with the performance heat treatment of the part, carrying out heat preservation for the same time, discharging the sample out of the furnace, transferring the sample to an end quenching platform, starting a water spraying device, carrying out end face quenching treatment, and taking down the sample after the sample is completely cooled.
In the embodiment, the end quenching sample is heated to 850 ℃ and is kept warm for 1h, then the sample is taken out of the furnace and is transferred to an end quenching platform, a water spraying device is started to carry out end face quenching treatment, and the sample is taken down after being completely cooled.
Step three: and sampling end quenching samples at certain intervals along the axial direction, and detecting and analyzing the mechanical properties.
In this example, a parallel plane was ground along the axial direction of the end-quenched sample to a depth of about 1.5mm. The grinding wheel grinding method is adopted, flowing water cooling is kept in the grinding process, and the situation that the structure is changed due to overhigh temperature of a sample is prevented. The rockwell hardness of the test specimens was measured on the grinding plane at positions 1.5mm, 5mm, 10mm, 15mm, 20mm, 25mm, 30mm, 35mm, 40mm, 45mm, 50mm, 55mm, 60mm, 65mm and 70mm from the quenched end face in the direction away from the quenched end face (as shown in fig. 2), and the hardness data was recorded.
Step four: finite element numerical simulation is carried out in the end quenching process, the temperature curve of the sampling position in the end quenching process is calculated, and the average cooling speed is obtained at intervals of certain temperature within the temperature range below the Ac3 of the material.
In this embodiment, the following is specifically mentioned:
and (3) establishing an end quenching geometric model by adopting a two-dimensional axisymmetric model, and dividing quadrilateral meshes of the model by using the mesh size of 0.5 mm.
Setting the material density, thermal conductivity and specific heat capacity data of a model material with the parameters of 42CrMo, setting the initial temperature of the model to be 850 ℃, and setting the heat exchange coefficient of a water quenching end to be 10000W/(m) 2 DEG C), setting the rest boundaries for air cooling heat exchange, wherein the heat exchange coefficient is 50W/(m) 2 ·℃)。
And performing transient temperature field calculation on the model to obtain temperature history data at nodes 1.5mm, 5mm, 10mm, 15mm, 20mm, 25mm, 30mm, 35mm, 40mm, 45mm, 50mm, 55mm, 60mm, 65mm and 70mm away from the quenching end face.
The Ac3 temperature of the 42CrMo material is 800 ℃, so that the data of 800-100 ℃ in the temperature data are extracted, and the average cooling rate in the temperature range is calculated every 20 ℃ and is used as the cooling rate array of the node. And corresponding the cold speed array of each node with the corresponding position performance data obtained in the third step. Table 1 shows partial data.
TABLE 1 neural network training data
Step five: designing a BP neural network, taking average cooling rate data at different positions as an input layer and corresponding mechanical property detection data as an output layer, selecting reasonable number of hidden layers, training methods and training parameters of the BP neural network to train the neural network, and obtaining a mapping relation between the cooling rate and the mechanical property.
In this example, there are 35 average cooling rate data per position, and one hardness. Setting the number of nodes of an input layer of the BP neural network to be 35 and the number of nodes of an output layer to be 1. The number of hidden layer neuron nodes should be in the range of 6-16 according to equation (2). The number of hidden layer neuron nodes was chosen here to be 10. The final designed BP neural network structure is a 35-10-1 three-layer structure. The hidden layer activation function is set to be a sigmoid function, and the training function is Bayesian regularization. The input data are 15 groups, and the data proportion of the training set, the verification set and the test set is approximately 8:1:1, designing 11 groups of training sets, 2 groups of verification sets and 2 groups of test sets. Fig. 3 is a linear regression plot of the training results, with the accuracy of the trained model being 99.8%.
Step six: establishing a finite element numerical model for heat treatment of the part to be predicted, obtaining cooling curves at different positions, and averaging the cooling speed at a certain temperature below Ac 3. And calculating to obtain performance data of different positions by using the neural network algorithm established in the steps and taking the average cooling speed data as input.
The embodiment specifically includes:
and (3) establishing a three-dimensional geometric model of the part, wherein the part is a cube and has symmetry, and the model can be set to be a 1/8 model for analysis. And adopting hexahedral meshes with the size of 1mm and carrying out mesh division.
Setting the material density, thermal conductivity and specific heat capacity data of the model material with parameters of 42CrMo, setting the initial temperature of the model to 850 ℃, and setting water quenchingThe heat exchange coefficient of the end is 10000W/(m) 2 And DEG C), setting the rest boundaries for air cooling heat exchange, wherein the heat exchange coefficient is 50W/(m < 2 >).
And (4) performing transient temperature field calculation on the model to obtain temperature history data on all nodes of the model.
And extracting 800-100 ℃ data in the temperature data, and calculating the average cooling rate in the temperature range every 20 ℃ for the temperature process data at different nodes to serve as a cooling rate array of the node.
And taking the cold velocity arrays on all the nodes as input layers, and calculating by adopting the BP neural network model constructed in the fifth step to obtain the performance data of different nodes. By importing the data into the model, the stiffness at different positions can be obtained as shown in fig. 4.
The predicted surface hardness of the part was 62.8HRC, and the core hardness was 57.2HRC. The measured outer surface hardness of the part is 62.3HRC, the core hardness is 56.7HRC, and the simulation result is basically consistent with the measured result.
Example 2
Fig. 5 is a schematic structural diagram of a system for predicting heat treatment performance of a part based on a BP neural network according to embodiment 2 of the present invention. As shown in fig. 5, the system for predicting heat treatment performance of a part based on a BP neural network in this embodiment includes:
the first average cooling rate obtaining module 201 is configured to obtain an average cooling rate of the part to be predicted at a plurality of sampling points to be predicted.
The prediction module 202 is used for inputting the average cooling speed of all sampling points to be predicted into the performance prediction model to obtain the performance of the part to be predicted; the performance prediction model is constructed based on a BP neural network; the performance includes: hardness, yield strength, tensile strength, elongation, reduction of area, fatigue strength, fracture toughness, impact toughness and abrasion resistance.
As an optional implementation manner, the first average cooling rate obtaining module 201 specifically includes:
and the first end quenching processing unit is used for carrying out end quenching processing on the part to be predicted and acquiring temperature time curves of the part to be predicted at a plurality of sampling points to be predicted in the end quenching processing process.
And the first average cooling rate determining unit is used for determining the average cooling rate of each sampling point to be predicted according to the temperature-time curve of each sampling point to be predicted.
As an alternative embodiment, the prediction module 202 includes: the performance prediction model training submodule comprises:
the end quenching sample determining unit is used for determining an end quenching sample; the end quenching sample and the part to be predicted are made of the same material.
And the second end quenching processing unit is used for carrying out end quenching processing on the end quenching sample and acquiring temperature-time curves and performance numerical values of the end quenching sample at a plurality of training sampling points in the end quenching processing process.
And the second average cooling rate determining unit is used for determining the average cooling rate of each training sampling point according to the temperature-time curve of each training sampling point.
And the training unit is used for training the BP neural network by taking the average cooling rate of all the training sampling points as input and the numerical values of the performance of all the training sampling points as output to obtain a performance prediction model.
In the present specification, the embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the foregoing, the description is not to be taken in a limiting sense.
Claims (6)
1. A method for predicting heat treatment performance of a part based on a BP neural network is characterized by comprising the following steps:
obtaining the average cooling speed of the part to be predicted at a plurality of sampling points to be predicted;
inputting the average cooling speed of all sampling points to be predicted into a performance prediction model to obtain the performance of the part to be predicted; the performance prediction model is constructed based on a BP neural network; the properties include: hardness, yield strength, tensile strength, elongation, reduction of area, fatigue strength, fracture toughness, impact toughness, and abrasion resistance.
2. The method for predicting the heat treatment performance of the part based on the BP neural network as claimed in claim 1, wherein the obtaining of the average cooling rate of the part to be predicted at a plurality of sampling points to be predicted specifically comprises:
carrying out end quenching treatment on the part to be predicted to obtain a temperature-time curve of the part to be predicted at a plurality of sampling points to be predicted in the end quenching treatment process;
and determining the average cooling rate of each sampling point to be predicted according to the temperature-time curve of each sampling point to be predicted.
3. The method for predicting the heat treatment performance of the part based on the BP neural network as claimed in claim 1, wherein the training method of the performance prediction model comprises:
determining an end quenching sample; the end quenching sample and the part to be predicted are made of the same material;
carrying out end quenching treatment on the end quenching sample to obtain temperature-time curves and performance numerical values of the end quenching sample at a plurality of training sampling points in the end quenching treatment process;
determining the average cooling rate of each training sampling point according to the temperature-time curve of each training sampling point;
and training the BP neural network by taking the average cooling rate of all the training sampling points as input and the numerical values of the performance of all the training sampling points as output to obtain the performance prediction model.
4. A BP neural network-based part heat treatment performance prediction system, characterized by comprising:
the first average cooling speed acquisition module is used for acquiring the average cooling speed of the part to be predicted at a plurality of sampling points to be predicted;
the prediction module is used for inputting the average cooling speed of all sampling points to be predicted into the performance prediction model to obtain the performance of the part to be predicted; the performance prediction model is constructed based on a BP neural network; the performance includes: hardness, yield strength, tensile strength, elongation, reduction of area, fatigue strength, fracture toughness, impact toughness, and abrasion resistance.
5. The system of claim 4, wherein the first average cooling rate obtaining module specifically comprises:
the first end quenching processing unit is used for carrying out end quenching processing on the part to be predicted and acquiring temperature-time curves of the part to be predicted at a plurality of sampling points to be predicted in the end quenching processing process;
and the first average cooling rate determining unit is used for determining the average cooling rate of each sampling point to be predicted according to the temperature-time curve of each sampling point to be predicted.
6. The BP neural network based part thermal processing performance prediction system of claim 4, wherein the prediction module comprises: a performance prediction model training sub-module, the performance prediction model training sub-module comprising:
the end quenching sample determining unit is used for determining an end quenching sample; the end quenching sample and the part to be predicted are made of the same material;
the second end quenching processing unit is used for carrying out end quenching processing on the end quenching sample to obtain temperature-time curves and performance numerical values of the end quenching sample at a plurality of training sampling points in the end quenching processing process;
the second average cooling rate determining unit is used for determining the average cooling rate of each training sampling point according to the temperature-time curve of each training sampling point;
and the training unit is used for training the BP neural network by taking the average cooling rate of all the training sampling points as input and the numerical values of the performance of all the training sampling points as output to obtain the performance prediction model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211660049.0A CN115985424A (en) | 2022-12-23 | 2022-12-23 | BP neural network-based part heat treatment performance prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211660049.0A CN115985424A (en) | 2022-12-23 | 2022-12-23 | BP neural network-based part heat treatment performance prediction method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115985424A true CN115985424A (en) | 2023-04-18 |
Family
ID=85971733
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211660049.0A Pending CN115985424A (en) | 2022-12-23 | 2022-12-23 | BP neural network-based part heat treatment performance prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115985424A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116449790A (en) * | 2023-06-16 | 2023-07-18 | 江苏省沙钢钢铁研究院有限公司 | Production control method of wide and thick plate |
-
2022
- 2022-12-23 CN CN202211660049.0A patent/CN115985424A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116449790A (en) * | 2023-06-16 | 2023-07-18 | 江苏省沙钢钢铁研究院有限公司 | Production control method of wide and thick plate |
| CN116449790B (en) * | 2023-06-16 | 2023-09-05 | 江苏省沙钢钢铁研究院有限公司 | Production Control Method of Wide and Thick Plate |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Investigation of the mechanisms by which hot isostatic pressing improves the fatigue performance of powder bed fused Ti-6Al-4V | |
| Alabort et al. | Superplasticity in Ti–6Al–4V: Characterisation, modelling and applications | |
| Li et al. | Modeling hot deformation behavior of low-cost Ti-2Al-9.2 Mo-2Fe beta titanium alloy using a deep neural network | |
| Sun et al. | Experimental-numerical study of laser-shock-peening-induced retardation of fatigue crack propagation in Ti-17 titanium alloy | |
| Cho et al. | A study on the hot-deformation behavior and dynamic recrystallization of Al–5 wt.% Mg alloy | |
| WO2014033928A1 (en) | Material organization predict device, product fabrication method, and material organization predict method | |
| JP2005315703A (en) | Method for predicting material in steel material | |
| Zhang et al. | Study of dynamic recrystallization behavior of T2 copper in hot working conditions by experiments and cellular automaton method | |
| Cui et al. | Microstructure distribution and mechanical properties prediction of boron alloy during hot forming using FE simulation | |
| Golling et al. | Characterization of ductile fracture properties of quench-hardenable boron steel: Influence of microstructure and processing conditions | |
| CN115985424A (en) | BP neural network-based part heat treatment performance prediction method and system | |
| Qiao et al. | A comparative study on arrhenius equations and BP neural network models to predict hot deformation behaviors of a hypereutectoid steel | |
| Feng et al. | Three different mathematical models to predict the hot deformation behavior of TA32 titanium alloy | |
| Bombač et al. | Amelioration of surface cracking during hot rolling of AISI D2 tool steel | |
| Kappatos et al. | Assessment of the effect of existing corrosion on the tensile behaviour of magnesium alloy AZ31 using neural networks | |
| Kang et al. | Corrosion fatigue behaviors of HSB800 and its HAZs in air and seawater environments | |
| Herbst et al. | Holistic consideration of grain growth behavior of tempering steel 34CrNiMo6 during heating processes | |
| Chen et al. | A comparative study of a back propagation artificial neural network and a Zerilli–Armstrong model for pure molybdenum during hot deformation | |
| Ghiabakloo et al. | Surrogate-based Pareto optimization of annealing parameters for severely deformed steel | |
| Cao et al. | Rolling force prediction for strip casting using theoretical model and artificial intelligence | |
| Meshkat et al. | A study on non-isothermal static recrystallization during hot rolling of carbon steels | |
| Anelli et al. | Microstructural Prediction of Heat Treated Steel Forgings for Severe Applications | |
| Prasad et al. | Constitutive modelling and formability prediction of Inconel-718 sheet at elevated temperatures for space applications | |
| Thankachan et al. | Artificial neural network modeling to evaluate and predict the mechanical strength of duplex stainless steel during casting | |
| Jagadish et al. | Prediction of mechanical properties of fe 415 steel in hot rolling process using artificial neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |