Gear heat treatment performance control method

Through finite element simulation and BP neural network optimization of gear heat treatment technology, the low efficiency and insufficient accuracy of trial and error methods in gear heat treatment technology are solved, precise control of hardened layer depth and distribution is achieved, and the anti-fatigue performance and processing of gears are improved. efficiency.

CN119066904APending Publication Date: 2024-12-03NANJING UNIV OF AERONAUTICS & ASTRONAUTICS
View PDF 0 Cites 8 Cited by

Patent Information

Application Number
CN202410910635.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-09
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

In the gear heat treatment process, there are problems such as high cost, low efficiency and insufficient intelligence in the trial and error method control performance, and it is difficult to accurately control the depth and distribution of the hardened layer, resulting in insufficient or excessive surface hardness and wear resistance, affecting the service life and accuracy of the gear.

Method used

The finite element method is used to simulate the heat treatment process of gears, combined with BP neural network and non-internal genetic algorithm, a gear heat treatment performance prediction model is constructed, the carburizing and quenching process parameters are optimized, and the hardening layer depth and distribution are achieved to improve the fatigue performance of gears. and heat treatment accuracy.

Benefits of technology

By precisely controlling the depth and distribution of the hardened layer, the gear deformation is reduced, the convex distribution of the hardened layer is improved, the fatigue resistance and processing efficiency of the gear are improved, the cost is reduced, and the load-bearing capacity and service life of the gear are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066904A_ABST
    Figure CN119066904A_ABST
Patent Text Reader

Abstract

The invention discloses a gear heat treatment performance control method, and relates to the field of gear heat treatment. The method is used for overcoming the defects that a heat treatment process test trial and error method is high in performance control cost, low in efficiency and not intelligent enough. The method specifically comprises the following steps that 1, a control target of gear heat treatment performance is determined, and main heat treatment process parameters influencing the performance are determined; 2, obtaining sample data; 3, obtaining a prediction model of the heat treatment performance of the gear; 4, an optimization function of hardened layer depth errors, carburizing time and deformation is constructed, and boundary conditions enabling hardened layer distribution to be in a convex shape are added; and 5, carrying out reinforcement learning on the prediction model of the gear heat treatment performance by adopting a non-dominated genetic algorithm NSGA-II to obtain an optimal set of heat treatment process parameters. According to the invention, the distribution curve of the carburized layer is in convex distribution by increasing the boundary conditions of the carburized layer depth and the carburizing time in the strong carburizing and diffusion processes, so that the anti-fatigue performance of the gear is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of gear heat treatment, and in particular to a gear heat treatment performance control method. Background Art

[0002] Gears are frequently used and widely used in various mechanical systems. The quality of gears greatly affects the stability and safety of the entire system of engineering machinery. Gears need to withstand large cyclic stress and impact loads during service, so gears are required to have good mechanical properties such as high surface hardness, high wear resistance and high core toughness.

[0003] Carburizing heat treatment is one of the most commonly used methods for gear surface strengthening. The gear is heated and kept warm in a carburized active medium to allow active carbon atoms to penetrate into the surface of the gear teeth to increase the carbon content on the surface of the gear teeth and obtain a certain carbon concentration gradient; after carburizing, the gear is quenched to obtain acicular martensite structure, acicular martensite and lath martensite mixed structure, and lath martensite structure from the surface to the core of the gear, so that the gear obtains a surface with high hardness and wear resistance and forms a good match with the high-toughness core structure, thereby achieving the goals of high load-bearing capacity, high fatigue performance, and high quality of the gear.

[0004] However, during the carburizing process of gears, it is very important to control the depth and distribution of the hardened layer. If the depth of the hardened layer is too shallow, the surface hardness and wear resistance of the gear cannot be effectively improved; if the depth of the hardened layer is too deep, it will cause problems such as surface cracks and deformation, affecting the service life and performance of the gear. Therefore, in the carburizing process, factors such as temperature, time and concentration need to be strictly controlled to ensure the accurate control of the depth of the hardened layer. In addition, studies have shown that the convex distribution of the hardened layer exhibits better fatigue resistance than the concave or linear distribution, and the relatively gentle hardness gradient can enhance the bonding strength between the carburized layer and the matrix, improving the reliability of the gear. At the same time, the uneven carbon content of the gear after carburizing and the drastic temperature changes in the quenching stage will cause thermal stress and structural stress inside the gear, which will cause the gear to deform significantly under the action of multiple stresses. The large deformation not only affects the gear accuracy, but also increases the difficulty of subsequent grinding, thereby reducing the bearing capacity and service life of the gear. Therefore, controlling the heat treatment performance of the gear has long been a technical focus worthy of attention and improvement.

[0005] How to find the appropriate heat treatment process parameters to control the heat treatment performance of gears is a complex nonlinear programming problem. In the traditional heat treatment process design, there is a problem of complex process and reliance on experience. Usually, the trial and error method is used to continuously conduct heat treatment process tests to adjust the process plan, which often leads to the difficulty of improving the heat treatment process, high cost, low preparation efficiency, and poor product quality. Therefore, such process design methods are relatively backward, and heat treatment intelligent optimization methods have become a hot spot for development. Summary of the invention

[0006] In view of the above problems, the present invention proposes a gear heat treatment performance control method to solve the defects of high cost, low efficiency and lack of intelligence in the trial and error method of heat treatment process test performance control.

[0007] The technical solution of the present invention is: specifically comprising the following steps:

[0008] Step 1: Determine the control target of gear heat treatment performance, and determine the main heat treatment process parameters that affect the performance based on the control target;

[0009] According to the gear load-bearing performance requirements, the carburizing time, gear hardening layer depth and deformation are controlled by optimizing the carburizing time t b and diffusion time t d , while comprehensive adjustments include strong carburizing potential C b and diffusion carbon potential C d The carbon potential and carburizing temperature T b , achieve the target carburized layer, realize the precise control of the depth and distribution of the hardened layer, improve the fatigue performance of the gear, and consider the quenching temperature T of the subsequent quenching process q and quenching oil temperature T m The influence of factors such as heat treatment can be controlled to control gear deformation and improve gear heat treatment accuracy;

[0010] Step 2: Select heat treatment process parameters for different gears, use finite element method to simulate the gear carburizing and quenching process, and obtain sample data; select multi-level gear heat treatment process parameters. Different from the traditional heat treatment process test method to obtain sample data, this method uses finite element method to simulate the gear carburizing and quenching process, obtain the gear carbon content distribution information and deformation under different gear heat treatment process parameter combinations after heat treatment, and obtain sample data;

[0011] Step 3: construct a BP neural network structure, and use the sample data to train the BP neural network structure to obtain a prediction model for gear heat treatment performance;

[0012] According to the sample data obtained by numerical simulation, a backward feedback BP neural network structure is constructed. A large amount of sample data is used to train the BP neural network structure, and a mapping relationship between gear heat treatment process parameters and performance is established to obtain a prediction model for gear heat treatment performance (carburizing time, gear hardening layer depth and deformation).

[0013] Construct a BP neural network structure, and use the sample data to train the BP neural network structure. During the training process, the BP neural network pushes the error between the sample result and the generated result back to the weight in the network, completes the input-output mapping, and obtains a prediction model for the heat treatment performance of the gear. Select process parameters that have a great impact on carburizing (strong carburizing potential and strong carburizing time) and process parameters that directly affect gear quenching deformation (quenching temperature and quenching oil temperature) for research, and obtain the neural network structure of the gear hardening layer depth and maximum deformation;

[0014] Step 4: Taking the hardened layer depth error, carburizing time and deformation after gear heat treatment as control targets, construct optimization functions of hardened layer depth error, carburizing time and deformation, and add boundary conditions that make the hardened layer distribution convex;

[0015] The optimization function constructed is shown in formula (1):

[0016]

[0017] In the formula, h represents the depth of hardened layer, h target Denotes the target value of the hardened layer depth. The objective function y1 indicates that the closer the hardened layer depth is to the target value, the better. The objective function y2 indicates that the shorter the carburizing time during the heat treatment process, the better. max represents the maximum deformation under a certain heat treatment process parameter, and the objective function y3 indicates that the smaller the maximum deformation, the better; X is the design variable, X = [t b ,T b ,C b ,t d ,C d ,T q ,T m ], X L , X U are the upper and lower limits of the design variables;

[0018] The boundary conditions are shown in formula (6):

[0019]

[0020] Among them, x b is the depth of the seepage layer produced in the strong seepage stage, x d is the depth of the permeation layer produced during the diffusion stage, t b is the penetration time, t d is the diffusion time;

[0021] Step 5: Use the non-dominated genetic algorithm NSGA-II to perform reinforcement learning on the prediction model of gear heat treatment performance to obtain the optimal set of heat treatment process parameters, so that the gear can achieve good performance under this set of process parameters and achieve the gear heat treatment performance control target.

[0022] In step 2, for the carburizing heat treatment method of "intensive carburizing + diffusion", the depth of the hardened layer, carburizing time and gear deformation are used as the control targets of the gear heat treatment performance. Among them, the carburizing time is generally proportional to the depth of the hardened layer. Therefore, the carburizing time is adjusted according to the requirements of the carburizing layer depth, and the carbon potential and carburizing temperature are adjusted comprehensively to achieve accurate control of the carburizing layer depth; the heat treatment process parameters finally selected include: intensive carburizing time t b , strong penetration temperature T b , strong carburizing potential C b , diffusion time t d , diffusion carbon potential C d , quenching temperature T q and quenching oil temperature T m .

[0023] In step 2, the finite element method is used to simulate the multi-level process parameter set to obtain results such as hardened layer depth and deformation. The simulation uses a single tooth model, and different heat treatment process parameters are input through DOS batch processing.inp files. The target value is extracted from the simulation result.Odb file using Python language, and the heat treatment process parameters are obtained as input sample data. The hardened layer depth and deformation are obtained as output sample data, so as to construct the initial training sample data set simply and efficiently.

[0024] Step 3 is as follows:

[0025] Constructing a BP neural network structure based on the obtained sample data, and using the sample data to train the BP neural network structure;

[0026] During the training process, the BP neural network pushes the error between the sample results and the generated results back to the weights in the network, completes the mapping from input to output, and obtains the prediction model of gear heat treatment process parameters and performance;

[0027] The gear heat treatment performance prediction model includes a neural network structure prediction model for the gear hardening layer depth and maximum deformation obtained by studying the process parameters that have a great influence on carburization, namely, strong carburization potential and strong carburization time, and the process parameters that directly affect the gear quenching deformation, namely, quenching temperature and quenching oil temperature;

[0028] Finally, find appropriate heat treatment process parameters according to the heat treatment performance control objectives.

[0029] The boundary conditions in step 4 are obtained by following the steps below:

[0030] The distribution of the hardened layer is consistent with the carbon concentration distribution. According to the carburizing diffusion process conforming to Fick's second law, its three-dimensional mathematical expression is obtained, as shown in formula (2):

[0031]

[0032] In the formula, t is the carburizing time; C is the carbon content; x, y, z represent the distance in the diffusion direction of carbon atoms; D is the diffusion coefficient of carbon atoms in iron-based alloys, which is generally taken as a function of temperature and carbon content, as shown in formula (3):

[0033]

[0034] The carbon concentration distribution law is analyzed from the one-dimensional form of the carburizing diffusion process, as shown in formula (4):

[0035]

[0036] In order to make the carbon concentration distribution curve convex, the second-order derivative of carbon concentration with respect to the diffusion distance of carbon atoms should be less than zero. In addition, during the carburizing diffusion process, the carbon concentration decreases with the increase of the diffusion distance. Combining the above factors, the boundary condition shown in formula (5) is obtained:

[0037]

[0038] For the carburizing heat treatment method of "strong carburizing + diffusion", the carbon concentration distribution can be regarded as the superposition of two stages. Therefore, the boundary condition for making the hardened layer convex under strong carburizing + diffusion treatment can be obtained as shown in formula (6):

[0039]

[0040] Among them, x b is the depth of the seepage layer produced in the strong seepage stage, x d is the depth of the permeation layer produced during the diffusion stage, t b is the penetration time, t d is the diffusion time.

[0041] In step 5, the boundary conditions and objective functions in step 4 are specifically combined, and the NSGA-II genetic algorithm based on the Pareto optimal concept is adopted to set the population size and the maximum evolutionary generation. Three optimization design parameter combinations are considered respectively, including: hardening layer depth error and carburizing time, hardening layer depth error and deformation, carburizing time and deformation. A multi-objective optimization model based on the NSGA-II genetic algorithm is established. First, the Pareto frontier optimal solution set of the dual-objective heat treatment process parameter optimization is obtained, and then the Pareto frontier optimal solution set of the three-objective heat treatment process parameter optimization is obtained. Finally, the optimal heat treatment process scheme is obtained, so as to control the gear to obtain good performance under this process.

[0042] The present invention adopts the finite element method instead of the traditional process test method to simulate the gear heat treatment process, efficiently obtains the simulation results under a large number of different heat treatment process parameters, and uses the sample data to train the BP neural network, establishes an agent model between the heat treatment process parameters and the performance, and uses the non-dominated sorting genetic algorithm NSGA-Ⅱ to determine the optimal process parameters to achieve the purpose of controlling the gear performance; finally, the optimal heat treatment process scheme is obtained, thereby controlling the gear to obtain good performance under the process.

[0043] The present invention increases the boundary conditions of the carburizing layer depth and carburizing time during the strong infiltration and diffusion process, so that the carburizing layer distribution curve is convex, thereby improving the performance of the gear, which is specifically manifested as follows: first, the deformation of the gear after heat treatment is small; second, the hardened layer, that is, the carburized layer, can reach the target carburizing layer; third, the carburizing time is shortened, which can greatly improve the processing efficiency, BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic flow chart of the gear heat treatment performance control method provided by the present invention;

[0045] Figure 2 is the depth of the single tooth hardening layer of the gear heat treatment simulation provided by the present invention;

[0046] Figure 3 It is the BP neural network topology structure of the gear hardening layer depth provided by the present invention;

[0047] Figure 4 The BP neural network topology structure of the maximum deformation of the gear heat treatment provided by the present invention;

[0048] Figure 5 It is the Pareto frontier optimal solution set for the dual-objective heat treatment process parameter optimization provided by the present invention;

[0049] Figure 6 It is the Pareto frontier optimal solution set for the three-objective heat treatment process parameter optimization provided by the present invention; DETAILED DESCRIPTION

[0050] In order to clearly illustrate the technical features of this patent, this patent is elaborated in detail below through a specific implementation method and in combination with its accompanying drawings.

[0051] Figure 1 is a schematic flow chart of a gear heat treatment performance control method provided by an embodiment of the present invention, such as Figure 1 As shown, including:

[0052] Step 1: First determine the control target of gear heat treatment performance, and determine the main heat treatment process parameters that affect the performance based on the control target;

[0053] Step 2: Select different gear heat treatment process parameters, use the finite element method to simulate the gear carburizing and quenching process, and obtain sample data;

[0054] Step 3: construct a BP neural network structure, and use the sample data to train the BP neural network structure to obtain a prediction model for gear heat treatment performance;

[0055] Step 4: Taking the hardened layer depth error, carburizing time and deformation amount after gear heat treatment as control targets, construct the optimization function of hardened layer depth error, carburizing time and deformation amount, and add boundary conditions that make the hardened layer distribution convex;

[0056] Step 5: Use the non-dominated genetic algorithm NSGA-II to perform reinforcement learning on the prediction model of gear heat treatment performance to obtain the optimal set of heat treatment process parameters.

[0057] This embodiment takes a certain type of face gear made of 9310 steel as an example. The forging is formed by hot forging process. After normalizing + tempering, the gear workpiece is processed by gear shaping. The carburizing heat treatment process technology requires an effective hardened layer depth of 1.1 to 1.3 mm. In order to leave a certain grinding allowance and facilitate the establishment of a multi-objective optimization function, the effective hardened layer depth is selected as 1.2 mm.

[0058] Based on the above implementation examples, the performance targets that need to be controlled during gear heat treatment are first determined, including: hardened layer depth, carburizing time and gear deformation;

[0059] Since the heat treatment process is affected by many factors, in order to reduce the amount of sample data that needs to be generated, the factors that have a great impact on the control target are selected according to the processing process, and the main heat treatment process parameters that affect the performance are determined;

[0060] Specifically, the main heat treatment process parameters determined in the embodiment of the present invention include: strong penetration time t b , strong penetration temperature T b , strong carburizing potential C b , diffusion time t d , diffusion carbon potential C d , quenching temperature T q and quenching oil temperature T m .

[0061] Secondly, different heat treatment process parameters of the gear are obtained, and the finite element method is used to simulate the process parameter set to obtain results such as hardening layer depth and deformation, and to construct training sample data.

[0062] Based on the above implementation example, a single tooth model is used to perform heat treatment simulation. Figure 2As shown in the figure. Since a large amount of simulated data is required to train the neural network model, the simulation files of ABAQUS are batch processed. This includes: batch processing of .inp files through DOS, inputting different sets of heat treatment process parameters; using Python language to extract the target value of the simulation result .Odb file to obtain the hardened layer depth and deformation.

[0063] The heat treatment process parameter set is used as input sample data, and the hardened layer depth and maximum deformation obtained by simulation post-processing are used as output sample data to obtain training sample data. Part of the obtained data is shown in Table 1 and Table 2:

[0064] Table 1 Depth of hardened layer under different heat treatment parameters

[0065]

[0066] Table 2 Maximum deformation under different heat treatment parameters

[0067] Serial number <![CDATA[C b (%)]]> <![CDATA[t b (s)]]> <![CDATA[T q (℃)]]> <![CDATA[T m (℃)]]> <![CDATA[De max (mm)]]> 1 1.05 21600 860 40 0.31161 2 1.05 21600 860 60 0.3114 3 0.95 43200 820 40 0.30414 4 0.95 43200 820 60 0.30403

[0068] Then, a BP neural network structure is constructed according to the obtained sample data, and the BP neural network structure is trained using the sample data;

[0069] During the training process, the BP neural network pushes the error between the sample results and the generated results back to the weights in the network, completes the mapping from input to output, and obtains the prediction model of gear heat treatment process parameters and performance;

[0070] The gear heat treatment performance prediction model includes a neural network structure prediction model for the gear hardening layer depth and maximum deformation obtained by studying the process parameters that have a great impact on carburizing (strong carburizing potential and strong carburizing time) and the process parameters that directly affect the gear quenching deformation (quenching temperature and quenching oil temperature).

[0071] Specifically, the BP neural network topology diagram of the gear hardening layer depth constructed in the embodiment of the present invention is as follows: Figure 3 As shown; the BP neural network topology structure of the maximum deformation of the gear heat treatment constructed in the embodiment of the present invention is as follows Figure 4 shown.

[0072] Finally, find the appropriate heat treatment process parameters according to the heat treatment performance control target;

[0073] Taking the gear heat treatment hardening layer depth error, carburizing time and maximum deformation as the control targets, a nonlinear optimization model is constructed.

[0074]

[0075] In the formula, h represents the depth of hardened layer, htarget Denotes the target value of the hardened layer depth. The objective function y1 indicates that the closer the hardened layer depth is to the target value, the better. The objective function y2 indicates that the shorter the carburizing time during the heat treatment process, the better. max represents the maximum deformation under a certain heat treatment process parameter, and the objective function y3 indicates that the smaller the maximum deformation, the better; X is the design variable, X = [t b ,T b ,C b ,t d ,C d ,T q ,T m ], X L , X U are the upper and lower limits of the design variables, where, in this specific embodiment, X L =[21600,850,0.95,10800,0.75,830,25],X U =[43200,950,1.05,18000,0.85,860,60].

[0076] Since the hardened layer distribution has better fatigue resistance, the boundary condition that makes the hardened layer distribution convex is added, so that the tooth surface is convexly distributed when it reaches the target carburized layer:

[0077] The distribution of the hardened layer is consistent with the carbon concentration distribution. According to the carburizing diffusion process conforming to Fick's second law, its three-dimensional mathematical expression is obtained, as shown in formula (2):

[0078]

[0079] In the formula, t is the carburizing time; C is the carbon content; x, y, z represent the distance in the diffusion direction of carbon atoms; D is the diffusion coefficient of carbon atoms in iron-based alloys, which is generally taken as a function of temperature and carbon content, as shown in formula (3):

[0080]

[0081] The carbon concentration distribution law is analyzed from the one-dimensional form of the carburizing diffusion process, as shown in formula (4):

[0082]

[0083] In order to make the carbon concentration distribution curve convex, the second-order derivative of carbon concentration with respect to the diffusion distance of carbon atoms should be less than zero. In addition, during the carburizing diffusion process, the carbon concentration decreases with the increase of the diffusion distance. Combining the above factors, the boundary condition shown in formula (5) is obtained:

[0084]

[0085] For the carburizing heat treatment method of "strong carburizing + diffusion", the carbon concentration distribution can be regarded as the superposition of two stages. Therefore, the boundary condition for making the hardened layer convex under strong carburizing + diffusion treatment can be obtained as shown in formula (6):

[0086]

[0087] Among them, x b is the depth of the seepage layer produced in the strong seepage stage, x d is the depth of the permeation layer produced during the diffusion stage, t b is the penetration time, t d is the diffusion time.

[0088] The NSGA-II genetic algorithm based on the Pareto optimal concept is used to set the population size and the maximum evolutionary number. In this embodiment, the NSGA-II genetic algorithm parameters are set as follows: the population size is 100, and the maximum evolutionary number is 200. Three optimization design parameter combinations are considered respectively, including: hardening layer depth error and carburizing time, hardening layer depth error and deformation, and carburizing time and deformation.

[0089] In this specific embodiment, the Pareto frontier optimal solution sets for the dual-objective heat treatment process parameter optimization, such as the hardening layer depth error and carburizing time, the hardening layer depth error and deformation, and the carburizing time and deformation, are respectively as follows: Figure 5 (a) Figure 5 (b) and Figure 5 The optimization results are shown in Table 3.

[0090] Table 3 Optimization results of dual-objective heat treatment process parameters

[0091] index Process parameter combination Optimization value A (1.05,0.80,11507,38681,934,845,40) (0.035,50189) B (1.05,0.82,14376,37535,942,830,55) (0.053,0.305) C (1.03,0.77,10801,21600,852,830,51) (32401,0.304)

[0092] The Pareto frontier optimal solution set for the optimization of the target heat treatment process parameters in this specific embodiment is as follows: Figure 6 As shown in Table 4, the optimal heat treatment process scheme is obtained, so as to control the gear to obtain good performance under this process.

[0093] Table 4 Optimal heat treatment process scheme

[0094] Carburizing temperature Strong carburizing potential Infiltration time Diffusion carbon potential Diffusion time Quenching temperature Quenching oil temperature 920℃ 1.045% 8h 0.75% 3h 820℃ 45℃

[0095] There are many specific implementation ways of the present invention. The above is only the preferred implementation mode of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the principle of the present invention. These improvements should also be regarded as the protection scope of the present invention.

Claims

1. A method for controlling gear heat treatment performance, characterized in that: The specific steps include: Step 1: Determine the control target of gear heat treatment performance, and determine the main heat treatment process parameters that affect the performance based on the control target; Taking carburizing time, gear hardening layer depth and deformation as control targets, the intensive carburizing time t b and diffusion time t d , while comprehensive adjustments include strong carburizing potential C b and diffusion carbon potential C d The carbon potential and carburizing temperature T b , achieve the target carburized layer, realize the precise control of the depth and distribution of the hardened layer, improve the fatigue performance of the gear, and consider the quenching temperature T of the subsequent quenching process q and quenching oil temperature T m The influence of factors such as heat treatment can be controlled to control gear deformation and improve gear heat treatment accuracy; Step 2: Select heat treatment process parameters for different gears, use finite element method to simulate the gear carburizing and quenching process, and obtain sample data; Step 3: construct a BP neural network structure, and use the sample data to train the BP neural network structure to obtain a prediction model for gear heat treatment performance; Step 4: Taking the hardened layer depth error, carburizing time and deformation amount after gear heat treatment as control targets, construct the optimization function of hardened layer depth error, carburizing time and deformation amount, and add boundary conditions that make the hardened layer distribution convex; The optimization function constructed is shown in formula (1): In the formula, h represents the depth of hardened layer, h target Denotes the target value of the hardened layer depth. The objective function y1 indicates that the closer the hardened layer depth is to the target value, the better. The objective function y2 indicates that the shorter the carburizing time during the heat treatment process, the better. max represents the maximum deformation under a certain heat treatment process parameter, and the objective function y3 indicates that the smaller the maximum deformation, the better; X is the design variable, X = [t b ,T b ,C b ,t d ,C d ,T q ,T m ], X L , X U are the upper and lower limits of the design variables; The boundary conditions are shown in formula (6): Among them, x b is the depth of the seepage layer produced in the strong seepage stage, x d is the depth of the permeation layer produced during the diffusion stage, t b is the penetration time, t d is the diffusion time; Step 5: Taking the hardened layer depth error, carburizing time and deformation after gear heat treatment as the control targets, the non-dominated genetic algorithm NSGA-II is used to perform reinforcement learning on the prediction model of gear heat treatment performance to obtain the optimal set of heat treatment process parameters.

2. A gear heat treatment performance control method according to claim 1, characterized in that: The heat treatment process parameters finally selected in step 2 include: strong penetration time t b , strong penetration temperature T b , strong carburizing potential C b , diffusion time t d , diffusion carbon potential C d , quenching temperature T q and quenching oil temperature T m .

3. A gear heat treatment performance control method according to claim 1, characterized in that: In step 2, the finite element method is used to simulate the multi-level process parameter set to obtain results such as hardened layer depth and deformation. The simulation uses a single tooth model, and different heat treatment process parameters are input through DOS batch processing. Inp files, the simulation result. Odb file is extracted from the target value using Python language, and the heat treatment process parameters are obtained as input sample data, and the hardened layer depth and deformation are obtained as output sample data.

4. A gear heat treatment performance control method according to claim 1, characterized in that: Step 3 is as follows: Constructing a BP neural network structure based on the obtained sample data, and using the sample data to train the BP neural network structure; During the training process, the BP neural network pushes the error between the sample results and the generated results back to the weights in the network, completes the mapping from input to output, and obtains the prediction model of gear heat treatment process parameters and performance; The gear heat treatment performance prediction model includes a neural network structure prediction model for the gear hardening layer depth and maximum deformation obtained by studying the process parameters that have a great influence on carburization, namely, strong carburization potential and strong carburization time, and the process parameters that directly affect the gear quenching deformation, namely, quenching temperature and quenching oil temperature; Finally, find appropriate heat treatment process parameters according to the heat treatment performance control objectives.

5. The gear heat treatment performance control method according to claim 1, characterized in that: The boundary conditions in step 4 are obtained by following the steps below: The distribution of the hardened layer is consistent with the carbon concentration distribution. According to the carburizing diffusion process conforming to Fick's second law, its three-dimensional mathematical expression is obtained, as shown in formula (2): In the formula, t is the carburizing time; C is the carbon content; x, y, z represent the distance in the diffusion direction of carbon atoms; D is the diffusion coefficient of carbon atoms in iron-based alloys, which is generally taken as a function of temperature and carbon content, as shown in formula (3): The carbon concentration distribution law is analyzed from the one-dimensional form of the carburizing diffusion process, as shown in formula (4): In order to make the carbon concentration distribution curve convex, the second-order derivative of carbon concentration with respect to the diffusion distance of carbon atoms should be less than zero. In addition, during the carburizing diffusion process, the carbon concentration decreases with the increase of the diffusion distance. Combining the above factors, the boundary condition shown in formula (5) is obtained: The carbon concentration distribution can be regarded as the superposition of two stages. The boundary condition for the hardened layer to be convex is shown in formula (6): Among them, x b is the depth of the seepage layer produced in the strong seepage stage, x d is the depth of the permeation layer produced during the diffusion stage, where t b is the penetration time, t d is the diffusion time.

6. A gear heat treatment performance control method according to claim 1, characterized in that: In step 5, the boundary conditions and objective functions in step 4 are specifically combined, and the NSGA-II genetic algorithm based on the Pareto optimal concept is adopted to set the population size and the maximum evolutionary generation. Three optimization design parameter combinations are considered respectively, including: hardening layer depth error and carburizing time, hardening layer depth error and deformation, carburizing time and deformation. A multi-objective optimization model based on the NSGA-II genetic algorithm is established. First, the Pareto frontier optimal solution set of the dual-objective heat treatment process parameter optimization is obtained, and then the Pareto frontier optimal solution set of the three-objective heat treatment process parameter optimization is obtained. Finally, the optimal heat treatment process scheme is obtained, so as to control the gear to obtain good performance under this process.

Citation Information

Cited By

  • Technological method for guaranteeing side taper angle of gear assembly

    CN121267565A

  • A process for ensuring side cone angle of gear assemblies

    CN121267565B

  • Micro-deformation adjustment control method for nitrocarburizing treatment of insert needle cylinder

    CN121411367A

  • High-precision machining method for complex curved surface gear

    CN121607717A

  • Heat treatment process optimization method of heavy-duty gear and product prepared by same

    CN122311013A