Quenching and tempering evaluation method for 42CrMo quenched and tempered steel crankshaft

Abstract

The invention discloses a hardening and tempering evaluation method for a 42CrMo hardened and tempered steel crankshaft. At present, an accurate evaluation method for the hardness of the 42CrMo hardened and tempered steel crankshaft does not exist. According to the hardening and tempering evaluation method, based on a risk quantitative evaluation normal distribution theory, early warning evaluation is carried out on crankshafts produced by different furnaces, crankshaft data produced by different furnaces are formed, and a crankshaft hardness field finite element analysis model and a backfire mechanical performance prediction model are established respectively; performing different tempering process center composite experiments on the crankshaft through a crankshaft hardness field finite element analysis model and a tempering mechanical performance prediction model, and fitting a linear regression equation of tempering hardness, tensile strength, yield strength, quenching temperature and tempering temperature according to response surface characteristics; and optimizing the linear regression equation through a Matlab genetic algorithm to obtain a standard equation, and comparing the standard equation with actually obtained crankshaft data of different furnace production to complete the evaluation process.

Description

Technical Field

[0001] The invention specifically relates to a tempering evaluation method for a 42CrMo tempered steel crankshaft, and belongs to the technical field of internal combustion engines. Background Art

[0002] As a core component of an internal combustion engine, the performance requirements of the crankshaft are directly related to the engine's reliability, durability, and efficiency. In particular, hybrid crankshafts for automobiles require ultra-high tensile strength under extreme operating conditions, while the hardness must be controlled within a certain range. This makes the selection of the tempering process extremely difficult, and single experiments are often costly. Some studies have combined machine learning with experiments to effectively control costs and maximize profits.

[0003] The crankshaft tempering process primarily consists of two stages: preliminary heat treatment and tempering. Preparatory heat treatment is typically performed before tempering to eliminate internal stresses and structural inhomogeneities generated during forging or casting, laying a good foundation for subsequent tempering. This step primarily involves normalizing or annealing. Normalizing involves heating the crankshaft to an appropriate temperature, holding it for a period of time, and then air or air cooling. Normalizing aims to eliminate forging stresses, achieve appropriate hardness, and improve the microstructure in preparation for tempering. Annealing may also be required to reduce hardness, improve plasticity and toughness, and create favorable conditions for subsequent machining and tempering. Tempering is a key step in crankshaft heat treatment and includes two main processes: quenching and tempering. Quenching involves heating the crankshaft to above the austenitizing temperature and holding it for a period of time to ensure homogenization of the austenite. Cooling involves rapidly cooling the crankshaft to room temperature or slightly higher to achieve a martensitic structure. The quenching medium can be water, oil, or a salt bath, depending on the crankshaft material and dimensions. Tempering is to heat the crankshaft after quenching to an appropriate temperature of 500-650℃ and keep it for a period of time. Then cool it to room temperature in air or furnace. The purpose of tempering is to eliminate the quenching stress and decompose the martensite into tempered martensite or tempered sorbite to obtain the required strength and toughness.

[0004] Current optimization strategies for quenching and tempering treatments lack a stable and systematic approach to guiding precise adjustments to temperature and time. During quenching and tempering, precise control of heating temperature, holding time, and cooling rate is crucial to minimize structural transformation inhomogeneities and internal stress. Only through standardized quenching and tempering can crankshafts achieve the optimal overall mechanical properties and fatigue strength required to meet the performance requirements of internal combustion engines. Currently, there is limited research on quenching and tempering process optimization using machine learning in the field of hot working, and even less so for crankshafts. Currently, there is no precise method for assessing the hardness of 42CrMo quenched and tempered steel crankshafts. Summary of the Invention

[0005] The present invention aims to solve the problem that there is no accurate evaluation method for the hardness of 42CrMo quenched and tempered steel crankshaft, and further proposes a quenching and tempering evaluation method for 42CrMo quenched and tempered steel crankshaft.

[0006] A quenching and tempering evaluation method for 42CrMo quenched and tempered steel crankshafts is disclosed. After obtaining crankshaft data produced in different furnaces, an early warning evaluation is performed on the crankshafts produced in different furnaces based on the normal distribution theory of risk quantification assessment. The crankshaft data produced in different furnaces are formed, and a finite element analysis model of the crankshaft hardness field and a tempering mechanical property prediction model are respectively established. Composite experiments of different quenching and tempering process centers of the crankshaft are carried out using the crankshaft hardness field finite element analysis model and the tempering mechanical property prediction model. A linear regression equation of tempering hardness, tensile strength, yield strength, quenching temperature, and tempering temperature is fitted according to the response surface characteristics. The linear regression equation is optimized by a Matlab genetic algorithm to obtain a standard equation. The standard equation is compared with the actual crankshaft data obtained from different furnaces to complete the evaluation process.

[0007] As a preferred solution: after obtaining the crankshaft data produced in different batches, the process of conducting early warning assessment on the crankshafts produced in different batches based on the normal distribution theory of risk quantification assessment is as follows: the crankshaft data is subjected to normal distribution assessment processing to screen out hardness values that conform to the normal distribution as target samples, the mathematical expectation μ value of the target samples is calculated, and the probability P of the data point being within a specific interval (μ-kσ, μ+kσ) is evaluated according to the Chebyshev inequality, where k is a constant and σ is the standard deviation of the crankshaft data. For the hardness test of the crankshaft production process, the interval (μ-kσ, μ+kσ) corresponding to the probability P is determined to be the reliable fluctuation range of the crankshaft hardness, and μ-kσ and μ+kσ are used as the lower limit value and upper limit value of the hardness warning, respectively;

[0008] The calculation process of the warning value is:

[0009] According to the normal distribution theory, the boundary value is calculated. When the data X conforms to the general normal distribution, that is, X~N(μ,σ 2 ), the probability that the data in the sample is within the interval is calculated as follows:

[0010]

[0011] This type of general normal distribution is standardized, so It turns out that:

[0012]

[0013] Further simplification yields:

[0014]

[0015] When the data is in the middle of (μ-kσ,μ+kσ), substitute the two end points into the above formula to get:

[0016] P={μ-kσ <X<μ+kσ}=Φ(k)-Φ(-k)=2Φ(k)-1 (5-4)

[0017] Where σ is the data standard deviation, μ is the mathematical expectation, Φ(x) is the standard normal distribution function, and the hardness data probability P is 70%;

[0018] After using spas software to calculate the data standard deviation σ and mathematical expectation μ, the k value is 1.04 according to the table when Φ(k)=0.85, and the hardness theoretical warning lower limit μ-kσ and hardness theoretical warning upper limit μ+kσ can be calculated; when the P value is 70%, 15% of the hardness data sorted by size is the actual warning lower limit, and 85% is the actual warning upper limit.

[0019] As a preferred solution: After calculating the hardness warning lower limit value and the hardness warning upper limit value based on the normal distribution theory, the crankshaft hardness data is evaluated. When the crankshaft hardness data is in the non-corner area, the surface crankshaft hardness is in a safe state. When the crankshaft hardness data is in the corner area, it indicates that the crankshaft hardness is outside the warning value, and the furnace process where the crankshaft is located is in a warning failure state.

[0020] As a preferred solution: a finite element analysis model of crankshaft hardness field and a tempering mechanical property prediction model are established respectively. A composite experiment of crankshaft quenching and tempering process center is carried out using the crankshaft hardness field finite element analysis model and the tempering mechanical property prediction model. The linear regression equation of tempering hardness, tensile strength, yield strength, quenching temperature, and tempering temperature is fitted according to the response surface characteristics as follows:

[0021] Based on the central composite design principle and combined with the quenching and tempering single-factor variable test, a two-factor, five-level design method was used to establish a quantitative assessment model for the crankshaft hardness failure risk. The results were verified based on the quantitative assessment model for the crankshaft hardness failure risk. The quenching temperature and tempering temperature were represented by T1 and T2, respectively, and -2, -1, 0, 1, and 2 were used to represent the variable levels, forming a two-factor, five-level experimental design coding value. The two-factor, five-level experimental design coding value was calculated based on the hardness simulation and tempering performance prediction formula to obtain the average value. The design expert software was used to design a quadratic multinomial fitting and perform a significance test. The fitting equation and the actual temperature formula were obtained as follows:

[0022] HRC=-7125.0142959912+10.816882183927T1+8.5991954023188T2+0.00024999999997561T1T2-0.006517241379313T12 -0.0072672413793098T2 2

[0023] Rm=-110415.65775901+164.88793103504T1+138.16172413843T2+0.005999999999396T1T2-0.10003448275874T1 2 -0.11803448275862T2 2

[0024] Re=-81624.416379443+209.7697988507T1-18.019137930769T2+0.0014999999997T1T2-0.12501724137927T1 2 +0.01248275862068T2 2 ;

[0025] The above calculations show that the tempering mechanical properties prediction model after response surface optimization shows that the various mechanical properties are clearly layered on the response surface projection surface. Combining the 3D surface model and the projection surface, the process range that meets the performance requirements of tempering hardness, tensile strength, yield strength, quenching temperature, and tempering temperature is obtained.

[0026] As a preferred solution: the linear regression equation is optimized by Matlab genetic algorithm to obtain a standard equation. The evaluation process is completed by comparing the standard equation with the actual crankshaft data obtained from different furnace production batches:

[0027] After determining the optimal range of various mechanical properties of the crankshaft, the optimal range of the crankshaft quenching temperature, and the optimal range of the crankshaft tempering temperature, the multi-objective optimization of the crankshaft mechanical properties and the heat treatment process optimization process are completed through the GA toolbox in the MATLAB genetic algorithm, combining the inhibition between the multiple objectives. That is, the minimum value of the function is the optimal value, so when setting the tensile strength and yield strength, the opposite values need to be set. Therefore, the multi-objective optimization function is obtained as follows:

[0028] minf(x)=[f1(x),-f2(x),-f3(x)] (5-5)

[0029] In the above formula, f1(x) is hardness, f2(x) is tensile strength, and f3(x) is yield strength;

[0030] According to the above function, the problem is solved using the Matlab genetic algorithm toolbox GA solver.

[0031] As the preferred solution: Matlab genetic algorithm optimizes the objective function through the Pareto method to ensure that all mechanical properties meet the target requirements, while ensuring that the heat treatment process range is within the result range obtained by the response surface. The optimization results are experimentally verified. When the hardness is reduced, the tensile strength is slightly reduced, while the cross-sectional shrinkage rate and elongation after fracture are slightly improved, indicating that the crankshaft impact toughness and fatigue strength are good, the metallographic grade is relatively higher, and the organizational clarity and uniformity are better than the initial process.

[0032] Beneficial effects of the present invention:

[0033] Based on the normal distribution theory of risk quantification assessment, early warning assessment of crankshafts produced in different furnaces was carried out to form crankshaft data produced in different furnaces, and a finite element analysis model of crankshaft hardness field and a tempering mechanical property prediction model were established respectively. A crankshaft data center composite experiment was carried out using the finite element analysis model of crankshaft hardness field and the tempering mechanical property prediction model. According to the response surface characteristics, the linear regression equation of tempering hardness, tensile strength, yield strength and quenching temperature, tempering temperature was fitted. The linear regression equation was then optimized by the Matlab genetic algorithm to obtain a standard equation. Based on the standard equation, subsequent corresponding evaluation conclusions were drawn, which can form a standardized and reasonable optimization strategy, which is conducive to the accurate evaluation and guidance of the performance and service life of the crankshaft, and is suitable for popularization and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a bar chart comparing the theoretical value and the actual value of the 85% probability interval for Heat 2;

[0035] Figure 2 This is a bar chart comparing the theoretical value and the actual value of the 85% probability interval for heat three;

[0036] Figure 3 This is a bar chart comparing the theoretical and actual values of the 85% probability interval for Heat 4;

[0037] Figure 4 This is a bar chart comparing the theoretical value and the actual value of the 85% probability interval of the heat order;

[0038] Figure 5 It is a structural diagram of the hardness response model in a 3D surface state;

[0039] Figure 6 It is a structural diagram of the hardness response model in the bottom projection state;

[0040] Figure 7 It is a structural diagram of the tensile strength response model in a 3D curved surface state;

[0041] Figure 8It is a structural diagram of the tensile strength response model in the bottom projection state;

[0042] Figure 9 It is a structural diagram of the yield strength response model in a 3D curved surface state;

[0043] Figure 10 It is a structural diagram of the yield strength response model in the bottom projection state;

[0044] Figure 11 Schematic diagram of Pareto solution for multi-objective optimization;

[0045] Figure 12a This is the physical picture of tempered troostite before optimization;

[0046] Figure 12b This is the actual picture of the optimized tempered martensite. DETAILED DESCRIPTION

[0047] To make the purpose, technical solutions and advantages of the invention clearer, the present invention is further described below with reference to specific embodiments and accompanying drawings, but the present invention is not limited to the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0048] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0049] Specific implementation method 1: Combination Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 、 Figure 5 、 Figure 6 、 Figure 7 、 Figure 8 、 Figure 9 、 Figure 10 、 Figure 11 、 Figure 12a and Figure 12bThis embodiment describes the tempering evaluation method for 42CrMo tempered steel crankshafts in this embodiment. After obtaining crankshaft data produced in different furnaces, the crankshafts produced in different furnaces are early warning evaluated based on the normal distribution theory of risk quantification assessment to form crankshaft data produced in different furnaces. The crankshaft data include crankshaft hardness early warning values. A finite element analysis model of the crankshaft hardness field and a tempering mechanical property prediction model are established respectively. A crankshaft data center composite experiment is carried out using the crankshaft hardness field finite element analysis model and the tempering mechanical property prediction model. The linear regression equations of tempering hardness, tensile strength, yield strength, quenching temperature and tempering temperature are fitted according to the response surface characteristics. The linear regression equations are optimized by the Matlab genetic algorithm to obtain a standard equation. The evaluation process is completed by comparing the standard equation with the actual crankshaft data obtained from different furnaces.

[0050] Specific implementation method 2: This implementation method is a further limitation of the specific implementation method 1, combined with Figures 1 to 4 As shown, in this embodiment, the crankshafts produced in four batches are mainly tested for normal distribution, and the statistical analysis data of the hardness of each batch is obtained, which is then used as a reference for the overall hardness warning value. Before applying this theory for statistical analysis, the first step is to perform a normality test on the data set. This test belongs to the category of non-parametric methods and mainly infers the distribution characteristics of the population through sample data. Researchers usually use two widely recognized methods: one is the Kolmogorov-Smirnov test, which is suitable for large-scale samples, often referred to as the KS test, and the other is the Shapiro-Wilk test, referred to as the SW test, which is more suitable for small-scale samples.

[0051] The scale sample size in this embodiment is 40, which is less than 50, making it suitable for SW testing. Each set of data was imported into SPASS software, and the results are shown in Table 1. When the P value is ≥ 0.05, the data set is considered to be normally distributed. The results show that among the four sets of crankshaft hardness data produced in the heat, only one set has a P value of 0.004 < 0.05, which does not conform to the strict normal distribution. The remaining sets have P values ≥ 0.05 that conform to the normal distribution. Therefore, the normal distribution theory calculation is first performed on the data that conform to the normal distribution.

[0052] Table 1 Normal distribution test of crankshaft hardness data from different furnaces

[0053]

[0054] Based on the above analysis, the experimental data from heats 2, 3, and 4 all conform to normal distribution characteristics. For hardness values that conform to a normal distribution, the mathematical expectation μ of the sample data must be calculated. Using Chebyshev's inequality, the probability P of a data point falling within a specific interval (μ - kσ, μ + kσ) can be estimated, where k is a constant and σ is the standard deviation of the sample data. For hardness testing during crankshaft production, industry standards and historical test data can be combined to determine, based on reliability requirements, the interval (μ - kσ, μ + kσ) corresponding to a given probability P as the "reliable" fluctuation range for crankshaft hardness. This range can be defined as the normal fluctuation range, with its upper and lower limits serving as the normal thresholds for quality control. According to reliability screening theory, the k value represents the tolerance, and the hardness threshold can be derived using empirical rules. In practical engineering applications, an 85% probability range is typically used as the normal fluctuation range, with the upper and lower limits, μ - kσ and μ + kσ, serving as hardness warning values.

[0055] In this embodiment, after obtaining the crankshaft data produced in different batches, the process of performing early warning assessment on the crankshafts produced in different batches based on the normal distribution theory of risk quantification assessment is as follows: performing normal distribution assessment on the crankshaft data to screen out hardness values that conform to the normal distribution as target samples, calculating the mathematical expectation μ value of the target samples, and evaluating the probability P of the data point within a specific interval (μ-kσ, μ+kσ) according to the Chebyshev inequality, where k is a constant and σ is the standard deviation of the crankshaft data. For the hardness test of the crankshaft production process, the interval (μ-kσ, μ+kσ) corresponding to the probability P is determined to be the reliable fluctuation range of the crankshaft hardness, and μ-kσ and μ+kσ are respectively used as the lower limit value and upper limit value of the hardness warning;

[0056] The calculation process of the warning value is:

[0057] According to the normal distribution theory, the boundary value is calculated. When the data X conforms to the general normal distribution, that is, X~N(μ,σ 2 ), the probability that the data in the sample is within the interval is calculated as follows:

[0058]

[0059] This type of general normal distribution is standardized, so It turns out that:

[0060]

[0061] Further simplification yields:

[0062]

[0063] When the data is in the middle of (μ-kσ,μ+kσ), substitute the two end points into the above formula to get:

[0064] P = {μ - kσ < X < μ + kσ} = Φ(k) - Φ(-k) = 2Φ(k) - 1 (5 - 4)

[0065] where σ is the standard deviation of the data, μ is the mathematical expectation, Φ(x) is the standard normal distribution function, and the probability P of the hardness data is 70%;

[0066] After calculating the standard deviation σ and the mathematical expectation μ of the data using spass software, the lower limit value μ - kσ and the upper limit value μ + kσ of the hardness warning can be calculated; after calculating the standard deviation σ and the mathematical expectation μ of the data using spass software, the value of k can be found by looking up the table according to Φ(k), and then the lower limit value μ - kσ and the upper limit value μ + kσ of the theoretical hardness warning can be calculated; when the P value is 70%, 15% of the hardness data sorted by size is the actual lower warning limit value, and 85% is the actual upper warning limit value.

[0067] The process of looking up the value of k according to Φ(k) in the above calculation process is an existing technology, and its derivation principle is the same as the existing process of looking up the value of k according to Φ(k), where looking up the table refers to the existing standard normal distribution table corresponding to Φ(k).

[0068] Taking a set of hardness data measured in the third furnace as an example, taking the probability P of the hardness data as 70%, the calculation is carried out according to the following steps:

[0069] Let P = {μ - kσ < X < μ + kσ} = 2Φ(k) - 1 = 0.70, Φ(k) = 0.85. Looking up the normal distribution table, k = 1.04 can be obtained. Using spass software to calculate, the standard deviation σ of the data is 0.89, the mathematical expectation μ is 30.61, μ - kσ = 29.68, and μ + kσ = 31.54. According to the above method, when the P value is 70%, the theoretical and actual warning values when the upper and lower limits of the data interval are 85% and 15% respectively can be calculated. In order to test the universality of this calculation and statistical method, that is, for the crankshafts produced in different furnace batches, whether the hardness warning values obtained by using this method have a certain accuracy, the test data of the second furnace batch and the fourth furnace batch are now statistically analyzed and a normal distribution test is carried out. The results are shown in Table 2 and Figure 2 as shown. The data errors between several groups of theoretical warning values and actual probability values obtained according to the calculation standard distribution method are all less than 3%, meeting the error requirements for the risk assessment of crankshaft hardness failure. <​​​​​​​​​Based on the above analysis, the normal distribution theory warning value calculation method is used to verify the non-normal distribution data in the test results. The hardness data with P value < 0.05 within the first heat are tested for normal distribution. The test results are shown in Table 3 and Figure 3 shown.

[0074] Table 3 Calculation results of hardness failure warning value for heat 1

[0075]

[0076] Similarly, 70% is taken as the reasonable interval for hardness failure. Table 3 shows that the error between the theoretical and actual warning values is less than 2%. This indicates that although the hardness data for Heat 1 does not strictly follow a normal distribution, its statistical characteristics are highly consistent with a normal distribution. The warning value calculation method based on normal distribution theory has high reliability and can provide an effective method for heat treatment failure risk assessment of crankshafts of this material type. In practical applications, this method can well meet the needs of risk assessment while maintaining theoretical rigor and practicality. The theoretical value is derived based on the normal distribution formula. Its range is defined by the upper and lower limits of the warning value and is described by the area enclosed by the normal distribution curve and the x-axis. The actual value is represented by the area enclosed by the normal distribution curve, the upper and lower warning values, and the x-axis, covering 85% of the area. The deviation between the theoretical and actual values is mainly due to the normal distribution theory's assumption of continuous data distribution. Actual test data may not fully conform to this characteristic and may contain data duplication. This discrepancy requires further analysis and correction in practical applications. In the actual production using this method, the warning value is defined as the average value of four furnaces, that is, after calculating the lower limit and upper limit of the hardness warning based on the normal distribution theory, the crankshaft hardness data is evaluated, and it is concluded that when the crankshaft hardness data is in the non-corner area, the surface crankshaft hardness is in a safe state. When the crankshaft hardness data is in the corner area, it indicates that the crankshaft hardness is outside the warning value, and the furnace process where the crankshaft is located is in a warning failure state. This is a classified evaluation form.

[0077] Specific implementation method three: This implementation method is a further limitation of specific implementation methods one or two. The response surface methodology is a statistical method used to optimize process parameters and product design, and is widely used in engineering, chemistry, biology and other fields. The optimal parameter combination can be found through a small number of experiments, significantly reducing the number and time of experiments. It is suitable for the optimization of complex systems with multiple factors and multiple levels. It can simultaneously analyze the impact of multiple factors and their interactions on the response, providing more comprehensive information. The model it establishes is more intuitive. By fitting the mathematical model between the target value and the independent variable, it more intuitively displays the relationship between the dependent variable and the independent variable, making it easier for users to understand and predict. In addition, its prediction accuracy is high, the error range is small, and the results of the system optimization method are reliable. Based on the experiments and mechanical property predictions in the previous chapter, in order to determine the quenching-tempering process for crankshaft heat treatment to obtain excellent mechanical properties, and to study the relationship between the quenching-tempering process and the tensile strength-hardness relationship, this section selects representative conditions, uses the response surface methodology and CCD central composite design method to formulate experiments and optimize the heat treatment process.

[0078] Specific embodiment 4: This embodiment is a further limitation of specific embodiment 3. In this embodiment, a crankshaft hardness field finite element analysis model and a tempering mechanical property prediction model are respectively established based on the crankshaft hardness warning values produced in different batches. A crankshaft data center composite experiment is carried out using the crankshaft hardness field finite element analysis model and the tempering mechanical property prediction model. The linear regression equation process of fitting tempering hardness, tensile strength, yield strength, quenching temperature, and tempering temperature according to the response surface characteristics is as follows:

[0079] Based on the central composite design principle and combined with single-factor variable experiments for quenching and tempering, a two-factor, five-level design approach was used to establish a quantitative assessment model for crankshaft hardness failure risk. Results were calculated based on this quantitative assessment model for crankshaft hardness failure risk. Quenching and tempering temperatures were represented by T1 and T2, respectively, and variable levels were represented by -2, -1, 0, 1, and 2. This formed the coding values for the two-factor, five-level experimental design. The coding values for the two-factor, five-level experimental design were then averaged using the hardness simulation and tempering performance prediction formulas. The specific coding levels are shown in Tables 4, 5, 6, and 7. The experimental results were calculated using the hardness simulation and tempering performance prediction formulas in the previous chapter, with multiple averages taken.

[0080] Table 4 Coding values for two-factor five-level experimental design

[0081]

[0082] Table 5 CCD design test results

[0083]

[0084]

[0085] Table 6

[0086]

[0087] The design expert software was used to design a quadratic multinomial fitting and perform a significance test. The fitting equation was as follows. The actual temperature formula used in this paper is:

[0088] HRC=-7125.0142959912+10.816882183927T1+8.5991954023188T2+0.00024999999997561T1T2-0.006517241379313T1 2 -0.0072672413793098T2 2

[0089] Rm=-110415.65775901+164.88793103504T1+138.16172413843T2+0.005999999999396T1T2-0.10003448275874T1 2 -0.11803448275862T2 2

[0090] Re=-81624.416379443+209.7697988507T1-18.019137930769T2+0.0014999999997T1T2-0.12501724137927T1 2 +0.01248275862068T2 2

[0091] Table 7 Response target value performance evaluation indicators

[0092]

[0093]

[0094] From the response indicators in Table 7, we can see that R 2 The values are all greater than 0.9, R 2 The larger the surface, the better the fit between the response surface and the experiment. 2All are close to 1, indicating that the predicted values are closer to the experimental values, and the degree of fit to the polynomial is higher. The P values of the three targets are all less than 0.05. The P value represents the significance of the target value to the model. A value less than 0.05 indicates that the response result is relatively significant. At the same time, the F values of 16.60, 31.98, and 29.40 are relatively large, further confirming the significance of the model, indicating a high correlation between the various mechanical properties and the heat treatment process. In addition, the AP value determines the accuracy of the response surface, requiring a value greater than 4. The AP values of the three target properties in the table above are much greater than 4, indicating that the response surface model is relatively accurate.

[0095] In this embodiment, the tempering mechanical property prediction model after response surface optimization shows that the various mechanical properties are clearly layered on the response surface projection surface. Combining the 3D surface model and the projection surface, the process range that meets the performance requirements of tempering hardness, tensile strength, yield strength, quenching temperature, and tempering temperature is obtained.

[0096] Specific implementation method 5: This implementation method is a further limitation of specific implementation methods 1, 2, 3 or 4. Figures 5 to 10 As shown, in this embodiment, the linear regression equation is optimized by the Matlab genetic algorithm to obtain a standard equation. The evaluation process is completed by comparing the standard equation with the actual crankshaft data obtained from different furnaces: after determining the optimal range of various mechanical properties of the crankshaft, the optimal range of the crankshaft quenching temperature, and the optimal range of the crankshaft tempering temperature, the multi-objective optimization of the crankshaft mechanical properties and the heat treatment process optimization process are completed by the GA toolbox in the Matlab genetic algorithm in combination with the inhibitory nature between multiple objectives. That is, the minimum value of the function is the optimal value, so when setting the tensile strength and yield strength, the opposite values need to be set. Therefore, the multi-objective optimization function is obtained as follows:

[0097] minf(x)=[f1(x),-f2(x),-f3(x)] (5-5)

[0098] In the above formula, f1(x) is hardness, f2(x) is tensile strength, and f3(x) is yield strength;

[0099] According to the above function, the GA solver of Matlab genetic algorithm toolbox was used to solve the problem. The specific process was as follows: the initial population was set to 100, the evolutionary generations were set to 200, the crossover probability was set to 0.85, the mutation probability was set to 0.04, and the optimal individualization coefficient was set to 0.3. After calculation, the multi-objective optimization results shown in Table 8 were obtained.

[0100] Table 8 Multi-objective optimization model operation results

[0101]

[0102]

[0103] The tensile strength and yield strength are negative in the results, but we take the absolute value of the optimization result. Since the essence of multi-objective optimization is that optimizing one objective function sacrifices the others, it's impossible to achieve optimal values for all three objectives. However, we can achieve the desired multi-objective values based on the conditions we set. Here, we require a hardness less than 34.4 HRC, a tensile strength greater than 900 MPa, and a yield strength greater than 800 MPa.

[0104] Combine Figure 11 As shown in the figure, the points that meet the requirements in the multi-objective optimization solution set are 1, 3, 8, 9, and 16. Among them, the heat treatment process of point 1 is quenching 845.6℃ + tempering 616.9℃, with a hardness of 32.0HRC, a tensile strength of 988.1MPa, and a yield strength of 826.2MPa; the heat treatment process of point 3 is quenching 842.3℃ + tempering 595.0℃, with a hardness of 30.0HRC, a tensile strength of 921.1MPa, and a yield strength of 856.7MPa; the heat treatment process of point 8 is quenching 843.7 ℃+tempering 603.5℃, hardness is 32.4HRC, tensile strength is 975.9MPa, yield strength is 821.5MPa; 9-point heat treatment process is quenching 842.2℃+tempering 598.8℃, hardness is 30.9HRC, tensile strength is 939MPa, yield strength is 848.8MPa, 16-point heat treatment process is quenching 841.9℃+tempering 591.0℃, hardness is 31.3HRC, tensile strength is 947.5MPa, yield strength is 844.5MPa.

[0105] Since the multi-objective genetic algorithm uses the Pareto method to optimize the objective function, the values of these five points are in a parallel range and there is no distinction between good and bad. And according to the optimal range obtained by the response surface in the previous section, the process of these five sets of solutions is exactly within the range of results obtained by the response surface, further proving the accuracy of the response surface model. During the solution process of the genetic algorithm, the randomness of the crossover and mutation operations will cause the results of each run to vary. However, these differences are usually small because the adjustment of each result is often achieved by compromising other objective functions. This shows that when the conditions are met, the results of the above five groups of experiments have no obvious distinction between good and bad, and are in a parallel relationship. Therefore, they can be selected according to actual requirements and specific needs.

[0106] Specific embodiment 6: This embodiment is a further limitation of specific embodiments 1, 2, 3, 4 or 5. The specific process of the result analysis of the response surface model in this embodiment is as follows:

[0107] Combine Figures 5 to 10As shown in the figure, the influence of quenching and tempering temperature on crankshaft hardness, tensile strength and yield strength is obtained by fitting regression equations to obtain the response surface model of various mechanical properties at different quenching and tempering temperatures. Figure 5 It can be seen that in the T1 direction, that is, the quenching temperature axis presents a bowl-shaped downward bend, indicating that the hardness increases first and then decreases as the quenching temperature increases, while in the T2 direction, that is, the tempering temperature axis, as the temperature increases, the hardness value decreases instead. According to the trend range, the quenching temperature has a greater impact on the hardness than the tempering temperature. The red area appears in the middle of the quenching temperature and the lowest tempering temperature position. This position is Figure 6 As shown in the figure, the red area is the position of maximum hardness, and the blue area is the position of minimum hardness. At this time, the quenching temperature is the minimum and the tempering temperature is the maximum. According to the crankshaft performance requirements of this article, the hardness requirement is lower rather than the lowest, so the green area is the temperature range required by this article.

[0108] Combine Figure 7 and Figure 8 As shown, in the response surface of tensile strength in this embodiment, in the T1 direction, that is, the quenching temperature direction, the tensile strength first increases significantly and then slowly decreases as the temperature increases, reaching a maximum value of 940 MPa at around 845°C. In the T2 direction, the tensile strength decreases as the temperature increases, reaching the lowest value at the highest temperature, and presents an arch shape as a whole. According to the 3D diagram, the yellow area appears at the quenching temperature of 845°C and the tempering temperature of 595°C. Figure 8 The bottom projection is more intuitive, with the minimum value at the lowest quenching temperature and the highest tempering temperature. Finally, according to the method of the present invention, the area with high tensile strength and the same change trend of yield strength and tensile strength is selected, that is, the yellow area is the selected range. This area can ensure that relevant data with high tensile strength and consistent change trend of yield strength and tensile strength are obtained.

[0109] Combine Figure 9 and Figure 10 As shown, based on the analysis of the 3D response diagram and the performance requirements of the crankshaft in this article, the hardness requirement is reduced but not the lowest, that is, the green area of the projection diagram, and the higher the tensile strength and yield strength, the better, that is, the yellow area of the projection diagram. Combining the three projection diagrams, it can be seen that the range that meets the three performance requirements is the quenching temperature of 835~850℃ and the tempering temperature of 590~620℃.

[0110] Specific embodiment seven: This embodiment is a further limitation of specific embodiments one, two, three, four, five or six. The Matlab genetic algorithm optimizes the objective function through the Pareto method to ensure that various mechanical properties meet the target requirements, while ensuring that the heat treatment process range is within the result range obtained by the response surface. The optimization results are experimentally verified. When the hardness is reduced, the tensile strength is slightly reduced, and the cross-sectional shrinkage rate and elongation after fracture are slightly improved, indicating that the crankshaft impact toughness and fatigue strength are good, the metallographic grade is relatively higher, and the organizational clarity and uniformity are better than the initial process.

[0111] The optimized value closest to the target process is selected from the Pareto solution of the optimization results of the present invention for verification. The optimal data of quenching and tempering determined by the present invention are: quenching 845.6℃ + tempering 616.9℃, and quenching and holding time is 60min. The mechanical properties and microstructure of the optimized crankshaft product are tested, and the results before optimization are shown in Table 9. Figure 12a and Figure 12b As shown in the figure, the hardness index is reduced from 34.5HRC to 32.0HRC, with a reduction of 7.25%, which meets the target range requirement of crankshaft hardness, and the tensile strength is reduced from 1022MPa to 988MPa, with a reduction of 3.33%, which meets the target requirement of this study. While reducing the hardness, the tensile strength is also slightly reduced, while the section shrinkage and elongation are slightly improved, indicating that the impact toughness and fatigue strength of the product under this target process are good. Figure 12a and Figure 12b As shown, the optimized tempered bainite has a clear morphology, finer and more evenly distributed carbide particles, and more complete precipitation of carbides from the supersaturated solid solution, without any aggregation. Comparison with standard microscope images shows that the optimized product is close to level 4, while the unoptimized product was level 3. The organizational clarity and uniformity are superior to those of the initial process. In summary, the crankshaft specimens after this optimized process have excellent overall performance, meet production requirements, and can significantly improve production yield.

[0112] Table 9 Comparison of mechanical properties before and after optimization

[0113]

[0114] The present invention can realize the quantitative process of failure risk early warning assessment of crankshaft hardness value using normal distribution theory. Based on quenching simulation and tempering mechanical property prediction, a two-factor, five-level test is established. Designexpert software is used to perform response surface analysis and fit the linear regression equation. The equation is optimized for multi-objective functions using the Matlab genetic algorithm toolbox. The results are verified and the following conclusions are drawn:

[0115] First: Based on the normal distribution theory, a crankshaft hardness failure risk assessment and early warning model was established. By looking up the normal distribution table, the K value was found to be 1.04. Analysis of the hardness of crankshafts produced with different furnace numbers showed that the error range between the theoretical warning value and the actual warning value was within 3%, which met the error requirements of the failure risk assessment. The failure warning value was successfully set, which played an early warning inspection role in the crankshaft production tempering process.

[0116] Second: A CCD central composite test was established to fit the linear regression equations between various mechanical properties and quenching and tempering temperatures. The response surface optimization model showed that the various mechanical properties were clearly layered on the response surface projection surface. Combining the 3D surface model and the projection surface, it was concluded that the process range that met the three performance requirements was quenching temperature: 835~850℃, and tempering temperature: 590~620℃.

[0117] The heat treatment process obtained after multi-objective optimization using the Matlab Genetic Algorithm Toolbox selected the first point, which is quenching at 845.6°C and tempering at 616.9°C. The hardness is 32.0 HRC, the tensile strength is 988.1 MPa, and the yield strength is 826.2 MPa. All mechanical properties meet the target requirements, and the heat treatment process range is within the range of results obtained by the response surface analysis. Experimental verification of the optimization results shows a strength reduction of 3.33%, which is far less than the hardness of 7.25%, meeting the target requirements. While reducing the hardness, the tensile strength is slightly reduced, while the cross-sectional shrinkage rate and elongation after fracture are slightly increased, indicating that the product has good impact toughness and fatigue strength under this target process. The metallographic grade is relatively higher, and the structural clarity and uniformity are better than those of the initial process.

Claims

1. A method for evaluating the quenching and tempering of a 42CrMo quenched and tempered steel crankshaft, characterized by: After obtaining the crankshaft data produced in different furnaces, an early warning assessment of the crankshafts produced in different furnaces was carried out based on the normal distribution theory of risk quantification assessment, and the crankshaft data produced in different furnaces were formed. The crankshaft hardness field finite element analysis model and the tempering mechanical property prediction model were established respectively. Composite experiments of different tempering process centers were carried out through the crankshaft hardness field finite element analysis model and the tempering mechanical property prediction model. The linear regression equation of tempering hardness, tensile strength, yield strength and quenching temperature, tempering temperature was fitted according to the response surface characteristics. The linear regression equation was optimized by Matlab genetic algorithm to obtain the standard equation. The evaluation process was completed based on the standard equation and the actual crankshaft data obtained from different furnaces.

2. The method for evaluating the quenching and tempering of a 42CrMo quenched and tempered steel crankshaft according to claim 1, wherein: After obtaining the crankshaft data produced in different batches, the process of early warning assessment of crankshafts produced in different batches based on the normal distribution theory of risk quantification assessment is as follows: the crankshaft data is subjected to normal distribution assessment processing to screen out hardness values that conform to the normal distribution as target samples, the mathematical expectation μ value of the target sample is calculated, and the probability P of the data point within a specific interval (μ-kσ, μ+kσ) is evaluated according to the Chebyshev inequality, where k is a constant and σ is the standard deviation of the crankshaft data. For the hardness test of the crankshaft production process, the interval (μ-kσ, μ+kσ) corresponding to the probability P is determined to be the reliable fluctuation range of the crankshaft hardness, and μ-kσ and μ+kσ are used as the lower limit and upper limit of the hardness warning, respectively; The calculation process of the warning value is: According to the normal distribution theory, the boundary value is calculated. When the data X conforms to the general normal distribution, that is, X~N(μ,σ 2 ), the probability that the data in the sample is within the interval is calculated as follows: This type of general normal distribution is standardized, so It turns out that: Further simplification yields: When the data is in the middle of (μ-kσ,μ+kσ), substitute the two end points into the above formula to get: P={μ-kσ <X<μ+kσ}=Φ(k)-Φ(-k)=2Φ(k)-1 (5-4) Where σ is the data standard deviation, μ is the mathematical expectation, Φ(x) is the standard normal distribution function, and the hardness data probability P is 70%; After using spas software to calculate the data standard deviation σ and mathematical expectation μ, the k value is 1.04 according to the table when Φ(k)=0.85, and the hardness theoretical warning lower limit μ-kσ and hardness theoretical warning upper limit μ+kσ can be calculated; when the P value is 70%, 15% of the hardness data sorted by size is the actual warning lower limit, and 85% is the actual warning upper limit.

3. The method for evaluating the quenching and tempering of a 42CrMo quenched and tempered steel crankshaft according to claim 2, wherein: After calculating the lower and upper limits of the hardness warning based on the normal distribution theory, the crankshaft hardness data is evaluated. When the crankshaft hardness data is in the non-corner area, the surface crankshaft hardness is in a safe state. When the crankshaft hardness data is in the corner area, it indicates that the crankshaft hardness is outside the warning value, and the furnace process where the crankshaft is located is in a warning failure state.

4. The method for evaluating the quenching and tempering of a 42CrMo quenched and tempered steel crankshaft according to claim 1, 2 or 3, wherein: The finite element analysis model of crankshaft hardness field and the tempering mechanical property prediction model were established respectively. The crankshaft data center composite experiment was carried out using the crankshaft hardness field finite element analysis model and the tempering mechanical property prediction model. The linear regression equation of tempering hardness, tensile strength, yield strength, quenching temperature and tempering temperature was fitted according to the response surface characteristics as follows: Based on the central composite design principle and combined with the quenching and tempering single-factor variable test, a two-factor, five-level design method was used to establish a quantitative assessment model for the crankshaft hardness failure risk. The results were verified based on the quantitative assessment model for the crankshaft hardness failure risk. The quenching temperature and tempering temperature were represented by T1 and T2, respectively, and -2, -1, 0, 1, and 2 were used to represent the variable levels, forming a two-factor, five-level experimental design coding value. The two-factor, five-level experimental design coding value was calculated based on the hardness simulation and tempering performance prediction formula to obtain the average value. The design expert software was used to design a quadratic multinomial fitting and perform a significance test. The fitting equation and the actual temperature formula were obtained as follows: HRC=-7125.0142959912+10.816882183927T1+8.5991954023188T2+0.00024999999997561T1T2-0.006517241379313T1 2 -0.0072672413793098T2 2 <h2 style=";text-align:left;direction:ltr">Rm=-110415.65775901+164.88793103504T1+138.16172413843T2+0.005999999999396T1T2-0.10003448275874T1<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> -0.11803448275862T2<h2 style=";text-align:left;direction:ltr"> 2 Re=-81624.416379443+209.7697988507T1-18.019137930769T2+0.0014999999997T1T2-0.12501724137927T1 2 +0.01248275862068T2 2 ; The above calculations show that the tempering mechanical properties prediction model after response surface optimization shows that the various mechanical properties are clearly layered on the response surface projection surface. Combining the 3D surface model and the projection surface, the process range that meets the performance requirements of tempering hardness, tensile strength, yield strength, quenching temperature, and tempering temperature is obtained.

5. The method for evaluating the quenching and tempering of a 42CrMo quenched and tempered steel crankshaft according to claim 4, wherein: The linear regression equation was optimized by Matlab genetic algorithm to obtain the standard equation. The evaluation process was completed by comparing the standard equation with the actual crankshaft data obtained from different furnace production batches: After determining the optimal range of various mechanical properties of the crankshaft, the optimal range of the crankshaft quenching temperature, and the optimal range of the crankshaft tempering temperature, the multi-objective optimization of the crankshaft mechanical properties and the heat treatment process optimization process are completed through the GA toolbox in the MATLAB genetic algorithm, combining the inhibition between the multiple objectives. That is, the minimum value of the function is the optimal value, so when setting the tensile strength and yield strength, the opposite values need to be set. Therefore, the multi-objective optimization function is obtained as follows: minf(x)=[f1(x),-f2(x),-f3(x)] (5-5) In the above formula, f1(x) is hardness, f2(x) is tensile strength, and f3(x) is yield strength; According to the above function, the problem is solved using the Matlab genetic algorithm toolbox GA solver.

6. The method for evaluating the quenching and tempering of a 42CrMo quenched and tempered steel crankshaft according to claim 5, wherein: The Matlab genetic algorithm optimizes the objective function through the Pareto method to ensure that all mechanical properties meet the target requirements. At the same time, the heat treatment process range is within the result range obtained by the response surface. The optimization results are verified experimentally. When the hardness is reduced, the tensile strength is slightly reduced, while the cross-sectional shrinkage rate and elongation after fracture are slightly improved, indicating that the crankshaft has good impact toughness and fatigue strength, a relatively higher metallographic grade, and better organizational clarity and uniformity than the initial process.

Images