Process parameter optimization method for reducing residual stress of cylinder cover of large-cylinder-diameter diesel engine

Abstract

The invention discloses a process parameter optimization method for reducing residual stress of a cylinder cover of a large-cylinder-diameter diesel engine. The process parameter optimization method comprises the following steps: establishing a J-C constitutive model of a cylinder cover material; performing multi-process machining process simulation analysis on the cylinder cover by using a life-death unit technology and the J-C constitutive model to obtain a residual stress simulation value; a blind hole method is adopted to obtain a test value of the residual stress, the test value is compared with the simulation value, and an error value is controlled within an allowable error range; establishing a nonlinear mapping relationship between the cutting parameters and the response target based on a data regression prediction model of a BP neural network to obtain a prediction model; the prediction model is adopted to optimize cutting parameters of three machining technologies of milling, turning and boring of the cylinder cover, and the minimum residual stress value and the optimal cutting force are obtained through prediction; and the optimal cutting force serves as a boundary condition, machining process simulation analysis is conducted, and the residual stress value optimization descending amplitude is obtained. The residual stress of the cylinder cover is greatly reduced.

Description

Technical Field

[0001] The invention relates to the technical field of process parameter optimization of a diesel engine cylinder head, and in particular to a process parameter optimization method for reducing residual stress of a large-bore diesel engine cylinder head. Background Art

[0002] As a key component of a diesel engine, the performance and reliability of the diesel engine cylinder head are directly related to the service life of the entire diesel engine. However, during the long-term service of the diesel engine, the area of ​​the large-diameter cylinder head fire surface is larger than that of the traditional cylinder head fire surface. It is more affected by temperature and pressure during operation, and the thermal load and mechanical load generated are also greater. The fire surface of the large-diameter cylinder head is very prone to fatigue damage such as cracks and fractures. During the production process of the cylinder head, large residual stress will be generated inside and on the surface of the cylinder head. Residual stress is an important factor that seriously affects the service life and safety of the cylinder head. During the production process of the cylinder head, by changing the tool and cutting parameters, the influence of process parameters and residual stress can be obtained through a large number of experiments. However, the cylinder head of a marine diesel engine is large in size and high in cost, and it is difficult to verify the changing relationship between processing parameters and residual stress through experimental methods. Therefore, simulation analysis will be used to predict the relationship between processing parameters and residual stress, which will greatly reduce manufacturing costs and production cycles; but the existing diesel engine cylinder head has the following problems in process parameter optimization analysis:

[0003] (1) Most studies use explicit dynamics to simulate dynamic processes such as high-speed cutting and high impact that occur in a short period of time. The calculation has good stability and high efficiency, but it is only suitable for smaller simulation models and only studies a single processing process. However, due to factors such as large volume, complex structure, and cumbersome processing procedures, it is impossible to obtain accurate simulation data for large-diameter cylinder heads.

[0004] (2) The method of using intelligent algorithms to optimize machining process parameters can reduce research costs. Intelligent algorithms have strong global search capabilities and can quickly find the globally optimal cutting parameter combination. However, most studies are focused on optimizing the process parameters of a single process, and lack multi-objective optimization analysis of the process parameters of multiple processes.

[0005] Therefore, it is urgent to solve the above problems. Summary of the invention

[0006] Purpose of the invention: The purpose of the present invention is to propose a process parameter optimization method for reducing residual stress in a large-diameter diesel engine cylinder head, which is beneficial to reducing residual stress and improving the fatigue life and overall performance of the cylinder head.

[0007] Technical solution: To achieve the above purpose, the present invention discloses a process parameter optimization method for reducing residual stress of a large-diameter diesel engine cylinder head, comprising the following steps:

[0008] (1) Establish the Johnson-Cook constitutive model of cylinder head material;

[0009] (2) Use Solidworks software to construct a cylinder head simulation model, and preset a "life and death unit" layer on the cylinder head surface; use the "life and death unit" technology and the Johnson-Cook constitutive model of the cylinder head material to perform a multi-process processing simulation analysis on the cylinder head, obtain the size and distribution of the maximum principal stress and minimum principal stress in the cylinder head cutting residual stress, and extract the simulation values ​​of the maximum principal stress and minimum principal stress at several measuring points on the cylinder head fire surface;

[0010] (3) using the blind hole method to measure the residual stress of the cylinder head fire surface after actual processing, and calculating the test values ​​of the maximum principal stress and the minimum principal stress of several measuring points on the cylinder head fire surface, comparing the test value of the maximum principal stress of each measuring point with the simulation value, and comparing the test value of the minimum principal stress of each measuring point with the simulation value, and controlling the error values ​​within the allowable error range. When the error value between the test value and the simulation value exceeds the allowable error, returning to step (2) to re-perform the processing technology simulation analysis of the multi-process processing flow of the cylinder head until the error value is controlled within the allowable error range;

[0011] (4) According to the cutting parameters of the three machining processes of milling, turning and boring, several groups of simulation analysis are carried out respectively. The nonlinear mapping relationship between the cutting parameters and the response targets is established based on the data regression prediction model of the BP neural network to obtain the prediction model, in which the cutting parameters include cutting speed, feed rate and cutting depth, and the response targets include cutting force and residual stress;

[0012] (5) The prediction model is used to optimize the cutting parameters of the three machining processes of cylinder head milling, turning, and boring, predict the minimum residual stress value and optimal cutting force corresponding to the optimal cutting parameters, and extract the mean cutting force in the simulation process;

[0013] (6) Taking the predicted optimal cutting force as the boundary condition, the multi-process machining process simulation analysis of the cylinder head is performed again. The simulation results after optimization of several measuring points are compared with the simulation results before optimization to obtain the optimized reduction amplitude of the residual stress value at each measuring point.

[0014] In the step (1), the material of the diesel engine cylinder head will generate high temperature, high strain and complex flow stress during cutting. The Johnson-Cook constitutive model is used to describe the dynamic behavior of the material from low strain rate to high strain rate. The material flow stress formula of the Johnson-Cook constitutive model is as follows:

[0015]

[0016] T * =(TT r ) / (T m -T r )

[0017] Where: e is the equivalent stress, ε e is the equivalent plastic strain, A is the initial yield stress, B is the hardening constant, n is the hardening exponent, C is the strain rate constant, m is the thermal softening exponent, ε is the current strain rate, ε0 is the reference strain rate, T is the room temperature, T r is the reference temperature, T m is the melting temperature of the material, f is the function, and In is the logarithmic sign.

[0018] Preferably, in step (2), firstly, according to the drawing of the cylinder head, a cylinder head simulation model is constructed using Solidworks software. After the construction is completed, a "life and death unit" layer is preset on the surface of the cylinder head using Solidworks software to make the appearance structure of the cylinder head in the blank state; according to the cutting force F in three directions in actual processing x 、F y 、F z , take the cutting force F in the actual processing process x 、F y 、F z Mean cutting force The cutting force mean is synthesized into a complete cutting force as the boundary condition in the cutting simulation process;

[0019] Secondly, the Johnson-Cook constitutive model obtained in step (1) is imported into the Ansys simulation software as a material property, and then the cylinder head simulation model is imported, the cylinder head simulation model is divided into tetrahedral grids, the grid size and the constraints of the cylinder head simulation model are set, and the cutting force is set as a boundary condition on the surface to be processed, and then the "life and death unit" technology is used to perform a multi-process processing process simulation analysis on the cylinder head. In the processing process simulation analysis, the cutting forces corresponding to the different processes in the actual processing are applied to each surface of the cylinder head. After the cutting process of the entire cylinder head is completed, a number of measuring points are taken on the cylinder head fire surface, the number of measuring points is>10, and the position range of the measuring points is selected in the area within 15mm of the nose bridge area of ​​the cylinder head fire surface and the intake and exhaust ports, and the simulation values ​​of the maximum principal stress and the minimum principal stress of each measuring point are extracted.

[0020] Furthermore, the preset "life and death unit" layer on the cylinder head surface in step (2) refers to the preset rough processing layer and fine processing layer of the "life and death unit" layer for the top surface, surrounding surfaces, outer circle, fire surface, valve hole and injection hole of the cylinder head respectively.

[0021] Furthermore, the multi-process processing flow of the diesel engine cylinder head cutting simulation in step (2) is sequentially planned as rough turning of the top surface, rough milling of the surrounding surfaces, fine milling of the surrounding surfaces, rough turning of the power surface, fine milling of the top surface, rough turning of the outer circle, fine turning of the outer circle, fine turning of the power surface, fine turning of the injection hole, rough boring of the valve hole and fine boring of the valve hole.

[0022] Preferably, when the blind hole method is used for measurement in step (3), a blind hole is firstly drilled on a cylinder head with residual stress. After the stress around the blind hole is released, strain is released. The strain release is measured using a strain rosette and substituted into formula (1) to calculate the residual stress at the measuring point:

[0023] Where: σ1 and σ2 are the minimum and maximum principal stresses, ε1, ε2, and ε3 are the release strains measured by strain gauges No. 1, No. 2, and No. 3, respectively; k1 and k2 are strain release coefficients; and θ is the angle between σ2 and strain gauge No. 1.

[0024] The strain release coefficients k1 and k2 of the material are related to the size of the blind hole and the mechanical properties of the material. When drilling a blind hole, the strain release coefficients k1 and k2 are obtained by the Kirsch theory:

[0025] Where: v is the Poisson's ratio of the material, E is the elastic modulus of the material, R is the punching radius, r1 and r2 are the inner and outer radii of the strain gauge;

[0026] The blind hole method is used to collect data from several measuring points on the cylinder head fire surface, and the residual stress values ​​of the measuring points are calculated using formula (1); several groups of residual stress values ​​are arranged in order of size, and the maximum principal stress and the minimum principal stress of at least three groups of values ​​with larger values ​​are compared, and the test value of the residual stress of the cylinder head fire surface measuring point is compared with the simulation value, and the error value is controlled within the allowable error range; when the error value between the test value and the simulation value exceeds the allowable error, return to step (2) to adjust the grid size in the cylinder head simulation model grid division, the constraint conditions of the cylinder head simulation model and / or increase the number of measuring points, and re-perform the multi-process processing flow simulation analysis of the cylinder head until the error value is controlled within the allowable error range.

[0027] Furthermore, in step (4), the cutting force and residual stress values ​​of the cylinder head at different cutting speeds, feed rates and cutting depths in the three machining processes of milling, turning and boring are first calculated by simulation, and the point values ​​of the simulation calculation are set according to the value range of the cutting parameters. Several groups of simulation analyses are performed according to the cutting parameters of the three processes to obtain corresponding simulation data; then the simulation data is input into a data regression prediction model based on a BP neural network, and a nonlinear mapping relationship between the cutting parameters and the cutting force and residual stress is established to obtain a prediction model.

[0028] Further, the point value setting for simulation calculation according to the value range of the cutting parameter in step (4) refers to:

[0029] In milling cutter processing: cutting speed v = 120 / 180 / 240 / 300m / min, feed rate a = 0.05 / 0.1 / 0.15 / 0.2mm, cutting depth d = 1 / 3mm; in turning tool processing: v = 100 / 200 / 300 / 400m / min, a = 0.05 / 0.1 / 0.15 / 0.2mm, d = 1 / 3mm; in boring tool processing: v = 30 / 50 / 70 / 90m / min, a = 0.05 / 0.1 / 0.15 / 0.2mm, d = 1 / 3mm.

[0030] Preferably, when the cutting parameters are optimized in step (5), the optimization variables include cutting speed v, feed rate a and cutting depth d. The three machining processes of milling, turning and boring correspond to different cutting parameter value ranges. The optimization targets are cutting force and residual stress. The minimum residual stress and optimal cutting force are predicted according to the prediction model constructed by MATLAB software.

[0031] Furthermore, the different cutting parameter value ranges in step (5) refer to:

[0032] In milling cutter processing: cutting speed v = 120-300m / min, feed rate a = 0.05-0.2mm / r, cutting depth d = 1-3mm; in turning tool processing: constraint cutting speed v = 100-400m / min, feed rate a = 0.05-0.2mm / r, cutting depth d = 1-3mm; in boring tool processing: constraint cutting speed v = 30-90m / min, feed rate a = 0.05-0.2mm / r, cutting depth d = 1-3mm.

[0033] Beneficial effect: Compared with the prior art, the present invention has the following significant advantages: the present invention comprehensively considers the optimization of the size and distribution of the residual stress of the cylinder head cutting by the cutting parameters of multiple processes, uses the "life and death unit" technology to perform a full-process cutting simulation analysis on the cylinder head, uses the blind hole method to verify and correct the simulation results of the cutting simulation analysis, and then performs multiple groups of simulation analyses to obtain the mapping relationship between the cutting parameters and the residual stress, establishes a nonlinear mapping relationship between the cutting parameters and the response target based on the BP neural network, and uses a high-precision prediction model to predict the optimal cutting parameters, which is beneficial to reducing the residual stress and improving the fatigue life and overall performance of the cylinder head. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the process of the present invention;

[0035] Figure 2It is a schematic diagram of the experiment of the 105A WANCE mechanical testing machine in the present invention;

[0036] Figure 3 This is a diagram of a split Hopkinson pressure bar experimental platform in the present invention;

[0037] Figure 4 It is a schematic diagram of the multi-step processing flow of the cylinder head in the present invention;

[0038] Figure 5 It is a platform diagram of the blind hole method experiment in the present invention;

[0039] Figure 6 Schematic diagram of the strain rosette structure used in the blind hole method of the present invention. DETAILED DESCRIPTION

[0040] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0041] like Figure 1 As shown, the present invention provides a process parameter optimization method for reducing residual stress of a large-bore diesel engine cylinder head, comprising the following steps:

[0042] (1) The Johnson-Cook constitutive model of the QT400 material of the cylinder head is obtained by using room temperature quasi-static tensile and split Hopkinson pressure bar tests;

[0043] In the study of the constitutive model of cylinder head materials, a 105A WANCE mechanical testing machine was used to conduct quasi-static tensile tests at room temperature to study the influence of plastic strain on the flow stress behavior of the material. Figure 2 As shown; In order to obtain the strain rate effect of the material at different temperatures, a split Hopkinson pressure bar is used to carry out impact tests on the cylinder head material at different strain rates, as shown in Figure 3 As shown in the figure, the Johnson-Cook constitutive model of the QT400 material of the cylinder head is fitted through the quasi-static tensile test at room temperature and the Hopkinson pressure bar test;

[0044] During the cutting process of the diesel engine cylinder head, the material will generate high temperature, high strain and complex flow stress. The Johnson-Cook constitutive model is used to describe the dynamic behavior of the QT400 material from low strain rate to high strain rate. The material flow stress formula of the Johnson-Cook constitutive model is as follows:

[0045]

[0046] T * =(TT r ) / (T m -T r )

[0047] Where:e is the equivalent stress, ε e is the equivalent plastic strain, A is the initial yield stress, B is the hardening constant, n is the hardening exponent, C is the strain rate constant, m is the thermal softening exponent, ε is the current strain rate, ε0 is the reference strain rate, T is the room temperature, T r is the reference temperature, T m is the melting temperature of the material, f is the function, and In is the logarithmic sign;

[0048] (2) Use Solidworks software to construct a cylinder head simulation model, and preset a "life and death unit" layer on the cylinder head surface; use the "life and death unit" technology and the Johnson-Cook constitutive model of the cylinder head QT400 material to perform a multi-process processing simulation analysis on the cylinder head, obtain the size and distribution of the maximum principal stress and minimum principal stress in the cylinder head cutting residual stress, and extract the simulation values ​​of the maximum principal stress and minimum principal stress at several measuring points on the cylinder head fire surface;

[0049] First, according to the cylinder head drawing, the cylinder head simulation model is constructed using Solidworks software. After the construction is completed, the "life and death unit" layer is preset on the cylinder head surface using Solidworks software to make the cylinder head in the blank state. The shape structure; according to the cutting force F in three directions in actual processing x 、F y 、F z , take the cutting force F in the actual processing process x 、F y 、F z Mean cutting force The cutting force mean is synthesized into a complete cutting force as the boundary condition in the cutting simulation process;

[0050] The "life and death unit" technology used: In the simulation analysis, a series of discrete analysis steps are combined into a dynamic cutting process, that is, the dynamic analysis process is simplified into a number of static analysis processes; the material actually removed by machining is removed in sequence through the life and death unit simulation and tool machining trajectory; the volume of the unit removed in each static analysis is as follows:

[0051] V max =a p ×a e ×f d ×t

[0052] Where: V max is the volume of material removed per analysis step, a p is the cutting depth, a e is the cutting width, f d is the feed rate, t is the time of unit analysis step;

[0053] Presetting the "life and death unit" layer on the cylinder head surface refers to respectively presetting the "life and death unit" layer of the rough machining layer and the fine machining layer for the top surface, surrounding surface, outer circle, fire surface, valve hole and injection hole of the cylinder head;

[0054] The multi-process machining process of the diesel engine cylinder head cutting simulation is planned in sequence as rough turning of the top surface, rough milling of the surrounding surface, fine milling of the surrounding surface, rough turning of the power surface, fine milling of the top surface, rough turning of the outer circle, fine turning of the outer circle, fine turning of the power surface, fine turning of the injection hole, rough boring of the valve hole and fine boring of the valve hole. The entire production process is as follows Figure 4 As shown;

[0055] Secondly, the Johnson-Cook constitutive model measured in step (1) is imported into the Ansys simulation software as a material property, and then the cylinder head simulation model is imported, and the cylinder head simulation model is divided into tetrahedral grids, the grid size is set to 10 mm, the constraints of the cylinder head simulation model are set, and the cutting force is set as a boundary condition on the surface to be processed, and then the "life and death unit" technology is used to perform a multi-process processing process simulation analysis on the cylinder head, and finally the size and distribution of the maximum principal stress and the minimum principal stress in the cylinder head cutting residual stress are obtained; in the processing process simulation analysis, the cutting force corresponding to the different processes in the actual processing is applied to each surface of the cylinder head, and the cutting process of the entire cylinder head can be divided into three parts: cutting force application, processing completion and cooling; after the cutting process of the entire cylinder head is completed, 17 measuring points are taken on the cylinder head firepower surface, and the position range of the measuring points is selected in the area within 15 mm between the nose bridge area of ​​the cylinder head firepower surface and the intake and exhaust ports, and 17 measuring points are evenly distributed to extract the simulation values ​​of the maximum principal stress and the minimum principal stress of each measuring point;

[0056] (3) The residual stress of the cylinder head fire surface after actual processing is measured by the blind hole method, and the test values ​​of the maximum principal stress and the minimum principal stress of several measuring points on the cylinder head fire surface are calculated, and the test value of the maximum principal stress of each measuring point is compared with the simulation value, and the test value of the minimum principal stress of each measuring point is compared with the simulation value, and the error value is controlled within the allowable error range of 20%, which proves the accuracy of the simulation. When the error value between the test value and the simulation value exceeds the allowable error of 20%, return to step (2) and re-perform the processing technology simulation analysis of the multi-process processing flow of the cylinder head until the error value is controlled within the allowable error range of 20%;

[0057] The blind hole method is used to measure the stress of the cylinder head fire surface. The accuracy of the simulation calculation results and whether the simulation model can truly reflect the actual physical phenomenon are verified through experiments, which is conducive to the use of simulation calculations to explore the influence of different cutting parameters on the processing quality in the later stage. Based on the stress test standard, the distance between each measuring point should be more than 6 times the diameter of the blind hole, that is, more than 12mm. During the test, the test position is arranged reasonably in combination with the specific size of the fuselage, ensuring that the measuring point and the edges of the measured surface are kept at more than 15mm, and the distance between each measuring point is kept at more than 12mm.

[0058] The experimental platform of the blind hole method is as follows Figure 5 As shown in the figure, the steps of measuring the residual stress at the firing surface by the blind hole method are as follows: use sandpaper to polish the measuring point and clean it; wear clean gloves to paste the strain gauge, the strain gauge model is TJ120-1.5, and wait for cooling for 10 to 15 minutes; use the solder-free butt terminal to connect the data cable; use the drill bit holder to drill the target area, the drill bit model is GB / T6135.2-2008; use the software to record the data, the software uses the matching USB-1608G dynamic signal acquisition and analysis system; finally process and analyze the data;

[0059] When measuring with the blind hole method, first drill a blind hole on a part with residual stress. After the stress around the blind hole is released, the strain is released. Figure 6 The strain rosette shown in the figure measures the release strain and substitutes it into formula (1) to calculate the residual stress at the measuring point:

[0060] Where: σ1 and σ2 are the minimum and maximum stresses, ε1, ε2, and ε3 are the release strains measured by strain gauges No. 1, No. 2, and No. 3, respectively; k1 and k2 are strain release coefficients; and θ is the angle between σ2 and strain gauge No. 1. Figure 6 As shown;

[0061] The strain release coefficients k1 and k2 of the material are related to the size of the blind hole and the mechanical properties of the material. In the case of drilling, the strain release coefficients k1 and k2 are obtained by Kirsch theory:

[0062] Where: v is the Poisson's ratio of the material, E is the elastic modulus of the material, R is the punch radius, r1 and r2 are the inner and outer radii of the strain gauge, such as Figure 6 As shown;

[0063] The blind hole method was used to collect data from 17 measuring points on the cylinder head fire surface, and the residual stress values ​​of the measuring points were calculated using formula (1). The 17 groups of residual stress values ​​were arranged in order of size, and the maximum principal stress and minimum principal stress of the four groups with larger values ​​were compared to verify the accuracy of the simulation. The test data of the cylinder head fire surface measuring points were compared with the residual stress values ​​of the simulation results. The error value was controlled within 20%. The error of the maximum stress value in the area prone to fatigue failure was within an acceptable range, indicating that the simulation calculation can more accurately reflect the real residual stress of the machined part.

[0064] When the error between the test data and the simulation result exceeds 20%, the error can be reduced by the following methods, then returning to step (2) to adjust the size of the grid in the cylinder head simulation model grid division, the constraint conditions of the cylinder head simulation model and / or increase the number of measuring points; re-performing the processing simulation analysis of the multi-process processing flow of the cylinder head until the error value is controlled within the allowable error range;

[0065] (4) According to the cutting parameter settings of the three machining processes of milling, turning, and boring, 32 sets of simulation analysis were performed for each process, for a total of 96 sets of simulation analysis. A nonlinear mapping relationship between cutting parameters and response targets was established based on the data regression prediction model of the BP neural network to obtain a high-precision prediction model, where the cutting parameters include cutting speed, feed rate, and cutting depth, and the response targets include cutting force and residual stress;

[0066] Firstly, the cutting force and residual stress values ​​of the cylinder head in the three machining processes of milling, turning and boring are calculated by simulation, and the point values ​​of simulation calculation are set according to the value range of cutting parameters; in milling cutter processing: cutting speed v = 120 / 180 / 240 / 300m / min, feed a = 0.05 / 0.1 / 0.15 / 0.2mm / r, cutting depth d = 1 / 3mm; in turning tool processing: v = 100 / 200 / 300 / 400m / min, a = 0. 05 / 0.1 / 0.15 / 0.2mm / r, d=1 / 3mm; in boring tool processing: v=30 / 50 / 70 / 90m / min, a=0.05 / 0.1 / 0.15 / 0.2mm / r, d=1 / 3mm, 32 sets of simulation analysis were performed according to the processing parameter settings of the three processes to obtain the corresponding simulation data; then the simulation data was input into the data regression prediction model based on BP neural network, and the nonlinear mapping relationship between cutting parameters and response targets was established to obtain a high-precision prediction model;

[0067] BP neural network is a multi-layer feedforward neural network. It calculates the error between the output layer and the true value and propagates the error back to each layer of the network, thereby updating the weights and bias items of the network, so that the output of the network gradually approaches the true value. In the BP neural network model, the number of nodes in the input layer and the output layer is determined based on actual research, and the number of nodes in the hidden layer is determined according to the following formula:

[0068]

[0069] Where: h is the number of nodes in the hidden layer, i is the number of nodes in the input layer, o is the number of nodes in the output layer, and c is a random constant with a value range of [1, 10];

[0070] In the forward propagation of the network model, the training set is input and the data set is first normalized using the following formula:

[0071]

[0072] In the formula: x is the normalized value, x1 is the original data, x max 、x min are the maximum and minimum values ​​of the original data respectively;

[0073] The output of each neuron in the hidden layer is calculated according to the following formula:

[0074]

[0075] Where: H i is the hidden layer output, f is the hidden layer activation function, ω is the weight of two nodes, h i is the hidden layer threshold, i is the number of hidden layer nodes, i=1, 2, ..., q, j is the number of output layer nodes, j=1, 2, ..., m;

[0076] The output of each neuron in the output layer is calculated according to the following formula:

[0077]

[0078] Where: O k is the output layer output, a k is the output layer threshold, k is the number of nodes in the output layer, k = 1, 2, ..., l;

[0079] In back propagation, the output error formula is as follows:

[0080] e k =d k -O k

[0081] Where: e k is the output error, d kis the expected output;

[0082] Adjust the connection weights and thresholds of the network according to the following formula:

[0083]

[0084] a k =a k +e k

[0085] ω ij =ω ij +ηH i e k

[0086] Where: η is the learning rate;

[0087] The BP neural network model was constructed using MATLAB software and the accuracy of the model was verified. First, 28 groups of 32 simulation results in the milling cutter processing simulation were extracted as training sets, and the other 4 groups of simulation results were used as test sets. Then, the data was normalized, and the normalized scaling interval was [0,1] to improve the stability and convergence speed of the model. Secondly, the feedforward neural network was called using the neural network toolbox in MATLAB. Finally, the error calculation and data denormalization were performed. After the 9th iteration, the iteration stopped when the error accuracy (RMSE<10) was met. The results obtained included training, verification, test and overall regression graphs. The RMSE of the training set of cutting force in the network model calculation was 4.4, and the RMSE of the test set was 7.3. The RMSE of the training set of residual stress in the network model calculation was 2.6, and the RMSE of the test set was 6.2. The accuracy of the prediction model was determined based on the calculated error value, and the model can be used to solve the optimal machining process solution in the later stage.

[0088] (5) The prediction model is used to optimize the cutting parameters of the three machining processes of cylinder head milling, turning, and boring, predict the minimum residual stress value and optimal cutting force corresponding to the optimal machining parameters, and extract the mean cutting force in the simulation process;

[0089] When optimizing cutting parameters, the optimization variables include cutting speed v, feed rate a and cutting depth d. The three processing technologies of milling, turning and boring correspond to different cutting parameter value ranges. In milling tool processing: cutting speed v = 120-300m / min, feed rate a = 0.05-0.2mm, cutting depth d = 1-3mm; in turning tool processing: constraint cutting speed v = 100-400m / min, feed rate a = 0.05-0.2mm, cutting depth d = 1-3mm; in boring tool processing: constraint cutting speed v = 30-90m / min, feed rate a = 0.05-0.2mm, cutting depth d = 1-3mm; the optimization targets are cutting force and residual stress. The minimum residual stress and the corresponding optimal cutting force are predicted according to the prediction model constructed by MATLAB software.

[0090] Taking cutting force and residual stress as optimization targets, the prediction formula of the objective function is as follows:

[0091] y=sim(net,A)

[0092] Where: y is the predicted value, net is the residual stress prediction model of BP surface after training, and A is the input variable matrix.

[0093] (6) Taking the predicted optimal cutting force as the boundary condition, the multi-process machining process simulation analysis of the cylinder head is carried out again. The simulation results after optimization of several measuring points are compared with the simulation results before optimization, and the optimization reduction amplitude of the residual stress value of each measuring point is obtained. The maximum optimization effect is a reduction of 20.5% in the residual stress value.

[0094] The present invention comprehensively considers the optimization of the size and distribution of the residual stress of the cylinder head cutting by the cutting parameters of multiple processes, uses the "life and death unit" technology to perform a full-process cutting simulation analysis on the cylinder head, uses the blind hole method to verify and correct the simulation results of the cutting simulation analysis, and then performs multiple groups of simulation analyses to obtain the mapping relationship between the cutting parameters and the residual stress. Based on the BP neural network, a nonlinear mapping relationship between the cutting parameters and the response target is established, and a high-precision prediction model is used to predict the optimal cutting parameters, which is beneficial to reducing the residual stress and improving the fatigue life and overall performance of the cylinder head.

Claims

1. A process parameter optimization method for reducing residual stress in a large-diameter diesel engine cylinder head, characterized in that: The steps include: (1) Establish the Johnson-Cook constitutive model of cylinder head material; (2) Use Solidworks software to construct a cylinder head simulation model, and preset a "life and death unit" layer on the cylinder head surface; use the "life and death unit" technology and the Johnson-Cook constitutive model of the cylinder head material to perform a multi-process processing simulation analysis on the cylinder head, obtain the size and distribution of the maximum principal stress and minimum principal stress in the cylinder head cutting residual stress, and extract the simulation values ​​of the maximum principal stress and minimum principal stress at several measuring points on the cylinder head fire surface; (3) using the blind hole method to measure the residual stress of the cylinder head fire surface after actual processing, and calculating the test values ​​of the maximum principal stress and the minimum principal stress of several measuring points on the cylinder head fire surface, comparing the test value of the maximum principal stress of each measuring point with the simulation value, and comparing the test value of the minimum principal stress of each measuring point with the simulation value, and controlling the error values ​​within the allowable error range. When the error value between the test value and the simulation value exceeds the allowable error, returning to step (2) to re-perform the processing technology simulation analysis of the multi-process processing flow of the cylinder head until the error value is controlled within the allowable error range; (4) According to the cutting parameters of the three machining processes of milling, turning and boring, several groups of simulation analysis are carried out respectively. The nonlinear mapping relationship between the cutting parameters and the response targets is established based on the data regression prediction model of the BP neural network to obtain the prediction model, in which the cutting parameters include cutting speed, feed rate and cutting depth, and the response targets include cutting force and residual stress; (5) The prediction model is used to optimize the cutting parameters of the three machining processes of cylinder head milling, turning, and boring, predict the minimum residual stress value and optimal cutting force corresponding to the optimal cutting parameters, and extract the mean cutting force in the simulation process; (6) Taking the predicted optimal cutting force as the boundary condition, the multi-process machining process simulation analysis of the cylinder head is performed again. The simulation results after optimization of several measuring points are compared with the simulation results before optimization to obtain the optimized reduction amplitude of the residual stress value at each measuring point.

2. The process parameter optimization method for reducing residual stress of a large-diameter diesel engine cylinder head according to claim 1, characterized in that: In the step (1), the diesel engine cylinder head will generate high temperature, high strain and complex flow stress during cutting. The Johnson-Cook constitutive model is used to describe the dynamic behavior of the material from low strain rate to high strain rate. The material flow stress formula of the Johnson-Cook constitutive model is as follows: T * =(T-T r ) / (T m -T r ) Where: e is the equivalent stress, ε e is the equivalent plastic strain, A is the initial yield stress, B is the hardening constant, n is the hardening exponent, C is the strain rate constant, m is the thermal softening exponent, ε is the current strain rate, ε0 is the reference strain rate, T is the room temperature, T r is the reference temperature, T m is the melting temperature of the material, f is the function, and In is the logarithmic sign.

3. The process parameter optimization method for reducing residual stress of a large-diameter diesel engine cylinder head according to claim 1, characterized in that: In the step (2), firstly, according to the drawing of the cylinder head, a cylinder head simulation model is constructed using Solidworks software. After the construction is completed, a "life and death unit" layer is preset on the surface of the cylinder head using Solidworks software to make the appearance structure of the cylinder head in the blank state; according to the cutting force F in three directions in actual processing x 、F y 、F z , take the cutting force F in the actual processing process x 、F y 、F z Mean cutting force The cutting force mean is synthesized into a complete cutting force as the boundary condition in the cutting simulation process; Secondly, the Johnson-Cook constitutive model obtained in step (1) is imported into the Ansys simulation software as a material property, and then the cylinder head simulation model is imported, the cylinder head simulation model is divided into tetrahedral grids, the grid size and the constraints of the cylinder head simulation model are set, and the cutting force is set as a boundary condition on the surface to be processed, and then the "life and death unit" technology is used to perform a multi-process processing process simulation analysis on the cylinder head. In the processing process simulation analysis, the cutting force corresponding to the different processes in the actual processing is applied to each surface of the cylinder head. After the cutting process of the entire cylinder head is completed, a number of measuring points are taken on the cylinder head fire surface, the number of measuring points is>10, and the position range of the measuring points is selected in the area within 15mm of the nose bridge area of ​​the cylinder head fire surface and the intake and exhaust ports, and the simulation values ​​of the maximum principal stress and the minimum principal stress of each measuring point are extracted.

4. The process parameter optimization method for reducing residual stress of a large-diameter diesel engine cylinder head according to claim 3 is characterized in that: In the step (2), the "life and death unit" layer is preset on the surface of the cylinder head, which means that the rough processing layer and the fine processing layer of the "life and death unit" layer are preset for the top surface, the surrounding surface, the outer circle, the fire surface, the valve hole and the injection hole of the cylinder head respectively.

5. The process parameter optimization method for reducing residual stress of a large-bore diesel engine cylinder head according to claim 3, characterized in that: The multi-process machining process of the diesel engine cylinder head cutting simulation in the step (2) is sequentially planned as rough turning of the top surface, rough milling of the surrounding surfaces, fine milling of the surrounding surfaces, rough turning of the power surface, fine milling of the top surface, rough turning of the outer circle, fine turning of the outer circle, fine turning of the power surface, fine turning of the injection hole, rough boring of the valve hole and fine boring of the valve hole.

6. The process parameter optimization method for reducing residual stress of a large-diameter diesel engine cylinder head according to claim 3 is characterized in that: When the blind hole method is used in step (3), a blind hole is first drilled on a cylinder head with residual stress. After the stress around the blind hole is released, strain is released. The strain release is measured using a strain rosette and substituted into formula (1) to calculate the residual stress at the measuring point: Where: σ1 and σ2 are the minimum and maximum principal stresses, ε1, ε2, and ε3 are the release strains measured by strain gauges No. 1, No. 2, and No. 3, respectively; k1 and k2 are strain release coefficients; and θ is the angle between σ2 and strain gauge No.

1. The strain release coefficients k1 and k2 of the material are related to the size of the blind hole and the mechanical properties of the material. When drilling a blind hole, the strain release coefficients k1 and k2 are obtained by the Kirsch theory: Where: v is the Poisson's ratio of the material, E is the elastic modulus of the material, R is the punching radius, r1 and r2 are the inner radius and outer radius of the strain gauge; The blind hole method is used to collect data from several measuring points on the cylinder head fire surface, and the residual stress values ​​of the measuring points are calculated using formula (1); several groups of residual stress values ​​are arranged in order of size, and the maximum principal stress and the minimum principal stress of at least three groups of values ​​with larger values ​​are compared, and the test value of the residual stress of the cylinder head fire surface measuring point is compared with the simulation value, and the error value is controlled within the allowable error range; when the error value between the test value and the simulation value exceeds the allowable error, return to step (2) to adjust the grid size in the cylinder head simulation model grid division, the constraint conditions of the cylinder head simulation model and / or increase the number of measuring points, and re-perform the multi-process processing flow simulation analysis of the cylinder head until the error value is controlled within the allowable error range.

7. A process parameter optimization method for reducing residual stress in a large-diameter diesel engine cylinder head according to claim 6, characterized in that: In the step (4), the cutting force and residual stress values ​​of the cylinder head at different cutting speeds, feed rates and cutting depths in the three machining processes of milling, turning and boring are first calculated by simulation, and the point values ​​of the simulation calculation are set according to the value range of the cutting parameters. Several groups of simulation analysis are performed respectively according to the cutting parameters of the three processes to obtain corresponding simulation data; Then the simulation data is input into the data regression prediction model based on BP neural network, and the nonlinear mapping relationship between cutting parameters and cutting force and residual stress is established to obtain the prediction model.

8. The process parameter optimization method for reducing residual stress of a large-diameter diesel engine cylinder head according to claim 7, characterized in that: The point value setting for simulation calculation according to the value range of the cutting parameter in step (4) refers to: In milling cutter processing: cutting speed v = 120 / 180 / 240 / 300m / min, feed rate a = 0.05 / 0.1 / 0.15 / 0.2mm / r, cutting depth d = 1 / 3mm; in turning tool processing: v = 100 / 200 / 300 / 400m / min, a = 0.05 / 0.1 / 0.15 / 0.2mm / r, d = 1 / 3mm; in boring tool processing: v = 30 / 50 / 70 / 90m / min, a = 0.05 / 0.1 / 0.15 / 0.2mm / r, d = 1 / 3mm.

9. The process parameter optimization method for reducing residual stress of a large-diameter diesel engine cylinder head according to claim 8, characterized in that: When the cutting parameters are optimized in step (5), the optimization variables include cutting speed v, feed rate a and cutting depth d. The three machining processes of milling, turning and boring correspond to different cutting parameter value ranges. The optimization targets are cutting force and residual stress. The minimum residual stress and optimal cutting force are predicted according to the prediction model constructed by MATLAB software.

10. The process parameter optimization method for reducing residual stress of a large-bore diesel engine cylinder head according to claim 9, characterized in that: The different cutting parameter value ranges in step (5) are: In milling cutter processing: cutting speed v = 120-300m / min, feed rate a = 0.05-0.2mm / r, cutting depth d = 1-3mm; in turning tool processing: constraint cutting speed v = 100-400m / min, feed rate a = 0.05-0.2mm / r, cutting depth d = 1-3mm; in boring tool processing: constraint cutting speed v = 30-90m / min, feed rate a = 0.05-0.2mm / r, cutting depth d = 1-3mm.

Images