White vehicle body lightweight design method based on relative sensitivity and radial neural network

Abstract

The invention provides a passenger vehicle body-in-white multi-target lightweight design method. The method integrates relative sensitivity analysis and a radial basis function neural network approximation model technology. Parts with significant lightweight potential are identified as design variables by means of relative sensitivity analysis. The thicknesses of the selected parts are used as input samples, and a radial basis function neural network approximation model is constructed through DOE test design. And based on the established approximation model, performing optimization processing through a multi-target genetic algorithm. In conclusion, the core of the method is that the advantages of the relative sensitivity analysis and the radial basis function neural network approximation model are closely combined, so that the calculation burden in the lightweight design process is remarkably reduced, and the optimization efficiency and accuracy are greatly improved; and an efficient and accurate new path is opened up for lightweight design of the body in white of the passenger car.

Description

Technical Field

[0001] The invention relates to the field of automobile design, and in particular to a body-in-white lightweight design method based on relative sensitivity and radial neural network. Background Art

[0002] As the number of cars in the world increases, the impact on energy, environment and traffic safety is becoming increasingly significant. Lightweighting of cars is one of the effective ways to reduce fuel consumption and emissions. As an important part of the weight of the whole vehicle, the lightweight design of the body in white is of great significance to the lightweighting of the whole vehicle. However, the lightweight design of the body in white is a complex system engineering, which needs to consider multiple performance requirements such as collision safety, stiffness and modality at the same time. The traditional body in white design method is usually based on trial and error, with a long design cycle, low efficiency and difficulty in achieving global optimization. Summary of the invention

[0003] The purpose of the present invention is to propose a multi-objective lightweighting method for a vehicle body that combines relative sensitivity analysis with radial basis function neural network approximate model technology. Relative sensitivity analysis is used to identify components with significant lightweighting potential as design variables. The thickness of these selected components is used as input samples, and a radial basis function neural network approximate model is constructed through DOE experimental design. Relying on the established approximate model, a multi-objective genetic algorithm is used for optimization processing.

[0004] The lightweight design method of the body in white based on relative sensitivity and radial neural network includes the following steps:

[0005] Step 1. Based on the BIW CAD data, a BIW finite element model is established using finite element pre-processing software;

[0006] Step 2. Perform modal, bending stiffness and torsional stiffness analysis on the BIW finite element model;

[0007] Step 3. Perform sensitivity analysis using the body sheet metal thickness as the design variable;

[0008] Step 4. Based on step 3, perform relative sensitivity analysis to screen out parts with smaller absolute values ​​of relative sensitivity, and use the thickness of the screened parts as the optimization design variable;

[0009] Step 5. Use the modified extensible lattice sequence method (MELS) to conduct experimental design on the design variables selected in step 4 to obtain sample distribution data;

[0010] Step 6. Establish a radial basis function neural network (RBF) model based on the sample distribution data obtained in step 5, simulate the body modal, torsional stiffness and bending stiffness conditions through the RBF model, and evaluate the accuracy of the approximate model. To evaluate the accuracy of the approximate model, The value range of is [0,1]. The closer it is to 1, the higher the accuracy of the approximate model.

[0011] Step 7. With the minimum body mass and maximum body torsional stiffness as the goal, and the body bending stiffness, first-order torsional frequency and first-order bending frequency as constraints, establish an optimization design mathematical model:

[0012] Variable:\, X=\left [ {{x}_{1},\, {x}_{2},\, ...\, ,\, {x}_{n}} \right ]^{T}

[0013] ;

[0014]

[0015] Where X is the design variable vector, To minimize the body mass, For maximum torsional stiffness, , are the first-order torsional modal frequencies before and after optimization, , are the first-order bending mode frequencies before and after optimization, , are the bending stiffness before and after optimization, For the The lower bound of the design variable, For the The upper limit of the design variable;

[0016] Step 8. Use multi-objective genetic algorithm (MOGA) to perform multi-objective optimization design on the body-in-white;

[0017] Step 9. Filter out the optimal solution from the optimized data and verify that the performance meets the standard;

[0018] Step 10. Determine the final optimization solution and set the optimal solution described in step 9 as the final optimization solution.

[0019] Furthermore, in the BIW finite element model established in the above step 1, all body parts are simulated using shell elements, and the mesh quality standards include: basic unit size is 8mm; 2mm≤unit size≤12mm; unit aspect ratio≤5; unit warpage≤15°; 40°≤inner angle of quadrilateral unit≤140°; 20°≤inner angle of triangular unit≤120°; unit skew angle≤40°; Jacobian≥0.6; and triangular unit ratio≤10%.

[0020] Furthermore, the calculation formulas for the torsional stiffness and bending stiffness described in step 2 are as follows,

[0021] The calculation formula of body-in-white torsional stiffness is:

[0022]

[0023]

[0024] Where, M is the loading torque value; is the body torsion angle; , are the absolute values ​​of Z-direction displacements of the left and right loading points, respectively; is the Y distance between the left and right loading points.

[0025] The calculation formula of the vehicle body bending stiffness is:

[0026]

[0027] In the formula, is the total load value applied to the vehicle body; is the absolute value of the maximum displacement in the Z direction at the right side door sill of the vehicle body; It is the absolute value of the maximum displacement in the Z direction at the left side door sill of the vehicle body.

[0028] Furthermore, the sensitivity analysis described in step 3 is: taking the plate thickness as the design variable, the sensitivity analysis is performed on the first-order torsional frequency, the first-order bending frequency, the torsional stiffness and the bending stiffness, and when performing the sensitivity analysis, the parts with very small mass, such as the mounting bracket, the connecting plate, etc. are ignored. In addition, many parts of the body-in-white are bilaterally symmetrical, and the thickness of these parts must be kept consistent, so the bilaterally symmetrical parts are placed in the same attribute.

[0029] Furthermore, the specific steps of the relative sensitivity analysis described in step 4 are as follows: the modal sensitivity and stiffness sensitivity are divided by the mass sensitivity to obtain the relative sensitivity. A positive relative sensitivity indicates that the plate thickness has the same change trend as the body-in-white mode, mass and stiffness, and a negative relative sensitivity indicates that the plate thickness has the opposite change trend as the body-in-white mode, mass and stiffness, and the larger the absolute value of the relative sensitivity, the greater the change in body performance when the same plate thickness is increased or decreased. Try to select parts with smaller absolute values ​​of relative sensitivity for optimization, so that the change in body performance is relatively small when the thickness of these parts changes.

[0030] The design variable described in step 7 is the thickness value of the body sheet metal. The thickness variable is set to a discrete value at an interval of 0.1 mm, and the range of variation is the initial thickness value. 30%, and the thickness of all parts shall not be less than 0.6mm.

[0031] The present invention has the following beneficial effects:

[0032] The present invention provides a body-in-white lightweight design method based on relative sensitivity and radial neural network, including a multi-objective lightweight optimization analysis of a passenger car body-in-white based on relative sensitivity and a radial basis neural network approximate model and an optimization scheme verification.

[0033] The multi-objective optimization analysis based on the radial basis function neural network RBF approximate model includes: based on the HyperStudy multi-disciplinary multi-objective optimization platform, the thickness of the body-in-white sheet metal with a relatively low absolute value of relative sensitivity is used as the design variable, the MELS method is used for DOE test design, and the finite element model of the body-in-white modal, bending stiffness and torsional stiffness is used as the sample model to establish the RBF approximate model, and the coefficient of determination is guaranteed. , taking the minimum body-in-white mass and the maximum torsional stiffness as the optimization objectives, and the first-order torsional frequency, first-order bending frequency, and bending stiffness as constraints, a multi-objective genetic algorithm (MOGA) was used for optimization analysis to obtain a design variable combination solution that meets the body-in-white performance requirements and has the minimum weight. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is the overall flow chart of the present invention. DETAILED DESCRIPTION

[0035] The present invention provides a lightweight design method for a body-in-white based on relative sensitivity and radial neural network, which can achieve a multi-objective lightweight design of a body. The present invention will be further described in detail below. It is understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

[0036] The lightweight design method of body-in-white based on relative sensitivity and radial neural network has the following specific steps:

[0037] Step 1: Based on the BIW CAD data, use finite element pre-processing software to establish a BIW finite element model.

[0038] 1) The quality of the shell element mesh must meet the following requirements: 2mm≤element size≤12mm; element aspect ratio≤5; element warpage≤15°; 40°≤inner angle of quadrilateral element≤140°; 20°≤inner angle of triangular element≤120°; element skew angle≤40°; Jacobian≥0.6; and the proportion of triangular elements≤10%.

[0039] 2) The body weld point unit adopts acm (equivalenced-(T1+T2) / 2) type; an even number of nodes ≥6 are arranged around the hole. If it is a bolt connection hole, at least one circle of washer units must be set around the hole. The bolt connection is simulated using RBE2 units; the body welds are connected point-to-point using RBE2 units.

[0040] Step 2: Perform modal, bending stiffness and torsional stiffness analysis on the body-in-white finite element model.

[0041] 1) The free modal analysis of the BIW is performed using the Lanczos method, that is, no constraints or loads are applied when calculating the BIW modes, and the first-order torsional vibration mode frequency and the first-order bending vibration mode frequency are extracted.

[0042] 2) The boundary condition for the torsional stiffness analysis of the body in white is to constrain all degrees of freedom at the center of the left and right rear shock absorber mounting points, and apply a torque of 2000Nm at the center of the left and right front shock absorber mounting points. The specific load application method is to apply a concentrated force in the Z direction upward at the center of the left front shock absorber mounting point. , apply a concentrated force downward in the Z direction at the center of the right front shock absorber mounting point ,in and ,in The Y distance between the center of the left and right front shock absorber installation points, in meters. The calculation formula for the torsional stiffness of the body in white is:

[0043]

[0044]

[0045] Where, M is the loading torque value; is the body torsion angle; , are the absolute values ​​of Z-direction displacements of the left and right loading points, respectively; is the Y distance between the left and right loading points.

[0046] 3) The boundary conditions for the body-in-white bending stiffness analysis are to constrain the x, y, z translational freedom of the center position of the front shock absorber mounting point and the z translational freedom of the center position of the rear shock absorber mounting point. A concentrated load of 1000N is applied vertically downward at the H points of the driver and co-driver, and the H points of the left and right passengers in the rear row, for a total of 4000N. The calculation formula for the body bending stiffness is:

[0047]

[0048] In the formula, is the total load value applied to the vehicle body; is the absolute value of the maximum displacement in the Z direction at the right side door sill of the vehicle body; It is the absolute value of the maximum displacement in the Z direction at the left side door sill of the vehicle body.

[0049] Step 3: Use the thickness of the body sheet metal as the design variable to perform sensitivity analysis on the body modal, torsional stiffness, and bending stiffness. When performing sensitivity analysis, ignore parts with very small mass, such as mounting brackets and connecting plates. Many parts of the body-in-white are symmetrical on the left and right, and the thickness of these parts must remain consistent, so the symmetrical parts are placed in the same attribute.

[0050] Step 4. Filter out parts with smaller absolute values ​​of relative sensitivity through relative sensitivity analysis, and use the thickness of the selected parts as the optimization design variable. Because the results of the sensitivity analysis in step 3 only show the relationship between plate thickness and mass, frequency and stiffness, and what we want to get is the relationship between mass and body performance. Therefore, the concept of relative sensitivity is introduced, and the sensitivity ratio obtained by dividing the modal sensitivity and stiffness sensitivity by the mass sensitivity is the relative sensitivity. A positive relative sensitivity indicates that the plate thickness has the same change trend as the body-in-white mode, mass and stiffness, and a negative relative sensitivity indicates that the plate thickness has the opposite change trend as the body-in-white mode, mass and stiffness, and the larger the absolute value of the relative sensitivity, the greater the change in body performance when the same plate thickness is increased or decreased. Try to select parts with smaller absolute values ​​of relative sensitivity for optimization, so that the change in body performance is relatively small when the thickness of these parts changes.

[0051] Step 5: Use the modified extensible lattice sequence (MELS) method to conduct experimental design on the design variables selected in step 4 to obtain sample distribution data.

[0052] Step 6: Establish a radial basis function neural network (RBF) model based on the sample distribution data obtained in step 5, simulate the vehicle body modal, torsional stiffness and bending stiffness conditions through the RBF model, and evaluate the accuracy of the approximate model. To evaluate the accuracy of the approximate model, The value range of is [0,1], The closer it is to 1, the higher the accuracy of the approximate model.

[0053] Step 7: With the minimum body mass and the maximum body torsional stiffness as the goal, and the body bending stiffness, first-order torsional frequency and first-order bending frequency as constraints, establish an optimization design mathematical model:

[0054] Variable:\, X=\left [ {{x}_{1},\, {x}_{2},\, ...\, ,\, {x}_{n}} \right ]^{T}

[0055] ;

[0056]

[0057] Where X is the design variable vector, To minimize the body mass, For maximum torsional stiffness, , are the first-order torsional modal frequencies before and after optimization, , are the first-order bending mode frequencies before and after optimization, , are the bending stiffness before and after optimization, For the The lower bound of the design variable, For the The upper limit of the design variable. The design variable is set to a discrete value with an interval of 0.1 mm, and the range of variation is the initial thickness value. 30%, and the thickness of all parts shall not be less than 0.6mm.

[0058] Step 8: Use multi-objective genetic algorithm (MOGA) to perform multi-objective optimization design on the body-in-white.

[0059] Step 9: Filter out the optimal solution from the optimized data and verify that the performance meets the standard. Select one of the Pareto Front solutions obtained in step 8 as the optimal solution, substitute the thickness combination of all parts in the optimal solution into the original body-in-white finite element model, calculate the body mode, torsional stiffness and bending stiffness respectively, and verify that the performance meets the standard. If the performance index cannot meet the requirements, select a point from the Pareto Front as the optimal solution for performance verification.

[0060] Step 10: Determine the final optimization plan, and set the optimal solution determined in step 9 as the final optimization plan.

Claims

1. A lightweight design method for a body-in-white based on relative sensitivity and radial neural network, characterized in that: The following steps are involved: S1. Establish the BIW finite element model using finite element pre-processing software based on BIW CAD data; S2. Conduct basic performance analysis on the BIW finite element model, including modal analysis, torsional stiffness analysis, and bending stiffness analysis. The first-order torsional frequency, first-order bending frequency, torsional stiffness, and bending stiffness values ​​of the BIW are obtained through finite element analysis as lightweight targets and constraints; The calculation formula of body-in-white torsional stiffness is: Where, M is the loading torque value; is the body torsion angle; , are the absolute values ​​of the Z-direction displacements of the left and right loading points, respectively; L is the Y-direction distance between the left and right loading points; The calculation formula of the vehicle body bending stiffness is: In the formula, is the total load value applied to the vehicle body; is the absolute value of the maximum displacement in the Z direction at the right side door sill of the vehicle body; is the absolute value of the maximum displacement in the Z direction at the left side door sill of the vehicle body; S3. Perform sensitivity analysis with the thickness of the body sheet metal as a variable; S4. Screen out parts with smaller absolute values ​​of relative sensitivity through relative sensitivity analysis, and use the thickness values ​​of the screened parts as optimization design variables; S5. Use the modified extensible lattice sequence (MELS) method to perform experimental design on the design variables described in S4 and obtain sample distribution data; S6. A radial basis function neural network (RBF) model is established based on the sample distribution data obtained in S5. The RBF model is used to simulate the vehicle body modal, torsional stiffness and bending stiffness conditions, and the accuracy of the approximate model is evaluated. The coefficient of determination is used To evaluate the accuracy of the approximate model, The value range of is [0,1], The closer it is to 1, the higher the accuracy of the approximate model; S7. With the minimum vehicle body mass and the maximum vehicle body torsional stiffness as the goal, and the vehicle body bending stiffness, first-order torsional frequency and first-order bending frequency as constraints, establish an optimization design mathematical model: In the formula, is the design variable vector, To minimize the body mass, For maximum torsional stiffness, , are the first-order torsional modal frequencies before and after optimization, , are the first-order bending mode frequencies before and after optimization, , are the bending stiffness before and after optimization, For the The lower bound of the design variable, For the The upper limit of the design variable; S8. Multi-objective genetic algorithm (MOGA) is used to perform multi-objective optimization design of the body-in-white; S9. Filter out the optimal solution from the optimized data and verify that the performance meets the standard; S10. Determine the final optimization solution, and set the optimal solution described in S9 as the final optimization solution.

2. The method for lightweight design of a body-in-white based on relative sensitivity and radial neural network according to claim 1, characterized in that: The sensitivity analysis described in S3 is: taking the plate thickness as the design variable, performing sensitivity analysis on the first-order torsional frequency, the first-order bending frequency, the torsional stiffness and the bending stiffness, and ignoring parts with extremely small mass, such as mounting brackets, connecting plates, etc., when performing the sensitivity analysis.

3. The method for lightweight design of a body-in-white based on relative sensitivity and radial neural network according to claim 1, characterized in that: The relative sensitivity described in S4 is the ratio of the modal sensitivity and stiffness sensitivity to the mass sensitivity, respectively.

4. The method for lightweight design of a body-in-white based on relative sensitivity and radial neural network according to claim 1, characterized in that: The design variable described in S7 is the thickness value of the body sheet metal. The thickness variable is set to a discrete value with an interval of 0.1 mm, and the range of variation is the initial thickness value. 30%, and the thickness of all parts shall not be less than 0.6mm.

5. The method for lightweight design of a body-in-white based on relative sensitivity and radial neural network according to claim 1, characterized in that: The specific method for S9 is as follows: Select one of the Pareto Front solutions obtained in S8 as the optimal solution, substitute the thickness combination of all parts in the optimal solution into the original body-in-white finite element model, calculate the body mode, torsional stiffness and bending stiffness respectively, and verify the performance compliance status. If the performance index cannot meet the requirements, select a point from the Pareto Front as the optimal solution for performance verification.

Images