Side collision simulation deviation correction system based on real vehicle waveform mapping

Abstract

This invention relates to the field of automotive passive safety engineering technology and discloses a side-impact simulation deviation correction system based on real vehicle waveform mapping. The system includes: acquiring the measurement point waveforms of the physical vehicle and the simulation model, establishing an isomorphic measurement point domain to eliminate spatial positional bias; extracting the intrusion velocity towards the passenger compartment and combining it with the input power calculated from the lateral contact force to obtain local energy absorption; using this local energy absorption to define the energy absorption progress and phase time, and extracting the phase misalignment amount characterizing the deviation in the order of component energy absorption; combining the phase misalignment amount, intrusion residual, and energy absorption ratio difference into a weighted residual to construct an objective function for constrained simulation deformation response; calculating the sensitivity matrix of the residual to model parameters and inverting to solve for the parameter correction amount to update the component parameter field; extracting the compliance matrix to propagate the measurement point waveform deviation to the unmeasured area to generate a full-field corrected waveform, and calculating the closure index and energy deviation output results.

Description

Technical Field

[0001] This invention relates to the field of automotive passive safety engineering technology, and more specifically, to a side collision simulation deviation correction system based on real vehicle waveform mapping. Background Technology

[0002] In automotive passive safety development, side impact is a highly stringent evaluation condition. Assessments are based on various mainstream safety evaluation systems (such as FMVSS214 or IIHS), evaluating lateral intrusion performance of the passenger compartment and dummy injury indicators, while also encompassing comprehensive dimensions such as side door strength, component deformation, and global acceleration. In real-world crash tests, because the barrier primarily loads on the vehicle's sidewalls, the collision energy is concentrated and absorbed by the B-pillar, doors, and lower sill, and the main load transfer path varies significantly under different dynamic conditions.

[0003] Currently, when CAE engineers benchmark collision finite element models, the common engineering practice is to fit pulse curves to the lateral acceleration of the entire vehicle or compare the maximum intrusion at a single critical moment. However, the overall lateral acceleration or global impulse only reflects the integral result of the global resultant force over time. In actual side impact tests, because the side components form a close-range series-parallel force absorption system, the load transfer sequence is usually: the door bends first, the middle of the B-pillar moves backward, and then the sill bears the load. However, in finite element simulations, due to limitations in the sensitivity of the contact stiffness mesh, deviations in weld failure conditions, or differences in the setting of material strain rate parameters, it is very easy for this to evolve into the sill bearing the load prematurely or the middle of the B-pillar intruding with a significant lag.

[0004] Because the timing misalignment of energy absorption between components varies layered with the Z-axis coordinate of the measurement point and the load transfer path, the aforementioned benchmarking methods based on global acceleration or a single peak value cannot capture such local load-bearing sequence deviations. Even if the overall pulse performance of the entire vehicle is similar, the local intrusion process and occupant survival space assessment are often severely distorted. Faced with complex waveform differences at measurement points, the engineering development phase mainly relies on the past experience of personnel to repeatedly adjust the thickness, yield stress, or failure limit of each weak component through manual trial and error. This approach makes it difficult to quantitatively map the timing misalignment directly into effective compensation for local stiffness or failure parameters, not only consuming huge computational resources but also making it difficult to ensure the consistency of the model in multi-path energy absorption distribution evaluation. Summary of the Invention

[0005] This invention provides a side collision simulation deviation correction system based on real vehicle waveform mapping, which solves the technical problems mentioned in the background art.

[0006] This invention provides a side-impact collision simulation deviation correction system based on real vehicle waveform mapping, comprising: The waveforms of the measurement points of the physical vehicle and the simulation model are obtained, and the waveforms of the measurement points are mapped to a unified vehicle coordinate system to establish an isomorphic measurement point domain that eliminates spatial position bias. The intrusion velocity toward the passenger compartment is extracted and combined with the input power calculated from the lateral contact force to calculate the local energy absorption used to characterize the spatial distribution of collision energy in the side structure. The phase time between the energy absorption progress and the target progress is defined by the local energy absorption, thereby extracting the phase misalignment amount that characterizes the deviation of the energy absorption sequence of the components; The phase misalignment amount, the intrusion residual characterizing the deformation difference, and the energy absorption ratio difference characterizing the force transmission path are combined into a weighted residual to construct the objective function of the constrained simulation structural response. Calculate the sensitivity matrix of the weighted residuals to the model parameters, and inversely solve for the parameter corrections used to correct local stiffness and failure characteristics; The component parameter field is updated based on the parameter correction amount, and the compliance matrix reflecting the structural mechanical transmission characteristics is extracted. The waveform deviation at the measurement point is propagated to the unmeasured area through the compliance matrix to generate a full-field corrected waveform. The closure index, which characterizes the alignment of the overall mechanical characteristics, the impulse deviation, which characterizes the deviation from the conservation of momentum, and the energy deviation, which characterizes the dissipation balance, are calculated using the full-field corrected waveform. The corrected result, which includes the corrected parameters, is then output.

[0007] The beneficial effects of this invention are as follows: This invention extracts local energy absorption to define the energy absorption progress and phase time, and obtains the phase misalignment amount reflecting the deviation of the energy absorption sequence of components; then, it integrates the intrusion residual and the difference in energy absorption ratio to construct an objective function, and uses this to invert the parameter correction amount to update the component parameter field, thereby generating a full-field correction waveform and deviation index. This invention effectively avoids the limitations of manual trial and error, and can objectively identify the temporal deviation of the B-pillar, door, and sill during the relay load-bearing process, directly converting the waveform error at the measurement point into the correction compensation value of the model material and failure parameters, thereby improving the consistency of the finite element model in local intrusion process, multi-path energy absorption ratio, and global impulse prediction. Attached Figure Description

[0008] Figure 1 This is a flowchart of the side collision simulation deviation correction system based on real vehicle waveform mapping of the present invention. Detailed Implementation

[0009] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0010] like Figure 1 As shown, the side collision simulation deviation correction system based on real vehicle waveform mapping includes: The waveforms of the measurement points of the physical vehicle and the simulation model are obtained, and the waveforms of the measurement points are mapped to a unified vehicle coordinate system to establish an isomorphic measurement point domain that eliminates spatial position bias. The intrusion velocity toward the passenger compartment is extracted and combined with the input power calculated from the lateral contact force to calculate the local energy absorption used to characterize the spatial distribution of collision energy in the side structure. The phase time between the energy absorption progress and the target progress is defined by the local energy absorption, thereby extracting the phase misalignment amount that characterizes the deviation of the energy absorption sequence of the components; The phase misalignment amount, the intrusion residual characterizing the deformation difference, and the energy absorption ratio difference characterizing the force transmission path are combined into a weighted residual to construct the objective function of the constrained simulation structural response. Calculate the sensitivity matrix of the weighted residuals to the model parameters, and inversely solve for the parameter corrections used to correct local stiffness and failure characteristics; The component parameter field is updated based on the parameter correction amount, and the compliance matrix reflecting the structural mechanical transmission characteristics is extracted. The waveform deviation at the measurement point is propagated to the unmeasured area through the compliance matrix to generate a full-field corrected waveform. The closure index, which characterizes the alignment of the overall mechanical characteristics, the impulse deviation, which characterizes the deviation from the conservation of momentum, and the energy deviation, which characterizes the dissipation balance, are calculated using the full-field corrected waveform. The corrected result, which includes the corrected parameters, is then output.

[0011] The side-impact collision simulation deviation correction system based on real vehicle waveform mapping provided in this embodiment is applied to the finite element model calibration scenario in automotive passive safety development. The system establishes a unified response analysis benchmark by synchronously collecting collision test data from actual vehicles and simulation model calculation results. It quantitatively identifies the deviations between the simulation and the actual vehicle in deformation timing, energy distribution, and mechanical transmission paths, automatically inverts and solves for model parameter corrections, and generates full-field corrected waveforms and comprehensive evaluation indicators.

[0012] S201: Perform coordinate transformation using the origin and rotation matrix to convert the global measurement point position in the global coordinate system to the local measurement point position that eliminates the interference of vehicle rigid body translation and rigid body rotation.

[0013] Where q represents the object identifier of the physical vehicle test or simulation model, t is the time variable, and i is the measurement point number. Let be the position vector of the i-th measurement point in the global coordinate system at time t. Let be the position vector of the origin of the vehicle's local coordinate system in the global coordinate system at time t. Let be the rotation matrix of the vehicle's local coordinate system relative to the global coordinate system at time t. It is the transpose of the rotation matrix. Let be the position vector of the i-th measurement point in the vehicle's local coordinate system after transformation.

[0014] The vehicle's local coordinate system is defined according to the SAE J211 standard. The origin is located at the vehicle's center of mass when the vehicle is stationary. The X-axis points in the vehicle's forward direction, the Y-axis points to the left side of the vehicle, and the Z-axis points vertically upward. The coordinate transformation reference points are selected from three non-collinear rigid regions on the vehicle: the undeformed area at the front end of the front longitudinal beam, the undeformed area at the rear end of the rear longitudinal beam, and the undeformed area at the midpoint of the roof crossbeam. All three reference points avoid the direct collision contact area and the main deformation area. The rotation matrix is ​​calculated using the global position vectors of the three reference points, and the optimal orthogonal rotation matrix is ​​solved using singular value decomposition.

[0015] The calculation formula is: ; Solving rotation matrices using singular value decomposition involves: given the initial position vectors of three reference points. and time position vector First, calculate the initial centroid. and time center of mass Then calculate the covariance matrix. ,right Singular value decomposition yields Rotation matrix ,like ,Will Recalculate after inverting the last column of elements .

[0016] S202: Combine the lateral vector pointing towards the inside of the crew compartment to extract the lateral intrusion amount that eliminates the coupling interference of longitudinal and vertical motion components.

[0017] in, Let be the unit lateral vector pointing inwards from the passenger compartment, along the negative Y-axis in the vehicle's local coordinate system. Let be the position vector of the i-th measuring point in the vehicle's local coordinate system at the initial time t=0. Let be the lateral intrusion amount of the i-th measuring point at time t.

[0018] The calculation formula is: ; S203: Compiles the lateral intrusion amount and the intrusion velocity, which characterizes the rate of change of intrusion displacement with time, and the intrusion acceleration, which characterizes the intrusion impact intensity, and generates the waveform at the measuring point.

[0019] in, Let be the intrusion velocity of the i-th measuring point at time t. Let be the intrusion acceleration of the i-th measuring point at time t. Let be the waveform vector of the i-th measurement point.

[0020] Lateral intrusion data is first preprocessed, using the 3σ criterion to remove outliers. Data points greater than the mean plus three standard deviations or less than the mean minus three standard deviations are considered outliers and replaced by linear interpolation between two adjacent normal data points. Then, a fourth-order Butterworth low-pass filter is used for filtering, with a cutoff frequency set to 100Hz. Zero-phase filtering is employed to avoid signal phase shift. Intrusion velocity is obtained by taking the first derivative of the filtered lateral intrusion with respect to time, and intrusion acceleration is obtained by taking the second derivative of the filtered lateral intrusion with respect to time. The derivatives are calculated using the central difference method.

[0021] The calculation formula is: ; The discrete calculation formula for the central difference method includes: for a sampling step size of... Discrete time series , The formula for calculating the first derivative, i.e., the invasion velocity, is: The formula for calculating the second derivative, i.e., the intrusion acceleration, is: For endpoints Using forward difference, , For endpoints Using backward difference, , To supplement the implementation method of zero-phase filtering, first perform a forward fourth-order Butterworth low-pass filter on the original signal, then invert the filter result and perform an inverse filter with the same parameters, and finally invert the second filter result again to obtain the zero-phase filtered signal.

[0022] S301: Obtain the lateral contact force and lateral velocity of the barrier as characterized by the barrier's dynamic response, and calculate the input power representing the total energy input boundary of the system.

[0023] in, Let t be the lateral contact force between the barrier and the vehicle body. Let be the lateral velocity of the barrier at time t. Let be the input power at time t. For barrier quality, For the lateral acceleration of the barrier, This refers to the contact area between the barrier and the vehicle body. Let be the lateral contact force of the c-th contact unit within the contact area.

[0024] Contact area The contact force is defined as the set of all units that come into contact with the barrier and the vehicle's side structure, including units that contact the barrier with the doors, B-pillars, door sills, and fenders, excluding contact units of non-primary energy-absorbing structures such as bumpers and rearview mirrors. The contact state judgment threshold is a unit penetration depth greater than 0.1 mm; units with a penetration depth less than this threshold are not included in the contact force calculation. In the actual vehicle test, the barrier's lateral velocity is obtained by integrating the signal from the acceleration sensor mounted on the barrier, with the initial velocity being the barrier's stable speed before the collision.

[0025] The formula for calculating input power is: ; The formula for calculating the lateral contact force of a solid vehicle is: ; The formula for calculating the lateral contact force in the simulation model is: ; S302: Calculate the local energy absorption ratio for the non-work component of the peeling structure rebound by combining the component length and the intrusion velocity towards the crew compartment.

[0026] in, Let be the characteristic length of the component to which the i-th measuring point belongs. The intrusion velocity is the velocity toward the crew compartment, i.e., the positive component of the intrusion velocity, and N is the total number of measuring points. Let be the local energy absorption ratio of the component to which the i-th measuring point belongs at time t.

[0027] The characteristic length of a component is defined according to its type: the characteristic length of the B-pillar is the vertical distance between its upper and lower mounting points; the characteristic length of the door is the horizontal distance from the center of the door hinge to the center of the door lock; the characteristic length of the sill is the horizontal distance between its front and rear mounting points; the characteristic length of the fender is the horizontal distance between its front and rear edges; and the characteristic length of the roof side beam is the horizontal distance between its front and rear ends. For irregularly shaped components, the characteristic length is the length of the longer side of the component's circumscribed rectangle.

[0028] The formula for calculating the positive component is: ; The formula for calculating the local energy absorption ratio is: ; S303: Based on the local energy absorption ratio, perform a time-domain cumulative summation operation on the input power to obtain a quantitative characterization of the local energy absorption share of each side component during the collision process.

[0029] in, Let be the local energy absorption of the component belonging to the i-th measuring point from the start of the collision to time t. This is the time variable for integration.

[0030] Numerical integration employs the trapezoidal rule, with the integration step size consistent with the data sampling step size, set to 1 ms. For discrete-time series data, local energy absorption is obtained by performing a trapezoidal integration on the product of the input power and the proportion of local energy absorption.

[0031] The calculation formula is: ; The discrete calculation formula for the trapezoidal integral method includes: for discrete time series , The integral result is ,in This is the sampling step size.

[0032] S401: Normalize the transient energy absorption by using the final energy absorption at the end of the collision process to obtain the energy absorption progress that eliminates the interference of the difference in absolute energy absorption amplitude of each component.

[0033] Where T is the total time for the collision process to end. Let be the final energy absorbed by the i-th component during the entire collision process. Let represent the energy absorption progress of the i-th component at time t.

[0034] The criteria for determining the collision end time T are as follows: when the relative speed between the barrier and the vehicle body is less than 0.1 m / s and the duration exceeds 5 ms, or reaches 150 ms as specified in the test standard, the earlier of the two times is taken as the collision end time.

[0035] The calculation formula is: ; S402: Map the energy absorption progress to the time domain, obtain the phase time that characterizes the sequence of structural deformation relay, and extract the phase difference between the simulation model and the actual vehicle response time axis.

[0036] Where e is a specific energy absorption rate parameter, with a value ranging from 0 to 1. The phase time required for the i-th component to reach the energy absorption progress e is the energy absorption progress function. inverse function, Let e ​​be the phase time at which the i-th component in the simulation model reaches the energy absorption progress e. Let e ​​be the phase time at which the i-th component in the physical vehicle reaches the energy absorption progress e. Let be the phase difference of the i-th component at the energy absorption rate e.

[0037] The inverse function is solved using the monotonically increasing cubic spline interpolation method. First, the discrete energy absorption progress time series is fitted with monotonically increasing cubic splines to ensure that the fitted function is strictly monotonically increasing. Then, the phase time corresponding to the energy absorption progress e is solved by the bisection method.

[0038] The formula for calculating phase time is: ; The formula for calculating phase difference is: ; S403: Based on the height of the measuring point and the path matrix representing the physical affiliation of the load transfer path, perform a spatial structure dimension decoupling operation on the phase difference to obtain the offset coefficient, and construct the phase misalignment quantity that removes the occasional error of a single sensor.

[0039] in, Let X be the height coordinates of the i-th measuring point, and let X be the path matrix, where each row corresponds to a measuring point and each column corresponds to a load transfer path. Let a(e) be the phase difference column vector of all measuring points under the energy absorption progress e, and let a(e) be the offset coefficient vector. Let be the phase misalignment at the i-th measuring point under the energy absorption progress e.

[0040] Path Matrix X Construction Rules: The side structure is divided into three main load transfer paths: the door path, the B-pillar path, and the sill path. Each measuring point is assigned to a corresponding path based on its installation location. Measuring points located on a single component have a path element value of 1, while those on other paths have an element value of 0. Measuring points located at the intersection of two components belong to both paths, with element values ​​of 0.5 for both paths. Measuring points located at the intersection of three components belong to all three paths, with element values ​​of 1 / 3 for all three paths. The number of columns in the path matrix equals the number of paths, and the number of rows equals the total number of measuring points.

[0041] The formula for calculating the offset coefficient is: ; The formula for calculating phase misalignment is: ; The path matrix includes the following: the indices of the three main paths of the side structure are, in order, door path 1, B-pillar path 2, and sill path 3. The row vector of the path matrix corresponding to the measuring point installed in the middle of the door is... The row vector corresponding to the measuring point installed on the upper part of the B-pillar is: The row vector corresponding to the measuring point installed in the middle of the threshold is... The row vector corresponding to the measuring point installed at the junction of the car door and the B-pillar is: The row vector corresponding to the measuring point installed at the intersection of the B-pillar and the door sill is: .

[0042] S501: Extract the displacement residual and velocity residual between the physical vehicle and the simulation model within the same energy absorption progress domain to form an intrusive residual used to characterize the local deformation follow-up deviation.

[0043] in, Let be the correction vector to be determined, containing all the model parameters that need to be corrected. For the i-th measurement point in the simulation model, in phase time Lateral intrusion below, For the i-th measurement point in the physical vehicle, in phase time Lateral intrusion below, For displacement residuals, For the i-th measurement point in the simulation model, in phase time The speed of intrusion below, For the i-th measurement point in the physical vehicle, in phase time The speed of intrusion below, This represents the velocity residual.

[0044] The energy absorption progress domain e ranges from 0 to 1, with a basic sampling interval of 0.01. Intensified sampling is performed in the region of rapid energy absorption change, from e=0.2 to e=0.8, with a sampling interval of 0.005. For each sampling point e, the displacement and velocity data of the physical vehicle and the simulation model at the corresponding phase time are extracted to calculate the residuals.

[0045] The formula for calculating displacement residuals is: ; The formula for calculating velocity residual is: ; S502: Quantify the transient distribution of the target structural path energy absorption rate in the total system energy absorption rate, and compare the difference in the distribution between the physical vehicle and the simulation model to obtain the energy absorption ratio difference used to characterize the deviation of the mechanical transmission path.

[0046] in, These are elements of the path attribution matrix, identical to the path matrix X in S403. A value of 1 indicates that the i-th measurement point belongs to the g-th structural path, and 0 indicates that it does not. Measurement points at the boundary retain their weight values. For the i-th component in phase time The energy absorption rate, i.e., the first derivative of local energy absorption with respect to time, Let g be the percentage of transient energy absorption of the g-th structural path at energy absorption progress e. The difference in the energy absorption ratio of the g-th structural path at energy absorption progress e.

[0047] The energy absorption rate is obtained by taking the first derivative of local energy absorption with respect to time, and is calculated using the central difference method.

[0048] The formula for calculating the percentage of transient energy absorption is: ; The formula for calculating the energy absorption ratio difference is: ; S503: The phase misalignment quantity, intrusion residual and energy absorption ratio difference are fused, and the dimensional difference interference of multi-source physical quantities is eliminated by the covariance matrix to generate a weighted residual.

[0049] Where C(e) is the covariance matrix under the energy absorption progress e, which is calculated from the variance and covariance of each residual component. For phase misalignment vectors, Let be the displacement residual vector. Let the velocity residual vector be... Let the difference in energy absorption ratio be the vector. This is the weighted residual vector.

[0050] The covariance matrix C(e) is calculated based on the measurement data of three repeated collision tests of the actual vehicle. For each energy absorption progress sampling point e, the variance and covariance of the phase misalignment, displacement residual, velocity residual and energy absorption ratio difference are calculated respectively to form a block diagonal matrix. Each block corresponds to a covariance submatrix of a type of residual.

[0051] The calculation formula is: ; The calculation of the inverse square root of the covariance matrix includes: for a symmetric positive definite covariance matrix First, perform eigenvalue decomposition to obtain ,in The eigenvector matrix, Let the eigenvalues ​​be the diagonal matrix, and then construct the inverse square root diagonal matrix. Its diagonal elements are the reciprocals of the square roots of the corresponding eigenvalues, and the inverse square root of the final covariance matrix is... .

[0052] S504: Introduces regularization coefficients and smoothing operators based on cross-validation to construct a penalty term to suppress sudden changes in parameters in local regions, and combines weighted residuals and the penalty term to generate the objective function.

[0053] in, This represents the energy absorption progress domain, with values ​​ranging from 0 to 1. Let L be the regularization coefficient determined by cross-validation, L be the smoothing operator, and J(ξ) be the objective function.

[0054] Regularity coefficient The method, determined through 10-fold cross-validation, involves randomly dividing all measurement points into 10 equal subsets. Each time, 9 subsets are used as the training set to calculate parameter corrections, and the remaining subset is used as the validation set to calculate the objective function value. This process is repeated 10 times, and the subset with the smallest average objective function value is selected. As the optimal regularization coefficient The smoothing operator L uses a second-order difference matrix. For two-dimensional structures, the Laplace difference scheme is used, and the boundary points are approximated by a first-order difference. The dimension of the difference matrix is ​​the same as the dimension of the correction vector to be determined.

[0055] The calculation formula is: ; Constructing a second-order difference smoothing operator includes: for the correction vector to be determined arranged in component number order, the second-order difference matrix... The dimension is the same as the dimension of the correction vector to be determined. For internal components... , , and components Adjacent components with physical connections corresponding All other elements are 0; for boundary components , Adjacent components The remaining elements are 0.

[0056] S601: Construct an exponential correction term to avoid interference from non-physical negative value calculations, and apply it to the correction vector to be determined to define the correction parameter within the physical value boundary range.

[0057] Where k is the solution step index, Let r be the original value of the r-th model parameter. Let r be the r-th component of the correction vector to be determined in the k-th iteration. This is the correction value for the r-th parameter during the k-th iteration.

[0058] The correction vector to be determined The model parameters include the following types: material yield strength, material elastic modulus, material failure strain, weld failure load, and contact stiffness coefficient. The parameter discretization granularity is a set of parameters for each body structural component. The body structural components are divided into 25 parts according to natural segmentation, including the inner door panel, outer door panel, B-pillar inner panel, B-pillar reinforcement panel, sill inner panel, sill reinforcement panel, fender, and roof side beam. Each component corresponds to 5 parameters to be corrected, and the total dimension of the correction vector is 125.

[0059] The calculation formula is: ; The components of the correction vector to be determined include: material yield strength, which is the stress value when the material begins to undergo plastic deformation; material elastic modulus, which is the ratio of stress to strain in the elastic deformation stage; material failure strain, which is the engineering strain value when the material fractures; weld joint failure load, which is the maximum axial load that the weld joint can withstand when it undergoes tensile fracture; and contact stiffness coefficient, which is the normal contact force required per unit penetration depth at the contact interface.

[0060] S602: Extract the partial derivatives of the weighted residuals with respect to the correction vector to obtain the sensitivity matrix that characterizes the system response's dependence on local material properties.

[0061] in, For the m-th component of the weighted residual vector, Let be the element in the m-th row and r-th column of the sensitivity matrix S.

[0062] The sensitivity matrix is ​​calculated using the forward finite difference method, for each component of the correction vector to be determined. Apply relative perturbation step size Calculate the weighted residual vector after perturbation. ,in Let r be the r-th unit vector, and then calculate the sensitivity coefficient using the finite difference formula.

[0063] The calculation formula is: ; S603: Solve the normal equation by combining the sensitivity matrix, iteratively obtain the parameter update amount, and update the correction vector to be determined.

[0064] in, This represents the parameter update amount in the k-th iteration. Let be the correction vector to be determined in the (k+1)th iteration.

[0065] The normal equations are solved using the preprocessed conjugate gradient method. The preprocessing matrix is ​​a diagonal matrix of the sensitivity matrix. The maximum number of iterations is set to 1000, and the convergence threshold is set to 1e-6. The same regularization coefficients as those used in cross-validation are introduced during the solution process. A stable solution process is achieved, avoiding numerical instability caused by ill-conditioned regular equations.

[0066] The formula for calculating parameter update amount is: ; The formula for updating the correction vector to be determined is: ; The iteration termination condition for the preprocessed conjugate gradient method includes: when the L2 norm of the iteration residual is less than... Alternatively, the solution process for the normal equation may be terminated when the number of iterations reaches 1000.

[0067] S604: The vector set that achieves convergence in minimizing the objective function is used as the final parameter correction for calibrating the local stiffness and failure characteristics of the simulation model.

[0068] in, This is the final parameter adjustment amount.

[0069] The convergence conditions for the iteration include: the relative change of the objective function is less than 1e-4, the objective function value does not decrease after 10 consecutive iterations, or the number of iterations reaches the preset maximum number of iterations, 50. The iteration terminates when any of the conditions is met, and the correction vector to be determined at this point is the optimal parameter correction.

[0070] The calculation formula is: ; S701: The discrete parameter correction values ​​are mapped to the structural continuum using a spatial interpolation function to construct a global correction field, which in turn updates and generates a component parameter field that reflects the physical consistency of the local mesh.

[0071] in, Let be the spatial interpolation function corresponding to the r-th parameter, and let x be the coordinates of any point on the structural continuum. For global correction field, For the original parameter field, This is the corrected component parameter field.

[0072] The spatial interpolation function is selected based on the parameter type: linear interpolation is used for material yield strength, elastic modulus, and failure strain; quadratic Lagrange interpolation is used for weld failure load and contact stiffness coefficient. The interpolation nodes are the geometric centers of each component. For complex curved surface structures, linear interpolation using triangular elements is used, with the three vertices of each triangular element being the geometric centers of the three adjacent components.

[0073] The formula for calculating the global correction field is: ; The formula for calculating component parameter field is: ; The calculation of the linear interpolation function for the triangular element includes: the coordinates of the three vertices of the triangular element. any point The three corresponding interpolation functions are as follows: ; ; ; S702: Compare the measurement point responses of the physical vehicle with those of the corrected simulation model, and extract the measurement point waveform deviations used to characterize the residual error in the known area.

[0074] Where I is the set of known measurement points, Let I be the measurement point response vector of the physical vehicle at the known measurement point set I. This is the corrected simulation model's response vector at the known set of measurement points I. This is the waveform deviation vector at the measurement point.

[0075] The measurement point response vector contains three components: displacement, velocity, and acceleration, and its structure is the same as the measurement point waveform vector defined in S203. For each energy absorption progress sampling point e, the deviation is calculated by extracting the response vectors of the physical vehicle and the corrected simulation model at the corresponding phase time.

[0076] The calculation formula is: ; S703: Extract the tangential stiffness characteristics of the corrected model under a specified energy absorption rate and convert them into a flexibility matrix that reflects the actual mechanical topological transmission capability.

[0077] in, G(e) is the tangent stiffness matrix of the corrected model at the energy absorption progress e, and G(e) is the compliance matrix. For the compliance submatrix of the measurement points, and For the cross-compliance submatrix, This is the blind zone compliance submatrix.

[0078] Tangent stiffness matrix extraction method: First, calculate the total energy absorption progress at each incremental step during the simulation, and find the two incremental steps that are closest to the target energy absorption progress e. and The corresponding total energy absorption rate is and Then, the tangent stiffness matrix under the energy absorption progress e is obtained by linear interpolation. For large tangent stiffness matrices, the substructure method is used to reduce the model order. The side structure is divided into 12 substructures, and the first 20 modes of each substructure are retained. The dimension of the reduced tangent stiffness matrix is ​​approximately 240. Then, the LU decomposition method is used to inverse the model to obtain the compliance matrix.

[0079] The calculation formula is: ; The specific rules for dividing the substructure include: the side structure is divided into 12 substructures, namely, the inner door panel, outer door panel, inner B-pillar panel, B-pillar reinforcement panel, outer B-pillar panel, inner sill panel, sill reinforcement panel, fender, roof side beam, lower A-pillar, lower C-pillar, and lower side crossbeam. The selection criteria for the supplementary modal truncation order are as follows: the first 20 modes contain more than 95% of the vibration energy of the side structure and can reflect the main mechanical transmission characteristics of the structure during a collision.

[0080] S704: Based on the structural strength correlation defined by the compliance matrix, the waveform deviation of the measuring point is projected along the physical force transmission path to the blind zone structure where no sensor is placed to calculate the propagation deviation.

[0081] in, The Moore-Penrose pseudoinverse of the compliance submatrix at the measurement point. For blind zone structure set, This is the propagation deviation vector of the blind zone structure.

[0082] The pseudo-inverse is calculated using the truncated singular value decomposition method. First, the compliance submatrix at the measurement point is calculated. Perform singular value decomposition ,in The singular value diagonal matrix is ​​formed by first creating a matrix of singular values, and then setting all singular values ​​less than 1e-6 times the largest singular value to zero, resulting in a truncated singular value matrix. Finally, calculate the pseudo-inverse. ,in for The pseudo-inverse matrix.

[0083] The calculation formula is: ; S705: The full-field corrected waveform with full structure coverage is reconstructed by fusing the original full-field response waveform with the propagation deviation after time-domain mapping.

[0084] in, This is the original, uncorrected full-field response waveform. This is a global energy absorption progress parameter. This is a local correction waveform for the blind zone structure. To correct the waveform at the measurement point, Corrects the waveform for the entire field.

[0085] Global energy absorption progress parameter Defined as the total energy absorption progress of the entire side structure, which is the sum of the local energy absorption of all components of the side structure divided by the total energy absorption of the side structure. The propagation deviation within the energy absorption progress domain is also considered. By mapping the global energy absorption progress parameter to the time domain, the time domain propagation deviation is obtained. Then, the waveform is superimposed with the original full-field response waveform to obtain the blind zone correction waveform. The measurement point correction waveform is the actual vehicle measurement point response waveform.

[0086] The formula for calculating the local correction waveform in the blind zone is: ; The formula for calculating the full-field corrected waveform is: ; S801: Extract the corrected phase misalignment, the corrected intrusion residual, and the corrected energy absorption ratio difference, and perform integral aggregation based on covariance weights within the energy absorption domain to form a closed index that eliminates the risk of local misjudgment of a single indicator and characterizes the convergence of the global state.

[0087] in, This is the corrected phase misalignment vector. This is the corrected displacement residual vector. This is the corrected energy absorption ratio difference vector. , , These are the covariance weight matrices for phase misalignment, displacement residual, and energy absorption ratio difference, respectively, with Q being the closure index.

[0088] The covariance weight matrix is ​​the integral result of the corresponding covariance matrix in S503 over the entire energy absorption domain, i.e. , , ,in , , These are the submatrices in the covariance matrix C(e) corresponding to the phase misalignment, displacement residual, and energy absorption ratio difference, respectively.

[0089] The calculation formula is: ; S802: Evaluate the dynamic boundary response of the system after the simulation is corrected, and extract the cumulative difference of the lateral contact force in the total time domain as the impulse deviation characterizing the deviation of the collision momentum conservation.

[0090] in, To correct the lateral contact force in the simulation model, For the lateral contact force of the physical vehicle, This is the impulse deviation.

[0091] Impulse calculation uses the trapezoidal rule for numerical integration, with the integration time range from the collision start time t=0 to the collision end time t=T, and the integration step size is consistent with the data sampling step size.

[0092] The calculation formula is: ; S803: Extract the cumulative difference in input power over the total time domain as the energy deviation characterizing the energy dissipation balance.

[0093] in, To correct the input power of the simulation model, For the input power of the physical vehicle, This refers to energy deviation.

[0094] Energy calculations employ the trapezoidal rule for numerical integration, with the integration time ranging from the collision start time t=0 to the collision end time t=T. The integration step size is consistent with the data sampling step size.

[0095] The calculation formula is: ; S804: Encapsulates the parameter corrections representing microscopic properties, the full-field correction waveform representing macroscopic response, along with the closure index, impulse deviation, and energy deviation, at the engineering file level, and outputs the correction results.

[0096] in, This is the corrected result for the final output.

[0097] The engineering file-level encapsulation includes: parameter correction values ​​saved in LS-DYNA .k format and Abaqus .inp format, mapped to the corresponding parts in the simulation model according to the part number, with the parameter correction values ​​for each part listed separately; the full-field correction waveform saved in CSV format, containing the number of each node, time series, and corresponding displacement, velocity, and acceleration values; and closure index, impulse deviation, and energy deviation saved in PDF format evaluation reports, including numerical values ​​of each index, comparison charts, and correction effect analysis.

[0098] The calculation formula is: .

[0099] The mapping between parameter corrections and simulation models includes: for material parameters, applying the parameter correction value corresponding to each component to all shell elements and solid elements contained in that component; for weld point parameters, applying the weld point failure load correction value corresponding to each component to all weld point elements on that component; and for contact parameters, applying the contact stiffness coefficient correction value corresponding to each contact pair to all contact elements of that contact pair.

[0100] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A side collision simulation deviation correction system based on real vehicle waveform mapping, characterized in that, include: The waveforms of the measurement points of the physical vehicle and the simulation model are obtained, and the waveforms of the measurement points are mapped to a unified vehicle coordinate system to establish an isomorphic measurement point domain that eliminates spatial position bias. The intrusion velocity toward the passenger compartment is extracted and combined with the input power calculated from the lateral contact force to calculate the local energy absorption used to characterize the spatial distribution of collision energy in the side structure. The phase time between the energy absorption progress and the target progress is defined by the local energy absorption, thereby extracting the phase misalignment amount that characterizes the deviation of the energy absorption sequence of the components; The phase misalignment amount, the intrusion residual characterizing the deformation difference, and the energy absorption ratio difference characterizing the force transmission path are combined into a weighted residual to construct the objective function of the constrained simulation structural response. Calculate the sensitivity matrix of the weighted residuals to the model parameters, and inversely solve for the parameter corrections used to correct local stiffness and failure characteristics; The component parameter field is updated based on the parameter correction amount, and the compliance matrix reflecting the structural mechanical transmission characteristics is extracted. The waveform deviation at the measurement point is propagated to the unmeasured area through the compliance matrix to generate a full-field corrected waveform. The closure index, which characterizes the alignment of the overall mechanical characteristics, the impulse deviation, which characterizes the deviation from the conservation of momentum, and the energy deviation, which characterizes the dissipation balance, are calculated using the full-field corrected waveform. The corrected result, which includes the corrected parameters, is then output.

2. The side collision simulation deviation correction system based on real vehicle waveform mapping according to claim 1, characterized in that, The establishment of the isomorphic measurement point domain that eliminates spatial positional bias includes: Coordinate transformation is performed using the origin and rotation matrix to convert the global measurement point position in the global coordinate system to the local measurement point position that eliminates the interference of vehicle rigid body translation and rigid body rotation. By combining the lateral vector pointing towards the inside of the crew compartment, the lateral intrusion amount that eliminates the coupling interference of longitudinal and vertical motion components is extracted; The lateral intrusion amount, the intrusion velocity (characterizing the rate of change of intrusion displacement with time), and the intrusion acceleration (characterizing the intrusion impact intensity) are compiled to generate the waveform at the measuring point.

3. The side collision simulation deviation correction system based on real vehicle waveform mapping according to claim 1, characterized in that, The calculations are used to characterize the local energy absorption of the spatial distribution of collision energy in the side structure, including: Obtain the lateral contact force and lateral velocity of the barrier as characterized by the barrier's dynamic response, and calculate the input power that characterizes the total energy input boundary of the system; By combining the component length with the intrusion velocity toward the crew compartment, the local energy absorption ratio for the non-work component of the peeling structure rebound is calculated. Based on the local energy absorption ratio, a time-domain cumulative summation operation is performed on the input power to obtain the local energy absorption, which quantitatively characterizes the actual energy absorption share of each side component during the collision process.

4. The side collision simulation deviation correction system based on real vehicle waveform mapping according to claim 1, characterized in that, The extraction of the phase misalignment amount, which characterizes the order of energy absorption deviation of the components, includes: The transient energy absorption is normalized by using the final energy absorption at the end of the collision process to obtain the energy absorption progress that eliminates the interference of the difference in absolute energy absorption amplitude of each component. Map the energy absorption progress to the time domain, obtain the phase time that characterizes the relay order of structural deformation, and extract the phase difference between the simulation model and the actual vehicle response time axis; Based on the height of the measuring point and the path matrix representing the physical affiliation of the load transfer path, a decoupling operation of the spatial structure dimension is performed on the phase difference to obtain the offset coefficient, and the phase misalignment amount is constructed by removing the occasional error of a single sensor.

5. The side collision simulation deviation correction system based on real vehicle waveform mapping according to claim 1, characterized in that, The construction of the weighted residuals and objective function includes: The displacement residual and velocity residual between the physical vehicle and the simulation model are extracted within the same energy absorption progress domain to form the intrusive residual used to characterize the local deformation follow-up deviation. The transient distribution of the target structural path energy absorption rate in the total system energy absorption rate is quantified, and the difference in the distribution between the actual vehicle and the simulation model is compared to obtain the energy absorption ratio difference used to characterize the deviation of the mechanical transmission path. The weighted residual is generated by integrating the phase misalignment quantity, the intrusion residual, and the energy absorption ratio difference, and eliminating the dimensional differences of multiple physical quantities via the covariance matrix. A penalty term is constructed by introducing a regularization coefficient based on cross-validation and a smoothing operator to suppress sudden changes in parameters in local regions. The objective function is generated by combining the weighted residual and the penalty term.

6. The side collision simulation deviation correction system based on real vehicle waveform mapping according to claim 1, characterized in that, The inversion solution is used to correct the parameter corrections for local stiffness and failure characteristics, including: An exponential correction term is constructed to avoid interference from the calculation of non-physical negative values, and it is applied to the correction vector to be determined to define the correction parameter within the physical value boundary range. Extract the partial derivative of the weighted residual with respect to the correction vector to be determined, and obtain the sensitivity matrix that characterizes the dependence of the system response on local material properties; Solve the normal equation using the aforementioned sensitivity matrix, and iteratively obtain the parameter update amount; The vector set that achieves the minimization and convergence of the objective function is used as the parameter correction amount for calibrating the local stiffness and failure characteristics of the simulation model.

7. The side collision simulation deviation correction system based on real vehicle waveform mapping according to claim 1, characterized in that, The generation of the full-field corrected waveform includes: The discrete parameter correction values ​​are mapped to the structural continuum using a spatial interpolation function to construct a global correction field, which is then used to update and generate the component parameter field that reflects the physical consistency of the local mesh. By comparing the measurement point responses of the physical vehicle with those of the corrected simulation model, the waveform deviation of the measurement points is extracted to characterize the residual error in the known area. Extract the tangential stiffness characteristics of the corrected model under the specified energy absorption rate and convert them into the compliance matrix that reflects the actual mechanical topological transmission capability; Based on the structural strength correlation defined by the compliance matrix, the waveform deviation at the measurement point is projected along the physical force transmission path to the blind zone structure where no sensors are placed to calculate the propagation deviation. The full-field corrected waveform is reconstructed by fusing the original full-field response waveform with the propagation deviation after time-domain mapping, thus covering the entire structure.

8. The side collision simulation deviation correction system based on real vehicle waveform mapping according to claim 1, characterized in that, The calculation and output of the closure index, impulse deviation, and energy deviation include: The corrected phase misalignment, the corrected intrusion residual, and the corrected energy absorption ratio difference are extracted and integrated and aggregated based on covariance weights in the energy absorption domain to form the closure index that eliminates the risk of local misjudgment of a single index and characterizes the convergence of the global state. The system dynamic boundary response after the correction simulation is evaluated, and the cumulative difference of the lateral contact force in the total time domain is extracted as the impulse deviation characterizing the deviation of the collision momentum conservation. The cumulative difference in the input power over the total time domain is extracted as the energy deviation characterizing the energy dissipation balance. The parameter correction amount, which characterizes the microscopic properties, the full-field correction waveform, which characterizes the macroscopic response, together with the closure index, the impulse deviation, and the energy deviation, are encapsulated at the engineering file level, and the correction result is output.

Images