Vehicle frame multi-objective optimization method based on rollover collision battery pack safety

Abstract

The invention relates to the technical field of vehicles, in particular to a vehicle frame multi-objective optimization method based on rollover collision battery pack safety, which comprises the following steps of: constructing a battery homogenization mechanical model to obtain the distance between current collectors of a positive electrode and a negative electrode of a battery under stress; establishing a battery electrical model through the equivalent circuit, and judging whether the battery is short-circuited in the battery electrical model according to the distance between the current collectors of the positive electrode and the negative electrode; establishing a mobile charging vehicle frame three-dimensional model to obtain an instant angular velocity, and analyzing the damage degree of the battery through the coupling model; and converting each unit of the finite element analysis model into independent cells, adjusting cell attributes according to a preset rule, and performing multi-round iteration by adjusting the cell attributes so as to complete optimization of the frame model. According to the method, aiming at the defect of safety optimization of the battery pack during rollover collision, the lightweight of the frame and the improvement of the safety of the battery pack can be realized at the same time, and the optimization efficiency of a complex model is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle technology, and in particular to a multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision. Background Art

[0002] In order to solve the problem that electric vehicles run out of power and cannot be moved before reaching the charging station, some people have designed a mobile charging vehicle. By loading a large number of rechargeable batteries in the vehicle body and configuring a distribution system and charging guns to form a mobile energy-carrying battery charging and distribution system, emergency charging of electric vehicles without power can be achieved.

[0003] The frame of the mobile charging vehicle is an important load-bearing component of the vehicle body. In order to meet the requirements of compact appearance and variable load conditions, the design of the frame structure needs to comprehensively consider multiple aspects such as strength and lightweight to ensure the safety of the battery pack when the mobile charging vehicle rolls over and collides. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a multi-objective optimization method for the frame based on the safety of the battery pack in a rollover collision. In view of the shortcomings of the optimization of the battery pack safety in a rollover collision, the lightweight of the frame and the improvement of the battery pack safety can be achieved at the same time, effectively improving the optimization efficiency of the complex model.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] A multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision comprises the following steps:

[0007] S1. Establishment of the electromechanical coupling model of the battery: by conducting quasi-static mechanical experiments on the positive electrode, negative electrode and separator of the battery, a homogenized mechanical model of the battery is constructed to obtain the distance between the current collectors of the positive electrode and the negative electrode of the battery under stress; a battery electrical model is established through an equivalent circuit, and according to the distance between the current collectors of the positive electrode and the negative electrode, whether the battery is short-circuited is determined in the battery electrical model;

[0008] S2. Establishment of rollover collision model: Establish a three-dimensional model of the mobile charging vehicle frame, and establish a finite element analysis model of the mobile charging vehicle frame based on the three-dimensional model to obtain the instantaneous angular velocity of the mobile charging vehicle frame at the position where it contacts the cement ground, and analyze the degree of battery damage based on the instantaneous angular velocity and through the coupling model of the battery homogenization mechanical model and the battery electrical model;

[0009] S3. Optimization of the frame model: convert each unit of the finite element analysis model into an independent cell, process the converted finite element analysis model through the solver LS-DYNA to extract key response parameters, and adjust the cell properties according to preset rules. Perform multiple rounds of iterations on the adjustment of the cell properties to complete the optimization of the frame model.

[0010] Further, in step S1, the load-displacement curve of the battery is obtained according to the positive and negative electrode and shell test, the positive and negative electrode and shell compression test, and the RS structure indentation test, and the stress-strain curve of the positive electrode, negative electrode and separator of the battery is calculated by the load-displacement curve:

[0011]

[0012] Where P is the load, A is the original cross-sectional area of ​​the battery; ε 1 is the experimental strain; L 0 is the original length of the battery; L is the length of the battery after deformation;

[0013] Furthermore, according to the relationship between stress and stress between each layer in the RS structure, the stress-strain curve of the RS structure is calculated through the electrode and the diaphragm, and the stress-strain curve of the RS structure is converted into a constitutive model through equivalent energy to obtain the increase in internal energy of the RS structure during the process:

[0014] σ average =σ a v a +σ c v c +σ s v s Formula (2)

[0015]

[0016] σ e =Aε 1 n Formula (4)

[0017] Among them, σ average is the average stress of RS structure; v a 、v c 、v s are the ratios of the volume of the negative electrode, positive electrode and separator to the volume of the RS structure; U represents the internal energy increased by the RS structure during the process, ε 1 represents the experimental strain, σ e represents the stress of the electrode, and A and n are fitting parameters.

[0018] Furthermore, the failure strain of the RS structure after homogenization is ε f, the energy absorbed by the homogenized material under failure strain is U 0 , then:

[0019]

[0020] Among them, ε f is the failure strain; U 0 is the failure strain ε f Energy absorbed by the homogenized material;

[0021] According to the energy absorbed by the RS structure under the same strain is the same as the energy absorbed by the homogenized material, the equivalent energy under failure strain is obtained:

[0022]

[0023] L′=ε 2 L 0 Formula (7)

[0024] According to the value of n, each failure strain value will correspond to the original cross-sectional area value A of a battery. By continuously changing the failure strain value, the original cross-sectional area value is obtained when the load-displacement curve and failure form of the fully homogenized model are close to the experimental results. The constitutive model of the homogenized material is determined by the original cross-sectional area of ​​the battery, and the original cross-sectional area of ​​the battery is simulated to obtain the simulated strain ε under different collision conditions. 2 , and the distance L′ between the current collectors of the positive electrode and the negative electrode is obtained according to the simulation strain.

[0025] Further, in step S1, the potential drop between the current collector nodes of the positive electrode and the negative electrode is:

[0026]

[0027] in, The negative electrode current collecting node N 1 The local potential of N is the positive electrode current collecting node 2 The local potential of the positive electrode is obtained according to the distance between the positive and negative current collectors. and I is the local current; R 0 is the internal resistance; U is the local open circuit voltage; V D is the local diffusion overpotential;

[0028] Obtained according to the distance between the positive and negative current collectors Whether the battery is short-circuited can be determined by judging the local diffusion overpotential.

[0029] Further, according to the composition of the equivalent circuit, the nodes corresponding to the current collectors of the positive electrode and the negative electrode are connected through the equivalent circuit, and the voltage across the RC pair is obtained as follows:

[0030]

[0031] Among them, C D is the capacitance of the RC pair in the equivalent circuit; R D is the resistance of the RC pair in the equivalent circuit;

[0032] The local current I is obtained by calculating the SOC of the local equivalent circuit:

[0033]

[0034] Among them, c Q is the conversion factor; Q is the electric quantity; R 0 , R D , C D It is obtained by identifying the pulse discharge experimental curves at different temperatures and different SOCs; the local open circuit voltage U is determined by the SOC.

[0035] Further, in step S2, a test bench is provided, on which the frame can be placed, and the test bench is provided with a rotating shaft so that the frame can be rolled over along the rotating shaft;

[0036] The vehicle frame is subjected to a rollover simulation on the test bench;

[0037] The initial velocity when the frame starts to roll over at the critical rollover point CG' is zero, and the initial energy is:

[0038]

[0039] The instantaneous energy at the impact surface position is:

[0040]

[0041] According to the initial energy and the instantaneous energy at the impact surface position, the instantaneous angular velocity of the frame at the position where it contacts the cement ground is obtained as:

[0042]

[0043] Where M is the total mass of the mobile charging vehicle; ω is the angular velocity; J is the moment of inertia, which is obtained by assigning a random angular velocity in the LS-dyna software. Get;h 0 h is the height between the critical rollover point and the bottom of the frame when the frame is on the test bench; 1 h is the height between the critical rollover point and the bottom of the frame when the frame is at the critical rollover of the test bench; 2is the height between the critical rollover point and the bottom of the frame after the frame rolls over; d is the height between the test bench and the cement floor;

[0044] The failure strain is obtained by performing a rollover collision test simulation and a modal simulation analysis on the frame of a mobile charging vehicle using the angular velocity. The damage degree of the battery is analyzed in a coupling model of the battery homogenization mechanical model and the battery electrical model through the failure strain calculation.

[0045] Furthermore, in step S3, the density parameter of the unit is mapped to the elastic modulus of the unit, and the elastic modulus calculation method is:

[0046] E l (x l )=x l p E 0 Formula (14)

[0047] Among them, l is the unit; x l is the density parameter of the unit; E 0 is the initial elastic modulus of the unit; E l is the elastic modulus of the unit;

[0048] The initial yield stress and strain hardening modulus are introduced into the elastic-plastic material constitutive model of the frame structure to obtain the yield stress and strain hardening modulus of the unit:

[0049] σ yl (x l )=x l p σ y0 Formula (15)

[0050] E hl (x l )=x l p E h0 Formula (16)

[0051] Among them, σ y0 is the initial yield stress; σ yl is the yield stress of the unit; E h0 is the initial strain hardening modulus; E hl is the strain hardening modulus of the element;

[0052] Taking the maximization of frame stiffness and minimization of volume as the optimization goal, the strain energy density is used as the local stiffness index, and the stiffness is quantified by strain energy. The maximization of stiffness is equivalent to the minimization of strain energy. Then:

[0053] C(X)=F T KF formula (17)

[0054]

[0055] Where C(X) is the structural strain energy function; X = {x l} is the unit density vector; K is the total stiffness matrix of the structure; F is the structural node displacement vector; V(X) is the structural volume function; v l is the volume of the unit;

[0056] With the unit density meeting the value limit and mechanical performance requirements and the maximum mechanical abuse of the battery pack being required to not cause internal short-circuit failure as constraints, the structural strain energy and the structural volume are constrained to obtain the constrained optimization target.

[0057] Furthermore, in step S3, each unit of the finite element analysis model is converted into an independent cell, and the cell state is:

[0058] S i =[x i U i S n ] Formula (19)

[0059] Among them, S i is the cell state; x i For x l The optimization variable is the density parameter of the cell; U i is strain energy; S n is the domain status; the calculation method is:

[0060]

[0061] Where n is the number of cells in the neighborhood; η i is the weight factor of cell i in the neighborhood, and the neighborhood strain energy of the cell is the average of the strain energies of each cell in the neighborhood;

[0062] By quantifying the center distance between the cells in the neighborhood and the central cell, the cells whose centers are within the neighborhood radius of a certain cell can be regarded as the neighborhood of the cell. According to the von Neumann type, the corresponding neighborhood radius is r = a, and the side length of the cell is equivalent to the diameter of the equal-area circle or the equal-volume sphere:

[0063]

[0064] According to the material properties of the frame, when the strain energy density of the cell rises to the peak value allowed by the material properties, the density of the cell will tend to be stable. At this time, the relative density of the cell is locked to 1, and the optimization objective function is expressed as:

[0065]

[0066]

[0067] in, are the average strain energy density and target strain energy density of unit i respectively; N is the total number of units; is the minimum value of the design variable, which is taken as 0.001 to avoid singular matrices;

[0068] The action of the agent is defined as the adjustment of the cell density value x. The adjustment value is the action space of the agent. The action space is set to an array containing n elements, where each element represents a possible adjustment. In each decision cycle, the agent selects a specific action a from the preset action space to perform the adjustment of the cell density:

[0069] x i+1 =x i +a formula (24)

[0070] a=[a 1 a 2 a 3 …a n ] Formula (25)

[0071] Among them, x i is the design variable for the cell, a is the action chosen by the cell, x i +1 is the state of the cell in the next iteration step;

[0072] Taking the maximization of frame stiffness and minimization of volume as the optimization goal, the optimization variable x i , strain energy U i , domain status S n Construct the itemized reward function:

[0073]

[0074] Among them, x i ',U i ',S' n are the states of the cells after the action, U max ,S max are the maximum value of the structural unit strain energy in the initial state and the maximum value of the neighborhood state respectively;

[0075] According to the optimization goal, each reward is negatively processed, and the sub-item reward function is unified to obtain the reward R(S,a) after the cell performs action a in state S:

[0076] R(S,a)=αR 1 +βR 2 +γR .3 Formula (27)

[0077] The agent learning strategy uses single-step temporal difference to obtain:

[0078]

[0079] Among them, Q(S i ,a i ) is the action value function of the agent; α is the learning factor; ε is the discount factor, and its value is 0.4;

[0080] The intelligent agent learns according to the action value function, and uses the learned action value function of the intelligent agent to optimize the frame. The cell iterates through the action value function of the intelligent agent for multiple rounds until the action value function of the intelligent agent converges to obtain the optimal cell state. The elastic modulus is optimized by optimizing the cell state to calculate the optimization target of the frame.

[0081] The beneficial effects of the present invention are as follows: by constructing a battery homogenization mechanical model, the distance between the current collectors of the positive electrode and the negative electrode of the battery after the collision force is obtained, and by using the distance between the current collectors of the positive electrode and the negative electrode, it is determined whether the battery has a short circuit through the battery electrical model, and by coupling the battery homogenization mechanical model with the battery electrical model, a close interaction between the models is ensured, so that the safety and performance of the battery under complex working conditions can be comprehensively and accurately analyzed; by establishing a three-dimensional model of the frame of the mobile charging vehicle, a finite element analysis of the frame rollover collision safety is performed, and the simulation and modal rollover collision test of the frame of the mobile charging vehicle has been realized. Through simulation analysis, the battery damage degree can be evaluated according to the data of the rollover collision model and the battery electromechanical coupling model, which is then convenient for the optimization of the subsequent model. In the optimization of the model, the finite element unit is used as the cell, the elastic modulus is used as the optimization object, the frame stiffness is maximized while the volume is minimized as the optimization goal, and the battery pack does not fail internally in a short circuit, the frame structure state and the maximum stress are used as constraints to realize the process of nonlinear topology optimization. The present invention addresses the deficiencies in the safety optimization of the battery pack during a rollover collision, and can simultaneously achieve lightweight frame and improved battery pack safety, effectively improving the optimization efficiency of complex models. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 It is a flow chart of a multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to a preferred embodiment of the present invention.

[0083] Figure 2 It is a schematic diagram of the structure of an equivalent circuit model of a multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to a preferred embodiment of the present invention.

[0084] Figure 3It is a schematic diagram of a rollover collision of a multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0085] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0086] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0087] Please also see Figure 1 and Figure 3 A multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to a preferred embodiment of the present invention comprises the following steps:

[0088] S1. Establishment of battery electromechanical coupling model: By conducting quasi-static mechanical experiments on the positive electrode, negative electrode and separator of the battery, a battery homogenization mechanical model is constructed to obtain the distance between the positive and negative current collectors of the battery under stress; a battery electrical model is established through an equivalent circuit, and based on the distance between the positive and negative current collectors, the battery electrical model is used to determine whether the battery is short-circuited.

[0089] In step S1, the load-displacement curve of the battery is obtained according to the positive and negative electrode and shell test, the positive and negative electrode and shell compression test, and the RS structure indentation test, and the stress-strain curve of the positive electrode, negative electrode and separator of the battery is calculated by the load-displacement curve:

[0090]

[0091] Where P is the load, A is the original cross-sectional area of ​​the battery; ε 1 is the experimental strain; L 0 is the original length of the battery; L is the length of the battery after deformation; the original cross-sectional area of ​​the battery is simulated to obtain the experimental strain, and according to the experimental strain, L and L 0 Obtain the distance between the current collectors of the positive and negative electrodes.

[0092] According to the relationship between stress and stress between layers in the RS structure, the stress-strain curve of the RS structure is calculated through the electrode and the diaphragm. The stress-strain curve of the RS structure is converted into a constitutive model through equivalent energy to obtain the increase in internal energy of the RS structure during the process:

[0093] σ average =σ a v a +σ c v c +σ s v s Formula (2)

[0094]

[0095] σ e =Aε 1 n Formula (4)

[0096] Among them, σ average is the average stress of RS structure; v a 、v c 、v s are the ratios of the volume of the negative electrode, positive electrode and separator to the volume of the RS structure; U represents the internal energy increased by the RS structure during the process, ε 2 represents the experimental strain, σ e represents the stress of the electrode, and A and n are fitting parameters.

[0097] The failure strain of the RS structure after homogenization is ε f , the energy absorbed by the homogenized material under failure strain is U 0 , then:

[0098]

[0099] Among them, ε f is the failure strain; U 0 is the failure strain ε f Energy absorbed by the homogenized material;

[0100] According to the energy absorbed by the RS structure under the same strain is the same as the energy absorbed by the homogenized material, the equivalent energy under failure strain is obtained:

[0101]

[0102] L′=ε 2 L 0 Formula (7)

[0103] According to the value of n, each failure strain value will correspond to the original cross-sectional area value A of a battery. By continuously changing the failure strain value, the original cross-sectional area value is obtained when the load-displacement curve and failure form of the fully homogenized model are close to the experimental results. The constitutive model of the homogenized material is determined by the original cross-sectional area of ​​the battery, and the original cross-sectional area of ​​the battery is simulated to obtain the simulated strain ε under different collision conditions. 2 , and the distance L′ between the current collectors of the positive electrode and the negative electrode is obtained according to the simulation strain.

[0104] In step S1, the potential drop between the current collector nodes of the positive electrode and the negative electrode is:

[0105]

[0106] in, The negative electrode current collecting node N 1 The local potential of N is the positive electrode current collecting node 2 The local potential of the positive electrode is obtained according to the distance between the positive and negative current collectors. and I is the local current; R 0 is the internal resistance; U is the local open circuit voltage; V D is the local diffusion overpotential;

[0107] Obtained according to the distance between the positive and negative current collectors Whether the battery is short-circuited can be determined by judging the local diffusion overpotential.

[0108] According to the composition of the equivalent circuit, the nodes corresponding to the current collectors of the positive electrode and the negative electrode are connected through the equivalent circuit, and the voltage across the RC pair is obtained as follows:

[0109]

[0110] Among them, C D is the capacitance of the RC pair in the equivalent circuit; R D is the resistance of the RC pair in the equivalent circuit;

[0111] The local current I is obtained by calculating the SOC of the local equivalent circuit:

[0112]

[0113] Among them, c Q is the conversion factor; Q is the electric quantity; R 0 , R D , C D It is obtained by identifying the pulse discharge experimental curves at different temperatures and different SOCs; the local open circuit voltage U is determined by the SOC.

[0114] In the battery electromechanical coupling model, the internal short circuit phenomenon is simulated by introducing a local short-circuit resistor. When the battery is deformed by external force, the distance between the positive and negative current collectors will decrease. Once this distance is reduced to below a preset threshold, a short-circuit resistor is used instead. Figure 1 middle and The equivalent circuit between the positive and negative current collectors is used to accurately simulate the short-circuit state inside the battery. The battery homogenization mechanical model can accurately solve the deformation of the battery under stress and pass the deformation information to the electrical model. The electrical model is used to determine whether there is a risk of internal short circuit based on the deformation data transmitted, especially the actual distance between the positive and negative current collectors. If a short circuit is determined to have occurred, the electrical model will further calculate the relevant electrical parameters, such as current, voltage, etc., to comprehensively evaluate the impact of the short circuit on battery performance.

[0115] The coupling relationship between the mechanical model and the electrical model is unidirectional, that is, the information flow only flows from the mechanical model to the electrical model. Among them, the distance between the positive and negative current collectors, as the key bridge connecting the two models, not only reflects the degree of physical deformation of the battery, but also directly determines the assessment result of the short circuit risk in the electrical model. The mechanical model is responsible for calculating the displacement information and transmitting the decisive parameter, the actual distance between the positive and negative current collectors, to the electrical model. The latter uses the internal short circuit model to determine whether the battery has a short circuit based on the actual distance between the positive and negative current collectors.

[0116] S2. Establishment of rollover collision model: Establish a three-dimensional model of the mobile charging vehicle frame, and establish a finite element analysis model of the mobile charging vehicle frame based on the three-dimensional model to obtain the instantaneous angular velocity of the mobile charging vehicle frame when it contacts the cement ground. According to the instantaneous angular velocity, the degree of battery damage is analyzed through the coupling model of the battery homogenization mechanical model and the battery electrical model.

[0117] like Figure 3 As shown, in step S2, a test bench is provided, the frame can be placed on the test bench, and the test bench is provided with a rotating shaft so that the frame can be rolled over along the rotating shaft;

[0118] The frame is subjected to rollover simulation on the test bench;

[0119] The initial velocity when the frame starts to roll over at the critical rollover point CG' is zero, and the initial energy is:

[0120]

[0121] The instantaneous energy at the impact surface position is:

[0122]

[0123] According to the initial energy and the instantaneous energy at the impact surface position, the instantaneous angular velocity of the frame at the position where it contacts the cement ground is obtained as:

[0124]

[0125] Where M is the total mass of the mobile charging vehicle; ω is the angular velocity; J is the moment of inertia, which is obtained by assigning a random angular velocity in the LS-dyna software. Get;h 0 h is the height between the critical rollover point and the bottom of the frame when the frame is on the test bench; 1 h is the height between the critical rollover point and the bottom of the frame when the frame is at the critical rollover of the test bench; 2 is the height between the critical rollover point and the bottom of the frame after the frame rolls over; d is the height between the test bench and the cement floor;

[0126] The angular velocity is used to simulate the rollover collision test of the mobile charging vehicle frame and the modal simulation analysis to obtain the failure strain. Through the failure strain calculation, the damage degree of the battery is analyzed in the coupling model of the battery homogenization mechanical model and the battery electrical model.

[0127] S3. Optimization of the frame model: Each unit of the finite element analysis model is converted into an independent cell. The converted finite element analysis model is processed by the solver LS-DYNA to extract key response parameters, and the cell properties are adjusted according to preset rules. Multiple rounds of iterations are performed to adjust the cell properties to complete the optimization of the frame model.

[0128] In step S3, the density parameter of the unit is mapped to the elastic modulus of the unit. The elastic modulus is calculated as follows:

[0129] E l (x l )=x l p E 0 Formula (14)

[0130] Among them, l is the unit; x l is the density parameter of the unit; E 0 is the initial elastic modulus of the unit; E l is the elastic modulus of the unit;

[0131] The initial yield stress and strain hardening modulus are introduced into the elastic-plastic material constitutive model of the frame structure to obtain the yield stress and strain hardening modulus of the unit:

[0132] σ yl (x l )=x l p σ y0 Formula (15)

[0133] E hl (x l )=x l p E h0 Formula (16)

[0134] Among them, σ y0 is the initial yield stress; σ yl is the yield stress of the unit; E h0 is the initial strain hardening modulus; E hl is the strain hardening modulus of the element;

[0135] Taking the maximization of frame stiffness and minimization of volume as the optimization goal, the strain energy density is used as the local stiffness index, and the stiffness is quantified by strain energy. The maximization of stiffness is equivalent to the minimization of strain energy. Then:

[0136] C(X)=F T KF formula(17)

[0137]

[0138] Where C(X) is the structural strain energy function; X = {x l} is the unit density vector; K is the total stiffness matrix of the structure; F is the structural node displacement vector; V(X) is the structural volume function; v l is the volume of the unit;

[0139] With the unit density meeting the value limit and mechanical performance requirements and the maximum mechanical abuse of the battery pack not causing internal short-circuit failure as constraints, the structural strain energy and structural volume are constrained to obtain the constrained optimization target.

[0140] Two parameters are selected to evaluate whether the structure meets the performance requirements: the morphological stability of the structure and the maximum stress level. The structural state and the maximum stress level of the structure are selected as the unit density x l Whether the value limits and mechanical performance requirements are met, the morphological stability is evaluated by monitoring the changes in the structure during the optimization process, and observing whether the structure evolves from the original rigid body state to the mechanism state. Once the structure is transformed into a mechanism, it means that it has lost its original bearing capacity and is therefore considered a failure state. This transformation can be intuitively reflected by monitoring the maximum displacement of the structure.

[0141] The maximum stress of the structure must be controlled within the allowable stress range of the material to ensure the safety and reliability of the structure. Consider the stresses that the structure may be subjected to in different directions, evaluate the impact of these stresses, and ensure that they do not exceed the limits of the material. The maximum mechanical abuse that the battery pack can withstand must not cause internal short circuit failure.

[0142] In step S3, each unit of the finite element analysis model is converted into an independent cell. The state of the cell should contain a record of feedback to its surrounding environment and the state information of its adjacent area. These complex information sets can be organized and stored in a multidimensional array for subsequent analysis and processing. In this way, each cell can not only perceive the changes in its own state, but also capture the impact of the environment on it and the state of neighboring cells, thereby realizing a more sophisticated and dynamic behavior pattern in the cellular automation system. The cell state is:

[0143] S i =[x i U i S n ] Formula (19)

[0144] Among them, S i is the cell state; x i For x l The optimization variable is the density parameter of the cell; U i is strain energy; S n is the domain status; the calculation method is:

[0145]

[0146] Where n is the number of cells in the neighborhood; η i is the weight factor of cell i in the neighborhood, and the neighborhood strain energy of the cell is the average of the strain energies of each cell in the neighborhood;

[0147] By quantifying the center distance between the cells in the neighborhood and the central cell, the cells whose centers are within the neighborhood radius of a certain cell can be regarded as the neighborhood of the cell. According to the von Neumann type, the corresponding neighborhood radius is r = a, and the side length of the cell is equivalent to the diameter of the equal-area circle or the equal-volume sphere:

[0148]

[0149] For each cell in the system, no matter where it is located, the update rules they follow are unified. The core of this rule is that it is necessary to collect and analyze the state information of other cells in the neighborhood around each cell so that the cell can adjust its own state according to this information. The concept of von Neumann neighborhood is used to collect and analyze the neighborhood state. By quantifying the center distance between the cells in the neighborhood and the central cell, the cells whose centers are within a certain cell neighborhood radius can be regarded as the neighborhood of the cell. The neighborhood radius corresponding to the von Neumann type is r = a, and the diameter of the equal-area circle or the equal-volume sphere is used to equate the side length of the cell.

[0150] According to the material properties of the frame, when the strain energy density of the cell rises to the peak value allowed by the material properties, the density of the cell will tend to be stable. At this time, the relative density of the cell is locked to 1, and the optimization objective function is expressed as:

[0151]

[0152]

[0153] in, are the average strain energy density and target strain energy density of unit i respectively; N is the total number of units; is the minimum value of the design variable, which is taken as 0.001 to avoid singular matrices;

[0154] The action of the agent is defined as the adjustment of the cell density value x. The adjustment value is the action space of the agent. The action space is set to an array containing n elements, where each element represents a possible adjustment. In each decision cycle, the agent selects a specific action a from the preset action space to perform the adjustment of the cell density:

[0155] x i+1 =x i +a formula (24)

[0156] a=[a 1 a 2 a 3 …a n ] Formula (25)

[0157] Among them, x i is the design variable for the cell, a is the action chosen by the cell, x i +1 is the state of the next iteration step of the cell;

[0158] The agent's reward function is the bridge between the optimization goal and the agent's behavior decision. In order to ensure that the agent's decision can accurately reflect the intention of topology optimization, both the main goal to be pursued and the constraints that must be followed must be explicitly encoded in the agent's reward function. Under the premise of ensuring the maximum stiffness of the structure, the amount of material used is reduced as much as possible, that is, the minimization of volume is pursued.

[0159] Taking the maximization of frame stiffness and minimization of volume as the optimization goal, the optimization variable x i , strain energy U i , domain status S n Construct the itemized reward function:

[0160]

[0161] Among them, xi ',U i ',S' n are the states of the cells after the action, U max ,S max are the maximum value of the structural unit strain energy in the initial state and the maximum value of the neighborhood state respectively;

[0162] According to the optimization goal, each reward is negatively processed, and the sub-item reward function is unified to obtain the reward R(S,a) after the cell performs action a in state S:

[0163] R(S,a)=αR 1 +βR 2 +γR .3 Formula (27)

[0164] The agent learning strategy uses single-step temporal difference to obtain:

[0165]

[0166] Among them, Q(S i ,a i ) is the action value function of the agent; α is the learning factor; ε is the discount factor, and its value is 0.4;

[0167] The intelligent agent learns according to the action value function, and uses the learned action value function of the intelligent agent to optimize the frame. The cell iterates through the action value function of the intelligent agent for multiple rounds until the action value function of the intelligent agent converges to obtain the optimal cell state. The elastic modulus is optimized by optimizing the cell state to calculate the optimization target of the frame.

[0168] When performing topological optimization of a structure, the original finite element model is first initialized. This process involves converting each unit in the model into an independent cell, each of which has unique properties, materials, and identification numbers. Subsequently, the converted model is submitted to the solver LS-DYNA for computational analysis, with the focus on extracting key response data such as strain energy density. In order to further optimize the structure, the relative density of each cell is dynamically adjusted based on the preset neighborhood definition and cell evolution criteria. This process is achieved through multiple rounds of iterations until the entire system reaches a stable state, completing an effective structural topological optimization cycle. By initializing the finite element model, decomposing it into independently operable cells, using the solver to analyze and extract important information, and then adjusting the cell properties according to specific rules, the precise topological optimization of the structure is achieved through repeated iterations until the optimization goal is achieved.

[0169] During the learning process, each cell first perceives its own state and the surrounding environment, and interacts with the agent module. The agent makes corresponding decisions based on the current state of the cell and obtains feedback from the environment, thereby obtaining a large amount of data for learning and updating. This cycle makes the entire training more efficient.

[0170] As two common topology optimization methods, SIMP and BESO rely on sensitivity analysis to evaluate the utilization of units. However, the sensitivity calculation is directly based on the mathematical model of the problem being solved, which means that when the problem model changes, the sensitivity needs to be recalculated, which limits the flexibility of these algorithms in dealing with diverse problems. The method in this embodiment establishes a connection between discrete individuals and the whole. When performing discrete calculations, the decision-making of the intelligent agent is more global, converges faster, and has a stronger ability to solve nonlinear problems.

Claims

1. A multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision, characterized in that: The steps include: S1. Establishment of the electromechanical coupling model of the battery: by conducting quasi-static mechanical experiments on the positive electrode, negative electrode and separator of the battery, a homogenized mechanical model of the battery is constructed to obtain the distance between the current collectors of the positive electrode and the negative electrode of the battery under stress; a battery electrical model is established through an equivalent circuit, and according to the distance between the current collectors of the positive electrode and the negative electrode, whether the battery is short-circuited is determined in the battery electrical model; S2. Establishment of rollover collision model: Establish a three-dimensional model of the mobile charging vehicle frame, and establish a finite element analysis model of the mobile charging vehicle frame based on the three-dimensional model to obtain the instantaneous angular velocity of the mobile charging vehicle frame at the position where it contacts the cement ground, and analyze the degree of battery damage based on the instantaneous angular velocity and through the coupling model of the battery homogenization mechanical model and the battery electrical model; S3. Optimization of the frame model: convert each unit of the finite element analysis model into an independent cell, process the converted finite element analysis model through the solver LS-DYNA to extract key response parameters, and adjust the cell properties according to preset rules. Perform multiple rounds of iterations on the adjustment of the cell properties to complete the optimization of the frame model.

2. The multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to claim 1, characterized in that: In step S1, the load-displacement curve of the battery is obtained according to the positive and negative electrode and shell test, the positive and negative electrode and shell compression test, and the RS structure indentation test, and the stress-strain curve of the positive electrode, negative electrode and separator of the battery is calculated by the load-displacement curve: Among them, P is the load, A is the original cross-sectional area of ​​the battery; ε1 is the experimental strain; L0 is the original length of the battery; and L is the deformed length of the battery.

3. The multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to claim 2, characterized in that: According to the relationship between stress and stress between layers in the RS structure, the stress-strain curve of the RS structure is calculated through the electrode and the diaphragm. The stress-strain curve of the RS structure is converted into a constitutive model through equivalent energy to obtain the increase in internal energy of the RS structure during the process: σ average = σ a v a + σ c v c + σ s v s Equation (2) σ e =Aε1 n Formula(4) Among them, σ average is the average stress of RS structure; v a 、v c 、v s are the ratios of the volume of the negative electrode, positive electrode and separator to the volume of the RS structure; U represents the internal energy increased by the RS structure during the process, ε1 represents the experimental strain, σ e represents the stress of the electrode, and A and n are fitting parameters.

4. The multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to claim 3, characterized in that: The failure strain of the RS structure after homogenization is ε f , the energy absorbed by the homogenized material under failure strain is U0, then: Among them, ε f is the failure strain; U0 is the failure strain ε f Energy absorbed by the homogenized material; According to the energy absorbed by the RS structure under the same strain is the same as the energy absorbed by the homogenized material, the equivalent energy under failure strain is obtained: L′=ε2L0 Formula (7) According to the value of n, each failure strain value will correspond to the original cross-sectional area value A of a battery. By continuously changing the failure strain value, the original cross-sectional area value is obtained until the load-displacement curve and failure form of the fully homogenized model are close to the experimental results. The constitutive model of the homogenized material is determined by the original cross-sectional area of ​​the battery, and the original cross-sectional area of ​​the battery is simulated to obtain the simulated strain ε2 under different collision conditions, and the distance L′ between the positive and negative current collectors is obtained based on the simulated strain.

5. The multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to claim 1, characterized in that: In step S1, the potential drop between the current collector nodes of the positive electrode and the negative electrode is: in, is the local potential of the negative current collecting node N1; is the local potential of the positive electrode current collecting node N2; obtained according to the distance between the positive and negative current collectors and I is the local current; R0 is the internal resistance; U is the local open circuit voltage; V D is the local diffusion overpotential; Obtained according to the distance between the positive and negative current collectors Whether the battery is short-circuited can be determined by judging the local diffusion overpotential.

6. The multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to claim 5, characterized in that: According to the composition of the equivalent circuit, the nodes corresponding to the current collectors of the positive electrode and the negative electrode are connected through the equivalent circuit, and the voltage across the RC pair is obtained as follows: Among them, C D is the capacitance of the RC pair in the equivalent circuit; R D is the resistance of the RC pair in the equivalent circuit; The local current I is obtained by calculating the SOC of the local equivalent circuit: Among them, c Q is the conversion factor; Q is the charge; R0, R D , C D It is obtained by identifying the pulse discharge experimental curves at different temperatures and different SOCs; the local open circuit voltage U is determined by the SOC.

7. The multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to claim 5, characterized in that: In step S2, a test bench is provided, on which the vehicle frame can be placed, and the test bench is provided with a rotating shaft so that the vehicle frame can be rolled over along the rotating shaft; The vehicle frame is subjected to a rollover simulation on the test bench; The initial velocity when the frame starts to roll over at the critical rollover point CG' is zero, and the initial energy is: The instantaneous energy at the impact surface position is: According to the initial energy and the instantaneous energy at the impact surface position, the instantaneous angular velocity of the frame at the position where it contacts the cement ground is obtained as: Where M is the total mass of the mobile charging vehicle; ω is the angular velocity; J is the moment of inertia, which is obtained by assigning a random angular velocity in the LS-dyna software. Obtained; h0 is the height between the critical rollover point and the bottom of the frame when the frame is on the test bench; h1 is the height between the critical rollover point and the bottom of the frame when the frame is at the critical rollover of the test bench; h2 is the height between the critical rollover point and the bottom of the frame after the frame rolls over; d is the height between the test bench and the cement floor; The failure strain is obtained by performing a rollover collision test simulation and a modal simulation analysis on the frame of a mobile charging vehicle using the angular velocity. The damage degree of the battery is analyzed in a coupling model of the battery homogenization mechanical model and the battery electrical model through the failure strain calculation.

8. The multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to claim 1, characterized in that: In step S3, the density parameter of the unit is mapped to the elastic modulus of the unit, and the elastic modulus calculation method is: E l (x l )=x l p E0 formula (14) Among them, l is the unit; x l is the density parameter of the unit; E0 is the initial elastic modulus of the unit; E l is the elastic modulus of the unit; The initial yield stress and strain hardening modulus are introduced into the elastic-plastic material constitutive model of the frame structure to obtain the yield stress and strain hardening modulus of the unit: σ yl (x l )=x l p σ y0 Formula (15) E hl (x l )=x l p E h0 Formula (16) Among them, σ y0 is the initial yield stress; σ yl is the yield stress of the unit; E h0 is the initial strain hardening modulus; E hl is the strain hardening modulus of the element; Taking the maximization of frame stiffness and minimization of volume as the optimization goal, the strain energy density is used as the local stiffness index, and the stiffness is quantified by strain energy. The maximization of stiffness is equivalent to the minimization of strain energy. Then: C(X)=F T KF formula(17) Where C(X) is the structural strain energy function; X = {x l } is the unit density vector; K is the total stiffness matrix of the structure; F is the structural node displacement vector; V(X) is the structural volume function; v l is the volume of the unit; With the unit density meeting the value limit and mechanical performance requirements and the maximum mechanical abuse of the battery pack being required to not cause internal short-circuit failure as constraints, the structural strain energy and the structural volume are constrained to obtain the constrained optimization target.

9. The multi-objective optimization method for a vehicle frame based on the safety of a battery pack in a rollover collision according to claim 8, characterized in that: In step S3, each unit of the finite element analysis model is converted into an independent cell, and the cell state is: S i =[x i U i S n ] Formula (19) Among them, S i is the cell state; x i For x l The optimization variable is the density parameter of the cell; U i is strain energy; S n is the domain status; the calculation method is: Where n is the number of cells in the neighborhood; η i is the weight factor of cell i in the neighborhood, and the neighborhood strain energy of the cell is the average of the strain energies of each cell in the neighborhood; By quantifying the center distance between the cells in the neighborhood and the central cell, the cells whose centers are within the neighborhood radius of a certain cell can be regarded as the neighborhood of the cell. According to the von Neumann type, the corresponding neighborhood radius is r = a, and the side length of the cell is equivalent to the diameter of the equal-area circle or the equal-volume sphere: According to the material properties of the frame, when the strain energy density of the cell rises to the peak value allowed by the material properties, the density of the cell will tend to be stable. At this time, the relative density of the cell is locked to 1, and the optimization objective function is expressed as: in, U * are the average strain energy density and target strain energy density of unit i respectively; N is the total number of units; is the minimum value of the design variable, which is taken as 0.001 to avoid singular matrices; The action of the agent is defined as the adjustment of the cell density value x. The adjustment value is the action space of the agent. The action space is set to an array containing n elements, where each element represents a possible adjustment. In each decision cycle, the agent selects a specific action a from the preset action space to perform the adjustment of the cell density: x i+1 =x i +a formula (24) a=[a1a2a3 …a n ] Official (25) Among them, x i is the design variable for the cell, a is the action chosen by the cell, x i +1 is the state of the cell in the next iteration step; Taking the maximization of frame stiffness and minimization of volume as the optimization goal, the optimization variable x i , strain energy U i , domain status S n Construct the itemized reward function: Among them, x i ',U i ',S' n are the states of the cells after the action, U max ,S max are the maximum value of the structural unit strain energy in the initial state and the maximum value of the neighborhood state respectively; According to the optimization goal, each reward is negatively processed, and the sub-item reward function is unified to obtain the reward R(S,a) after the cell performs action a in state S: R(S,a) = αR1 + βR2 + γR .3 Formula (27) The agent learning strategy uses single-step temporal difference to obtain: Among them, Q(S i ,a i ) is the action value function of the agent; α is the learning factor; ε is the discount factor, and its value is 0.4; The intelligent agent learns according to the action value function, and uses the learned action value function of the intelligent agent to optimize the frame. The cell iterates through the action value function of the intelligent agent for multiple rounds until the action value function of the intelligent agent converges to obtain the optimal cell state. The elastic modulus is optimized by optimizing the cell state to calculate the optimization target of the frame.

Images