NSGA-II optimization method based on semi-active suspension hybrid coefficient

Abstract

The application relates to an NSGA-II optimization method based on a semi-active suspension hybrid coefficient, the range of the hybrid coefficient is determined, two smoothness indexes of vehicle body acceleration and suspension dynamic deflection are simultaneously optimized through a multi-objective optimization method NSGA-II, the optimization effects of the two indexes under different alpha are compared, and the optimal set of alpha is obtained. The application adopts a harmonic superposition method to establish a random road input model, so that the simulation result is more random and close to the actual road; three automobile smoothness evaluation indexes are taken as output vectors to establish a state space equation, so that the simulation result is less likely to be distorted; the selection of the hybrid coefficient is combined with a multi-objective genetic algorithm, two mutually contradictory objective functions are taken as optimization targets, and the optimization purpose is closer to the actual application; the hybrid coefficient selected by using the NSGA-II multi-objective optimization method can compromise various evaluation indexes, and the comprehensive performance of the semi-active suspension about ride comfort is optimal.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of automobile suspensions, and particularly relates to an NSGA-II optimization method based on a hybrid coefficient of a semi-active suspension. BACKGROUND

[0002] With the development of suspension technology, people have higher requirements for the performance of automobile suspensions, not only focusing on the functionality of automobiles in daily work and life, but also starting to pay attention to the safety and ride comfort of automobiles on various road surfaces. The suspension system is a mechanism for realizing the elastic connection of the wheel and the vehicle body, mainly used for relieving the vibration impact of the road on the vehicle body and conducting various forces and torques between the wheel and the vehicle body. Generally, the suspension system is divided into active suspension, semi-active suspension and passive suspension. Among them, the active suspension can significantly improve the ride comfort of the automobile at the cost of energy consumption, but its structure is complex and the cost is high. While the passive suspension has poor performance. Therefore, the semi-active suspension with low cost and good performance advantage has high theoretical research value.

[0003] There are many control strategies for semi-active suspensions, and the classic control strategies can be roughly divided into three types: skyhook control strategy, groundhook control strategy and sky-groundhook hybrid control strategy. Among them, the sky-groundhook hybrid control combines the advantages of skyhook control and groundhook control, and adjusts the proportion of the damping force of skyhook control and groundhook control in hybrid control through the control of the hybrid coefficient alpha, which plays a great role in the vibration attenuation of the sprung mass and the unsprung mass of the suspension. However, due to the limitation of experimental cost, there is no clear method for selecting the hybrid coefficient alpha at the present stage, and most of them use empirical method or single-objective optimization method for a certain ride comfort index, which has the disadvantage that one index is optimized, but other indexes may be deteriorated. Therefore, it is necessary to design a multi-objective optimization algorithm that can compromise various optimization indexes. SUMMARY

[0004] The purpose of the application is to provide an NSGA-II optimization method based on the hybrid coefficient of a semi-active suspension, taking two mutually contradictory indexes affecting ride comfort as the starting point, using the NSGA-II optimization method in the multi-objective optimization method to complete the selection of the hybrid coefficient, to solve the problem of making the optimization purpose more realistic and making the comprehensive performance of the semi-active suspension on ride comfort optimal.

[0005] The purpose of the application is achieved by the following technical solutions:

[0006] A kind of NSGA-II optimization method based on semi-active suspension hybrid coefficient, first determine the range of hybrid coefficient α, two smoothness indexes of body acceleration and suspension deflection are optimized simultaneously by multi-objective optimization method NSGA-II, and the optimization effect of two indexes under different α is compared, and the optimal set of α is obtained.

[0007] Further, specifically include the following steps:

[0008] A, select B-grade road surface as reference, utilize harmonic superposition method to establish random road surface model, as the input of subsequent suspension model;

[0009] B, the semi-active suspension model based on skyhook and groundhook hybrid control is modeled, the state space equation is established, and the selection range of the initial value of hybrid coefficient α is determined;

[0010] C, with body acceleration and suspension deflection as optimization target, with hybrid coefficient α as optimization variable, the semi-active suspension model under different α is optimized by NSGA-II multi-objective optimization, and the α with the best optimization effect is selected to form the optimal set.

[0011] Further, step A, based on harmonic superposition method, the random process of discrete spectrum approximation target is used, the road roughness is expressed by a large number of sine or cosine with random phase, and the road surface model expression of harmonic superposition method is as follows:

[0012]

[0013] In the formula, n1 and n2 are lower limit and upper limit of spatial frequency respectively, Δn i It is frequency interval, the frequency interval is divided into m intervals, then the power spectral density G m,i (n q ) at the center frequency n m,i Of the i interval is used to approximate the value of the power spectral density of the entire interval;θ i It is a random number in the interval [0, 2 π], and each θ i Is independent of each other;L is the road longitudinal length of automobile travel;

[0014] The road roughness q is expressed by a large number of sine values with random phase, which can be used as the input of subsequent semi-active suspension state space equation.

[0015] Further, step B, a hybrid damping two-degree-of-freedom quarter semi-active suspension model is established, the hybrid control suspension model has two dampers c sky And c gud Connected to the outside, in the skyhook control, a virtual skyhook damper c skyThe part of the sprung mass m2 is connected to simulate the skyhook control; and the groundhook control is introduced between the wheel and the road surface to simulate the groundhook control. gud to simulate the groundhook control.

[0016] Further, in step B, the initial value of the mixed damping coefficient a is selected in the range of 0-1, and when a is 1, the suspension control strategy is pure skyhook control, and when a is 0, the skyhook control is not effective, and the suspension becomes pure groundhook control.

[0017] Further, in step B, when establishing the motion differential equation of the suspension model, the linear description of the damping forces of the skyhook damper and the groundhook damper in this case is determined in combination with the mixed damping coefficient a, and according to Newton's second law, the damping forces of the two dampers are:

[0018]

[0019] In the formula, is the vertical vibration speed of the unsprung mass (m / s); is the vertical vibration speed of the sprung mass (m / s).

[0020] According to the Lagrange equation rule, the entire mechanical model is described in the form of energy conversion, and F sky and F gud are non-potential forces in addition to gravity and elastic force, and the motion differential equation of the suspension model is:

[0021]

[0022] In the formula, m1 is the unsprung mass, m2 is the sprung mass, k is the passive spring stiffness; k t is the tire stiffness;

[0023] According to the motion differential equation, the state space equation of the system is constructed; let the state vector , the input vector is u=q, and the output vector is The following single-input state space equation can be constructed:

[0024]

[0025] y=Cx+Du (8)

[0026] In the formula, A is the system matrix; B is the input (or control) matrix; C is the output matrix; D is the direct transfer matrix;

[0027]

[0028] By changing the value of the mixing coefficient α, different output vectors y can be obtained, and by comparing the three output quantities in y, the performance of the semi-active suspension in this case can be judged.

[0029] Further, step C specifically comprises the following steps:

[0030] C1, for the purpose of optimizing the comfort of the vehicle, taking the vertical acceleration of the vehicle body and the suspension deflection z2-z1 as the optimization objectives of the multi-objective optimization, namely:

[0031]

[0032] C2, taking the mixing coefficient α as the optimization variable, using the NSGA-II fast non-dominated sorting optimization method for multi-objective optimization, and selecting the best α with the best optimization effect to form the optimal set.

[0033] Further, in the NSGA-II fast non-dominated sorting method of step C2, the specific steps of NSGA-II are as follows: set a set Pop containing all populations, and make a loop for each point p(n) in Pop; set sets S p and integer n p , respectively containing the individuals dominated by the individual p(n) and representing the number of individuals capable of dominating the current individual p(n); then search for other individuals q(n), if p(n)πq(n), then put the individual q(n) into the set S p , if it is a non-dominated relationship, then n p +1; if n p =0, then put the individual p(n) into the first sequence solution set, i.e. rank=1; then, set a set ζ, and also put p(n) into the set ζ; after traversing all individuals in Pop, proceed to the next operation, i.e. for each first sequence solution in set ζ, take these points as the first batch of output results to form the first layer of Pareto optimal front, and after removing these points, for each individual q(n) in set S p , its corresponding n p -1, then judge if n p =0, then put the individual q(n) into the second layer of Pareto optimal front as the second batch of output points, then n p -1 for the next batch of individuals q(n), then judge if n p =0, then put the individual q(n) into the third layer of Pareto optimal front as the third batch of output points, and so on, until all individuals are output, and the first layer or other layers of Pareto optimal front are displayed according to requirements.

[0034] Further, the optimization method of NSGA-II is adopted, different mixed coefficients alpha in the value range are optimized, and multiple Pareto optimal frontiers can be obtained; the crowded distance d of the individual i on the Pareto optimal frontier is defined i :

[0035] d i =|x i+1 -x i-1 |+|y i+1 -y i-1 | (12)

[0036] In the formula, x i+1 and x i-1 are the horizontal coordinates of the i+1th point and the i-1th point respectively, y i+1 and y i -1 are the vertical coordinates of the i+1th point and the i-1th point respectively.

[0037] One of the great requirements for the optimal range selection of the Pareto optimal frontier is uniformity, points with too small crowded distance d i are filtered, and each individual with crowded distance d i sufficiently large is selected, so that the optimal range of the body vertical acceleration and the suspension dynamic deflection z2-z1 is obtained. Finally, by comparing the root mean square values of the two optimization objectives in the range, multiple mixed coefficients alpha meeting the optimization purpose can be selected to form the optimal set of alpha values.

[0038] Compared with the prior art, the present application has the following beneficial effects:

[0039] 1. The present application adopts the harmonic superposition method to establish a random road input model, so that the simulation result is more random and close to the actual road;

[0040] 2. The state space equation is established by taking three major vehicle ride comfort evaluation indexes as output vectors, so that the simulation result is less likely to be distorted;

[0041] 3. The selection of the mixed coefficient alpha is combined with the multi-objective genetic algorithm NSGA-II, two mutually contradictory objective functions are taken as optimization objectives, and the optimization purpose is closer to the actual application;

[0042] 4. The mixed coefficient alpha selected by the NSGA-II multi-objective optimization method can compromise each evaluation index, so that the comprehensive performance of the semi-active suspension on the ride comfort is optimal. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation on the scope. Other related drawings can also be obtained by those skilled in the art without creative labor on the basis of the drawings.

[0044] Figure 1 Hybrid damping control two-degree-of-freedom quarter suspension model

[0045] Figure 2 Pareto optimal frontier when the mixing coefficient is 0.5

[0046] Figure 3 Step flow chart of NSGA-II optimization method based on semi-active suspension mixing coefficient. DETAILED DESCRIPTION

[0047] The present application will be further described below in conjunction with the embodiments:

[0048] The present application will be further described below in conjunction with the embodiments:

[0049] It should be noted that: similar signs and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present application, the terms "first", "second" and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0050] The present application proposes a NSGA-II optimization method based on semi-active suspension mixing coefficient, first, the range of mixing coefficient a is determined, two smoothness indexes of vehicle body acceleration and suspension dynamic deflection are simultaneously optimized by multi-objective optimization method NSGA-II, and the optimization effects of two indexes under different a are compared, finally, the optimal set of a is obtained.

[0051] Specifically, the following steps are included:

[0052] Step one, select B-grade road surface as reference, use harmonic superposition method to establish random road surface model as input of subsequent suspension model;

[0053] Step two, model the semi-active suspension model based on skyhook and groundhook hybrid control, establish state space equation, and determine the selection range of initial value of mixing coefficient a;

[0054] Step three, taking the body acceleration and suspension deflection as the optimization target, taking the mixed coefficient alpha as the optimization variable, the NSGA-II multi-objective optimization is carried out on the semi-active suspension model under different alpha, and the best alpha is selected to form the optimal set.

[0055] International standard ISO8608:2016 takes the power spectral density G q (n0) of the road surface as the classification standard of the road surface roughness, and the fitting expression of G q (n0) is as follows:

[0056]

[0057] In the formula, W is the frequency index; n is the spatial frequency (m -1 ); n0 is the reference spatial frequency (m -1 ), n0=0.1m -1 ; G q (n0) is the road surface roughness coefficient (m 3 ); G q (n) is the power spectral density function of the road surface roughness (m 3 ). Generally, W is 2, and G q (n0) is obtained by table lookup.

[0058] International standard ISO8608:2016 divides the value of G q (n0) to define different levels of road surfaces, and also clearly defines the upper and lower limit values of the road surface roughness coefficient under each level of road surface. Based on the harmonic superposition method, the present application uses a discrete spectrum to approximate the target random process, and represents the road surface roughness as a sum of a large number of sinusoids or cosines with random phases. The expression of the road surface model of the harmonic superposition method is as follows:

[0059]

[0060] In the formula, n1 and n2 are the lower limit and upper limit of the spatial frequency respectively, Δn i is the frequency interval, the frequency interval is divided into m intervals, and then the power spectral density G m,i (n q ) at the center frequency n m,i of the i-th interval is used to approximate the value of the power spectral density of the entire interval; θ i is a random number in the interval [0, 2π], and each θ i is independent of each other; l is the road longitudinal length of the vehicle driving.

[0061] The road surface roughness q is represented as a sum of a large number of sinusoidal values with random phases, which can be used as the input of the subsequent semi-active suspension state space equation.

[0062] Establish a hybrid damping two-degree-of-freedom quarter-semi-active suspension model as follows: Figure 1 As shown. Figure 1 In the middle, c sky The damping coefficient of the ceiling damper (Ns / m) will be used thereafter. sky This represents the ceiling damper; c gud Here, c represents the damping coefficient (Ns / m) of the ground cover damper. gud This represents the damper for the ground; α is the mixed damping coefficient; m1 is the unsprung mass (kg); m2 is the sprung mass (kg); c is the damper damping coefficient (Ns / m); k is the passive spring stiffness (N / m); k t This represents tire stiffness (N / m).

[0063] The main difference between the hybrid control suspension model and the traditional two-degree-of-freedom quarter-suspension model is that the hybrid control suspension model has two externally connected dampers. sky and c gud In ceiling control, a virtual ceiling damper c is introduced. sky It is connected to the sprung mass m2 to simulate roof control; while ground control introduces a virtual ground damper c between the wheel and the road surface. gud To simulate ground-mounted control, the hybrid damping control suspension model integrates these two control strategies. In short, the model requires not only a sky-mounted damper between the vehicle body and the surrounding environment, such as the sky, but also a ground-mounted damper between the wheels and the road surface to achieve hybrid damping control.

[0064] The initial value of the hybrid damping coefficient α can be set to the range of 0-1. When α is 1, the suspension control strategy is pure sky control. When α is 0, sky control is ineffective and the suspension becomes pure ground control.

[0065] Pre-establishment Figure 1 The differential equations of motion for the suspension model first require determining the linear description of the damping forces of the ceiling damper and the floor damper under this condition, based on the combined damping coefficient α. According to Newton's second law, the damping forces of the two dampers are:

[0066]

[0067] In the formula, The vertical vibration velocity of the unsprung mass (m / s); The vertical vibration velocity of the spring-loaded mass is denoted as (m / s).

[0068] Combining the Lagrange equations, the entire mechanical model is described using energy conversion, with F... sky and F gudAs a non-potential force in addition to gravity and elastic force, the motion differential equation of the suspension model is:

[0069]

[0070] According to the motion differential equation, the state space equation of the system can be constructed. Let the state vector be The input vector is u=q, and the output vector is The following single-input state space equation can be constructed:

[0071]

[0072] y=Cx+Du (8)

[0073] In the formula, A is the system matrix; B is the input (or control) matrix; C is the output matrix; and D is the direct transfer matrix.

[0074]

[0075]

[0076] By changing the value of the mixing coefficient α, different output vectors y can be obtained, and by comparing the three output quantities in y, the performance of the semi-active suspension in this case can be judged.

[0077] Multi-objective optimization is an improved version of single-objective optimization, which is used for simultaneous optimization of multiple objective functions, and the increase and decrease relationship between these objective functions conflicts with each other. If one objective function is optimized, other objective functions will be degraded. For the present invention, the reduction of the root mean square value of the body vertical acceleration will cause the maximum value of the suspension deflection to rise, and vice versa. Therefore, the multi-objective optimization method can be used to optimize these two objectives for the purpose of compromise. For the purpose of optimizing the vehicle ride comfort, the body vertical acceleration and the suspension deflection z2-z1 are used as the optimization objectives of multi-objective optimization, that is:

[0078]

[0079] The mixing coefficient α is used as the optimization variable, and the NSGA-II fast non-dominated sorting optimization method is used for multi-objective optimization.

[0080] Firstly, for the concept of dominance, two parameters x and y can be set. If all the solutions of x satisfy that they are not worse than y, and x contains at least one objective function that is better than y, then x dominates y, denoted as xπy, otherwise it is non-dominated.

[0081] NSGA-II fast non-dominated sorting method is an improved version of the traditional non-dominated sorting method of NSGA-II. The specific steps of NSGA-II are as follows: set a set Pop containing all populations, then make a loop for each point p(n) in Pop, and set sets S and n containing the dominated individuals of individual p(n) and the number of individuals capable of dominating the current individual p(n) respectively, then search for other individuals q(n), if p(n)πq(n), then put individual q(n) into set S p and integer n p , respectively, containing the dominated individuals of individual p(n) and the number of individuals capable of dominating the current individual p(n), then search for other individuals q(n), if p(n)πq(n), then put individual q(n) into set S p , if it is a non-dominated relationship, then n p +1. If n p =0, then put the individual p(n) into the first sequence solution set, i.e. rank=1. Then, set a set ζ, and put p(n) into set ζ. After traversing all individuals in Pop, the next step is to perform the following operation: for each first sequence solution in set ζ, take these points as the first batch of output results to form the first layer of Pareto optimal front, and after excluding these points, for each individual q(n) in set S p , its corresponding n p -1, and then judge if n p =0, then put the individual q(n) into the second layer of Pareto optimal front as the second batch of output points, and then n p -1 for the next batch of individuals q(n), and then judge if n p =0, then put the individual q(n) into the third layer of Pareto optimal front as the third batch of output points, and so on, until all individuals are output, and the first layer or other layers of Pareto optimal front are displayed as required. The worst complexity of this method is O(MN 2 ), M is the number of optimization objectives, and N is the number of population individuals, which is better than the worst complexity O(MN 3 ) of the traditional non-dominated sorting method of NSGA.

[0082] Using the optimization method of NSGA-II, different mixing coefficients α in the value range can be optimized, and multiple Pareto optimal fronts can be obtained. Taking α=0.5 as an example, taking the vertical acceleration of the vehicle body as the first optimization objective 1 st Objective, and taking the suspension deflection z2-z1(m) as the second optimization objective 2 nd Objective, the Pareto optimal front thereof is shown in the following formula. Figure 2

[0083] Then, set the crowding distance d i of individual i on the Pareto optimal front: ​

[0084] d i = |x i+1 - x i-1 | + |y i+1 - y i-1 | (12)

[0085] where x i+1 and x i-1 are the horizontal coordinates of the i+1th point and the i-1th point respectively, and y i+1 and y i-1 are the vertical coordinates of the i+1th point and the i-1th point respectively.

[0086] One of the requirements for the optimal range selection of the Pareto optimal front is uniformity. Therefore, the points with too small crowding distance d i are filtered out, and the individuals with large enough crowding distance d i are selected. Then, the optimal ranges of the body vertical acceleration and the suspension deflection z2-z1 can be obtained. For different mixing coefficients α, there are different Pareto optimal fronts, and the optimal ranges of the optimization objectives are also different. Finally, by comparing the root mean square values of the two optimization objectives in the ranges, the multiple mixing coefficients α that meet the optimization objectives can be selected to form the optimal set of α values.

[0087] The present application uses the multi-objective optimization method of NSGA-II to complete the selection of the mixing coefficients, and only takes the body vertical acceleration and the suspension deflection z2-z1 as the optimization objectives. This idea can also be used for multi-objective optimization with all the ride comfort indicators as the optimization objectives.

[0088] Note that the above is only the preferred embodiment of the present application and the technical principles used. Those skilled in the art will understand that the present application is not limited to the specific embodiments described herein, and those skilled in the art can make various obvious changes, re-adjustments and substitutions without departing from the scope of the present application. Therefore, although the present application has been described in more detail through the above embodiments, the present application is not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the appended claims.

Claims

1. An NSGA-II optimization method based on the hybrid coefficient of semi-active suspension, characterized in that: First, determine the range of the mixing coefficient α. Then, use the multi-objective optimization method NSGA-II to simultaneously optimize two ride comfort indices: vehicle acceleration and suspension dynamic deflection. Compare the optimization effects of the two indices under different α values ​​to obtain the optimal set of α values. Specifically, the following steps are included: A. Select Class B road surface as a reference and use the harmonic superposition method to establish a random road surface model as the input for the subsequent suspension model; B. Model the semi-active suspension model based on hybrid control of top and bottom, establish the state-space equations, and determine the selection range of the initial value of the hybrid coefficient α. C. Using vehicle acceleration and suspension deflection as optimization objectives and hybrid coefficient α as optimization variable, perform NSGA-II multi-objective optimization on semi-active suspension models under different α values, and select the α with the best optimization effect to form the optimal set. Step B: The initial value of the mixed damping coefficient α is set to a range of 0-1. When α is 1, the suspension control strategy is pure roof control; when α is 0, roof control is ineffective, and the suspension becomes pure ground control. When establishing the motion differential equations of the suspension model, the linear description of the damping forces of the roof damper and ground damper under this condition must be determined first in conjunction with the mixed damping coefficient α. According to Newton's second law, the damping forces of the two dampers are: In the formula, The vertical vibration velocity of the unsprung mass is m / s; The vertical vibration velocity of the spring-loaded mass is given in m / s. Combining the Lagrange equations, the entire mechanical model is described using energy conversion, with F... sky and F gud As non-forces other than gravity and elasticity, the differential equation of motion for the suspension model is: In the formula, m1 is the unsprung mass, m2 is the sprung mass, and k is the passive spring stiffness; k t For tire stiffness; Based on the equations of motion, construct the state-space equations of the system; let the state vectors be... The input vector is u = q, and the output vector is The following single-input state-space equation is then constructed: y=Cx+Du (8) In the formula, A is the system matrix; B is the input matrix; C is the output matrix; and D is the direct transfer matrix. By changing the value of the mixing coefficient α, different output vectors y can be obtained. By comparing the three output quantities in y, the performance of the semi-active suspension under this condition can be judged. Step C specifically includes the following steps: C1. To optimize vehicle ride comfort, the vertical acceleration of the vehicle body... The suspension dynamic deflection z2-z1 is used as the optimization objective in the multi-objective optimization, namely: C2. Using the mixing coefficient α as the optimization variable, the NSGA-II fast non-dominated sorting optimization method is used for multi-objective optimization, and the optimal set is formed by selecting the α with the best optimization effect.

2. The NSGA-II optimization method based on the semi-active suspension hybrid coefficient according to claim 1, characterized in that: Step A: Based on the harmonic superposition method, a stochastic process of approximating the target using discrete spectrum is employed. The road surface roughness is represented by a sum of a large number of sine or cosine waves with random phases. The road surface model expression of the harmonic superposition method is shown below: In the formula, n1 and n2 are the lower and upper limits of the spatial frequency, respectively, and Δn i Given a frequency range, divide the frequency range into m intervals, and then use the center frequency n of the i-th interval. m,i Power spectral density G at q (n m,i θ is used to approximate the power spectral density value over the entire interval; i It is a random number on the interval [0, 2π], and each θ i They are mutually independent; l represents the longitudinal length of the road through which the car travels. The road surface unevenness q can be represented by a large number of sinusoidal values ​​with random phases, which can then be used as the input for the subsequent semi-active suspension state space equations.

3. The NSGA-II optimization method based on the semi-active suspension hybrid coefficient according to claim 1, characterized in that: Step B: Establish a hybrid damping two-degree-of-freedom quarter-semi-active suspension model. The hybrid control suspension model has two externally connected dampers c. sky and c gud In ceiling control, a virtual ceiling damper c is introduced. sky It is connected to the sprung mass m2 to simulate roof control; while ground control introduces a virtual ground damper c between the wheel and the road surface. gud To simulate the control of the ground canopy.

4. The NSGA-II optimization method based on the semi-active suspension hybrid coefficient according to claim 1, characterized in that, Step C2, in the NSGA-II fast non-dominated sorting method, the specific steps of NSGA-II are as follows: Let a set Pop contain all populations, perform a loop for each point p(n) in Pop, and then set a set S p and integer n p Let S contain the individuals controlled by individual p(n) and the number of individuals that can control the current individual p(n). Then, search for other individuals q(n). If p(n) ≠ q(n), then add individual q(n) to set S. p In the case of a non-dominant relationship, then n p +1; if n p If rank = 0, then the individual p(n) is placed in the first sequence solution set, i.e., rank = 1; Then, let's define another set ζ, and add p(n) to set ζ as well. After traversing all individuals in Pop, we proceed to the next step: for each first sequence solution in set ζ, we take these points as the first batch of output results, forming the first Pareto optimal front. After removing these points, for set S... p For each individual q(n), its corresponding n p -1, then determine if n p If n = 0, then the individual q(n) is placed in the second Pareto optimal frontier as the point of the second batch of outputs, and then the next batch of individual q(n) is processed. p -1, and then determine if n p If q(n) = 0, then the individual q(n) is placed in the third Pareto optimal frontier as the third batch of output points, and so on, until all individuals are output. The first layer or other layers of Pareto optimal frontier are displayed as required.

5. The NSGA-II optimization method based on the semi-active suspension hybrid coefficient according to claim 4, characterized in that: By employing the NSGA-II optimization method, multiple Pareto optimal fronts can be obtained by optimizing the mixing coefficient α within different value ranges; the crowding distance d of individual i on the Pareto optimal front is defined. i : d i =|x i+1 -x i-1 |+|y i+1 -y i-1 | (12) In the formula, x i+1 and x i-1 Let y be the x-coordinates of the (i+1)th point and the (i-1)th point, respectively. i+1 and y i -1 represents the ordinates of the (i+1)th and (i-1)th points, respectively; A key requirement for selecting the optimal range of the Pareto optimal front is uniformity, especially for the crowding distance d. i Filter out points that are too small and select the crowding distance d. i Sufficiently large individual units yield the vertical acceleration of the vehicle body. The optimal range of suspension dynamic deflection z2-z1 is determined; finally, by comparing the root mean square values ​​of these two optimization objectives within this range, multiple mixing coefficients α that meet the optimization purpose can be selected to form the optimal set of α values.

Images