Construction method of semi-active suspension control system based on adaptive LQR wheel hub direct drive

Abstract

The invention discloses a construction method of a semi-active suspension control system based on an adaptive LQR wheel hub direct drive in the automotive field. The method sets the vehicle body vertical acceleration, stator-rotor eccentricity, vehicle body pitch angle acceleration, front wheel suspension dynamic travel, rear wheel suspension dynamic travel, front wheel tire dynamic load, rear wheel tire dynamic load, front wheel optimal suspension force and rear wheel optimal suspension force of 12 typical working conditions, obtains corresponding 12 target weight matrices and control weight matrices, optimizes them respectively by using a genetic algorithm to obtain 12 groups of optimal target weight matrices and optimal control weight matrices, and inputs them into an LQR controller for storage to form an adaptive LQR controller. The adaptive LQR controller calculates a cost function and solves the optimal suspension forces of the front and rear wheels. When the vehicle runs to a certain typical working condition, the optimal weight matrix coefficient is selected and the optimal suspension force is solved, so as to solve the problem that the traditional suspension control cannot obtain the optimal suspension force under different continuous operating conditions.

Description

Technical Field

[0001] The invention belongs to the field of automobiles, and relates to a semi-active suspension system therein, and specifically to a construction method of a wheel hub direct-driven semi-active suspension control system using an LQR controller. Background Art

[0002] On the one hand, the introduction of the hub motor increases the unsprung mass, causing the vehicle tire ground contact and riding smoothness to deteriorate; on the other hand, the introduction of the hub motor couples multiple components to form a complex electromechanical system. When working, the electric wheel is excited by the road surface, causing the motor stator and rotor to be eccentric, generating unbalanced electromagnetic force, and causing vertical vibration and noise problems in the electric vehicle.

[0003] The automobile suspension system has a great influence on the vehicle's ride smoothness, handling stability and other performance. At present, the suspension system can be divided into three categories: passive suspension, active suspension and semi-active suspension. The spring stiffness and shock absorber damping coefficient of the passive suspension cannot be changed once they are selected during the design, and cannot adapt to the different requirements of the vehicle for ride smoothness and handling stability under different working conditions. The active suspension can achieve the requirements of both vehicle ride smoothness and handling stability through a series of complex mechanical and electrical structures, but its cost is high, the structure is complex, and the vehicle economy is low. The semi-active suspension replaces the active elements of the actuator with passive but parameter-adjustable passive elements (generally damping elements), which can improve ride comfort and handling stability. Its performance is better than that of the passive suspension, and its control quality is also close to that of the active suspension. It has a simple structure, low cost, and low energy consumption. It is currently a widely used technology.

[0004] In the field of optimal control, the Linear Quadratic Regulator (LQR) has become one of the most representative optimal control strategies due to its simple structure, simple control method and easy closed-loop optimization. As a full-state feedback controller, the LQR controller evaluates the effectiveness of the solution through the cost function and the gain matrix, thereby ensuring the good performance of the system. The cost function is established according to the expected value or demand of the system, and the gain matrix is ​​obtained by solving the Algebraic Riccati equation. However, when the existing LQR controller is applied to a wheel hub motor direct-drive semi-active suspension vehicle, it is unable to adaptively adjust the suspension force for different driving conditions, so the control performance is poor.

[0005] Genetic algorithm is a computational model that simulates Darwin's theory of evolution and Mendel's genetics. It obtains the optimal solution to the problem by continuously replicating, crossing over and mutating a group of chromosomes according to the principle of survival of the fittest. Different from other optimization algorithms, the genetic algorithm needs to encode the initial population from decimal to binary, that is, to form a chromosome population. Then, according to the designed fitness function and crossover mutation probability, the individuals in the chromosome population are passed on to the next generation through replication, crossover and mutation. In the process of continuous repetition, the new generation of chromosome populations generated will meet the requirements of the fitness function better than the previous generation. The optimal solution in the last generation of population is decoded and becomes the approximate optimal solution to the problem being solved. Summary of the invention

[0006] The purpose of the present invention is to provide a construction method of a control system based on an adaptive LQR wheel hub direct-drive semi-active suspension. Starting from the top-level design, the influence of the damping coefficient on the performance is always attributed to the suspension force provided by the adjustable damping shock absorber. The control objectives are to optimize the suspension performance, reduce the vehicle body pitch angle acceleration and reduce the motor eccentricity, and extend the motor life. The constructed control system adopts an adaptive LQR controller to achieve high-performance control of the wheel hub direct-drive semi-active suspension and improve the comprehensive performance of the wheel hub direct-drive suspension system.

[0007] The technical solution adopted by the construction method of the adaptive LQR wheel hub direct drive semi-active suspension control system of the present invention is: establishing a vertical vibration control model of a semi-vehicle seven-degree-of-freedom wheel hub direct drive semi-active suspension system to obtain a suspension dynamics differential equation, and further comprising the following steps:

[0008] Step (1): Simulate road excitation for 12 typical working conditions and set the corresponding 12 vehicle body vertical accelerations The weight coefficient is m1, the eccentricity between the stator and rotor is e, the weight coefficient is m2, and the vehicle body pitch angle acceleration The weight coefficient is m3, the front wheel suspension travel f d1 The weight coefficient is m4, the rear wheel suspension travel f d2 The weight coefficient is m5, the front tire dynamic load F d1 The weight coefficient is m6, the rear tire dynamic load F d2 The weight coefficient is m7, the optimal suspension force of the front wheel is F u1 The weight coefficient is p1 and the optimal suspension force F of the rear wheel u2 The weight coefficient is p2, and the corresponding 12 target weight matrices Q=diag[m1,m2,m3,m4,m5,m6,m7] and 12 control weight matrices R=diag[p1,p2] are obtained. diag is a diagonal matrix, which corresponds to a set of target weight matrices Q and control weight matrices R under a typical working condition.

[0009] Step (2): Use genetic algorithm to optimize the 12 sets of target weight matrices Q and control weight matrices R respectively, and obtain 12 sets of optimal target weight matrices Q b and the optimal control weight matrix R b And input into the LQR controller for saving, forming an adaptive LQR controller;

[0010] Step (3): The adaptive LQR controller is based on the suspension dynamics differential equation and the optimal target weight matrix Q b The various weight coefficients in are used to calculate the cost function J;

[0011] Step (4): The goal is to find the minimum cost function min(J) and solve for the optimal suspension force F of the front wheel. u1 and the optimal suspension force F of the rear wheel u2 , and input into the damping coefficient controller;

[0012] Step (5): The sensor module, the extended observer module, the adaptive LQR controller, the damping coefficient controller, and the damping coefficient actuator are connected in series in sequence to form a semi-active suspension control system based on an adaptive LQR wheel hub direct drive.

[0013] The significant effects of the present invention after adopting the above technical solution are:

[0014] 1. The present invention runs 12 typical working conditions of low speed, medium speed and high speed on four road surfaces A, B, C and D, and uses genetic algorithm to offline optimize and store the weight matrix coefficients of the LQR controller under different typical working conditions. When the vehicle runs to a certain typical working condition, the adaptive LQR controller selects the optimal weight matrix coefficient and solves the optimal suspension force of the system. The optimal damping force is obtained by solving the optimal suspension force through the optimal damping coefficient, which solves the problem that the traditional suspension control cannot obtain the optimal suspension force under different continuous operating conditions.

[0015] 2. The constructed control system aims at the optimal control problem of wheel hub direct drive semi-active suspension. The adaptive LQR controller based on genetic algorithm is used to calculate the optimal suspension force of the front wheel and the optimal suspension force of the rear wheel. The required suspension force is provided by the damping coefficient actuator, that is, the optimal damping force of the front wheel provides the required optimal suspension force of the front wheel, and the optimal damping force of the rear wheel provides the required optimal suspension force of the rear wheel, thereby reducing the vertical acceleration of the vehicle body, the dynamic travel of the suspension, the dynamic load of the tire, the pitch angular acceleration of the vehicle body and the eccentricity between the stator and the rotor, thereby improving the comprehensive performance of the wheel hub direct drive suspension system, improving ride comfort and improving road friendliness. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a schematic diagram of the control model of the seven-degree-of-freedom wheel hub direct-drive semi-active suspension system for a semi-car.

[0017] Figure 2 is a schematic block diagram of the composition of the adaptive LQR controller;

[0018] Figure 3 The present invention is a schematic block diagram of a control system constructed using the method of the present invention. DETAILED DESCRIPTION

[0019] See also Figure 1 Taking the half-wheel hub direct drive suspension model as an example, the vehicle's forward direction is Figure 1 As shown by the arrow of the vehicle's forward speed v, the semi-wheel hub direct-drive suspension model adopts rear-wheel drive. The hub motor is installed on the rear wheel rim and drives the wheel to rotate. The introduction of the hub motor causes the electric vehicle to have coupled vertical vibration problems. During operation, the electric wheel is excited by the road surface, causing the motor stator and rotor to be eccentric. The eccentricity of the motor stator and rotor will produce unbalanced electromagnetic force. The unbalanced electromagnetic force acts on the motor structure to aggravate the eccentricity, causing vibration and noise problems. The rear wheel is divided into the inner stator and outer rotor of the motor with the hub motor air gap as the dividing interface, and the rear wheel is divided into the stator core mass block, the rim and rotor core mass block, and the tire mass block; the front wheel is used as the driven wheel, and the wheel is divided into the front wheel tire mass block and the rim mass block. The vertical vibration control model of the semi-vehicle seven-degree-of-freedom hub direct-drive semi-active suspension system is established, and the suspension dynamics differential equation is obtained as follows:

[0020]

[0021] In the formula, F u1 is the optimal suspension force of the front wheel, F u2 is the optimal suspension force of the rear wheel, F spr1 is the front wheel air spring force, F spr2 is the rear wheel air spring force, c 13 is the front suspension damping, c 23 is the rear suspension damping, z s1 is the vertical displacement of the front wheel suspension, is the vertical speed of the front wheel suspension, z s2 is the vertical displacement of the rear wheel suspension, is the vertical speed of the rear wheel suspension, z 12 is the vertical displacement of the front wheel rim, is the vertical velocity of the front wheel rim, z 23 is the vertical displacement of the stator core, is the vertical velocity of the stator core, m s is the sprung mass, is the vertical acceleration of the vehicle body, I p is the body moment of inertia, is the pitch angular acceleration of the half vehicle, l1 is the distance from the front wheel to the center of mass of the vehicle, l2 is the distance from the rear wheel to the center of mass of the vehicle, m 11is the mass of the front tire, m 12 is the front wheel rim mass, m 21 is the mass of the rear tire, m 22 is the rear wheel hub motor rim and rotor mass, m 23 is the stator core mass block of the rear wheel hub motor, is the vertical speed of the front tire, is the vertical acceleration of the front tire, is the vertical acceleration of the front wheel rim, z 21 is the vertical displacement of the rear tire, is the vertical speed of the rear tire, is the vertical acceleration of the rear tire, z 22 is the vertical displacement of the rim and rotor core, is the vertical velocity of the rim and rotor core, is the vertical acceleration of the rim and rotor core, is the vertical acceleration of the stator core, k 11 is the residual stiffness of the front tire, k 12 is the radial stiffness of the front tire, k 21 is the residual stiffness of the rear tire, k 22 is the radial stiffness of the rear tire, k 23 is the vertical bearing stiffness of the hub motor, c 12 is the radial damping of the front tire, c 22 is the radial damping of the rear tire, z 11 is the vertical displacement of the front tire, q1 is the front road excitation, q2 is the rear road excitation, F ez is the vertical unbalanced electromagnetic force of the hub motor, and e is the eccentricity between the stator and rotor.

[0022] Based on the suspension dynamics differential equation of formula (1), the following state space equation is established.

[0023]

[0024] in: represents the first-order derivative of the control model state quantity x(t). The control model state quantity x(t) consists of 14 state variables, which are the vertical displacement z of the vehicle body, s , vertical displacement of front wheel rim z 12 , vertical displacement of the front tire z 11 , stator core vertical displacement z 23 , rim and rotor core vertical displacement z 22 , vertical displacement of rear tire z 21 , half vehicle pitch angle θ, vehicle body vertical speed Front wheel rim vertical speed Front tire vertical speed Stator core vertical speed Vertical speed of rim and rotor core Rear tire vertical speed And the pitch angular velocity of the half vehicle These 14 state variables are:

[0025]

[0026] y(t) is the control output, u(t) is the control variable, ω(t) is the disturbance input, A is the state space system matrix, B u With B ω is the input matrix, C is the output matrix, D u With D ω is the feedforward matrix. By solving the Jacobi matrix, we can obtain the state space system matrix A and the input matrix B u With B ω , output matrix C, feedforward matrix D u With D ω as follows:

[0027]

[0028] in, Represents the first derivative of y(t).

[0029] The control variable u(t) is the optimal suspension force F of the front wheel u1 , optimal rear wheel suspension force F u2 .

[0030] u(t)=[F u1 F u2 ] (5)

[0031] The other nonlinear inputs of the system are taken as the disturbance input ω(t) of the system, specifically the front wheel road excitation q1, the rear wheel road excitation q2 and the vertical unbalanced electromagnetic force F of the wheel hub motor. ez :

[0032] ω(t)=[q1 q2 F ez ] T (6)

[0033] Where T is the matrix transpose.

[0034] Filtered white noise is used to simulate random road excitation q1 and q2, and its mathematical model can be expressed as:

[0035]

[0036] In the formula, represents the vertical velocity of the road surface (m / s), q(t) represents the vertical displacement of the road surface (m), q(t) is the front and rear wheel road surface excitation q1, q2, f0 is the lower cutoff frequency (HZ), Gd(n0) is the road surface roughness coefficient, according to the road surface grade from A to H, then G d The corresponding values ​​of (n0) are also different. The road roughness classification standard is shown in Table 1 below. l(t) represents Gaussian white noise, and v is the vehicle speed, km / h:

[0037] Table 1

[0038]

[0039]

[0040] It can be seen from formula (7) that the road surface excitation is related to the road surface grade and the vehicle speed. Therefore, the simulated road surface excitation is divided into 12 typical working conditions, namely, 30 km / h on A-grade road surface, 60 km / h on A-grade road surface, 90 km / h on A-grade road surface, 30 km / h on B-grade road surface, 60 km / h on B-grade road surface, 90 km / h on B-grade road surface, 30 km / h on C-grade road surface, 60 km / h on C-grade road surface, 90 km / h on C-grade road surface, 30 km / h on D-grade road surface, 60 km / h on D-grade road surface, and 90 km / h on D-grade road surface. These conditions cover low speed, medium speed, and high speed under A-grade road surface, B-grade road surface, and C-grade road surface. Among them, the speed of 30 km / h represents the low speed condition, the speed of 60 km / h represents the medium speed condition, and the speed of 90 km / h represents the high speed condition. These conditions represent the general operating conditions of the vehicle to a certain extent. After simulation of road surface excitation, 12 different road surface excitations were obtained.

[0041] The control output y(t) of the wheel hub direct drive semi-active suspension system has seven evaluation indicators, namely, the vertical acceleration of the vehicle body Eccentricity between stator and rotor e, dynamic load of front tire F d1 , dynamic load of rear tire F d2 , front wheel suspension dynamic travel f d1 , rear wheel suspension dynamic travel f d2 and the vehicle pitch angular acceleration Right now:

[0042]

[0043] Therefore, the control goal of the wheel hub direct drive semi-active suspension system is to reduce the vertical acceleration of the vehicle body. Eccentricity between stator and rotor e, dynamic load of front tire F d1 , dynamic load of rear tire F d2 , front wheel suspension dynamic travel f d1 , rear wheel suspension dynamic travel fd2 , vehicle pitch angular acceleration

[0044] For 12 typical working conditions, i.e. 12 different road excitations, 12 corresponding vehicle body vertical accelerations are set respectively. The weight coefficient is m1, the eccentricity between the stator and rotor is e, the weight coefficient is m2, and the vehicle body pitch angle acceleration The weight coefficient is m3, the front wheel suspension travel f d1 The weight coefficient is m4, the rear wheel suspension travel f d2 The weight coefficient is m5, the front tire dynamic load F d1 The weight coefficient is m6, the rear tire dynamic load F d2 The weight coefficient is m7, the optimal suspension force of the front wheel is F u1 The weight coefficient is p1 and the optimal suspension force F of the rear wheel u2 The weight coefficient is p2; then, 12 different groups of seven weight coefficients m1, m2, m3, m4, m5, m6, m7 respectively obtain the corresponding 12 target weight matrices Q = diag [m1, m2, m3, m4, m5, m6, m7], and 12 different groups of two weight coefficients p1, p2 respectively obtain the corresponding 12 control weight matrices R = diag [p1, p2], where diag refers to the mathematical meaning of a diagonal matrix, a square matrix whose elements not on the diagonal are all zero; therefore, under a typical working condition, there is a group of target weight matrices Q and control weight matrices R, a total of 12 groups of target weight matrices Q and control weight matrices R.

[0045] Genetic algorithm is used to optimize the above 12 groups of target weight matrices Q and control weight matrices R respectively, and 12 groups of optimal target weight matrices Q are obtained. b and the optimal control weight matrix R b , the specific optimization method is as follows:

[0046] First, generate the initial population, encode the initial population, and form a chromosome population. Assign the individuals generated by the population to m1, m2, m3, m4, m5, m6, m7 in the target weight matrix Q and p1, p2 in the control weight matrix R one by one, and establish the initial target weight matrix Q0 = [m1', m2', m3', m4', m5', m6', m7'] and the initial control weight matrix R0 = [p1', p2']. In the process of continuous repetition, the new generation of chromosome populations generated needs to meet the requirements of the fitness function better than the previous generation, and solve the global optimal solution of the fitness function.

[0047] Then, select the vertical acceleration of the vehicle body Eccentricity e between stator and rotor, vehicle body pitch angle acceleration Front wheel suspension travel f d1, rear wheel suspension dynamic travel f d2 , front wheel tire dynamic load F d1 , dynamic load of rear tire F d2 The optimization goal of the genetic algorithm is that due to the differences in units and magnitudes between the performance indicators, it is necessary to normalize the components of the objective function when constructing the fitness function. The minimum value of the sum of the ratios of the root mean square values ​​of the evaluation indicators under LQR control and without control is selected as the fitness function:

[0048]

[0049] Where, RMS is the root mean square value mathematical calculation, J w is the fitness function value; the subscript "p" represents uncontrolled under the same working condition, and no subscript p represents LQR control under the same working condition. Since the suspension performance under LQR control is compared with the passive suspension under uncontrolled, the fitness function is the normalized sum of the suspension performance indicators under uncontrolled and LQR control under the same working condition.

[0050] Finally, the following suspension system constraints must be met when searching for the global optimal solution to the problem. The new generation of chromosome populations in the genetic algorithm optimization process needs to better meet the requirements of the fitness function than the previous generation. The constraints are as follows:

[0051]

[0052] In the above formula, the vertical acceleration of the vehicle body is Weight coefficient m1, stator-rotor eccentricity e, weight coefficient m2, vehicle body pitch angle acceleration Weight coefficient m3, front wheel suspension travel f d1 Weight coefficient m4, rear wheel suspension travel f d2 Weight coefficient m5, front tire dynamic load F d1 Weight coefficient m6, rear tire dynamic load F d2 The weight coefficient m7 is constrained to range from [1,1e 10 ]; optimal front wheel suspension force F u1 The weight coefficient p1 and the optimal suspension force F of the rear wheel u2 The constraint condition of the weight coefficient p2 is [1e -5 ,1].

[0053] If the output parameters satisfy the above constraints, the fitness function value (J w ), determine whether the termination condition min(J w), if it is not satisfied, a new population is generated through genetic operations such as selection, crossover, and mutation to ensure that the population evolves in the direction of satisfying the constraints. After continuous iteration of the genetic algorithm, the optimal fitness function min(J w ), thus obtaining the optimal target weight matrix Q under this condition b and the optimal control weight matrix R b .

[0054] like Figure 2 As shown, the 12 groups of optimal target weight matrices Q b and the optimal control weight matrix R b Input into the LQR controller and save it to form an adaptive LQR controller.

[0055] The adaptive LQR controller is based on the suspension dynamics differential equation (1) and the optimal target weight matrix Q b The various weight coefficients in , the control target of the wheel hub direct drive air suspension system is converted into the cost function J:

[0056]

[0057] After rewriting into a standard quadratic form, we get the following cost function J:

[0058]

[0059] The optimization problem of the system is described as follows. The goal of the adaptive LQR controller is to obtain the minimum cost function, that is:

[0060] min(J) (13)

[0061] In the formula, Q c , R c and N c is the weight matrix of the cross product terms:

[0062] Q c =C T Q b C,N c =C T Q b D u ,R c =D u T Q b D u +R b .

[0063] The adaptive LQR controller solves the algebraic Riccati equation A T P+PA+Q c -PBu R c -1 B u T P = 0, we can get the solution of the semi-positive definite matrix P, and get the state feedback matrix K = R b -1 B u T P, A are the state space system matrices A, B u is the input matrix. Therefore, according to the state feedback matrix K and equation (5), u(t) = [F u1 F u2 ]The optimal control variable u(t) is:

[0064] u(t)=[F u1 F u2 ]= -Kx(t) (14)

[0065] To minimize the value of the LQR cost function J in equation (13), it depends on the selection of the state feedback matrix K, which is determined by the optimal target weight matrix Q b and the optimal control weight matrix R b Therefore, the adaptive LQR controller controls the control variable u(t) of the half-wheel hub direct drive suspension system = [F u1 F u2 ] depends on the optimal target weight matrix Q b and the optimal control weight matrix R b The adaptive LQR controller finally outputs the optimal control variable u(t) under the corresponding working condition.

[0066] like Figure 2 and Figure 3 As shown in FIG. 1 , the sensor module, the extended observer module, the adaptive LQR controller, the damping coefficient controller, and the damping coefficient actuator are connected in series in sequence to form a control system based on the adaptive LQR wheel hub direct drive semi-active suspension. The output end of the sensor module is connected to the input end of the extended observer module and the adaptive LQR controller respectively. The output end of the extended observer module is connected to the input end of the adaptive LQR controller and the damping coefficient controller respectively. The output end of the adaptive LQR controller is connected to the input end of the damping coefficient controller. The adaptive LQR controller outputs the optimal suspension force F of the front wheel. u1 and the optimal suspension force F of the rear wheel u2 to the damping coefficient controller.

[0067] When the constructed adaptive LQR wheel hub direct drive semi-active suspension control system is working, the sensor module consists of seven acceleration sensors, which are installed at the mass of the semi-vehicle center of mass, the front wheel rim, the front wheel tire, the rear wheel motor stator, the rear wheel rim and the motor rotor, and the rear wheel tire to measure the vertical acceleration of the vehicle body. Vehicle pitch angle acceleration Front wheel rim mass vertical acceleration Vertical acceleration of the front tire mass Vertical acceleration of the stator core mass of the rear wheel motor Vertical acceleration of rear wheel rim and rotor core mass and the vertical acceleration of the rear tire mass The sensor module measures the vertical acceleration of the vehicle body Vehicle pitch angle acceleration Vertical acceleration of front wheel rim mass Vertical acceleration of the front tire mass Vertical acceleration of the stator core mass of the rear wheel motor Vertical acceleration of rear wheel rim and rotor core mass and the vertical acceleration of the tire mass Transmitted to the extended observer module, the vertical acceleration of the vehicle body and the vehicle pitch angular acceleration Transmitted to the adaptive LQR controller. The input of the extended observer module is the vertical acceleration of the vehicle body output by the sensor module Vehicle pitch angle acceleration Front wheel rim mass vertical acceleration Vertical acceleration of the front tire mass Vertical acceleration of the stator core mass of the rear wheel motor Vertical acceleration of rear wheel rim and rotor core mass and the vertical acceleration of the tire mass It is used to estimate the information required by this system but cannot be measured by ordinary sensors. The estimated information is the vertical displacement z of the vehicle body. s and its speed Front wheel suspension vertical displacement z s1 and its speed Front wheel rim vertical displacement z 12 and its speed Front tire vertical displacement z 11 and its speed Vehicle pitch angular velocity Rear wheel suspension vertical displacement z s2 and its speed Vertical displacement of stator core z 23 and its speed Vertical displacement z of rim and rotor core 22 and its speed Rear tire vertical displacement z 21 and its speed The estimated vertical displacement of the vehicle body is transmitted to the adaptive LQR controller. Front wheel suspension vertical speed and the vertical speed of the rear wheel suspension Transmitted to the damping coefficient controller. This system has certain nonlinearity, and the extended Kalman filter algorithm is better at solving nonlinear problems. Therefore, the extended Kalman filter algorithm is used to build the system state observer. For the specific algorithm of the extended observer module, see (Yu Wenhao. Research on collaborative control method of vehicle interconnected air suspension system [D]. Jiangsu University, 2020.).

[0068] The adaptive LQR controller directly calls a set of corresponding optimal target weight matrices Q stored therein according to different typical driving conditions of the vehicle. b and the optimal control weight matrix R b , the current front wheel optimal suspension force F is calculated by formulas (11) to (14) u1 and the optimal suspension force F of the rear wheel u2 And output to the damping coefficient controller.

[0069] The input of the damping coefficient controller is the optimal front wheel suspension force F output by the adaptive LQR controller u1 , optimal suspension force F of the rear wheel u2 And the vertical speed of the front wheel suspension output by the extended observer module Rear wheel suspension vertical speed Front wheel rim vertical speed Stator core vertical speed The damping control force is used to provide the required optimal suspension force, that is, the front wheel suspension damping force F D1 Provide the required optimal suspension force F for the front wheel u1 , through the rear wheel suspension damping force F D2 Provides the required optimal suspension force F for the rear wheel u2 , the required damping force is provided by the damping coefficient controller. Based on the adjustable damping shock absorber model, in order to ensure that the optimal damping control force is always within the actual output force range of the shock absorber, it is necessary to limit the solution range when solving the damping control force. During the extension or compression of the suspension, the front wheel suspension damping force F D1 and rear wheel suspension damping force F D2 The constraints are as follows:

[0070]

[0071] In the formula, c minis the minimum damping coefficient of the damping coefficient controller, c max is the maximum damping coefficient of the damping coefficient controller, is the rate of change of front wheel suspension travel, is the front wheel suspension dynamic travel change rate. When the front wheel suspension damping force F D1 <0, indicating that the damping force of the front wheel is in the downward direction, and the damping shock absorber is in a stretched state to provide damping force; when the rate of change of the dynamic travel of the front wheel suspension is When the front wheel suspension damping force F D1 >0, indicating that the damping force of the front wheel is in the upward direction, and the damping shock absorber is in a compressed state to provide damping force. Similarly, the rear wheel suspension dynamic travel change rate and the rear wheel suspension damping force F D2 relationship.

[0072] The damping coefficient controller integrates the damping coefficient control method. The front wheel suspension damping force F is used in the semi-active suspension control. D1 Provides the required optimal suspension force F for the front wheel u1 , rear wheel suspension damping force F D2 Provides the required optimal suspension force F for the rear wheel u2 , solve for the optimal damping coefficient c of the front wheel 13 * and the optimal damping coefficient c for the rear wheel 23 * :

[0073]

[0074] The damping coefficient controller sets the optimal damping coefficient c of the front wheel 13 * and the optimal damping coefficient c for the rear wheel 23 * The damping coefficient actuator is composed of an adjustable damping shock absorber, which has a significant impact on the vehicle's handling stability and driving smoothness. The adjustable damping shock absorber can change the damping coefficient according to the vehicle's performance requirements to improve the vibration reduction performance of the suspension system under different driving conditions. By adjusting the damping valve opening, the front and rear wheel damping coefficients are adjusted to the corresponding front wheel optimal damping coefficient c. 13 * and the optimal damping coefficient c for the rear wheel 23 * .

Claims

1. A construction method based on an adaptive LQR wheel hub direct drive semi-active suspension control system is proposed. A vertical vibration control model of a semi-car seven-degree-of-freedom wheel hub direct drive semi-active suspension system is established to obtain the suspension dynamics differential equation. Its characteristics are The following steps are involved: Step (1): Simulate road excitation for 12 typical working conditions and set the corresponding vehicle body vertical acceleration for the 12 typical working conditions. The weight coefficient is m 1 , the weight coefficient of the eccentricity e between the stator and rotor is m 2 , vehicle pitch angular acceleration The weight coefficient is m 3 , front wheel suspension dynamic travel f d1 The weight coefficient is m 4 , rear wheel suspension dynamic travel f d2 The weight coefficient is m 5 , front wheel tire dynamic load F d1 The weight coefficient is m 6 , dynamic load of rear tire F d2 The weight coefficient is m 7 , optimal suspension force of the front wheel F u1 The weight coefficient is p 1 and the optimal suspension force F of the rear wheel u2 The weight coefficient is p 2 , and obtain the corresponding 12 target weight matrices Q = diag[m 1 ,m 2 ,m 3 ,m 4 ,m 5 ,m 6 ,m 7 ] and 12 control weight matrices R = diag[p 1 ,p 2 ], diag is a diagonal matrix, which corresponds to a set of target weight matrix Q and control weight matrix R under a typical working condition; Step (2): Use genetic algorithm to optimize the 12 sets of target weight matrices Q and control weight matrices R respectively, and obtain 12 sets of optimal target weight matrices Q b and the optimal control weight matrix R b And input into the LQR controller for saving, forming an adaptive LQR controller; When optimizing the 12 sets of target weight matrices Q and control weight matrices R, the individuals generated by the population are assigned to m one by one. 1 , m 2 , m 3 , m 4 , m 5 , m 6 , m 7 With p 1 , p 2 , select the minimum value of the sum of the ratios of the root mean square values ​​of each evaluation index under LQR control and without control as the fitness function J w : RMS is the root mean square value mathematical calculation. The subscript "p" represents no control under the same working conditions. No subscript p represents the same working conditions with LQR control. Solve the global optimal solution of the fitness function to obtain the optimal target weight matrix Q b and the optimal control weight matrix R b ; When solving the global optimal solution of the fitness function, the suspension system constraints must be met: Step (3): The adaptive LQR controller is based on the suspension dynamics differential equation and the optimal target weight matrix Q b The cost function J is calculated by the weight coefficients in is the vertical acceleration of the vehicle body, e is the eccentricity between the stator and rotor, is the vehicle body pitch angular acceleration, f d1 is the front wheel suspension travel, f d2 is the rear wheel suspension travel, F d1 is the front tire dynamic load, F d2 is the rear tire dynamic load, F u1 is the optimal suspension force of the front wheel, F u2 Optimal suspension force for the rear wheels; Step (4): The goal is to find the minimum cost function min(J) and solve for the optimal suspension force F of the front wheel. u1 and the optimal suspension force F of the rear wheel u2 , and input into the damping coefficient controller; Rewrite the cost function J into a standard quadratic form: Among them, the control model state quantity are the vertical displacement of the vehicle body z s , vertical displacement of front wheel rim z 12 , vertical displacement of the front tire z 11 , vertical displacement of stator core z 23 , rim and rotor core vertical displacement z 22 , vertical displacement of rear tire z 21 , half vehicle pitch angle θ, vehicle body vertical speed Front wheel rim vertical speed Front tire vertical speed Stator core vertical speed Vertical speed of rim and rotor core Rear tire vertical speed And the pitch angular velocity of the half vehicle Control variable u(t) = [F u1 F u2 ]; Q c , R c and N c is the weight matrix of the cross product terms, Q c =C T Q b C,N c =C T Q b D u ,R c =D u T Q b D u +R b , solve equation A T P+PA+Q c -PB u R c -1 B u T P = 0 to obtain the solution of the semi-positive definite matrix P, and the state feedback matrix K = R b -1 B u T P, A is the state space system matrix, B u is the input matrix, and the control variable u(t)=[F u1 F u2 ]=-Kx(t), solve for the optimal suspension force F of the front wheel u1 and the optimal suspension force F of the rear wheel u2 ; The damping coefficient controller is based on the formula Calculate the optimal damping coefficient c of the front wheel 13 * and the optimal damping coefficient c for the rear wheel 23 * And enter the damping coefficient actuator, F D1 and F D2 are the front wheel suspension damping force and the rear wheel suspension damping force, is the vertical speed of the front wheel suspension, is the vertical speed of the front wheel rim, is the vertical speed of the rear wheel suspension, is the vertical velocity of the stator core; During the extension or compression of the suspension, the front wheel suspension damping force F D1 and rear wheel suspension damping force F D2 for: c min is the minimum damping coefficient of the damping coefficient controller, c max is the maximum damping coefficient of the damping coefficient controller, is the rate of change of front wheel suspension travel, is the rate of change of front wheel suspension dynamic travel; Step (5): The sensor module, the extended observer module, the adaptive LQR controller, the damping coefficient controller, and the damping coefficient actuator are sequentially connected in series to form a semi-active suspension control system based on an adaptive LQR wheel hub direct drive; the output end of the sensor module is respectively connected to the input end of the extended observer module and the adaptive LQR controller, the output end of the extended observer module is respectively connected to the input end of the adaptive LQR controller and the damping coefficient controller, the output end of the adaptive LQR controller is connected to the input end of the damping coefficient controller, and the adaptive LQR controller outputs the optimal suspension force F of the front wheel u1 and the optimal suspension force F of the rear wheel u2 to the damping coefficient controller.

2. The method for constructing a control system based on an adaptive LQR wheel hub direct drive semi-active suspension according to claim 1, Its characteristics are: The suspension dynamics differential equation is: F u1 is the optimal suspension force of the front wheel, F u2 is the optimal suspension force of the rear wheel, F spr1 is the front wheel air spring force, F spr2 is the rear wheel air spring force, c 13 is the front suspension damping, c 23 is the rear suspension damping, z s1 is the vertical displacement of the front wheel suspension, is the vertical speed of the front wheel suspension, z s2 is the vertical displacement of the rear wheel suspension, is the vertical speed of the rear wheel suspension, z 12 is the vertical displacement of the front wheel rim, is the vertical velocity of the front wheel rim, z 23 is the vertical displacement of the stator core, is the vertical speed of the stator core, m s is the sprung mass, is the vertical acceleration of the vehicle body, I p is the body moment of inertia, is the pitch angular acceleration of the half vehicle, l 1 is the distance from the front wheel to the center of mass of the vehicle, l 2 is the distance from the rear wheel to the center of mass of the vehicle, m 11 is the mass of the front tire, m 12 is the front wheel rim mass, m 21 is the mass of the rear tire, m 22 is the rear wheel hub motor rim and rotor mass, m 23 is the stator core mass block of the rear wheel hub motor, is the vertical speed of the front tire, is the vertical acceleration of the front tire, is the vertical acceleration of the front wheel rim, z 21 is the vertical displacement of the rear tire, is the vertical speed of the rear tire, is the vertical acceleration of the rear tire, z 22 is the vertical displacement of the rim and rotor core, is the vertical velocity of the rim and rotor core, is the vertical acceleration of the rim and rotor core, is the vertical acceleration of the stator core, k 11 is the residual stiffness of the front tire, k 12 is the radial stiffness of the front tire, k 21 is the residual stiffness of the rear tire, k 22 is the radial stiffness of the rear tire, k 23 is the vertical bearing stiffness of the hub motor, c 12 is the radial damping of the front tire, c 22 is the radial damping of the rear tire, z 11 is the vertical displacement of the front tire, q1 is the front road excitation, q2 is the rear road excitation, F ez is the vertical unbalanced electromagnetic force of the hub motor, and e is the eccentricity between the stator and rotor.

3. The method for constructing a control system based on an adaptive LQR wheel hub direct drive semi-active suspension according to claim 1, Its characteristics are: The 12 typical operating conditions are: Class A road speed of 30km / h, Class A road speed of 60km / h, Class A road speed of 90km / h, Class B road speed of 30km / h, Class B road speed of 60km / h, Class B road speed of 90km / h, Class C road speed of 30km / h, Class C road speed of 60km / h, Class C road speed of 90km / h, Class D road speed of 30km / h, Class D road speed of 60km / h and Class D road speed of 90km / h.

Images