Disclosure of Invention
The invention aims to provide a damping adjustable semi-active suspension hybrid model predictive control method which comprehensively balances comfort and operation stability and is based on damping force constraint.
In order to solve the technical problems, the invention provides a damping force constraint-based damping adjustable semi-active suspension hybrid model predictive control method, which comprises the following steps of establishing a semi-active suspension hybrid system model, carrying out finite time domain optimization control on a semi-active suspension hybrid system based on a model predictive control theory, converting the finite time domain optimization control into a mixed integer quadratic programming containing a real-valued variable and a binary variable, and realizing hybrid model predictive control on a damping continuously adjustable semi-active suspension, wherein the finite time domain optimization control comprises the following steps:
step 1: dynamic modeling of a damping-adjustable semi-active suspension hybrid system;
step 2: optimizing nonlinear constraint of the damping adjustable semi-active suspension;
and step 3: semi-active suspension hybrid model predictive control;
and 4, step 4: and (3) solving the rolling time domain optimization of the semi-active suspension hybrid system.
The dynamic modeling of the damping-adjustable semi-active suspension hybrid system in the step 1 comprises the following steps: a) introducing auxiliary variables, and converting logical and nonlinear constraint conditions into a group of threshold values and logical conditions; b) and establishing a hybrid system model containing system equations and constraint conditions by using a hybrid system description language, and converting the hybrid system model into an MLD model by using a self-contained HYSDEL compiler.
The logical constraint condition refers to a passive constraint condition of the semi-active suspension, namely when the sign of the damping force demand is the same as the relative speed of the suspension, the required control force is equal to the optimal control force.
The nonlinear constraint condition means that the damping force of the shock absorber must meet a certain boundary condition, namely the optimized damping force is located in the force action range of the actual shock absorber.
The semi-active suspension hybrid model predictive control of the
step 3 comprises the following steps: a) in the prediction time domain
Keeping the function type and weight function of the objective function unchanged to the first
kThe predicted input and output of the point are independent variables, the first one is established
kThe discrete target function of the points, and then the linear weighting of the target function of each point is carried out to obtain the target function in the prediction time domain; b) expanding the constraint condition to the whole prediction time domain, namely constraining each prediction point to obtain a state space equation and the constraint condition in the prediction time domain; c) and constructing model predictive control of the semi-active suspension hybrid system based on an objective function and a constraint condition in a prediction time domain, solving the optimization through a rolling time domain at each sampling moment, taking a first element of an optimal control sequence as control input of the current moment, and repeating the optimization process at the next sampling moment.
The rolling time domain optimization solution of the semi-active suspension hybrid system in the step 4 comprises the following steps: a) loosening the integral constraint limit of part or all of the binary variables in the optimized variables to generate a series of quadratic plans corresponding to the original MIQP; b) and solving the series of quadratic programming to obtain a suboptimal solution or a global optimal solution of the MIQP which meets the integer constraint condition.
The invention has the following advantages:
1) the invention solves the contradiction between comfort and operation stability by constructing a multi-objective optimization method;
2) the invention solves the nonlinear constraint condition existing in the optimization problem of the semi-active suspension by a hybrid logic dynamic modeling method;
3) the invention adopts a hybrid model predictive control method to convert the semi-active suspension control problem into an optimization control method with constraint in a limited time domain;
4) the invention converts the optimization problem with nonlinear constraint into mixed integer quadratic programming containing real-valued variables and binary variables, and solves the problem by means of a branch-and-bound algorithm, thereby effectively solving the rolling time domain optimization problem with nonlinear constraint.
Detailed Description
The invention is described in detail below with reference to the accompanying drawings, taking a model of a semi-active suspension single-wheel vehicle as an example, but it can be implemented in many different ways, as defined and covered by the claims.
The suspension optimization control objective function comprehensively balancing comfort and operation stability at least comprises two parts of contents: a) the vertical vibration acceleration of the vehicle body and the two-norm of the dynamic stroke of the suspension are used as comfort performance indexes; b) the two-norm of the tire dynamic displacement is used as an operation stability performance index.
The invention provides a damping force constraint-based damping adjustable semi-active suspension hybrid model predictive control method, which comprises the following steps of establishing a semi-active suspension hybrid system model, carrying out finite time domain optimization control on a semi-active suspension hybrid system based on a model predictive control theory, converting the finite time domain optimization control into mixed integer quadratic programming containing real-valued variables and binary variables, and realizing hybrid model predictive control on a damping continuously adjustable semi-active suspension, wherein the mixed integer quadratic programming comprises the following real-valued variables and binary variables:
step 1: dynamic modeling of a damping-adjustable semi-active suspension hybrid system;
step 2: optimizing nonlinear constraint of the damping adjustable semi-active suspension;
and step 3: semi-active suspension hybrid model predictive control;
and 4, step 4: and (3) solving the rolling time domain optimization of the semi-active suspension hybrid system.
The dynamic modeling of the damping-adjustable semi-active suspension hybrid system in the step 1 comprises the following steps: a) introducing auxiliary variables, and converting logical and nonlinear constraint conditions into a group of threshold values and logical conditions; b) and establishing a hybrid system model containing system equations and constraint conditions by using a hybrid system description language, and converting the hybrid system model into an MLD model by using a self-contained HYSDEL compiler.
The logical constraint condition refers to a passive constraint condition of the semi-active suspension, namely when the sign of the damping force demand is the same as the relative speed of the suspension, the required control force is equal to the optimal control force.
The nonlinear constraint condition means that the damping force of the shock absorber must meet a certain boundary condition, namely the optimized damping force is located in the force action range of the actual shock absorber.
Step 1: dynamic modeling of a damping-adjustable semi-active suspension single-wheel vehicle:
fig. 1a shows a passive suspension of a model of a single wheel vehicle, and fig. 1b shows a semi-active suspension of a model of a single wheel vehicle equipped with a damper adjustable shock absorber, wherein:
in order to obtain a sprung mass,
is the unsprung mass of the spring,
in order to be the spring rate, the spring,
in order to provide a passive suspension damping coefficient,
in order to be the rigidity of the tire,
、
and
respectively the sprung mass, the unsprung mass and the vertical displacement of the road surface. Suspension stiffness in FIG. 1a
And damping
Is a constant value; the damping force of fig. 1b can be continuously adjustable according to the actual driving road conditions.
For the damping continuously adjustable semi-active suspension, the dynamic equation is
In the formula:
、
the vertical acceleration of the sprung and unsprung masses respectively,
a control damping force is applied to damp the shock absorber.
Selection state variable
Output variable
Control quantity of
Obtaining a system state equation of
Wherein the content of the first and second substances,
step 2: optimizing nonlinear constraint of the damping adjustable semi-active suspension:
comprehensively considering the requirements of the semi-active suspension on the smoothness and the operation stability of the vehicle, defining performance indexes as follows:
wherein the content of the first and second substances,
,
and
weighting coefficients of the vertical vibration acceleration of the vehicle body, the dynamic stroke of the suspension and the dynamic displacement of the tire are selected according to the relative importance of the smoothness and the operation stability;
weighting coefficients are input for control.
According to the formulas (3) and (4), the objective function (5) can be expressed as
Wherein the content of the first and second substances,
,
the required control force of the semi-active suspension is subject to certain constraints, i.e.
(7) The equation gives the passive constraint condition of the semi-active suspension that the required control force is equal to the optimal control force when the damping force demand sign is the same as the relative velocity of the suspension. (8) The formula gives the boundary condition that the damping force of the shock absorber must satisfy, namely the optimized damping force is located in the force action range of the actual shock absorber. Fig. 2 is a damping characteristic curve obtained by a characteristic test of an actual shock absorber, and the characteristic curve of the present embodiment is drawn only under partial current, and the actual current action range of the shock absorber is 0.29-1.6A. The damping force at the maximum current (1.6A) and the minimum current (0.29A) are taken as control quantity boundaries, which are related to the relative speed of the suspension and exhibit strong nonlinearity.
For the purpose of description of optimization, the boundary of the expression (8) is expressed by a set of linear inequalities of the control quantity and the state quantity, that is
Wherein the content of the first and second substances,
,
and
in order to be a coefficient of fit,
。
combining equations (6) - (9), the optimal control of the semi-active suspension can be regarded as hybrid system optimal control under the nonlinear constraint condition.
And step 3: semi-active suspension hybrid model predictive control:
1) dynamic modeling of a semi-active suspension hybrid system:
for the nonlinear constraint problem of the semi-active suspension, in this embodiment, a Mixed Logical Dynamic (MLD) modeling method proposed by Bemporad a is adopted, and a system equation and constraint conditions are described under a unified model framework in consideration of logic, dynamics, constraints and the like of the system.
Introducing binary auxiliary variables
And
the constraint (7) is converted into a set of thresholds and logic conditions, i.e.
In the formula: "
"(equivalent) and"
"(implication) is a logical operator.
Introducing binary auxiliary variables
And
continuous auxiliary variable
、
、
、
、
、
、
、
The constraints (8) (9) are converted into a set of thresholds and logic conditions, i.e.
Establishing a Hybrid System model containing System equations (3) (4) and constraints (10) (11) (12) (13) (14) by using Hybrid System Description Language (HYSDEL), and converting the Hybrid System model into an MLD model by using an own HYSDEL compiler, namely converting the Hybrid System model into the MLD model
Wherein the content of the first and second substances,
,
,
,
,
,
,
。
2) semi-active suspension hybrid system model predictive control
The present embodiment is based on Hybrid Model Predictive Control (HMPC) theory, and researches the finite time domain optimization Control of the semi-active suspension Hybrid system. In the prediction time domain
Keeping the function type and weight function of the objective function (6) unchanged, and
kthe predicted input and output of the point are independent variables, the first one is established
kA discrete objective function of points; then linearly weighting the objective function of each point to obtain the objective function in the prediction time domain, namely
Similarly, the two types of (15) and (16) are maintained, and are extended to the whole prediction time domain
I.e. constraining each predicted point, obtaining the state space equation and constraint condition in the prediction time domain
Wherein the content of the first and second substances,
,
,
,
,
。
in that
tAt the moment of time, the time of day,
for the state values of the MLD system, model predictive control of the semi-active suspension hybrid system is regarded as solving the following optimization problem:
wherein the variables are optimized
;
At each sampling instant
tSolving the above optimization by rolling time domain, from the optimal control sequence
First element of (1)
As control input for the current time, i.e.
At the next sampling instant
t+1, repeating the optimization process. It can be concluded that the optimized variables of the quadratic program (20) include real-valued variables and binary variables, which can be further converted into a Mixed Integer Quadratic Program (MIQP) solution.
The rolling time domain optimization solution of the semi-active suspension hybrid system in the step 4 comprises the following steps: a) loosening the integral constraint limit of part or all of the binary variables in the optimized variables to generate a series of quadratic plans corresponding to the original MIQP; b) and solving the series of quadratic programming to obtain a suboptimal solution or a global optimal solution of the MIQP which meets the integer constraint condition.
And 4, step 4: and (3) solving the rolling time domain optimization of the semi-active suspension hybrid system:
the optimization solution is convenient, and the output vector in the prediction time domain is defined
Controlling the input vector
Weighting matrix
And
namely:
the objective function (17) is written in the form:
meanwhile, equation of state space (18) is written in the form:
wherein the content of the first and second substances,
the formula (22) is substituted into the formula of the objective function (21) to obtain
Order to
The objective function (23) is expressed as
The constraint (19) is written as follows:
wherein
Is further written as
To this end, a mixed integer quadratic programming MIQP of the semi-active suspension system is obtained according to an objective function (24) formula and a constraint condition (26) formula, namely
In the formula:
in order to optimize the vector, the vector is optimized,
i.e. by
Both real and binary variables.
For the problem posed by equation (27), Branch-and-bound (Branch) is based on&Bound) algorithm, first relaxing the optimization variables
Medium two-value variable
Generating a series of quadratic plans corresponding to the original MIQP by limiting part or all of the integer constraints; and then solving the series of quadratic programming to obtain a suboptimal solution or a global optimal solution of the MIQP which meets the integer constraint condition, wherein the solving of the quadratic programming is completed by means of a Matlab optimization tool box.
In this embodiment, simulation verification is performed in a Matlab \ Simulink environment, and main parameters are shown in fig. 3.
Two simulation working conditions are set: 1) inputting working conditions on a random road surface; 2) and inputting the working condition of the road surface with the sine wave protrusions. The control effect is compared with a passive suspension and an LQR control semi-active suspension, wherein the LQR control parameter is selected to be the same as the hybrid model predictive control algorithm of the invention, and the optimal control law is
In the formula:
the gain is feedback controlled for LQR.
Derived from LQR control algorithms due to nonlinear constraints of semi-active suspensions
Cannot be completely realized, and additional control laws are required
For the convenience of later simulation comparison, a Passive suspension simulation result is represented by 'Pasive', an LQR algorithm control result is represented by 'LQR', and a hybrid model predictive control algorithm for research is represented by 'HMPC'. Fig. 4 to 10 show simulation effect diagrams of the present embodiment.
Random road input conditions are widely used to test vehicle comfort and handling stability. A, B, C three-level road surfaces are respectively adopted, the vehicle speeds are 40km/h, 60km/h, 80km/h and 100km/h in sequence, and the simulation time is 20 s. Fig. 11 shows the root mean square values of A, B, C systems at three different road surface levels, and it can be seen that the root mean square values of the evaluation indexes (vehicle body vertical acceleration, suspension dynamic stroke and tire dynamic load) of the Passive suspension, the LQR suspension and the HMPC suspension all show an ascending trend from level a to level C, while on the road surface at different levels, the HMPC suspension can effectively reduce the numerical values of the indexes relative to the Passive suspension and the LQR suspension, and improve the comprehensive performance of the suspension. In order to test the vibration response at different vehicle speeds, the root mean square values of the vibration response quantity at different vehicle speeds are shown in fig. 4 to 6, and it can be seen that the root mean square values of various evaluation indexes of the Passive suspension, the LQR suspension and the HMPC suspension, such as the vehicle body vertical acceleration, the suspension dynamic stroke and the tire dynamic load, all show an increasing trend along with the increase of the vehicle speed. Under different vehicle speeds, the LQR suspension and the HMPC suspension can effectively reduce each evaluation index value, and the control effect of the HMPC suspension is better than that of the LQR suspension.
The sine wave protrusion road surface input is used for testing the suspension bounce performance, in order to simulate the suspension performance more clearly, the amplitude and the width of the sine wave protrusion are respectively set to be 0.05m and 6m, the amplitude and the width exceed the actual vehicle wheelbase, the vehicle speed is 30km/h, 60km/h and 90km/h in sequence, and the simulation time is 10 s.
FIGS. 7-10 show sinusoidal wave raised road surface shapes and corresponding vibration response time history curves, wherein the vehicle speed is
And only intercepting the simulation result when 2-5 s is given. As seen from fig. 8, the HMPC suspension can effectively control the vertical acceleration of the vehicle body, and can rapidly attenuate the vibration acceleration of the vehicle body after the wheel passes the sinusoidal bump, as compared to the Passive suspension and the LQR suspension. As can be seen from fig. 9 and 10, the Passive suspension and the LQR semi-active suspension both fluctuate to different degrees at the later stage of the sine wave bulge, while the HMPC suspension can rapidly attenuate the dynamic stroke of the suspension and the dynamic load of the tire at the later stage of the sine wave bulge, and has a better vibration attenuation effect.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.