Extension coordination fault-tolerant control method for semi-active suspension of whole vehicle

Abstract

The invention discloses an extension coordination fault-tolerant control method for a semi-active suspension of a whole vehicle, and belongs to the technical field of vehicle control systems. The method comprises the following steps: establishing a whole vehicle semi-active suspension dynamical model containing an actuator fault; designing an H / H-infinity robust controller as a basic controller; the fault information of the actuator is estimated in real time through a fault observer, and it is guaranteed that the suspension actuator obtains the fault-tolerant control capability after a fault occurs through a self-compensation control mode; designing a mutual compensation controller based on a TD3 algorithm; designing a gain fault factor representing a fault state of the actuator and a pre-output factor representing an output state; constructing a two-dimensional extension set based on an extension theory to divide an extension state domain; and adaptively selecting a self-compensation control strategy, a mutual compensation control strategy or a coordinated compensation strategy according to the extension state domain judgment result. According to the method, the problem that the fault-tolerant control method is insufficient in adaptability when facing actuator faults of different degrees is solved, and the riding comfort and the handling stability of the vehicle under various fault working conditions are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle control systems, in particular to a whole vehicle semi-active suspension extensible coordination fault-tolerant control method considering the fault degree of actuators and output state. BACKGROUND

[0002] As a key component of modern vehicles, semi-active suspension can adjust the damping and stiffness of the suspension, effectively improve the vehicle dynamics performance and ensure the safety of vehicle driving, and has become a current research hotspot. The actuator of semi-active suspension often uses a continuously variable damping (CDC) shock absorber, which has been widely used in various types of vehicles due to its rapid response, low energy consumption and low cost.

[0003] However, due to harsh environments and component aging factors, actuators are prone to gain, bias, and stuck faults, which can deteriorate the performance of the suspension system and even cause the vehicle system to be unstable. At present, the research on fault observation and fault compensation of vehicle suspension fault actuators is relatively sufficient, such as control rate reconstruction or compensation based on fault observation for fault actuators (self-compensation control strategy). Under certain degree of fault, self-compensation control can directly act on the fault actuator to restore the performance of the suspension to normal level. However, due to the different fault degrees of the suspension actuators and the road input, the output state of the actuator may reach its limit under certain fault conditions, at which time the actuator can only output the maximum value instead of the expected value, reducing the fault-tolerant control effect of the suspension system; and when the fault actuator is in a completely failed state, the suspension self-compensation control will not be applicable.

[0004] In the face of the completely failed actuator that the self-compensation control method cannot be applied to, the whole vehicle suspension can adjust the output of the normal actuator to compensate for the performance loss of the fault actuator (complementary compensation control strategy). However, the traditional whole vehicle suspension fault complementary compensation control method has the problems of causing the normal actuator to reach the output limit and the deterioration of some suspension indicators in order to compensate for the performance loss of the fault actuator, affecting the effect of complementary compensation fault-tolerant control.

[0005] Both self-compensation control and complementary compensation control may cause the output state of the actuator to reach the limit due to the large fault degree of the actuator and the high original control output demand in these harsh fault conditions, resulting in the fault-tolerant control performance not meeting the expectation, that is, there is a contradiction between the vehicle fault-tolerant control performance and the output state limit of the suspension actuator in these harsh fault conditions.

[0006] Therefore, there is an urgent need for a fault-tolerant control method that can adaptively adjust according to the fault degree of the actuator and the output state to cope with different degrees of actuator fault conditions. SUMMARY

[0007] In order to solve the existing problems, the application provides a whole vehicle semi-active suspension extension coordination fault-tolerant control method, and the specific scheme is as follows:

[0008] A whole vehicle semi-active suspension extension coordination fault-tolerant control method comprises the following steps:

[0009] S1: a seven-degree-of-freedom whole vehicle semi-active suspension dynamics model containing an actuator fault is established;

[0010] S2: an H2 / H∞ robust controller is designed as a basic controller to generate a basic control force of the suspension system;

[0011] S3: a fault observer is designed to detect and estimate the actuator fault in real time, and a self-compensation control mode is used to ensure that the suspension actuator has fault-tolerant control capability after the fault;

[0012] S4: a double-delay deep deterministic policy gradient algorithm (TD3 algorithm) is used to learn and train a complementary compensation controller through the interaction between an agent and an environment, which is used to generate a complementary compensation control force borne by a normal actuator when any actuator fails;

[0013] S5: a gain fault factor and a pre-output factor are designed to construct a two-dimensional extension set;

[0014] S6: the extension state domain is divided, the associated function is calculated, the coordination compensation control strategy is constructed, and the fault-tolerant control strategy is adaptively selected according to the current system fault state;

[0015] S7: according to the selected fault-tolerant control strategy, the final fault-tolerant control force is calculated and output.

[0016] Preferably, the seven-degree-of-freedom whole vehicle semi-active suspension dynamics model in S1 comprises body vertical, pitch and roll movements and vertical movements of four non-sprung masses, and when the actuator fails, the state space expression is:

[0017]

[0018] wherein X is a state vector, W is an interference input, U is a control input, W is an interference input, Y is a measurement output, A is a 14 × 14 matrix; is a 14 × 4 matrix; is a 14 × 4 matrix; is an 11 × 14 matrix; is an 11 × 4 matrix; is a 7 × 14 matrix; .

[0019] Preferably, the fault observer design in S3 comprises:

[0020] S31: Constructing state observer equation:

[0021] ;

[0022] wherein, is the estimation of state ; is the estimation of Y; ; L is the observer gain matrix;

[0023] S32: Designing fault estimation algorithm: ;

[0024] wherein, is the fault learning rate;

[0025] S33: Designing observer gain matrix L and fault learning rate by solving linear matrix inequality, so that the state estimation error and fault estimation error converge gradually.

[0026] Preferably, the complementary compensation controller design based on TD3 algorithm in S4 comprises:

[0027] S41: Defining state space: including suspension relative displacement, tire dynamic deformation and vehicle body motion state;

[0028] S42: Defining action space: the compensation control force of normal actuator;

[0029] S43: Designing reward and punishment function: including comfort reward and punishment, steering stability safety reward and punishment and segmented auxiliary reward and punishment;

[0030] S44: Training the intelligent agent to learn the optimal complementary compensation control strategy by constantly interacting with the environment.

[0031] Preferably, the design of gain fault factor α and pre-output factor β in S5 is as follows:

[0032] , wherein is the actuator gain coefficient estimated by the fault observer;

[0033] Pre-output factor characterizes the actuator output state under different road input, by sensing the road excitation information in advance, the pre-output force of each actuator can be calculated by pre-inputting the road excitation information into the H2 / H∞ basic controller, and after normalization operation, the pre-output factor of the fault actuator is obtained .

[0034] Preferably, the extension state domain in S6 is divided into three regions:

[0035] Classical region: the region of α + |β| ≤ 1, the part of the actuator pre-output that does not exceed the post-fault output range of the actuator is taken as the classical region, and a self-compensation control strategy is adopted;

[0036] Extensible region: the region of α + |β| > 1 and α < 1, the part of the actuator pre-output that exceeds the post-fault performance range of the actuator is taken as the extensible region, and a coordination compensation strategy is adopted;

[0037] Non-region: the region of α ≥ 1, the part of the actuator adjustable damping that is completely disabled is taken as the non-region, and a complementary compensation control strategy is adopted.

[0038] Preferably, the control force of the coordination compensation strategy is calculated as follows:

[0039] ;

[0040] wherein is a self-compensation control force, is a complementary compensation control force, and are coordination weight coefficients based on a correlation function.

[0041] Preferably, the H2 / H∞ robust controller in S2 is obtained by solving a linear matrix inequality, and simultaneously meets the H2 and H∞ performance index requirements, thereby ensuring the basic control performance of the system in the fault-free case.

[0042] The application further discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method according to any one of the above methods when executing the program.

[0043] The application further discloses a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement the steps of the method according to any one of the above methods.

[0044] The application has the following beneficial effects:

[0045] The application establishes an adaptive mapping relationship between fault states and fault-tolerant strategies by the extension theory, so that the system can automatically select the optimal fault-tolerant control strategy according to the fault severity and the actuator output state; combining the advantages of self-compensation control and mutual-compensation control, the performance deficiency of single fault-tolerant control method in severe fault conditions is solved by coordinating the compensation strategy; the strong learning ability of TD3 algorithm is used to enable the mutual-compensation controller to find the optimal compensation strategy in the case of multi-index conflict; through the design of the fault observer and the pre-output module, real-time monitoring of the actuator fault and the output state is realized, providing accurate basis for fault-tolerant decision; under different degrees of actuator fault conditions, good vehicle ride comfort and steering stability can be maintained. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings described below are some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0047] Figure 1 is a seven-degree-of-freedom whole vehicle semi-active suspension dynamics model;

[0048] Figure 2 is a two-dimensional extension set division;

[0049] Figure 3 is a C-level simulation road surface;

[0050] Figure 4 is a suspension actuator fault observation result;

[0051] Figure 5 (a) is a extension region discrimination;

[0052] Figure 5 (b) is the calculation result of the correlation function;

[0053] Figure 6 is the fault actuator output force of self-compensation control. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0055] A whole vehicle semi-active suspension fuzzy coordination fault-tolerant control method, comprising the following steps:

[0056] S1: a seven-degree-of-freedom whole vehicle semi-active suspension dynamics model containing actuator faults is established.

[0057] 1. Seven-degree-of-freedom whole vehicle semi-active suspension dynamics model:

[0058] Considering the vertical, pitch, roll motion of the vehicle body and the vertical motion of the unsprung mass, a seven-degree-of-freedom whole vehicle semi-active suspension dynamics model is established. For example, Figure 1 is the seven-degree-of-freedom whole vehicle semi-active suspension dynamics model. In the figure is the pitch angle of the vehicle body; a and b are the distances from the vehicle body mass center to the front and rear axles, respectively; L is one-half of the wheelbase; is the unsprung mass; is the tire stiffness; is the road vertical displacement; is the fixed passive damping coefficient; is the suspension stiffness; is the adjustable damping force; is the displacement of the vehicle body suspension; the vertical displacement of the wheel , subscript i takes 1, 2, 3, 4, representing the left front, right front, left rear, and right rear vehicle body positions or tire positions.

[0059] Due to the limitations of the specific structure (bumper block) of the vehicle body, the maximum stroke of the suspension deflection is limited; in addition, due to the differences in the structure and type of the automobile suspension actuator itself, the size of the suspension control output force is also limited. Considering the above factors, the control output is selected as:

[0060]

[0061] wherein, is the suspension deflection, is the vertical displacement of the unsprung mass; is the mass center displacement at the four ends of the vehicle body; i takes 1, 2, 3, 4, representing the left front, right front, left rear, and right rear vehicle body positions. is the maximum limit value of the suspension deflection, is the maximum limit value of the actuator control output force, i takes 1, 2, 3, 4, representing the left front, right front, left rear, and right rear tire positions.

[0062] In addition, considering the ride comfort of the whole vehicle, it is desirable that the vertical acceleration of the vehicle body , the pitch angle acceleration , and the roll angle acceleration be as small as possible, so the control output is selected as:

[0063] ;

[0064] The state vector X, the disturbance input W, the control input U, the control output Z and the measurement output Y are selected to obtain the 7-DOF state space model of the semi-active suspension as follows:

[0065] ;

[0066] Wherein:

[0067]

[0068]

[0069]

[0070]

[0071] In the formula, X is the state vector, W is the disturbance input, U is the control input, W is the disturbance input, Y is the measurement output, A is a 14 × 14 matrix; is a 14 × 4 matrix; is a 14 × 4 matrix; is an 11 × 14 matrix; is an 11 × 4 matrix; is a 7 × 14 matrix; .

[0072] Vehicle body mass center velocity Vehicle body pitch angle velocity Vehicle body roll angle velocity Wheel vertical velocity Suspension deflection rate i is 1, 2, 3, 4, representing the left front, right front, left rear and right rear tire positions respectively.

[0073] 2. Actuator fault suspension model:

[0074] The output state limit of the actuator and its fault will cause the control effect of the ASS (Active Suspension System, ASS) to attenuate. Considering that the CDC damper generally uses an electromagnetic hydraulic valve to adjust the damping force, the leakage of hydraulic oil will cause the gain loss of the controlled output damping force of the hydraulic mechanism, so that the output state of the actuator is easy to reach the limit.

[0075] When the jth actuator has a gain fault, its output form can be uniformly expressed as:

[0076] ;

[0077] where j = 1, 2, 3, 4 represent the left front, right front, left rear, right rear actuator of semi-active suspension respectively; is the output of the jth actuator when it fails; is the output of the jth actuator when it is normal; is the gain coefficient of the jth actuator (when , it represents that the actuator is normal; when , it represents that the actuator completely fails in adjustable damping).

[0078] When the suspension actuator fails as described above, its output is:

[0079] ;

[0080] In summary, the state space model of the whole vehicle suspension system when the actuator fails is:

[0081] .

[0082] The whole vehicle semi-active suspension fault-tolerant control system designed in this paper contains three modules: controlled object, decision layer and control layer.

[0083] The controlled object in this paper is the whole vehicle semi-active suspension with actuator failure.

[0084] The fault-tolerant decision layer is composed of fault observation module, pre-output module and extension state domain division module. The fault observer in the fault observation module observes the fault information of the suspension actuator, and obtains the gain fault factor of the fault actuator after processing. The pre-output module is used to receive road information, and the pre-output information of the actuator can be obtained through the basic controller. After normalization processing, the pre-output information of the fault actuator is obtained. The extension state domain division module divides different extension state domains according to the gain fault factor and pre-output factor of the fault actuator, decides the fault-tolerant control strategy used in the current extension state, and calculates the corresponding correlation function.

[0085] The fault-tolerant controller is composed of self-compensation control and TD3 agent complementary compensation control. The self-compensation control is formed by adjusting the control rate of the faulty actuator according to the actuator fault information obtained by the fault observer and the control output of the basic controller. The training environment is built by combining the TD3 reinforcement learning theory. The agent is trained in the process of continuous interaction with the fault environment by expanding the state space and designing a segmented auxiliary reward and punishment function. The well-trained agent is used for TD3 complementary compensation control. The self-compensation control and TD3 agent complementary compensation control are coordinated and distributed by the correlation function output by the decision layer to obtain the final fault-tolerant control force of the whole vehicle semi-active suspension. In different extension state domains, the advantages of each fault-tolerant control strategy are fully utilized to improve the fault-tolerant control performance of the whole vehicle suspension.

[0086] S2: Design an H2 / H∞ robust controller as a basic controller to generate the basic control force of the suspension system.

[0087] An all-state feedback controller is designed by combining the H2 and H∞ performance indexes to realize the control of the output force of each actuator of the suspension. The state space system equation of the whole vehicle H2 / H∞ control suspension is:

[0088]

[0089] In the formula: , , , , , , , and are coefficient matrices; is the state variable; is the road excitation disturbance input; is the control variable; is the constraint output variable describing the control output part of the suspension system; is the expected output variable of the control output part of the suspension system.

[0090] The designed H2 / H∞ state feedback closed-loop control system needs to satisfy that the H∞ norm of the closed-loop transfer function T∞(s) from the external disturbance input i.e. the road excitation signal to the controlled output is less than . The smallest H2 performance index is solved to make the upper limit of the H2 norm of the closed-loop transfer function T2(s) as small as possible.

[0091] A common Lyapunov matrix To ensure that the system's H∞ and H2 performances are under the same controller, the H2 / H∞ controller must have a positive definite symmetric matrix. and Under the premise that the following necessary and sufficient condition is met:

[0092]

[0093] In the formula: For the identity matrix; Trace( ) is a matrix The traces.

[0094] The optimal solution that satisfies the design objective can be obtained by using the mincx function in MATLAB to solve the above linear matrix inequality. , If the closed-loop system state feedback control problem has a feasible region solution, then the state feedback control law H2 / H∞ can be obtained. .

[0095] S3: Design a fault observer to detect and estimate actuator faults in real time, and ensure that the suspension actuator has fault-tolerant control capability after a fault through self-compensation control.

[0096] To detect the system status and fault information after an actuator failure, the following actuator fault observer is established in the fault observation module:

[0097]

[0098] in, For state The estimate; An estimate of Y; L is the observer gain matrix.

[0099] definition Design a fault estimation algorithm using fault learning factors:

[0100] Then an augmented error system can be constructed:

[0101] in:

[0102] ; .

[0103] To ensure that the fault diagnosis observer can accurately estimate system state and fault information, an appropriate observer gain matrix must be designed. and fault learning rate Define the H∞ performance index and stability margin index When there is a positive definite matrix P and a matrix Q that make the augmented error system satisfy the following LMI equation

[0104]

[0105] The state estimation error and the fault estimation error converge with a stability margin index , and satisfy , and the system state vector gradually tends to 0.

[0106] The self-compensation control is as follows:

[0107] The actuator fault estimation value obtained by the fault observer and the normal control force of the H2 / H∞ robust controller are processed to obtain the actuator gain coefficient under the condition of gain loss fault of the suspension actuator , and the control rate of the faulty actuator is adjusted based on the actuator gain coefficient , so as to ensure that the suspension actuator has a certain fault-tolerant control ability after the fault through the self-compensation control.

[0108]

[0109] Since the output of the actuator has an upper limit value in practice, the output of the faulty actuator under the self-compensation control strategy is as follows:

[0110]

[0111] In the formula: is the output under the condition of actuator fault; is the reconstructed output of the faulty actuator under the self-compensation control strategy.

[0112] S4: A complementary compensation controller is obtained based on the double-delay deep deterministic policy gradient algorithm (TD3 algorithm) and through the interaction learning and training of the agent and the environment, which is used to generate the complementary compensation control force borne by the normal actuator when any actuator fails.

[0113] Specifically, the DRL method is used herein for whole vehicle suspension fault complementary compensation fault-tolerant control, that is, in the case of gain loss fault of a certain suspension actuator of the whole vehicle and complete failure of the adjustable damping force, the intelligent agent is trained to constantly interact with the environment to learn, the normal actuators are controlled to coordinate multiple conflicting vehicle indicators within the output range, and the reward function is set to guide the intelligent agent to explore better fault-tolerant control effect. Based on the above requirements, the double-delay deep deterministic policy gradient algorithm (TD3) combining the deep deterministic policy gradient algorithm (DDPG) and deep double Q-learning is adopted, the algorithm structure of which can effectively reduce the estimation bias, insufficient exploration efficiency and convergence problem in training, and enhance the stability of the policy.

[0114] The complementary compensation controller design based on the TD3 algorithm in S4 includes:

[0115] S41: Define the state space: including the suspension relative displacement, tire dynamic deformation and vehicle body motion state. Specifically: the following state observations are selected as the state space of the intelligent agent:

[0116]

[0117] Wherein: represents the suspension relative displacement; represents the tire dynamic deformation, i = 1, 2, 3, 4.

[0118] In order to prevent the magnitude difference between the input states from being too large, causing the intelligent agent to be difficult to converge during training, the states are normalized before being input to the deep reinforcement learning network:

[0119]

[0120] In the formula: represents the normalized value of the i-th state variable; represents the i-th state variable; represents the maximum absolute value of the i-th state variable.

[0121] S42: Define the action space: the compensation control force of the normal actuator. Specifically, the compensation control force of a certain actuator is taken as an example, then the action space is:

[0122]

[0123] Wherein: represents the compensation force of the remaining three normal actuators controlled by the intelligent agent, represents the sum of the compensation force and the basic control force of a certain actuator, , .

[0124] S43: design reward and punishment functions: including comfort reward, steering stability and safety reward, and segmented auxiliary reward. Specifically:

[0125] The reward and punishment functions are set as follows

[0126] Comfort reward function:

[0127]

[0128] Vehicle steering stability and safety reward function:

[0129]

[0130] wherein, (i=1, 2, …, 6) is the reward and punishment coefficient of each physical quantity.

[0131] The segmented auxiliary reward function is set as follows:

[0132]

[0133]

[0134]

[0135]

[0136] In the above segmented auxiliary reward function, , for giving additional positive rewards to better exploration, for giving additional negative rewards to poor exploration, then for giving greater punishment to the situation where the action result exceeds the safety boundary range, so as to improve the convergence ability of the agent training and guarantee the requirements of safe action and smooth driving.

[0137] Based on the above analysis, the comprehensive reward and punishment function of the reinforcement learning controller is represented as:

[0138] .

[0139] S44: train the agent to learn the optimal control strategy through the network.

[0140] S5: design gain fault factor and pre-output factor, and build two-dimensional extension set.

[0141] Specifically, the extension division theory is applied to the fault-tolerant control of the whole vehicle suspension. When any one of the actuators fails, the corresponding self-compensation control, TD3 intelligent agent complementary compensation control and coordinated compensation strategy are obtained through extension coordination decision in different extension state domains to meet different fault-tolerant control requirements. The extension characteristic quantity design first designs the gain fault factor The gain coefficient of the faulty actuator is estimated by the fault observer in the fault observation module The gain fault factor of the faulty actuator is obtained after calculation and processing The pre-output factor representing the output state of the actuator under different road inputs is designed from the perspective of the controller In the pre-output module, the road excitation information is obtained in advance through the sensing sensor The pre-input of the road excitation information into the H2 / H∞ basic controller can calculate the pre-output force of each actuator in advance After normalization operation, the pre-output factor of the faulty actuator is obtained The gain fault factor and the pre-output factor are used as the characteristic quantities of the two-dimensional extension set, and the decision relationship between the fault working condition and the fault-tolerant control strategy is established.

[0142] S6: Divide the extension state domain and calculate the associated function, and adaptively select the fault-tolerant control strategy according to the current system state.

[0143] Specifically, according to the selected characteristic quantities, the two-dimensional extension set of the gain fault factor with a change range of [0, +∞) and the pre-output factor with a change range of [-1, 1] is constructed, the fault working condition is converted into extension set division information, and the two-dimensional extension set division is shown in Figure 2 .

[0144] Effective division of the classical domain, the extension domain and the non-domain is one of the key links of the application of the extension set division theory. The invention combines the actuator fault state and the output state to divide three regions for discrimination.

[0145] Classical domain: the region of α + |β| ≤ 1, the part of the actuator pre-output that does not exceed the output range after the fault is taken as the classical domain, and the self-compensation control strategy is adopted;

[0146] Extension domain: the region of α + |β| > 1 and α < 1, the part of the actuator pre-output that exceeds the performance range after the fault is taken as the extension domain, and the coordinated compensation strategy is adopted;

[0147] Non-domain: the region of a ≥ 1, the part of the actuator adjustable damping complete failure as a non-domain, using complementary compensation control strategy.

[0148] S7: according to the selected fault-tolerant control strategy, calculate and output the final fault-tolerant control force.

[0149] Specifically, the coordination compensation fault-tolerant control needs to coordinate and distribute the self-compensation control and TD3 intelligent agent complementary compensation control through the correlation function to obtain the final fault-tolerant control force of the whole vehicle semi-active suspension.

[0150] The extension distance is the distance from the value point to the domain boundary, and the correlation function is a measure of the deviation of the current state of the system from the domain boundary. In this paper, the gain fault factor representing the degree of actuator failure and the pre-output factor representing the state of actuator output under different road inputs are taken as characteristic quantities for the calculation of the correlation function. Assuming that the current state P of the vehicle is in the extension domain, a straight line is drawn from the point P to the stable node O, which intersects the classical domain and the extension domain at points P1 and P2 respectively. The classical domain is represented as <O, P1>, and the extension domain is represented as <P1, P2>. The extension distance of P to the classical domain and the extension domain is and The calculation formula is shown in the formula.

[0151]

[0152]

[0153] According to the formula, the correlation function of point P can be calculated as

[0154]

[0155]

[0156] The output of the coordination compensation fault-tolerant control strategy in the extension domain is

[0157]

[0158] where is the self-compensation control force, is the complementary compensation control force, and is the coordination weight coefficient based on the correlation function.

[0159] Embodiment:

[0160] Table 1 Parameters of the whole vehicle suspension system

[0161]

[0162] 1. Fault observation simulation analysis

[0163] Setting appropriate H performance index for fault observation module in MATLAB / Simulink and stability margin index , and solving the observer gain matrix L and fault learning rate G.

[0164] Set the simulation road surface to C-class road surface, the simulation time is 10s, the actuator occurs gain loss fault at t = 2s, and the adjustable damping force completely fails at t = 6s. The C-class simulation road surface is shown in . The suspension actuator fault observation results are shown in Fig. 4, in which the output damping force of the actuator in good condition is Figure 3 and the damping force of the actuator fault observation is .

[0165] From the simulation results, it can be seen that the observer can effectively observe the actuator gain loss fault and the complete failure of the adjustable damping force. The gain loss fault is observed within 2-6s, and the complete failure of the adjustable damping force of the actuator is observed within 6-10s.

[0166] 2. Extension state domain discrimination

[0167] Set C-class road surface fault working condition.

[0168] Figure 5 (a) shows that when the gain fault factor is on the C-class road surface, the extension state switches between the classical domain and the extension domain, and at this time the fault actuator exists the output reaches the limit, which needs to enhance the suspension fault-tolerant control performance through the coordinated fault-tolerant control strategy. Figure 5 (b) shows the calculation results of the correlation function.

[0169] 3. Extension coordinated fault-tolerant control simulation analysis

[0170] The vehicle comfort index and its root mean square value for the fault working condition are shown in Table 2.

[0171] Table 2 Root mean square value of comfort index in extension domain simulation

[0172]

[0173] The output force of the fault actuator of the self-compensation control is shown in Figure 6 ​​The fault-tolerance effects of self-compensation control, AFC mutual compensation control, TD3 mutual compensation control, and coordinated control were compared in the extension domain. Among them, AFC mutual compensation control: in order to ensure that the semi-active suspension system can still maintain stability and acceptable performance when the actuator fails, based on the online estimation of actuator failure, the damping force of the faulty actuator is compensated by the intact actuator for the loss of vehicle performance by the vehicle dynamics relationship.

[0174] Without loss of generality, consider the case where only one actuator fails. j = 1 indicates that the left front actuator has malfunctioned, and the malfunction value is... Taking an example, the compensating damping force of each intact actuator is derived. Here, the AFC method is a mutual compensation control method derived from the vehicle dynamics relationship of the semi-active suspension:

[0175]

[0176] in, Indicates the actuator fault value. Indicates the compensating force of a normal actuator

[0177] Considering the boundedness of the suspension actuator output, the output of the k-th (k≠j) intact actuator under AFC mutual compensation fault-tolerant control... for

[0178]

[0179] in, Indicates the basic output force. Indicates compensatory force. This is the maximum output limit.

[0180] The AFC mutual compensation fault-tolerant control method does not require changes to the controller structure and avoids the complexity of conventional fault-tolerant control algorithms. It only needs to estimate the actuator fault amplitude and draw on the basic idea of ​​fault compensation to handle the fault-tolerant control problem in common actuator fault modes. However, its fault-tolerant effect is reduced by the actuator output limitation under severe fault conditions, which has certain limitations.

[0181] Figure 6 It can be seen that the output state of the faulty actuator reaches its limit under self-compensation control; TD3 mutual compensation, benefiting from the training of the agent, demonstrates its fault-tolerant control potential by controlling other normal actuators; the coordinated compensation strategy simultaneously leverages the efficient fault-tolerant advantage of self-compensation control and the fault-tolerant exploration potential of TD3 mutual compensation control, making its fault-tolerant effect the best compared to other strategies. Under severe fault conditions in the extended domain, the coordinated compensation strategy designed in this paper exhibits the best fault-tolerant control effect for vehicle ride comfort and has good practical engineering application value.

[0182] The application further discloses a computer readable storage medium and a computer system, wherein the computer readable storage medium has a computer program stored thereon, and the computer program performs the method described above after being executed. The computer system comprises a processor and a storage medium, and the storage medium has a computer program stored thereon. The processor reads and executes the computer program from the storage medium to perform the method described above.

[0183] Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

[0184] In one or more exemplary embodiments, the functions described can be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functions can be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media can be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Also, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. Disk and disc, as used herein, includes compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and blu-ray disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0185] Although the present application has been described in detail with reference to the foregoing embodiments, it should be understood that modifications can be made to the foregoing embodiments, or additional implementations can be implemented, without departing from the spirit and scope of the inventive subject matter. Accordingly, the present application is not limited to the implementations described herein, but is intended to be defined by the claims set forth below, and equivalents thereof.

Claims

1. A whole vehicle semi-active suspension fuzzy coordination fault-tolerant control method, characterized in that, The method comprises the following steps: S1: establishing a seven-degree-of-freedom vehicle semi-active suspension dynamics model containing actuator faults; S2: design The robust controller, as a base controller, generates a basic control force of the suspension system; S3: designing a fault observer to detect and estimate the actuator faults in real time, and ensuring that the suspension actuator has fault-tolerant control capability after the fault through self-compensation control; S4: learning and training a complementary compensation controller through the interaction between an agent and an environment based on a double-delay deep deterministic policy gradient (TD3) algorithm, so that the complementary compensation controller generates a complementary compensation control force borne by a normal actuator when any actuator fails; S5: designing a gain fault factor and a pre-output factor, and constructing a two-dimensional extension set; S6: dividing an extension state domain and calculating an associated function, constructing a coordinated compensation control strategy, and adaptively selecting a fault-tolerant control strategy according to a current system fault state; S7: calculating and outputting a final fault-tolerant control force according to the selected fault-tolerant control strategy.

2. The method of claim 1, wherein, The seven-degree-of-freedom vehicle semi-active suspension dynamics model in S1 comprises body vertical, pitch and roll movements, and vertical movements of four non-sprung masses, and when the actuator fails, the state space expression is: ; where X is the state vector, W is the disturbance input, U is the control input, W is the disturbance input, Y is the measurement output, A is a 14x14 matrix; is a 14x4 matrix; is a 14x4 matrix; is a 11x14 matrix; is a 11x4 matrix; is a 7x14 matrix; .

3. The method of claim 1, wherein, The fault observer design in S3 comprises: S31: constructing a state observer equation: ； wherein is an estimate of the state is an estimate of Y; ; L is an observer gain matrix;​ S32: design a fault estimation algorithm ; wherein, is the fault learning rate; S33: Design the observer gain matrix L and the fault learning rate by solving a linear matrix inequality such that the state estimation error and the fault estimation error converge asymptotically.

4. The method of claim 1, wherein, The complementary compensation controller design based on the TD3 algorithm in S4 comprises: S41: defining a state space: including suspension relative displacement, tire dynamic deformation and body movement state; S42: defining an action space: a compensation control force of a normal actuator; S43: designing a reward and punishment function: including comfort reward and punishment, steering stability and safety reward and punishment, and segmented auxiliary reward and punishment; S44: training the agent to learn an optimal complementary compensation control strategy through interaction with the environment.

5. The method of claim 1, wherein, The gain fault factor α and the pre-output factor β in S5 are designed as follows: wherein is the actuator gain coefficient estimated by the fault observer; Pre-output factor The pre-output factor of the fault actuator is obtained by characterizing the actuator output state under different road inputs, obtaining the road excitation information in advance by the sensing sensor, pre-inputting the road excitation information into the H2 / H∞ basic controller, pre-calculating the pre-output force of each actuator, and performing normalization operation .

6. The method of claim 1, wherein, The extension state domain in S6 is divided into three regions: The classic domain: the region of α + |β| ≤ 1, the part of the actuator pre-output that does not exceed the output range after the fault is taken as the classic domain, and the self-compensation control strategy is adopted; The extension domain: the region of α + |β| > 1 and α < 1, the part of the actuator pre-output that exceeds the performance range after the fault is taken as the extension domain, and the coordinated compensation strategy is adopted; The non-domain: the region of α ≥ 1, the part of the actuator adjustable damping that is completely invalid is taken as the non-domain, and the complementary compensation control strategy is adopted.

7. The method of claim 6, wherein, The control force calculation of the coordinated compensation strategy is as follows: ； wherein is a self-compensating control force, is a mutual-compensating control force, and is a coordination weight coefficient based on a correlation function.

8. The method of claim 1, wherein, The H2 / H∞ robust controller in S2 is obtained by solving a linear matrix inequality, which meets the requirements of H2 and H∞ performance indicators and ensures the basic control performance of the system under fault-free conditions.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the method in any one of claims 1-8.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the method in any one of claims 1-8.

Images