Fault diagnosis and fault-tolerant method for vehicle electronic stability system based on analytical redundancy

Abstract

The application discloses a vehicle electronic stability system fault diagnosis and fault-tolerant method based on analytic redundancy, and comprises the following steps: S1, acquiring vehicle state parameters by using related sensors; S2, fusing and estimating a mass center side slip angle of the vehicle by using a Kalman filtering algorithm and a direct integration method to obtain a mass center side slip angle fusion estimation result; S3, establishing a sensor redundancy observer according to the mass center side slip angle fusion estimation result and in combination with the vehicle state parameters; and S4, establishing a fault diagnosis and fault diagnosis compensation mechanism under different driving states of the vehicle based on the sensor redundancy observer. The application can accurately detect and analyze yaw rate and lateral acceleration sensor faults, and can provide available sensor signals for an ESP system under the condition of sensor faults, so that the functional safety performance of the ESP system of the intelligent electric vehicle is improved.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and more specifically to a fault diagnosis and fault-tolerance method for automotive electronic stability systems based on analytical redundancy. Background Technology

[0002] Electronic Stability Program (ESP), a typical automotive electronic control unit, effectively improves the handling stability of intelligent electric vehicles, but it also faces the critical issue of safe and reliable operation. ESP controls vehicle stability through a closed-loop sensor signal system; if a sensor signal fails to acquire the signal, the vehicle cannot return to a stable driving state. Therefore, fault diagnosis and fault-tolerance methods for ESP sensors are crucial for the driving safety of intelligent electric vehicles.

[0003] Currently, deep learning-based methods have emerged as a novel approach for sensor fault diagnosis in recent years. However, these methods require the analysis of large amounts of experimental data, resulting in high complexity. Fault diagnosis based on a specific mathematical model of the system, comparing information obtained from the model with the actual process state, is a way to simplify complexity. However, fault diagnosis and fault tolerance based on a single model cannot improve the accuracy of fault detection or the redundancy of the sensor system, which also increases the risk of vehicle instability.

[0004] Therefore, how to provide a fault diagnosis and fault tolerance method for automotive electronic stability systems based on analytical redundancy is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a fault diagnosis and fault tolerance method for automotive electronic stability systems based on analytical redundancy, which can provide usable sensor signals for the ESP system in the event of sensor failure, thereby improving the functional safety performance of the ESP system in intelligent electric vehicles.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A fault diagnosis and fault-tolerance method for automotive electronic stability systems based on analytical redundancy includes the following steps:

[0008] S1. Obtain vehicle status parameters using relevant sensors;

[0009] S2. The Kalman filter algorithm and the direct integration method are used to perform a fusion estimation of the vehicle's center of gravity sideslip angle to obtain the fusion estimation result of the center of gravity sideslip angle.

[0010] S3. Based on the centroid sideslip angle fusion estimation result and the vehicle state parameters, establish a sensor redundancy observer;

[0011] S4. Based on the aforementioned sensor redundancy observer, establish a fault diagnosis and fault tolerance compensation mechanism for different vehicle driving states.

[0012] Furthermore, step S2 specifically includes:

[0013] S21. Based on the Kalman filter algorithm of the linear two-degree-of-freedom vehicle dynamics model, establish a centroid sideslip angle observer and obtain the centroid sideslip angle observation value.

[0014] S22. Establish the kinematic relationship related to the center of gravity sideslip angle using the direct integration method, and use integration to estimate the current center of gravity sideslip angle of the vehicle to obtain the center of gravity sideslip angle estimation result.

[0015] S23. The centroid sideslip angle observation and the centroid sideslip angle estimation result are fused using a first-order low-pass filtering algorithm to obtain the centroid sideslip angle fusion estimation result.

[0016] Furthermore, step S21 specifically includes:

[0017] Assuming the vehicle's longitudinal speed remains constant, a two-degree-of-freedom model incorporating lateral and yaw forces is established, and the vehicle dynamics equations are as follows:

[0018]

[0019] In the formula, m is the total mass of the vehicle, and v y C represents the lateral speed of the vehicle. αf v is the lateral stiffness of the vehicle's front wheels. x l represents the longitudinal speed of the vehicle. f This is the distance from the vehicle's center of gravity to the front axle. C is the yaw angle of the vehicle. αr For the rear wheel lateral stiffness of the vehicle, l r I is the distance from the vehicle's center of gravity to the rear axle. z Let δ be the yaw moment of inertia of the vehicle. f The steering angle of the vehicle's front wheels;

[0020] Transforming equation (1), we obtain the vehicle dynamics equation with the sideslip angle and yaw rate as state variables:

[0021]

[0022] In the formula, β is the sideslip angle of the vehicle's center of gravity, and ω is the yaw rate of the vehicle.

[0023] Furthermore, the calculation formula for step S22 includes:

[0024]

[0025] In the formula, a is the current estimated value of the centroid sideslip angle. y For lateral acceleration, β kin This is the estimated value of the centroid sideslip angle at the next moment, obtained based on the direct integration method.

[0026] Furthermore, the formula for fusing the centroid sideslip angle estimation results using a first-order low-pass filtering algorithm in step S23 is as follows:

[0027]

[0028] In the formula, β obs The result of the centroid sideslip angle fusion estimation is τ(a) y ) represents the adaptive filter parameters, s kf For the fusion weight coefficients based on the centroid side-slip angle of the Kalman filter, s kin β is the fusion weighting coefficient for the centroid sideslip angle based on the direct integration method. kf The observed centroid sideslip angle is obtained based on Kalman filtering.

[0029] Furthermore, step S3 specifically includes:

[0030] S31. By combining various vehicle motion states and overall vehicle dynamics characteristics, a yaw rate redundancy observer is established, and the trigger probability of the yaw rate redundancy observer is determined.

[0031] S32. Combining the yaw rate redundancy observer and the centroid sideslip angle fusion estimation results, establish a lateral acceleration redundancy observer and determine the trigger probability of the lateral acceleration redundancy observer.

[0032] Furthermore, step S4 specifically includes:

[0033] S41. Based on the sensor redundancy observer, calculate the estimated residual values ​​of the vehicle under different driving conditions using the redundant observation values ​​and the actual measurement values ​​of the sensor.

[0034] S42. Establish a fault diagnosis threshold. Combined with the estimated residual value, when the estimated residual value is greater than the threshold, determine that the sensor is faulty.

[0035] S43. Based on the sensor fault determination results under different vehicle driving conditions, establish a corresponding fault diagnosis and compensation mechanism.

[0036] Furthermore, in step S43, the process of establishing the fault diagnosis and compensation mechanism is as follows:

[0037] A fault exit time constant t is preset. cThe timing begins from the point where the estimated residual at the previous time step is greater than or equal to the threshold, and the estimated residual at the current time step is less than the threshold. If at time t... c If the estimated residuals are all less than the threshold within the time period, the sensor is determined to be not faulty.

[0038] As can be seen from the above technical solution, compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. The redundant observer design is based on the Kalman filter observer, the linear two-degree-of-freedom vehicle model, the simplified vehicle model, and the model that considers the front and rear wheel steering angles and longitudinal vehicle speed, which fully reflects the overall vehicle dynamics characteristics exhibited by the intelligent electric vehicle during driving.

[0040] 2. A first-order low-pass filtering algorithm is designed to fuse the centroid sideslip angle estimation results based on the Kalman filter algorithm and the direct integration method. This can solve the cumulative error caused by noise interference or sensor failure and effectively improve the centroid sideslip angle estimation accuracy.

[0041] 3. Design the estimation residuals, trigger probability functions, and sensor fault tolerance values ​​of redundant observers under different vehicle driving conditions to provide accurate judgment basis for sensor fault diagnosis and fault tolerance algorithms. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0043] Figure 1 This is a schematic diagram of the ESP sensor fault diagnosis and fault tolerance method based on analytical redundancy provided by the present invention.

[0044] Figure 2 The present invention provides a linear two-degree-of-freedom vehicle dynamics model.

[0045] Figure 3 The experimental results of the serpentine trajectory following fault injection provided by this invention.

[0046] Figure 4 The experimental results of sinusoidal hysteresis fault injection provided by this invention.

[0047] Figure 5 The experimental results of fault injection for dual-track trajectory following provided by this invention.

[0048] Figure 6 The experimental results of the steering wheel step fault injection provided by this invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] This invention discloses a fault diagnosis and fault tolerance method for automotive electronic stability systems based on analytical redundancy, such as... Figure 1 As shown, it includes the following steps:

[0051] S1. Use relevant sensors to obtain vehicle state parameters, including longitudinal vehicle speed, road surface adhesion coefficient, vehicle front wheel steering angle and rear wheel steering angle, etc.

[0052] S2. The Kalman filter algorithm and the direct integration method are used to perform a fusion estimation of the vehicle's center of gravity sideslip angle to obtain the fusion estimation result of the center of gravity sideslip angle.

[0053] To achieve fault diagnosis and fault tolerance for ESP sensors in intelligent electric vehicles, the vehicle's center of gravity sideslip angle needs to be estimated before establishing a sensor redundancy observer. This embodiment applies a fusion estimation algorithm for the center of gravity sideslip angle based on a Kalman filter algorithm and a direct integration method, as detailed below:

[0054] S21. Based on the Kalman filter algorithm of the linear two-degree-of-freedom vehicle dynamics model, establish a centroid sideslip angle observer to obtain the centroid sideslip angle observation value.

[0055] Assuming the vehicle's longitudinal speed remains constant and neglecting the slip ratio between the tires and the road surface, establish a two-degree-of-freedom model that includes lateral and yaw movements, such as... Figure 2 As shown, the vehicle dynamics equations can be obtained as follows:

[0056]

[0057] In the formula, m is the total mass of the vehicle; v y C represents the lateral speed of the vehicle. αf For the lateral stiffness of the vehicle's front wheels; v x The longitudinal speed of the vehicle; l f This is the distance from the vehicle's center of gravity to the front axle. C is the yaw angle of the vehicle. αr For the rear wheel lateral stiffness of the vehicle; l r I is the distance from the vehicle's center of gravity to the rear axle. z Let δ be the yaw moment of inertia of the vehicle. fLet cos(δ) be the front wheel steering angle of the vehicle. Considering that the front wheel steering angle is relatively small, we can approximate it as cos(δ). f ) = 1.

[0058] Transforming equation (1), we obtain the vehicle dynamics equation with the sideslip angle and yaw rate as state variables:

[0059]

[0060] In the formula, β is the sideslip angle of the vehicle's center of gravity; ω is the yaw rate of the vehicle, and

[0061] The state equations and measurement equations of a Kalman filter observer based on a two-degree-of-freedom vehicle dynamics model are known to be:

[0062]

[0063] in:

[0064]

[0065]

[0066]

[0067]

[0068] In the formula; x k Let x be the system state at time k; A is the system state transition matrix; x k-1 Let B be the system state at time k-1; B is the system control matrix; u k-1 w is the input variable of the system at time k-1. k Input white noise into the system; y k Let be the system output variable at time k; C be the system observation matrix; D be the system feedforward matrix; u k v is the input variable of the system at time k; k To measure noise in the system. In Kalman filtering, w k and v k Gaussian white noise with zero mean and variances of Q and R.

[0069] The time update equation is obtained by predicting the prior estimate of the current time step from the previous time step state of the Kalman filter observer:

[0070]

[0071] In the formula, To derive the prior estimate of the system at time k from time k-1, P k|k-1To calculate the system error state covariance at time k from time k-1, P k-1 Let Q be the system error state covariance at time k-1. k-1 The system noise matrix at time k-1;

[0072] The prior estimate is corrected by the current measurement value of the Kalman filter observer to obtain the optimal estimate value of the Kalman filter observer at the current time. The measurement update equation is as follows:

[0073]

[0074] In the formula, P k Let I be the system error state covariance at time k, and let K be the identity matrix. k Let R be the Kalman filter gain matrix at time k. k Let be the measurement noise matrix at time k.

[0075] The Kalman filter parameters Q and R are set as follows:

[0076]

[0077] S22. Establish the kinematic relationship between yaw rate, lateral acceleration, longitudinal acceleration, and sideslip angle using the direct integration method, and estimate the current sideslip angle of the vehicle using integration to obtain the sideslip angle estimation result. The calculation formula is:

[0078]

[0079] In the formula, a is the current estimated value of the centroid sideslip angle. y For lateral acceleration, β kin This is the estimated value of the centroid sideslip angle at the next moment, obtained based on the direct integration method.

[0080] S23. The observed centroid sideslip angle and the estimated centroid sideslip angle are fused using a first-order low-pass filtering algorithm to obtain the fused centroid sideslip angle estimate. The calculation formula is as follows:

[0081]

[0082] In the formula, β obs The result of the centroid sideslip angle fusion estimation is τ(a) y ) represents the adaptive filter parameters, s kf For the fusion weight coefficients based on the centroid side-slip angle of the Kalman filter, s kin β is the fusion weighting coefficient for the centroid sideslip angle based on the direct integration method. kf The observed centroid sideslip angle is obtained based on Kalman filtering.

[0083] S3. Based on the centroid sideslip angle fusion estimation results and combined with vehicle state parameters, establish a sensor redundancy observer, including the following implementation steps:

[0084] S31. By combining various vehicle motion states and overall vehicle dynamics characteristics, a yaw rate redundancy observer is established, and the trigger probability of the yaw rate redundancy observer is determined.

[0085] S32. Combining the yaw rate redundancy observer and the centroid sideslip angle fusion estimation results, establish a lateral acceleration redundancy observer and determine the trigger probability of the lateral acceleration redundancy observer.

[0086] Specifically:

[0087] (1) Sensor Redundancy Observer 1

[0088] The yaw rate observation ω obtained based on the Kalman filter algorithm kf As a redundant observation value of yaw rate ω obs1 :

[0089] ω obs1 =ω kf (8)

[0090] The trigger probability P(g) of the yaw rate redundant observer 1 ω1 ) is represented as:

[0091]

[0092] In the formula, μ is the road surface adhesion coefficient; μ th δ is the threshold value for road surface adhesion coefficient; fth The front wheel steering angle threshold; a x For longitudinal acceleration; a xth λ is the longitudinal acceleration threshold. 11 , λ 12 and λ 13 They are respectively the corresponding μ th δ fth and a xth The weighting factor, and λ 11 +λ 12 +λ 13 =1.

[0093] When the sideslip angle of the car's center of gravity is small, its lateral acceleration a y Represented as:

[0094]

[0095] Combined with redundant observations of yaw rate ω obs1 And the estimated value of the centroid sideslip angle β obsEstablish redundant observations a from the lateral acceleration sensor. yobs1 :

[0096] a yobs1 =v x ω yobs1 +(v x β obs )′ (11)

[0097] The trigger probability P(g) of the lateral acceleration redundant observer 1 ay1 ) is represented as:

[0098]

[0099] In the formula, β th Centroid sideslip angle threshold; η 11 η 12 η 13 and η 14 They are respectively the corresponding μ th δ fth a xth and β th The weighting factor, and η 11 +η 12 +η 13 +η 14 =1.

[0100] (2) Sensor Redundancy Observer 2

[0101] Referring to the linear two-degree-of-freedom vehicle model under steady-state conditions, the yaw rate is a constant under steady-state conditions. and All are zero, and the resulting steady-state yaw rate ω obs2 As redundant observations of yaw rate:

[0102]

[0103] In the formula, K is the stability factor, and K = m(l f / Car-lr / Ca f ) / L 2 L is the wheelbase of the vehicle, and L = l f +l r .

[0104] The condition for the steady-state linear two-degree-of-freedom vehicle model to hold is met, and the trigger probability of the yaw rate redundant observer 2 is P(g ω2 )=P(g ω1 ).

[0105] Combined with redundant observations of yaw rate ω obs2 And the estimated value of the centroid sideslip angle β obsEstablish redundant observations a from the lateral acceleration sensor. ybos2 :

[0106]

[0107] The condition for the steady-state linear two-degree-of-freedom vehicle model to hold is met, and the trigger probability of the yaw rate redundant observer 2 is P(g ay2 )=P(g ay1 ).

[0108] (3) Sensor Redundancy Observer 3

[0109] Considering only the simplified front axle state of the vehicle model, the calculated yaw rate ω obs3 As redundant observations of yaw rate:

[0110]

[0111] In the formula, B l The wheelbase of a car; u fr The speed of the car's left front wheel; u fl This refers to the wheel speed of the car's right front wheel.

[0112] When the front wheel steering angle is small, the slip rate between the two front wheels and the ground is small, and the wheel speed sensors of the two front wheels are not faulty, the trigger probability P(g) of the yaw rate redundant observer 3 is... ω3 )for:

[0113]

[0114] In the formula, s fr The slip ratio of the right front wheel of the car; s fl λ represents the slip ratio of the left front wheel of the car. 31 , λ 32 and λ 33 λ is the weighting factor corresponding to the threshold. 31 +λ 32 +λ 33 =1.

[0115] Combined with redundant observations of yaw rate ω obs3 And the estimated value of the centroid sideslip angle β obs Establish redundant observations a from the lateral acceleration sensor. yobs3 :

[0116]

[0117] Similarly, if the front wheel steering angle is small, the slip ratio between the two front wheels and the ground is small, and the wheel speed sensors of the two front wheels are not faulty, then the trigger probability P(g) of the lateral acceleration redundant observer 3 is... ay3 )for:

[0118]

[0119] In the formula, s rr The slip ratio of the right rear wheel of the car; s rl η is the slip ratio of the left rear wheel of the car. 31 η 32 η 33 and η 34 η is the weighting factor corresponding to the threshold, and η 31 +η 32 +η 33 +η 34 =1.

[0120] (4) Sensor Redundancy Observer 4

[0121] Considering only the simplified rear axle state of the vehicle model, the calculated yaw rate value ω obs4 As redundant observations of yaw rate:

[0122]

[0123] If the slip rate between the two rear wheels of the car and the ground is small, and the wheel speed sensors of the two rear wheels are not faulty, then the trigger probability P(g) of the yaw rate redundant observer 4 is... ω4 )for:

[0124] P(g ω4 )=λ 41 (1-|s rr |)+λ 42 (1-|s rl |) (20)

[0125] In the formula, λ 41 and λ 42 λ is the weighting factor corresponding to the threshold. 41 +λ 42 =1.

[0126] Combined with redundant observations of yaw rate ω obs4 And the estimated value of the centroid sideslip angle β obs Establish redundant observations a from the lateral acceleration sensor. yobs4 :

[0127]

[0128] Similarly, given that the slip ratio between the two rear wheels of the car and the ground is small, and the wheel speed sensors of the two rear wheels are not faulty, the trigger probability P(g) of the lateral acceleration redundant observer 4 is... ay4 )for:

[0129]

[0130] In the formula, η 41 η 42 and η 43 η is the weighting factor corresponding to the threshold, and η 41 +η 41 +η 43 =1.

[0131] (5) Sensor Redundancy Observer 5

[0132] Considering that the vehicle's tilt and pitch during driving affect the operation of the ESP system, and given that this embodiment uses a front-wheel steering structure, the rear wheel steering angle is ignored. Instead, the front wheel steering angle and longitudinal vehicle speed are introduced as inputs. The calculated yaw rate value ω is then used... obs5 As redundant observations of yaw rate:

[0133]

[0134] If the front wheel steering angle of the car is small and the steering wheel angle sensor is not faulty, then the trigger probability P(g) of the yaw rate redundant observer 5 is... ω5 )for:

[0135] P(g ω5 )=1-|δ f | / δ fth (twenty four)

[0136] Combined with redundant observations of yaw rate ω obs5 And the estimated value of the centroid sideslip angle β obs Establish redundant observations a from the lateral acceleration sensor. yobs5 :

[0137]

[0138] The trigger probability P(g) of the lateral acceleration redundant observer 5 ay5 )for:

[0139]

[0140] In the formula, η 51 and η 52 η is the weighting factor corresponding to the threshold, and η 51 +η 52 =1.

[0141] S4. Based on the aforementioned sensor redundancy observer, establish a fault diagnosis and fault tolerance compensation mechanism for different vehicle driving states, including the following implementation steps:

[0142] S41. Based on the sensor redundancy observer, calculate the estimated residual values ​​of the vehicle under different driving conditions using the redundant observation values ​​and the actual measurement values ​​of the sensor.

[0143] S42. Establish a fault diagnosis threshold. Combined with the estimated residual value, when the estimated residual value is greater than the threshold, determine that the sensor is faulty.

[0144] S43. Based on the sensor fault determination results under different vehicle driving conditions, establish a corresponding fault diagnosis and compensation mechanism: preset a fault exit time constant t. c The timing begins from the point where the estimated residual at the previous time step is greater than or equal to the threshold, and the estimated residual at the current time step is less than the threshold. If at time t... c If the estimated residuals are all less than the threshold within the time period, the sensor is determined to be not faulty.

[0145] Specifically, this embodiment, based on the sensor redundancy observer, constructs a residual signal using the difference between redundant observations and actual sensor measurements, and estimates the residual based on the current vehicle driving conditions. The residual refers to the actual sensor measurement value y. out Compared with observed values The difference is denoted by e. For fault diagnosis under different vehicle driving conditions, corresponding estimated residuals are designed. This embodiment classifies the vehicle driving state CS into three categories: stable state SB, substable state NSB, and unstable state USB. The specific classification criteria for the three vehicle driving states and the method for generating the estimated residuals are as follows:

[0146] (1) Steady state SB: All sensor redundant observers are effective and the influence of uncertainties is small, that is, the number of effective sensor redundant observers k = 5. At this time, CS∈SB, CS∈E1∩E2∩…∩E n E n This is the set of vehicle driving states that makes the nth redundant observer effective. Under steady-state conditions, the uncertainty of the redundant sensor observers has a relatively small impact on the residuals. Furthermore, to avoid the influence of signal noise, the average of all the maximum and second largest residuals is taken as the estimated residual for fault diagnosis.

[0147]

[0148] In the formula, e k1 e represents the maximum residual value. k2 The second largest residual value; y out This is the actual measured value from the sensor; This is the maximum observed value of the sensor; This is the second largest observation from the sensor; e vis the set of residual values; n is the nth effective redundant observer; v is the set of effective redundant observers; k1 is the redundant observer that makes the sensor reach the maximum actual measurement value; k2 is the redundant observer that makes the sensor reach the second largest actual measurement value.

[0149] (2) Substeady state NSB: Most sensor redundant observers are effective and the influence of uncertain factors is limited, that is, the number of effective sensor redundant observers k = 2, 3, 4.

[0150] At this point, CS∈NSB, assuming j sensors with redundant observers are effective, then the condition is satisfied: CS∈E1∩E2∩…∩E k and In the substeady state, the uncertainty of the redundant observer of the sensor has a certain impact on the residual, but the impact is limited. It can be controlled within an acceptable range by the averaging method and used for the estimation of residuals for fault diagnosis.

[0151] The formula for calculating the mean of sensor redundancy observers is:

[0152]

[0153] In the formula, The mean of the sensor redundancy observers; For the observations of the sensor redundancy observer; y (k+1)out This is the (k+1)th actual measured value of the sensor.

[0154] The difference between redundant observations and their mean is calculated as follows: i = 1, 2, ..., k+1, and choose i1, i2, i3 such that for And j≠i1, i2, i3, the condition for this to hold is:

[0155]

[0156] In the formula, y jout , These are sensor measurements from different sources.

[0157] The estimated residual obtained from the design is:

[0158]

[0159] (3) Unstable state USB: A small number of sensor redundant observers are effective or have a large impact from uncertain factors, i.e., the number of effective sensor redundant observers k = 1.

[0160] In this case, CS∈USB. Under unstable conditions, the number of effective redundant observers is small or the uncertainty of the observers has a significant impact. Therefore, the minimum residual value is directly selected as the estimated residual for fault diagnosis.

[0161]

[0162] In the formula, e j Let be the residual value of the sensor under condition j.

[0163] Based on the estimated residuals obtained from the analytical redundant observer, a fault diagnosis threshold needs to be determined. When the estimated residuals exceed the set threshold, i.e., e > R... th If the sensor value becomes zero for a certain period during a poor vehicle operating condition or a constant gain fault, the estimated residual may fluctuate around the threshold, reducing the accuracy of sensor fault detection. To address this issue, this embodiment adds a fault exit time constant t. c That is, timing begins when the residual at the previous time step is greater than or equal to the threshold and the residual at the current time step is less than the threshold. If at time t c If the residual is less than the threshold within the time period, the sensor is determined to be not faulty. Adding the fault flag bit after fault exit time constant correction and the fault tolerance value can reduce the jitter of the sensor signal.

[0164] Based on the different states of the reference vehicle and the varying true values ​​of the sensors, a fault-tolerant algorithm was designed that includes stable, substable, and unstable states.

[0165] (1) Steady state: CS∈SB

[0166] The mean of the redundant observations of the sensor was calculated:

[0167]

[0168] By calculating the difference between the sensor's redundant observations and the mean... The maximum error is selected as:

[0169]

[0170] In the formula, Let k0 be the observations of the sensor with redundant observers, k0∈[1,n]. For ease of calculation, without loss of generality, we assume that k0=n, that is, we assume that the nth redundant observation has the largest error with the mean of all redundant observers, where n=5.

[0171] Remove the largest error and then weight and fuse the remaining signals to obtain the true sensor value:

[0172]

[0173] In the formula, p represents the sensor's true value. i The weights of the redundant observers, and

[0174] Where, p i The calculation depends on the trigger probability P(g) of each redundant observer. ωi ) or P(g ayi ):

[0175]

[0176] (2) Substable state: CS∈NSB

[0177] At this point, the weighted fusion of effective redundant observations is directly used as the true sensor value:

[0178]

[0179] In the formula, the number of redundant observers k = 2, 3, 4.

[0180] Where, p i The calculation also depends on the trigger probability P(g) of each redundant observer. ωi ) or P(g ayi ):

[0181]

[0182] (3) Unstable state: CS∈USB

[0183] At this point, the number of effective redundant observers is very small, only one. This redundant observation is directly used as the sensor's true value, and the trigger probability P(g) of this effective observer is selected. i To measure the accuracy and reliability of a signal, assuming only the i-th redundant observer is effective, the true sensor value can be expressed as:

[0184]

[0185] When the probability of triggering the i-th redundant observer is P(g) i When the value of the redundant observer is ≥0.95, it is considered that the redundant observer can replace the true value of the sensor; otherwise, when the sensor fails and the effective redundant observer trigger probability P(g) is less than 0.95, it is considered that the redundant observer can replace the true value of the sensor. i If the value is less than 0.95, redundant observation signals cannot be provided for the sensor, and the ESP instability judgment and stability control functions related to the signal will fail.

[0186] This embodiment will be verified and illustrated using a specific case:

[0187] A co-simulation platform based on MATLAB / Simulink and the vehicle dynamics software CarSim was built to test the fault diagnosis and fault tolerance algorithms of this embodiment.

[0188] The first test condition involved an initial vehicle speed of 72 km / h and a serpentine trajectory following test with a road adhesion coefficient of 0.4. A constant gain fault was injected into the yaw rate sensor between 6 and 10 seconds, with a gain coefficient of 1.9. The experimental results are as follows: Figure 3 As shown in (a), (b), (c) and (d).

[0189] from Figure 3 As can be seen, due to the low road surface adhesion coefficient, the actual yaw rate fluctuates significantly relative to the expected yaw rate. Even after a constant-gain fault occurs, the expected yaw rate can still be tracked through analytical redundancy. Between 6.9s and 7.4s and between 8.6s and 9s, the yaw rate residual is less than the set threshold for a short period. By introducing a fault exit time constant, it can be seen that the corrected fault flag effectively reduces the fluctuation of the yaw rate, and the fault diagnosis algorithm can quickly complete the constant-gain fault detection response.

[0190] The second test condition involved an initial vehicle speed of 60 km / h and a road surface adhesion coefficient of 0.8. The accelerator pedal was depressed from 0 to 4 seconds, accelerating to 80 km / h and then maintaining a constant speed. At 5 seconds, a sinusoidal hysteresis signal with a steering wheel angle peak of 4A was used as input. From 6.2 to 8.2 seconds, a constant gain fault was injected into the yaw rate sensor, with a gain coefficient of 1.5. The experimental results are as follows... Figure 4 As shown in (a), (b), (c) and (d).

[0191] from Figure 4 As can be seen, when a fault occurs, the established fault diagnosis and fault-tolerance algorithm can quickly complete the constant gain fault detection response and effectively detect the fault of the yaw rate sensor. While achieving the desired yaw rate tracking, it also improves the tracking accuracy of the sensor's fault-tolerance value.

[0192] The third working condition was a double-track following condition with an initial speed of 80 km / h and a road surface adhesion coefficient of 0.8. A constant deviation fault was injected into the yaw rate sensor between 6 and 10 seconds, with an error value of 0.1 rad / s. The experimental results are as follows: Figure 5 As shown in (a), (b), (c) and (d).

[0193] from Figure 5 As can be seen, after a fault occurs, the yaw rate tolerance value generated by multiple redundant observers can also accurately follow its true value, and can quickly identify the sensor constant deviation fault and correct the fault position in a short time.

[0194] The fourth test case involves a hardware-in-the-loop test of injecting a steering wheel step fault. Under a steering wheel angle step input condition with an initial speed of 80 km / h and a road surface adhesion coefficient of 0.8, a constant deviation fault was injected into the lateral acceleration sensor from 2 to 5 seconds, with a deviation value of 1 m / s². 2 The experimental results are as follows Figure 6 As shown in (a), (b), (c) and (d).

[0195] from Figure 6 As can be seen, thanks to the design of multiple redundant observers, the generated lateral acceleration fault-tolerant value can also accurately follow the real sensor signal.

[0196] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0197] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fault diagnosis and fault-tolerance method for automotive electronic stability systems based on analytical redundancy, characterized in that, Includes the following steps: S1. Obtain vehicle status parameters using relevant sensors; S2. The Kalman filter algorithm and the direct integration method are used to perform a fusion estimation of the vehicle's center of gravity sideslip angle to obtain the fusion estimation result of the center of gravity sideslip angle. S3. Based on the centroid sideslip angle fusion estimation result and the vehicle state parameters, establish a sensor redundancy observer; S4. Based on the aforementioned sensor redundancy observer, establish a fault diagnosis and fault diagnosis compensation mechanism for different vehicle driving states. Specifically, step S2 includes: S21. Based on the Kalman filter algorithm of the linear two-degree-of-freedom vehicle dynamics model, establish a centroid sideslip angle observer to obtain the observed centroid sideslip angle values. Specifically, this includes: Assuming the vehicle's longitudinal speed remains constant, a two-degree-of-freedom model incorporating lateral and yaw forces is established, and the vehicle dynamics equations are as follows: (1) In the formula, For the total mass of the vehicle. The lateral speed of the vehicle. For the lateral stiffness of the vehicle's front wheels, The longitudinal speed of the vehicle. This is the distance from the vehicle's center of gravity to the front axle. The yaw angle of the vehicle. For the rear wheel lateral stiffness of the vehicle, This is the distance from the vehicle's center of gravity to the rear axle. Let the vehicle's yaw moment of inertia be... The steering angle of the vehicle's front wheels; Transforming equation (1), we obtain the vehicle dynamics equation with the sideslip angle and yaw rate as state variables: (2) In the formula, The sideslip angle is the angle at the vehicle's center of gravity. The yaw rate of the vehicle; S22. Establish the kinematic relationship related to the sideslip angle of the center of gravity using the direct integration method, and estimate the current sideslip angle of the center of gravity of the vehicle using integration to obtain the estimated results of the sideslip angle of the center of gravity. The calculation formula is: (6) In the formula, This is the current estimated value of the centroid sideslip angle. It is lateral acceleration. This is the estimated value of the centroid sideslip angle at the next moment, obtained based on the direct integration method; S23. The centroid sideslip angle observation values ​​are processed using a first-order low-pass filtering algorithm. and the estimated centroid sideslip angle The centroid sideslip angle fusion estimation results are obtained by performing fusion. The formula is: (7) In the formula, For adaptive filter parameters, These are the fusion weight coefficients based on the centroid side-slip angle of the Kalman filter. The fusion weighting coefficients are based on the centroid side deflection angle using the direct integration method. Step S3 specifically includes: S31. Combining various vehicle motion states and overall vehicle dynamics characteristics, establish a redundant yaw rate observer and determine the trigger probability of the redundant yaw rate observer; the redundant yaw rate observer includes at least five observers based on different principles: an observer based on Kalman filtering. Observer based on steady-state two-degree-of-freedom model Observer based on front wheel speed difference Observer based on rear wheel speed difference and observers based on Ackerman turning geometry ; S32. Combining the yaw rate redundancy observer and the centroid sideslip angle fusion estimation results Establish a lateral acceleration redundant observer and determine the trigger probability of the lateral acceleration redundant observer; Step S4 specifically includes: S41. Based on the sensor redundancy observer, calculate the estimated residual values ​​of the vehicle under different driving states using the redundant observation values ​​and the actual measurement values ​​of the sensor; the vehicle driving states are divided into stable state SB, substable state NSB and unstable state USB. Under steady state SB, the effective number of redundant observers k=5, and the mean of all the maximum and second largest residuals is taken as the estimated residual e: (27) In the formula, This represents the maximum residual value. This is the second largest residual value; This is the actual measured value from the sensor; This is the maximum observed value of the sensor; This is the second largest observation from the sensor; is the set of residual values; n is the nth effective redundant observer; v is the set of effective redundant observers; A redundant observer to enable the sensor to achieve the maximum actual measurement value; To enable the sensor to achieve the second largest actual measurement value, a redundant observer is required; Under the substeady state NSB, the number of effective redundant sensor observers k=2, 3, 4. The difference between the redundant observations and their mean is calculated, and the residuals of a specific order are selected as the estimated residuals e. In the unstable USB state, the number of effective redundant observers k=1, and the smallest residual value is selected as the estimated residual e: (31) In the formula, Let be the residual value of the sensor under condition j; This represents the estimate of the sensor output by the j-th redundant observer; S42. Establish a fault diagnosis threshold. Combined with the estimated residual value, when the estimated residual value is greater than the threshold, determine that the sensor is faulty. S43. Based on the sensor fault determination results under different vehicle driving conditions, establish a corresponding fault diagnosis and compensation mechanism. In the steady state SB, the mean of redundant sensor observations is calculated, the observation with the largest error from the mean is removed, and the remaining signals are weighted and fused according to their trigger probabilities to obtain the true sensor value. : (34) In the formula, This is the actual value from the sensor; The weights of the redundant observers, and ; In the substeady state NSB, the effective redundant observations are directly weighted and fused according to their trigger probabilities to obtain the true sensor values. In the unstable USB state, when the trigger probability of the effective redundant observer... If the redundant observation is used as the true value of the sensor, then it is determined that the redundant signal cannot be provided.

2. The method for fault diagnosis and fault tolerance of automotive electronic stability systems based on analytical redundancy according to claim 1, characterized in that, In step S43, the process of establishing the fault diagnosis and compensation mechanism is as follows: Preset a fault exit time constant The timing begins when the estimated residual at the previous time step is greater than or equal to the threshold, while the estimated residual at the current time step is less than the threshold. If... If the estimated residuals are all less than the threshold within the time period, the sensor is determined to be not faulty.

Images