Integrated fault diagnosis and estimation method for active suspension system of electric vehicle

Abstract

The application discloses an integrated fault diagnosis and estimation method for an electric vehicle active suspension system, and comprises the following steps: firstly, collecting vehicle signal inputs required for control; then, estimating a sensor fault vector through a first unknown input observer, and estimating an actuator fault value through a second unknown input observer; for the coupling problem of the two fault estimations, introducing an actuator temporary shutdown mechanism; through monitoring the suspension dynamic stroke sensor fault estimation value, triggering the actuator shutdown mechanism after exceeding a preset threshold value and lasting for a specific time, so as to identify the fault source; finally, generating a fault flag and masking and correcting the estimation value, and outputting accurate fault estimation. The application solves the problem that the actuator and the sensor fault coupling cannot be diagnosed simultaneously, and realizes the synchronous and accurate diagnosis and estimation of various faults of the electric vehicle active suspension system.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology for electric vehicle electronic control systems, and in particular to fault detection and estimation for electric vehicle active suspension systems, specifically an integrated fault diagnosis and estimation method and system for electric vehicle active suspension systems. Background Technology

[0002] Modern electric vehicle active suspension systems deliver active force in real time through actuators, improving vehicle ride comfort and handling stability while also impacting energy recovery efficiency. The system's performance is highly dependent on the precise perception of vehicle status by sensors and the accurate tracking of control commands by the actuators.

[0003] However, as key components, sensors and actuators can experience performance degradation or even sudden malfunctions under complex in-vehicle environments and long-term mechanical and electrical stress. Sensors may experience signal drift, abnormal gain, or jamming, resulting in severely distorted output signals; actuators may experience efficiency decline, output force saturation, or jamming, causing them to be unable to accurately track control commands. These malfunctions reduce the control performance of the active suspension system, affecting not only comfort but also, in extreme cases, potentially causing vehicle instability and seriously threatening driving safety.

[0004] Fault-Tolerant Control (FTC) is a key technology that ensures a system can maintain basic performance and safety after component failure. The effectiveness of FTC largely depends on the accurate and timely fault information provided by the Fault Detection and Diagnosis (FDD) module. This means that it is necessary not only to detect the occurrence of a fault, but also to locate the source of the fault and quantify its magnitude.

[0005] Fault diagnosis methods based on analytical models are currently the mainstream research approach. Among them, observer methods, such as the Unknown Input Observer (UIO), have attracted much attention due to their ability to decouple specific disturbances and directly estimate fault values. Existing research is mostly limited to handling single types of faults; for example, designing a dedicated observer to detect actuator faults or designing another residual generator to detect sensor faults. However, this strategy encounters a fundamental theoretical bottleneck when facing scenarios where both actuators and sensors may fail simultaneously: when fault terms from both actuators and all sensors are introduced into the system model, their number often exceeds the number of independently measured parameters, causing the system to fail to meet the strong detectability condition. This means that no single linear observer can uniquely and unbiasedly estimate all states and faults.

[0006] In particular, in active suspension systems, actuator failures and suspension deformation sensor failures exhibit a strong coupling effect in the system's dynamic response. Their output response characteristics in the time and frequency domains are very similar, making it difficult to distinguish the fault source using residual analysis or direct estimation methods based on a single observer. Misdiagnosis will lead to incorrect compensation behaviors in fault-tolerant control strategies; for example, misjudging an actuator failure as a sensor failure and correcting the sensor signal accordingly. This not only fails to restore system performance but may also introduce additional errors or even amplify the impact of the fault, leading to system instability.

[0007] Existing methods focus on actuator failures in semi-active suspension shock absorbers. Although they have achieved the estimation of additive and multiplicative failures by using an unknown input observer and recursive least squares method respectively, and verified their effectiveness on real vehicle models, they do not consider sensor failures at all. Furthermore, since they do not address the coupling problem between actuator and sensor signals in the suspension system, they cannot effectively identify sensor malfunctions and may even produce false positives (FaultDetection for Automotive Shock Absorber).

[0008] While existing methods can handle multiple faults occurring simultaneously (such as leaks, blockages, sensor and actuator faults), and achieve residual identification and discrete mode isolation of multiple faults in a three-tank system using an adaptive unscented Kalman filter bank, their application is limited to the field of chemical process control. They do not model and process the unique physical coupling characteristics of vehicle active suspension systems (i.e., the strong coupling between actuator faults and displacement sensor faults). Therefore, they cannot solve the problem of difficulty in distinguishing fault sources due to the lack of strong detectability conditions (Fault diagnosis of anonlinear hybrid system using adaptive unscented Kalman filter bank).

[0009] Therefore, there is an urgent need in this field for an integrated solution that can solve fault coupling at the system level and achieve synchronized, accurate diagnosis and estimation of actuator and sensor faults. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system that can effectively solve the problem of actuator and sensor fault coupling. The core of this invention is to adopt a parallel estimation-active discrimination-integrated output architecture to realize the integration of multiple fault synchronization, accurate diagnosis and estimation in the active suspension system of electric vehicles under a unified framework.

[0011] The present invention is achieved by at least one of the following technical solutions.

[0012] An integrated fault diagnosis and estimation method for active suspension systems in electric vehicles includes the following steps: Step S1: Acquire the output signal of the active suspension system; Step S2: Based on the signal acquired in step S1, estimate the sensor fault using the first unknown input observer; Step S3: Based on the signal acquired in step S1, estimate the actuator fault using the second unknown input observer; Step S4: Monitor the estimated value of the suspension dynamic travel sensor fault by the first unknown input observer. When the estimated value exceeds the threshold and continues for a period of time, trigger the actuator shutdown mechanism. Step S5: During the actuator shutdown period, based on the change in the estimated value of the suspension dynamic stroke sensor fault, determine whether the fault originates from the actuator or the sensor, and take corresponding system state switching actions based on the determination result: if the fault is determined to be an actuator fault, generate an actuator fault pre-flag and trigger the system fault-tolerant control mode; if the fault is determined to be a sensor fault, generate a sensor fault pre-flag and restore the actuator to normal operation. Step S6: Generate the final fault diagnosis flag based on the pre-flag signal and the output of each unknown input observer; Step S7: Using the final fault diagnosis flag, integrate and mask the original fault estimation vectors of the first unknown input observer and the second unknown input observer, and output the corrected final fault estimation value.

[0013] Furthermore, the first unknown input observer and the second unknown input observer are based on The theoretical design process includes: extending the system state vector to include the fault to be estimated and its derivative, constructing extended system equations, and determining the observer gain by solving an optimization problem based on linear matrix inequalities, so that the estimation error is asymptotically stable and meets the disturbance attenuation performance index.

[0014] Furthermore, the system state vector is represented as , For suspension travel, For tire dynamic deformation, and These represent the vertical velocities of the sprung mass and the unsprung mass, respectively.

[0015] Furthermore, the first unknown input observer is described by an equation of the following form:

[0016] in, The first unknown input is the internal state of the observer. For extended state The estimated value, For the sensor in The active suspension system continuously collects signals. Let be the observer gain matrix to be determined. For the force of the actuator, This is a system matrix related to the vehicle's physical parameters.

[0017] Furthermore, the second unknown input observer is described by an equation of the following form:

[0018] in, The second unknown input is the internal state of the observer. For extended state The estimated value, For the sensor in The active suspension system continuously collects signals. Let be the observer gain matrix to be determined. For the force of the actuator, This is a system matrix related to the vehicle's physical parameters.

[0019] Furthermore, in step S5, the logic for determining the source of the fault is as follows: if, after the actuator is turned off, the estimated value of the suspension dynamic stroke sensor fault falls below the threshold, then the fault source is determined to be the actuator, and an actuator fault pre-flag is generated. And trigger the system fault-tolerant control mode; if the estimated value of the suspension dynamic travel sensor fault remains above the preset threshold after the actuator is turned off, it is determined that the fault originates from the suspension dynamic travel sensor, and a sensor fault pre-flag is generated. And restore the actuator to normal operation.

[0020] Furthermore, in step S6, the pre-flag signal is smoothed using an exponential moving average filter, and the final Boolean fault flag is generated based on a hysteresis threshold with upper and lower limits.

[0021] Furthermore, in step S7, the integration and masking process is implemented in the following way: the final fault diagnosis flags are used to form a diagonal masking matrix, and the diagonal masking matrix is ​​multiplied by a vector composed of the original fault estimates of the first unknown input observer and the second unknown input observer, thereby filtering and outputting the integrated and corrected fault estimation vector.

[0022] The system for implementing the integrated fault diagnosis and estimation method for active suspension systems of electric vehicles includes: The signal acquisition module is used to acquire suspension dynamic travel, sprung mass vertical acceleration, and unsprung mass vertical acceleration signals. The fault diagnosis system model module is used to characterize the dynamic behavior of the active suspension system through the fault diagnosis system model. The sensor fault observation and estimation module is used to estimate sensor faults through a first unknown input observer. The actuator fault observation and estimation module is used to estimate actuator faults through a second unknown input observer. The actuator switching and fault diagnosis module is used to trigger the actuator shutdown mechanism and identify the source of the fault. An integrated fault flag generation module is used to generate the final fault diagnosis flags; The fault estimation integration and correction module is used to integrate and mask the processing to generate a corrected fault estimation vector.

[0023] A computer device according to the present invention includes a memory and a processor, the memory being electrically connected to the processor, the memory storing a computer program, which, when executed by the processor, causes the processor to implement the method described herein.

[0024] The present invention provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor implements the method described herein.

[0025] Compared with the prior art, the present invention has the following advantages: (1) By using an integrated solution, the fault diagnosis signs and accurate estimates of actuators and all sensors are output synchronously in one system, solving the problem of simultaneous diagnosis of multiple faults.

[0026] (2) By actively switching the actuator, a physical means, combined with logical judgment, coupled faults that are difficult to separate mathematically can be effectively distinguished, thus fundamentally reducing the misdiagnosis rate.

[0027] (3) Automatically switch the system operating mode according to the fault diagnosis results to ensure system safety in the event of component failure.

[0028] (4) Observer based on Theoretical design optimizes disturbance attenuation performance. To address persistent road surface irregularities It has strong suppression capabilities, ensuring the accuracy of fault estimation under real and complex working conditions.

[0029] (5) A stability flag is generated by using EMA filtering and hysteresis threshold, and a clean estimate is output through masking logic. This result can be directly and reliably used for fault compensation and controller reconfiguration of the upper-level fault-tolerant controller.

[0030] (6) This method has relatively low dependence on model accuracy. The main observer is designed based on a linear model, but the model mismatch and nonlinear effects are compensated by an active discrimination mechanism. Attached Figure Description

[0031] Figure 1 This is an architecture diagram of an integrated fault diagnosis and estimation system for an active suspension system in an electric vehicle, as illustrated in the embodiment. Figure 2 This is a flowchart of an embodiment of an integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle; Figure 3 This is a schematic diagram of a linear quarter-vehicle active suspension model; Figure 4 This is a flowchart of the actuator switching and fault diagnosis module in the embodiment; Figure 5 This is a schematic diagram of the fault flag generation logic in an embodiment. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0033] This invention designs two independent, based on Theoretically, robust Unknown Input Observers (UIOs) are constructed in parallel. The first UIO (UIO-SF) is dedicated to estimating the sensor fault vector, assuming the actuator is fault-free at design time; the second UIO (UIO-AF) is dedicated to estimating actuator faults, assuming the sensor is fault-free at design time. This functional separation reduces the dimensionality of the unknown input (fault) for each observer at design time, thus satisfying its respective detectability conditions and obtaining a preliminary estimate of the corresponding fault. However, this parallel estimation architecture introduces a new problem: misestimation under coupled faults. That is, when the actuator actually fails, UIO-SF will be biased due to its invalid model assumptions, incorrectly estimating the sensor fault value; and vice versa.

[0034] To address this, the present invention introduces a second core technology: a mechanism for the temporary shutdown of the actuator and fault detection. This mechanism acts as an intelligent switch, activated when the system detects a potential coupling fault. Specifically, it manifests as the UIO-SF's fault estimate of the suspension dynamic travel sensor. Exceeding a preset threshold and persisting for a specific time. A transient test condition is created by actively and briefly shutting down the actuator: if the limit is exceeded... If the error disappears after shutdown, it indicates that the exception originated from the actuator, because the source of the fault has been removed; if If the fault persists, it indicates that the anomaly originates from the sensor itself. This mechanism leverages the dynamic differences in the system under different operating modes to separate and identify coupled faults. Based on the identification result, the system adopts different state switching strategies: if the fault is determined to be an actuator, the system's fault-tolerant control mode is triggered to isolate or restrict the faulty actuator; if the fault is determined to be a sensor, the actuator is restored to normal operation, and only the faulty sensor signal is processed. Finally, stable fault flags are generated based on the identification results, and these flags are used to construct a dynamic masking matrix to integrate and filter the original estimation outputs of the two parallel observers. Only the estimated value of a fault that is marked as confirmed will be finally output, thereby effectively eliminating false estimates caused by coupling effects and ensuring the accuracy of the final diagnosis and estimation results.

[0035] like Figure 2 As shown in this embodiment, an integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle specifically includes the following steps: S1. Collect signals from the vehicle's active suspension system: In one embodiment, the active suspension system signal includes: suspension travel. Vertical acceleration of the sprung mass Unsprung mass vertical acceleration These signals together constitute the input vector of the fault diagnosis system model. .

[0036] S2. Establish a fault diagnosis system model.

[0037] like Figure 3 As shown in the figure, a discrete-time state-space model for fault diagnostic tool design is established based on a linear quarter-vehicle active suspension model. and For both sprung and unsprung mass, and For the vertical displacement of the sprung mass and the unsprung mass, Powering the actuator and These are the suspension spring stiffness and suspension damping coefficient. and For tire stiffness and tire damping coefficient, For road surface excitation, define the state vector of the active suspension system. .in, Indicates the discrete-time sampling time. For active suspension systems in The state vector at time t, for Suspension travel at any moment for The tire deforms constantly. and They are respectively The vertical velocities of the sprung and unsprung masses at any given time. Control input. for The actuator force at any given moment. The road disturbance input at time t is , for The road surface speed at any given moment is a driving force.

[0038] Considering additive faults, the fault diagnosis system model can be expressed as:

[0039] in, +1 indicates the discrete-time sampling time. For active suspension systems in The state vector at time t, For the sensor in The active suspension system continuously collects signals. for The vertical acceleration of the sprung mass at time t. for The vertical acceleration of the unsprung mass at time t. for Actuator failure at any moment for Sensor fault vector at time step for The suspension travel sensor at all times. for The sprung mass acceleration sensor malfunctioned at any given time. for The unsprung mass acceleration sensor malfunctioned at that moment. To match vehicle physical parameters and sampling time Related system matrices.

[0040] in, , , , It is the identity matrix. Sampling time. , , , , , .

[0041] S3. Design a sensor fault observer (UIO-SF).

[0042] Design a first unknown input observer (UIO-SF) specifically for estimating... Sensor malfunction at any time .

[0043] System extension: The original system state vector is extended to... Accordingly, the extended system equations are constructed:

[0044] in, For UIO-SF in The system state vector at time t. for Time sensor malfunction The derivative of This is the extended system matrix. , , , , .

[0045] Observer dynamics: UIO-SF is described by equations of the following form:

[0046] in, for The internal state of the UIO-SF observer at time t. for Always The estimated value, Let be the observer gain matrix to be determined.

[0047] Gain Calculation and Robust Design: By Defining The estimation error at time is Substituting the observer equations and system equations, the error dynamics are derived.

[0048] definition And take the error at the next time step.

[0049] Substitution and ,Will Decomposed into and substitute Organized

[0050] The observer gain must satisfy the following constraints:

[0051] make Substituting these equations into the discrete system equations and performing a series of algebraic operations, the error dynamics can be simplified to:

[0052] in, To augment the perturbation vector, for Constant road surface disturbances. .

[0053] At this point, the error dynamics consist only of the system matrix, the observer gain, and the perturbation. Decision. In one embodiment, employing Performance metrics, i.e., designing the observer to make To suppress the effects of disturbances. Among them, This is the level of perturbation attenuation to be minimized. This is determined by choosing the Lyapunov function. And using the Linear Matrix Inequality (LMI) tool, the design of the observer gain is transformed into a convex optimization problem:

[0054] Right now Under the constraint that the matrix is ​​symmetric positive definite (i.e., ensuring observer stability), minimize the performance index. .in, Let Lyapunov be the matrix. Solving this problem yields the result that makes the observer stable and satisfies... Gain matrix of performance metrics .

[0055] S4. Design an actuator fault observer (UIO-AF).

[0056] Design a second unknown input observer (UIO-AF) specifically for estimating... Actuator failure at any time Its design process is similar to that of UIO-SF.

[0057] System extension: Extend the state vector to Construct extended system equations:

[0058] in, For UIO-AF in The system state vector at time t. for The derivative of the actuator failure at any given moment, .

[0059] Observer dynamics:

[0060] in, for The internal state of the UIO-AF observer at time t. for Always The estimated value Let be the observer gain matrix to be determined.

[0061] Gain calculation: definition The estimation error at time is Using the same design flow as step S3, solve for the observer gain. And make it meet the performance indicators. .in, It is the level of disturbance attenuation to be minimized. To augment the perturbation vector S5. Actuator switching and fault diagnosis.

[0062] This step resolves the fault coupling problem by switching actuators, and the process is as follows: Figure 4 As shown.

[0063] (S5-1) Monitoring: Continuously monitor the suspension dynamic travel sensor fault estimates output by UIO-SF. .

[0064] (S5-2) Trigger judgment: When Exceeding the preset fault threshold And this over-limit state lasts for a predetermined time window. When a credible fault indication is identified, a trigger signal is generated.

[0065] The time window is used to avoid false triggering caused by a single impact on the road surface or a noise spike.

[0066] (S5-3) Actuator shutdown: The trigger signal causes the system to send a zero-force command to the actuator, which lasts for a short period of time. During this period, the suspension system temporarily switches to passive mode to isolate the effects of the actuators.

[0067] (S5-4) Fault source identification: Continue monitoring during actuator shutdown .

[0068] Scenario A (Actuator Failure): If Rapidly dropped and below the threshold This indicates that the previous abnormal estimate was caused by an actuator malfunction. After the actuator is shut down, the fault source is removed, and the estimated value returns to normal. At this point, an actuator fault pre-flag is generated. and reset This triggers the system's fault-tolerant control mode, limiting or completely isolating the faulty actuator, and simultaneously sending an actuator fault alarm to the upper-level system.

[0069] Scenario B (sensor failure): If It remains at the threshold even after the actuator is turned off. This indicates that the anomaly originates from the suspension dynamic travel sensor itself and is unrelated to the actuator. At this point, a fault pre-signal for the suspension dynamic travel sensor is generated. and reset At this point, the actuator is confirmed to be fault-free, and the controller takes over again, restoring the actuator to normal operation.

[0070] (S5-5) End diagnosis: During the duration After the diagnostic process is completed, the system enters the corresponding state based on the assessment results. If the fault is a sensor, the system resumes normal fully active control. If the fault is an actuator, the system enters a degraded fault-tolerant operating state.

[0071] S6, Fault flag generation.

[0072] This step receives estimates from UIO-SF and UIO-AF, as well as the pre-flag generated in step S5, to generate a stable and reliable final fault flag. The process is as follows: Figure 5 As shown.

[0073] (1) Initial marker generation: For the malfunction of the sprung mass acceleration sensor Unsprung mass acceleration sensor failure They are weakly coupled with actuator faults, and their corresponding UIO-SF estimates can be directly used. and With threshold Comparison, generating preliminary markers and .

[0074] For suspension travel sensor malfunction and actuator failure Their initial flags directly adopt the pre-flags output in step S5. and .

[0075] (2) EMA filtering: For the above four preliminary flag signals Applying Exponential Moving Average (EMA) filtering:

[0076] in, For fault type index, the value is... . The actuator is faulty. The problem is a faulty suspension travel sensor. The problem is a faulty sprung mass acceleration sensor. The fault is due to a non-sprung mass acceleration sensor. For the first Class of faults at any time The initial sign signal, For the first Class of faults at any time The exponential moving average, For the first Class of faults at any time The exponential moving average, For the first Smoothing factor for fault class The smaller the value, the stronger the smoothing effect. EMA can effectively suppress rapid signal jitter and isolated false signals caused by noise.

[0077] (3) Hysteresis threshold decision: Apply hysteresis comparison logic to the filtered EMA value to generate the final Boolean fault flag. : like (Upper threshold), then The fault has been confirmed.

[0078] like (Lower threshold), and ),but Troubleshooting.

[0079] like ,but Keep the value from the previous moment.

[0080] in, for Type of fault at time The final Boolean fault flag. For fault type index, the value is... . for Upper limit threshold for this type of fault for The lower threshold for this type of fault.

[0081] Hysteresis logic effectively prevents frequent flag transitions at the fault occurrence / disappearance boundary, improving the stability of diagnostic results.

[0082] S7, Fault estimation integration and correction.

[0083] The original estimated vectors output by UIO-SF and UIO-AF This includes coupling and potential error estimation. A diagonal masking matrix is ​​constructed using the final fault flags generated in step S6. :

[0084] Final, integrated and corrected fault estimation vector for:

[0085] in, For a moment A diagonal masking matrix constructed using the final fault flag. For a moment The final, integrated and corrected fault estimation vector, For the actuator at time The final Boolean fault flag, For the suspension dynamic travel sensor at any time The final Boolean fault flag, For the sprung mass acceleration sensor at time The final Boolean fault flag, For unsprung mass accelerometers at time The final Boolean fault flag, For UIO-SF and UIO-AF at time The output raw estimates form a vector. For UIO-AF at time Output actuator fault estimate.

[0086] This operation means that the estimated value is only output when a fault is officially marked as occurring; otherwise, the output is zero. This effectively eliminates spurious estimates caused by coupling effects, ensuring the final output... The accuracy.

[0087] like Figure 1 An integrated fault diagnosis and estimation system for an active suspension system of an electric vehicle, shown, includes: Signal acquisition module: used to implement step S1, acquire suspension dynamic travel and sprung / unsprung mass vertical acceleration signals.

[0088] System Model Module: Used to encapsulate the fault diagnosis system model in step S2.

[0089] Sensor Fault Observation and Estimation Module: Used to estimate sensor faults through a first unknown input observer.

[0090] Actuator Fault Observation and Estimation Module: Used to estimate actuator faults through a second unknown input observer.

[0091] Actuator switching and fault diagnosis module: used to implement the active diagnosis logic of step S5.

[0092] Integrated fault flag generation module: used to implement EMA filtering and flag generation in step S6, and generate the final fault diagnosis flag.

[0093] Fault estimation integration and correction module: used to perform masking calculation in step S7 and generate the corrected fault estimation vector.

[0094] These modules can be integrated into one or more electronic control units of the vehicle. Preferably, the system state estimate output by the sensor fault observation and estimation module... It can be provided to the active suspension controller to calculate the control force of the active suspension. Because UIO-SF takes sensor fault compensation into account when estimating the state, its state estimation is more accurate than that of UIO-AF, making it more suitable for feedback control.

[0095] This invention discloses an integrated fault diagnosis and estimation system for an active suspension system in electric vehicles. This system can monitor the operating status of key suspension components in real time and provide accurate fault information when a fault occurs, providing a basis for subsequent alarms, degraded operation, or active fault-tolerant control, thereby improving vehicle reliability and safety.

[0096] This invention effectively solves the coupling and detectability problems in multi-fault diagnosis and estimation of active suspension systems through a parallel estimation-active discrimination-integrated output architecture, and provides a comprehensive, robust and reliable integrated solution.

[0097] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, enabling those skilled in the art to better understand and utilize the invention.

Claims

1. An integrated fault diagnosis and estimation method for active suspension systems of electric vehicles, characterized in that, Includes the following steps: Step S1: Acquire the output signal of the active suspension system; Step S2: Based on the signal acquired in step S1, estimate the sensor fault using the first unknown input observer; Step S3: Based on the signal acquired in step S1, estimate the actuator fault using the second unknown input observer; Step S4: Monitor the estimated value of the suspension dynamic travel sensor fault by the first unknown input observer. When the estimated value exceeds the threshold and continues for a period of time, trigger the actuator shutdown mechanism. Step S5: During the actuator shutdown period, based on the change in the estimated value of the suspension dynamic stroke sensor fault, determine whether the fault originates from the actuator or the sensor, and take corresponding system state switching actions based on the determination result: if the fault is determined to be an actuator fault, generate an actuator fault pre-flag and trigger the system fault-tolerant control mode; if the fault is determined to be a sensor fault, generate a sensor fault pre-flag and restore the actuator to normal operation. Step S6: Generate the final fault diagnosis flag based on the pre-flag signal and the output of each unknown input observer; Step S7: Using the final fault diagnosis flag, integrate and mask the original fault estimation vectors of the first unknown input observer and the second unknown input observer, and output the corrected final fault estimation value.

2. The integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle according to claim 1, characterized in that, The first unknown input observer and the second unknown input observer are based on The theoretical design process includes: extending the system state vector to include the fault to be estimated and its derivative, constructing extended system equations, and determining the observer gain by solving an optimization problem based on linear matrix inequalities, so that the estimation error is asymptotically stable and meets the disturbance attenuation performance index.

3. The integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle according to claim 2, characterized in that, The system state vector is represented as , For suspension travel, For tire dynamic deformation, and These represent the vertical velocities of the sprung mass and the unsprung mass, respectively.

4. The integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle according to claim 1, characterized in that, The first unknown input observer is described by an equation of the following form: in, The first unknown input is the internal state of the observer. For extended state The estimated value, For the sensor in The active suspension system continuously collects signals. Let be the observer gain matrix to be determined. For the force of the actuator, This is a system matrix related to the vehicle's physical parameters.

5. The integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle according to claim 1, characterized in that, The second unknown input observer is described by an equation of the following form: in, The second unknown input is the internal state of the observer. For extended state The estimated value, For the sensor in The active suspension system continuously collects signals. Let be the observer gain matrix to be determined. For the force of the actuator, This is a system matrix related to the vehicle's physical parameters.

6. The integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle according to claim 1, characterized in that, In step S5, the logic for determining the fault source is as follows: if the estimated value of the suspension dynamic stroke sensor fault falls below the threshold after the actuator is turned off, then the fault source is determined to be the actuator, and an actuator fault pre-flag is generated. And trigger the system fault-tolerant control mode; if the estimated value of the suspension dynamic travel sensor fault remains above the preset threshold after the actuator is turned off, it is determined that the fault originates from the suspension dynamic travel sensor, and a sensor fault pre-flag is generated. And restore the actuator to normal operation.

7. The integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle according to claim 1, characterized in that, In step S6, the pre-mark signal is smoothed by exponential moving average filtering, and the final Boolean fault mark is generated based on the hysteresis threshold with upper and lower limits.

8. The integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle according to claim 1 or 6, characterized in that, In step S7, the integration and masking process is implemented as follows: the final fault diagnosis flags are used to form a diagonal masking matrix, and the diagonal masking matrix is ​​multiplied by a vector composed of the original fault estimates from the first unknown input observer and the second unknown input observer, thereby filtering and outputting the integrated and corrected fault estimation vector.

9. A system for implementing the integrated fault diagnosis and estimation method for an active suspension system of an electric vehicle as described in any one of claims 1-8, characterized in that, include: The signal acquisition module is used to acquire suspension dynamic travel, sprung mass vertical acceleration, and unsprung mass vertical acceleration signals. The fault diagnosis system model module is used to characterize the dynamic behavior of the active suspension system through the fault diagnosis system model. The sensor fault observation and estimation module is used to estimate sensor faults through a first unknown input observer. The actuator fault observation and estimation module is used to estimate actuator faults through a second unknown input observer. The actuator switching and fault diagnosis module is used to trigger the actuator shutdown mechanism and identify the source of the fault. An integrated fault flag generation module is used to generate the final fault diagnosis flags; The fault estimation integration and correction module is used to integrate and mask the processing to generate a corrected fault estimation vector.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor implements the method as described in any one of claims 1 to 7.

Images