A dual-mode fault-tolerant filtering fault diagnosis method for magnetic levitation vehicle sensor failures

Abstract

The present invention provides a dual-modal fault-tolerant filtering fault diagnosis method for maglev vehicle sensor faults, comprising: establishing a nonlinear coupled dynamic model of the multi-magnet suspension system of the high-speed maglev vehicle's overlapped structure based on the maglev vehicle's electromagnet parameters; reducing the nonlinear coupled dynamic model to a two-way linear low-order time-varying differential equation using characteristic modeling theory and the projected gradient method; performing gap information filtering estimation in a dual-modal fault-tolerant filter based on the two-way linear low-order time-varying differential equation and a Kalman filtering method to obtain filtered values ​​and predicted values; performing fault diagnosis based on the filtered values ​​and predicted values ​​to obtain a fault diagnosis result; and dynamically adjusting the information fusion weight coefficient based on the fault diagnosis result to achieve gap information reconstruction in the event of a fault. This method improves the robustness and reliability of the maglev system.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic levitation of vehicles, and in particular to a dual-mode fault-tolerant filtering fault diagnosis method for magnetic levitation vehicle sensor faults. Background Art

[0002] Because maglev vehicles lack wheels, they operate without contact between the vehicle and the track, resulting in frictionless operation. This increases speed while reducing energy consumption and noise, making them more environmentally friendly. They also offer superior climbing and cornering performance, making them a strong contender in the future rail transit sector. Currently, China has several low- and medium-speed maglev lines in operation, and the technology is nearing saturation. In recent years, driven by national policies and technological advancements, research has shifted its focus to high-speed maglev technology. The suspension control system, a fundamental and key component of high-speed maglev technology, is a highly nonlinear, open-loop unstable coupled system, making it difficult to establish a precise mathematical model. Vehicle operation under complex operating conditions can introduce additional disturbances to the system. Over extended periods of operation, the system's suspension gap sensors may malfunction. This faulty suspension gap information is fed back into the system, potentially causing rapid system divergence. Furthermore, empirically pinpointing the faulty component in the system is difficult, compromising safe operation and control performance. In practice, solutions often involve hardware redundancy or purchasing more reliable components. However, this approach does not necessarily improve overall system reliability and can actually increase maintenance costs and complexity. Furthermore, given the compact size of the bogie, excessive sensor placement can disrupt electromagnetic field distribution, hindering real-time data detection and timely dynamic coordination. Furthermore, when multiple sensor data conflict, hardware redundancy makes it difficult to quickly locate the fault. Therefore, the ability to detect suspension gap sensor faults in real time through software, and to reconstruct reliable gap feedback values ​​when a fault occurs, thereby achieving stable suspension control, is crucial for improving the reliability and fault tolerance of the suspension system.

[0003] Currently, a common approach to fault management in maglev vehicle suspension systems is to introduce partial hardware redundancy. For example, if a single electromagnet (including the suspension chopper) is detected to have failed, the system immediately deactivates the suspension electromagnet, allowing the vehicle to be supported by the lapped structure of the high-speed maglev vehicle suspension system. However, this approach can lead to model mismatch and shifts in the control equilibrium point, significantly degrading overall suspension performance and making it prone to "drop points" and "iron strike" problems. Existing maglev train fault diagnosis methods rely on the accuracy of established mathematical mechanism models, making their design relatively difficult. Furthermore, most methods focus on single-point suspension systems and employ linear simplifications. Summary of the Invention

[0004] In view of this, the present invention provides a dual-modal fault-tolerant filtering fault diagnosis method for magnetic levitation vehicle sensor faults to solve the above problems.

[0005] The present invention provides a dual-modal fault-tolerant filtering fault diagnosis method for magnetic levitation vehicle sensor faults, comprising: establishing a nonlinear coupling dynamic model of a multi-magnet suspension system of a high-speed magnetic levitation vehicle overlap structure based on the electromagnet parameters of the magnetic levitation vehicle; reducing the nonlinear coupling dynamic model to a two-way linear low-order time-varying differential equation based on characteristic modeling theory and a projected gradient method; performing gap information filtering estimation in a dual-modal fault-tolerant filter based on the two-way linear low-order time-varying differential equation and a Kalman filtering method to obtain a filtered value and a predicted value; performing fault diagnosis based on the filtered value and the predicted value to obtain a fault diagnosis result; and dynamically adjusting an information fusion weight coefficient based on the fault diagnosis result to achieve gap information reconstruction under fault conditions.

[0006] In another implementation of the present invention, the nonlinear coupling dynamics model is expressed as:

[0007]

[0008] in, , is the voltage of the left and right electromagnets; is the resistance of the magnet coil; is the electromagnetic force coefficient, ; is the suspension gap between the left and right electromagnets; , is the first-order derivative of the left and right suspension gaps; , is the second-order derivative of the left and right suspension gaps; is the electromagnet excitation current; , is the first-order derivative of the excitation current; , is the electromagnetic force of the left and right electromagnets; is the laminated spring pressure; is the spring stiffness; is the displacement of the suspension support arm; is the first-order derivative of the suspension arm displacement; is the second-order derivative of the suspension arm displacement; is the mass of the left and right electromagnet modules, is the acceleration due to gravity, is the coupling stiffness of the left and right modules; It is the quality of the support arm.

[0009] In another implementation of the present invention, the two-way linear low-order time-varying difference equations are expressed as:

[0010]

[0011] in, Represents the time step The characteristic model output of the gap value between the left and right suspension electromagnet modules; , The time steps The characteristic model output of the gap value between the left and right suspension electromagnet modules; , The time steps The characteristic model output of the gap value between the left and right suspension electromagnet modules; is the characteristic parameter, obtained through parameter identification; It is the input quantity of the left and right electromagnet modules.

[0012] In another implementation of the present invention, the method further includes: before performing the parameter identification, designing a closed-loop controller using a full-coefficient adaptive control method based on a characteristic model, which is expressed as:

[0013]

[0014] in, It is The system input of the step; For the Step maintenance or tracking control quantity; For the The golden section adaptive control quantity of the step; For the The logical integral control quantity of the step; For the The full-coefficient adaptive controller includes golden section adaptive control, maintenance or tracking control, logical differential control, and logical integral control.

[0015] In another implementation of the present invention, the golden section adaptive control is expressed as:

[0016]

[0017] in, , is the golden ratio coefficient; Respectively The difference between the gap of the step feature output and the expected output gap value; Respectively The gap value of the expected output of the step; It is The characteristic parameter estimator of the step.

[0018] Maintaining or tracking control is expressed as:

[0019]

[0020] in, is a positive number for adjustment; For the The gap value of the expected output.

[0021] In another implementation of the present invention, the logic differential control is expressed as:

[0022]

[0023] in, is the parameter to be adjusted; is the differential coefficient.

[0024] The logic integral control is expressed as:

[0025]

[0026] in, is the integration coefficient, is a small positive number, .

[0027] In another implementation of the present invention, the gap information is estimated by a gap information prediction model; the gap information prediction model is expressed as:

[0028]

[0029] in, They are respectively , The predicted value vector of the step gap information.

[0030]

[0031] is the introduced system process noise, which is independent of the sensor noise.

[0032] In another implementation of the present invention, the fault diagnosis result is expressed as:

[0033]

[0034] in, is the fault detection function, which has 2 degrees of freedom. distribution, i.e. ; is a Gaussian random vector 、 A linear function of for The transposed amount of is the covariance matrix.

[0035] In another implementation of the present invention, the fault diagnosis result is determined by:

[0036]

[0037] in, is the fault judgment threshold.

[0038] In another implementation of the present invention, the method further includes: using a fuzzy membership function to continuously quantify the fault degree in the fault diagnosis result. The specific form of the fuzzy membership function is as follows:

[0039]

[0040] Among them, the lower limit The 90% quantile of the chi-square value under normal working conditions, the upper limit It is the 99% quantile of the chi-square value under normal operating conditions.

[0041] The dual-modal fault-tolerant filtering fault diagnosis method for magnetic levitation vehicle sensor failure of the present invention introduces characteristic modeling theory for the high-speed magnetic levitation overlap structure suspension control system in order to avoid the impact of inaccurate system model establishment, establishes an output decoupled low-order linear time-varying difference equation equivalent to the original system, and designs an adaptive dual-modal fault-tolerant filter based on the established characteristic model. By monitoring the system status and fault information in real time, the suspension gap value is reconstructed through data fusion and state estimation methods in the event of a fault, ensuring that the system can still maintain stable operation when a fault occurs, thereby improving the robustness and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. By reading the detailed description of the embodiments below, the advantages and benefits of the solutions will become clear to those skilled in the art. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. In the drawings:

[0043] Figure 1 The figure is a flow chart of a dual-modal fault-tolerant filtering fault diagnosis method for a magnetic levitation vehicle sensor fault according to an embodiment of the present invention.

[0044] Figure 2 This is a control block diagram of a high-speed magnetic levitation vehicle suspension system based on a characteristic model and dual-mode fault-tolerant filtering according to an embodiment of the present invention.

[0045] Figure 3This is a force diagram of the suspension system of the high-speed maglev vehicle joint structure according to one embodiment of the present invention.

[0046] Figure 4 Schematic diagram of the suspension system characteristic parameter identification simulation structure according to one embodiment of the present invention.

[0047] Figure 5 Schematic diagram of a characteristic parameter identification curve according to an embodiment of the present invention.

[0048] Figure 6 A schematic diagram of a characteristic modeling error curve according to an embodiment of the present invention.

[0049] Figure 7 This is a schematic diagram of the gap sensor fault diagnosis results under scenario (1) of an embodiment of the present invention.

[0050] Figure 8 This is a schematic diagram of the gap sensor fault diagnosis results under scenario (2) of an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and detailedly described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments in the embodiments of the present invention should fall within the scope of protection of the embodiments of the present invention.

[0052] Figure 1 A schematic flow chart of a dual-mode fault-tolerant filtering fault diagnosis method for a magnetic levitation vehicle sensor fault provided by an embodiment of the present invention is shown as follows: Figure 1 As shown, this embodiment mainly includes:

[0053] S101. Based on the electromagnet parameters of the maglev vehicle, a nonlinear coupling dynamic model of the multi-magnet suspension system of the high-speed maglev vehicle lap structure is established.

[0054] S102. According to characteristic modeling theory and projected gradient method, the nonlinear coupled dynamic model is reduced to a two-way linear low-order time-varying differential equation.

[0055] S103 , performing filtering estimation of gap information in a dual-modal fault-tolerant filter based on the two-way linear low-order time-varying difference equation and the Kalman filtering method to obtain a filtered value and a predicted value.

[0056] S104: Perform fault diagnosis based on the filtered value and the predicted value to obtain a fault diagnosis result.

[0057] S105 , dynamically adjusting the information fusion weight coefficient according to the fault diagnosis result to achieve gap information reconstruction in a fault situation.

[0058] The dual-modal fault-tolerant filtering fault diagnosis method for magnetic levitation vehicle sensor failure of the present invention introduces characteristic modeling theory for the high-speed magnetic levitation overlap structure suspension control system in order to avoid the impact of inaccurate system model establishment, establishes an output decoupled low-order linear time-varying difference equation equivalent to the original system, and designs an adaptive dual-modal fault-tolerant filter based on the established characteristic model. By monitoring the system status and fault information in real time, the suspension gap value is reconstructed through data fusion and state estimation methods in the event of a fault, ensuring that the system can still maintain stable operation when a fault occurs, thereby improving the robustness and reliability of the system.

[0059] In another implementation of the present invention, Figure 2 and 3 As shown, the suspension system of a high-speed maglev vehicle utilizes a modular design. Its core unit is a stacked structure consisting of two control units (each containing an electromagnet, a controller, and two sensors). In this suspension system, the left and right electromagnet modules are connected to the suspension support arms via laminated springs, forming a limited degree of freedom system. The vehicle body is then supported by air springs. The maglev vehicle's suspension control process utilizes a chopper circuit to regulate the current in the electromagnet windings, generating a variable electromagnetic attraction to balance the vehicle's weight. The control system dynamically adjusts the current using a closed-loop regulation strategy based on real-time monitoring of parameters such as gap and acceleration. When the measured gap deviates from the set value, the controller increases or decreases the current output accordingly, achieving precise control of the gap through changes in electromagnetic attraction. The suspension electromagnet is primarily subjected to vertical force, achieving a stable suspension state when vertical equilibrium is achieved. If external disturbances cause the electromagnet to deviate from its equilibrium position, the suspension controller adjusts the actuator output control current accordingly to achieve stable suspension, continuously adjusting the electromagnet to maintain a stable equilibrium position.

[0060] according to Figure 3 The force condition in the figure is based on the lower surface of the track as the reference plane, and the vertical downward direction is defined as the positive direction. According to Maxwell's equations and Biot-Savart theorem, the electromagnetic force can be obtained as The expression is:

[0061] (1)

[0062] Where, is the electromagnetic force coefficient, , is the unilateral magnetic area of ​​the magnet, is the magnetic permeability of air, is the number of coil turns. is the suspension gap between the left and right electromagnets. is the electromagnet excitation current.

[0063] According to Kirchhoff's voltage law, the relationship between the electromagnet winding current and the voltage across the electromagnet is:

[0064] (2)

[0065] Where, is the magnet coil resistance, is the voltage across the left and right electromagnets.

[0066] According to Newton's second law, the left and right electromagnet modules contain the following forces in the vertical direction: gravity, electromagnetic coupling force, laminated spring force, and electromagnetic suspension force, as shown in the following formula:

[0067] (3)

[0068] Where, is the mass of the left and right electromagnet modules, is the acceleration due to gravity, is the coupling stiffness of the left and right modules, is the pressure of the laminated spring, which is considered as a spring damper. The force it generates is:

[0069] (4)

[0070] Where, is the spring stiffness, is the spring damping coefficient, is the displacement of the suspension support arm.

[0071] The dynamic equation of the suspension support arm includes the spring force on both sides and the deadweight load:

[0072] (5)

[0073] Where, It is the quality of the support arm.

[0074] Combining equations (1) to (5) can obtain the complete dynamic model of the multi-electromagnet suspension system of high-speed maglev vehicles, that is, the nonlinear coupling dynamic model is expressed as:

[0075] (6)

[0076] in, , is the voltage of the left and right electromagnets; is the resistance of the magnet coil; is the electromagnetic force coefficient, ; is the suspension gap between the left and right electromagnets; , is the first-order derivative of the left and right suspension gaps; , is the second-order derivative of the left and right suspension gaps; is the electromagnet excitation current; , is the first-order derivative of the excitation current; , is the electromagnetic force of the left and right electromagnets; is the laminated spring pressure; is the spring stiffness; is the displacement of the suspension support arm; is the first-order derivative of the suspension arm displacement; is the second-order derivative of the suspension arm displacement; is the mass of the left and right electromagnet modules, is the acceleration due to gravity, is the coupling stiffness of the left and right modules; It is the quality of the support arm.

[0077] The multiple overlapping structures of the carriage are decoupled, and the coupling system connecting two suspension electromagnets and the suspension support arm through laminated springs is taken as the object. Fault diagnosis and gap information reconstruction methods under single gap sensor failure are designed.

[0078] In another implementation of the present invention, based on the characteristic modeling theory, a characteristic model of the output gap of the left and right electromagnet modules is established, and the original nonlinear coupling dynamic model is reduced to a two-way linear low-order time-varying differential equation. That is, the two-way electromagnet suspension gap in the dynamic equation (6) is Through parameter identification, it is transformed into a two-way linear low-order time-varying difference equation, which can be expressed as:

[0079] (7)

[0080] in, Represents the time step The characteristic model output of the gap value between the left and right suspension electromagnet modules; , The time steps The characteristic model output of the gap value between the left and right suspension electromagnet modules; , The time steps The characteristic model output of the gap value between the left and right suspension electromagnet modules; is the characteristic parameter, obtained through parameter identification; It is the input quantity of the left and right electromagnet modules.

[0081] It should be understood that this method can effectively characterize the key characteristics of a complex system by constructing a model with a relatively simple structure, and flexibly adjust as the system behavior changes, so that the designed controller can ensure that the closed-loop system achieves the expected performance.

[0082] Next, we establish the characteristic model of the dynamic equation (6) based on the characteristic modeling theory. First, we define:

[0083]

[0084] in, , is the state column vector of the system state space equation, Include Two state variables, Include Six state variables, are the input and output column vectors of the system, Include Two input quantities.

[0085] Then the second-order affine nonlinear state space equation of the suspension system in equation (6) can be written as:

[0086] (8)

[0087] in:

[0088]

[0089]

[0090] in, , , Is the state quantity , nonlinear function.

[0091] Then, according to the feature modeling theory, first calculate about The partial derivatives and The upper bound of the norm :

[0092] (9)

[0093] Then calculate the bounds of the system's nonlinear functions and their derivatives:

[0094] (10)

[0095] Defining the time scale :

[0096] (11)

[0097] In the feature-based modeling method, the sampling step size is generally ,in Generally a large positive number. Define the function:

[0098] (12)

[0099] Assume that when hour, ,in The modeling accuracy of the given feature. for The upper bound of , according to the inequality:

[0100] (13)

[0101] Can be determined According to the principle of feature modeling, it can be obtained that the feature parameters satisfy the inequality:

[0102] (14)

[0103] Therefore, the projection range satisfied by the characteristic parameters can be obtained:

[0104] (15)

[0105] in, is the projection range of the characteristic parameters.

[0106] In order to establish the characteristic model shown in formula (7), it is necessary to identify the characteristic parameters online. Formula (7) can be converted into the corresponding error characteristic model:

[0107] (16)

[0108] Where, , For the The expected gap value of the step, Separate step The difference from the expected value, It is The difference between the step output value and the expected value, It is The difference between the step output value and the expected value. The characteristic coefficient is recorded as a vector In the form of:

[0109]

[0110] Data vector for:

[0111]

[0112] Then the system model (16) can be rewritten as

[0113] (17)

[0114] The identification method of feature parameters adopts the commonly used projected gradient method, which is famous for its simple structure and is often used for feature model identification. Its recursive formula is as follows:

[0115] (18)

[0116] Where, It is The characteristic parameter estimation vector of the step, It is The characteristic parameter estimation vector of the step, It is step data vector, Is an adjustable positive real number, representing the learning rate, In the compact set Orthogonal projection on the set It is given by formula (15).

[0117] Based on the characteristic modeling theory, the characteristic parameters of a high-order, coupled nonlinear suspension system are identified through the projected gradient method and converted into a characteristic model of two low-order linear time-varying difference equations. During the identification process, a full-coefficient adaptive control method based on the characteristic model is used to achieve closed-loop stable suspension.

[0118] In another implementation of the present invention, the golden section adaptive control is expressed as:

[0119] (19)

[0120] in, , is the golden ratio coefficient; Respectively The difference between the gap of the step feature output and the expected output gap value; Respectively The gap value of the expected output of the step; It is The characteristic parameter estimator of the step.

[0121] Maintaining or tracking control is expressed as:

[0122] (20)

[0123] in, is a positive number for adjustment; Respectively The gap value of the expected output.

[0124] In another implementation of the present invention, the logic differential control is expressed as:

[0125] (twenty one)

[0126] in, is the parameter to be adjusted; is the differential coefficient.

[0127] The logic integral control is expressed as:

[0128] (twenty two)

[0129] in, is the integration coefficient, is a small positive number, .

[0130] In another implementation of the present invention, the method further includes: before performing the parameter identification, designing a closed-loop controller using a full-coefficient adaptive control method based on a characteristic model, that is, the four parts above are summed to form a full-coefficient adaptive controller based on a characteristic model as shown below, which is expressed as:

[0131] (twenty three)

[0132] in, It is The system input of the step; For the Step maintenance or tracking control quantity; For the The golden section adaptive control quantity of the step; For the The logical integral control quantity of the step; For the The full-coefficient adaptive controller includes golden section adaptive control, maintenance or tracking control, logical differential control, and logical integral control.

[0133] For example, because the multi-electromagnet suspension model is an open-loop unstable system, direct signal input will cause the system to become unstable and unable to perform proper identification. Therefore, closed-loop feedback control must be introduced to stabilize the system for effective identification. Therefore, before identification can be performed, a suitable control method must be designed to achieve stability. Therefore, a closed-loop controller is designed using a full-coefficient adaptive control method based on a characteristic model.

[0134] The accuracy and effectiveness of the characteristic model of the multi-electromagnet suspension system with overlapping structure and the parameter identification algorithm used are verified by MATLAB / Simulink numerical simulation. The structural block diagram of the simulation model for parameter identification is shown in the figure. Figure 4 As shown. Figure 5 and Figure 6 The parameter identification curve and modeling error curve show that the absolute value of the maximum modeling error is 0.045, which stabilizes to 0 after about 1.25 seconds. The characteristic parameters can be identified normally and stabilize within 0.8 seconds. The sum of the three parameters approaches 1, which conforms to the principle of characteristic modeling.

[0135] In another implementation of the present invention, the gap information is estimated by a gap information prediction model.

[0136] For example, based on the Kalman filtering method, a characteristic model is combined to establish the gap information filter estimation portion of the dual-modal fault-tolerant filter, which is used to estimate gap information and reduce sensor noise. The system characteristic model established in this step estimates gap information through the information prediction module. Because the characteristic model is a mathematical model based on input and output data, a discrete model of the system can be established first. By collecting system information, the system gap value is calculated in real time, which is unaffected by the actual sensor data. At the same time, to reduce noise interference and the influence of the measurement sensor, the estimated data and the actual sensor data are corrected in real time using the Kalman filtering method.

[0137] First, define the gap information state quantity of the left and right modules of the system as:

[0138]

[0139] Each gap sensor can measure a gap information, and its measurement equation is:

[0140] (twenty four)

[0141] in, Indicates the gap sensor output value, is the measurement matrix, represents the introduced sensor measurement noise, whose mean and covariance matrix are:

[0142] (25)

[0143] in, is the variance of the measurement noise.

[0144] Based on the feature model and the real-time input data of the system, a gap information prediction algorithm is constructed. The advantage of this method is that it can reduce the interference of sampling noise on the prediction results. The gap information prediction model is obtained by transforming Equation (7):

[0145] (26)

[0146] in, They are respectively , The predicted value vector of the step gap information.

[0147] (27)

[0148] is the introduced system process noise, which is independent of the sensor noise, and its mean and covariance matrix are:

[0149] (28)

[0150] in, is the variance of the process noise.

[0151] Assume that the output gap of the information predictor module is , the state covariance matrix is , the algorithm flow of the prediction link is as follows:

[0152] (29)

[0153] in, are the initial random information prediction output and the random state covariance matrix respectively.

[0154] In order to filter out the measurement noise and process noise of the system, the Kalman filter is used for sampling filtering. Combined with the above predictor, the definition is is the state quantity The filtered value is defined as is the covariance matrix of the previous estimated state. The filtering update process is as follows:

[0155] (30)

[0156] Where, is the Kalman filter gain, is the initial Gaussian random vector. From Equations (29) and (30), we can see that are also Gaussian random vectors.

[0157] Aiming at the problem of complete or partial failure of the gap sensor in the suspension system, a reliable state estimation model is constructed to realize the dynamic reconstruction of the suspension gap and ensure the stable operation of the suspension control system in the case of partial sensor failure.

[0158] In another implementation of the present invention, the fault diagnosis result is expressed as:

[0159]

[0160] in, is the fault detection function, which has 2 degrees of freedom. distribution, i.e. ; is a Gaussian random vector 、 A linear function of for The transposed amount of is the covariance matrix.

[0161] Exemplarily, a fault diagnosis module and a data fusion module combining filtered values ​​and predicted values ​​are constructed.

[0162] The estimated error is defined as:

[0163] (31)

[0164] Where, is a Gaussian random vector 、 A linear function of . Then its covariance is:

[0165] (32)

[0166] When the sensor is normal, By Gaussian random vector Linear structure, so Obeys a Gaussian distribution with zero mean, and its covariance matrix Satisfies the expression: , from which we can get:

[0167] (33)

[0168] because It is the state prediction obtained by recursion of the prior state model and is not affected by the measurement information. Therefore, when the sensor fails, it still satisfies .and It is the information measured by the sensor and filtered by the filter. When the sensor fails, it is no longer an unbiased estimate of the state quantity, that is, ,at this time , so you can choose As a fault detection amount, whether a sensor fault occurs is detected.

[0169] right Make the following binary assumptions:

[0170] : No fault, , ;

[0171] : There is a fault, , .

[0172] According to the above definition, the following fault detection function can be obtained

[0173] (34)

[0174] The fault detection function Obey the degree of freedom of 2 distribution, i.e. .

[0175] In another implementation of the present invention, the fault diagnosis result is determined by:

[0176]

[0177] in, is the fault judgment threshold.

[0178] In another implementation of the present invention, the method further includes: using a fuzzy membership function to continuously quantify the fault degree in the fault diagnosis result. The specific form of the fuzzy membership function is as follows:

[0179]

[0180] Among them, the lower limit The 90% quantile of the chi-square value under normal working conditions, the upper limit It is the 99% quantile of the chi-square value under normal operating conditions.

[0181] For example, traditional threshold diagnosis methods can lead to missed detections if the threshold is too high, while false alarms can occur if the threshold is too low. Therefore, this study uses a fuzzy membership function to continuously quantify the fault severity, calculating the sensor's effective probability instead of a binary decision to improve diagnostic reliability.

[0182] The fault detection mechanism of the gap sensor of the high-speed maglev vehicle suspension system, which integrates feature modeling, Kalman filtering, chi-square state detection and data fusion, can accurately determine the fault type and severity of the gap sensor fault in the multi-electromagnet suspension system of the high-speed maglev vehicle lap structure, thereby effectively improving the real-time and accuracy of fault diagnosis.

[0183] First, the collected data is divided into: (1) by the information predictor ( ) data generated by the left and right sensors ( The bimodal fault-tolerant filter weight function is defined as:

[0184] (36)

[0185] This function indicates that the output number is When the sensor data is The proportion of sensor data can be used to design weight distribution strategies under different working conditions:

[0186] (1) When the sensor is normal, the chi-square test verifies that its data is reliable, and the filter update module can effectively reduce noise interference. At this time, the left and right sensors mainly use their own data, supplemented by other information. At this time, their own data accounts for a larger proportion, and other data accounts for a smaller proportion. As shown in Table 1, is the weight adjustment coefficient of the gap sensor, which can be a smaller value such as .

[0187] Table 1 Data fusion weights of gap sensors under normal conditions Value Table

[0188]

[0189] (2) In the event of a gap sensor failure, the reliability of its data is significantly reduced, and the system will significantly reduce the weight of the sensor. At this time, it is necessary to combine the data of other normal sensors or prediction models to compensate, and at the same time increase the weight distribution ratio of information prediction data to ensure the final output accuracy. As shown in Table 2, is the weight adjustment coefficient of the gap sensor, which can be .

[0190] Table 2 Data fusion weights under gap sensor failure Value Table

[0191]

[0192] Based on the above adaptive weight adjustment strategy, a dual-modal fault-tolerant filter function can be established As shown below:

[0193] (37)

[0194] At this point, a dual-modal fault-tolerant filter that can realize gap sensor fault diagnosis and suspension gap value reconstruction has been constructed.

[0195] The above fault diagnosis method is applied to the suspension system of high-speed maglev vehicle, and the system parameters and simulation initial values ​​are configured to simulate the gap sensor of the left electromagnet module. The effect of the dual-mode fault-tolerant filter is simulated and verified in MATLAB / Simulink. The simulation experiment considers the following two typical cases: (1) the gap sensor complete failure mode. The gap sensor output value is set to zero during the fault period; (2) the gap sensor partial failure mode. The gap sensor output signal is set to be contaminated by strong noise during the fault period.

[0196] The simulation results of case (1) are as follows Figure 7 As shown. Figure 7 (a) It can be seen that the normal sensor Not affected. Figure 7 As can be seen in (b), the faulty gap sensor has a sampling output of 0 at 3s, indicating that the sensor is completely faulty at this time. The sampling output of the sensor is subject to noise interference, which is filtered by the adaptive dual-mode fault-tolerant filter to output smoother and more accurate data. From the comparison between (c) and (d), it can be seen that at 3s, the sensor When a fault occurs, the chi-square detection function drops rapidly from 1 to 0, indicating a complete failure, while a normal sensor Always output 1. At about 4 seconds, the sensor Return to normal, at this time the chi-square test function returns to 1.

[0197] The simulation results of case (2) are as follows Figure 8 As shown in (a) and (b), the faulty sensor The sampling output of the sensor itself contains the external noise interference introduced. At 3-4s, the sensor The sampled data showed a large amount of random noise with larger amplitude. However, the adaptive dual-mode fault-tolerant filter effectively filtered out the larger noise, making the output smoother. Therefore, the feedback value can be applied to closed-loop control. As can be seen from (d), the chi-square state detection function suddenly changed from 1 at 3s. Due to the setting of the fuzzy membership function, It fluctuates greatly between 0 and 1, indicating that the reliability of the gap sensor has changed. Everything has returned to normal and the sampling data can be trusted again.

[0198] The verification of the dual-modal fault-tolerant filter in the above two situations shows that the filter can not only effectively filter out the noise introduced by the sensor, but also, when the faulty sensor fails completely or partially, calculate the chi-square state detection result of the fault diagnosis function through information prediction and filter estimation to determine whether the sensor has failed. Combined with the chi-square detection result, the available gap value is fed back to the control end through data fusion to realize the fault diagnosis and fault-tolerant filtering functions of the gap sensor. The characteristic model established by the present invention can simplify and decouple the gap output of the left and right electromagnet modules. Based on this characteristic model, the gap sensor fault condition can be detected through the chi-square state, and the effective gap information can be estimated through fusion filtering.

[0199] Without adding additional sensors, a highly fault-tolerant online diagnostic system is constructed by combining feature modeling and adaptive filtering algorithms to reduce system cost and complexity. Aiming at the strong nonlinearity and coupling characteristics of the maglev train suspension control system, a robust and generalizable fault diagnosis filter framework is designed, which can be applied to fault monitoring and fault-tolerant control of high-speed maglev systems in complex operating environments.

[0200] Compared with the prior art, the present invention has the following advantages:

[0201] Based on the software fault-tolerant mechanism, the hardware redundant design is abandoned, thereby significantly reducing the hardware cost and complexity of the maglev vehicle suspension system, reducing the difficulty of maintenance, and adapting to the needs of compact equipment space. By combining feature modeling with adaptive filtering, it is possible to monitor the sensor status in real time and perform dynamic data fusion in the event of a sensor failure, accurately provide available gap values, and effectively improve the system's fault tolerance to ensure the reliability and stability of the suspension control system under complex working conditions. The fault diagnosis method proposed in the present invention does not rely on an accurate system mathematical model, but adopts a feature modeling method, which has strong adaptability and flexibility. This method is not only suitable for the suspension control system of high-speed maglev vehicles, but can also be extended to other maglev systems with similar fault diagnosis requirements.

[0202] Another aspect of the present invention provides a dual-mode fault-tolerant filtering fault diagnosis system for magnetic levitation vehicle sensor faults, comprising:

[0203] Model building module: Based on the electromagnet parameters of the maglev vehicle, a nonlinear coupled dynamic model of the multi-magnet suspension system of the high-speed maglev vehicle overlap structure is established; according to the characteristic modeling theory and the projected gradient method, the nonlinear coupled dynamic model is reduced to a two-way linear low-order time-varying difference equation.

[0204] Fault diagnosis module: Based on the two-way linear low-order time-varying difference equation and Kalman filtering method, gap information filtering estimation is performed in the dual-modal fault-tolerant filter to obtain filtered values ​​and predicted values; fault diagnosis is performed according to the filtered values ​​and the predicted values ​​to obtain a fault diagnosis result.

[0205] Information reconstruction module: dynamically adjusts the information fusion weight coefficient according to the fault diagnosis result to achieve gap information reconstruction in the event of a fault.

[0206] In order to avoid the impact of inaccurate system model establishment, the dual-modal fault-tolerant filtering fault diagnosis system for magnetic levitation vehicle sensor failure of the present invention introduces characteristic modeling theory for the high-speed magnetic levitation overlap structure suspension control system, establishes an output decoupled low-order linear time-varying difference equation equivalent to the original system, and designs an adaptive dual-modal fault-tolerant filter based on the established characteristic model. By monitoring the system status and fault information in real time, the suspension gap value is reconstructed through data fusion and state estimation methods in the event of a fault, ensuring that the system can still maintain stable operation when a fault occurs, thereby improving the robustness and reliability of the system.

[0207] In another aspect of the present invention, an electronic device includes a processor, a memory, a communication bus, and a communication interface.

[0208] in:

[0209] The processor, memory and communication interface communicate with each other through a communication bus.

[0210] Communication interface, used to communicate with other electronic devices or servers.

[0211] The processor is used to execute the program, and specifically can execute the steps of the dual-modal fault-tolerant filtering fault diagnosis method for magnetic levitation vehicle sensor failure in any of the above embodiments.

[0212] Specifically, the program may include program codes including computer operation instructions.

[0213] The processor may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the smart device may be processors of the same type, such as one or more CPUs, or processors of different types, such as one or more CPUs and one or more ASICs.

[0214] Memory, used to store programs. Memory may include high-speed RAM (RAM) or non-volatile memory, such as at least one disk drive.

[0215] The program can be specifically configured to cause a processor to execute the steps of any of the dual-modal fault-tolerant filtering fault diagnosis methods for magnetic levitation vehicle sensor faults described in the embodiments. The specific implementation of each step in the program can be found in the corresponding descriptions of the steps and units executed in any of the aforementioned dual-modal fault-tolerant filtering fault diagnosis methods for magnetic levitation vehicle sensor faults, and will not be repeated here. Those skilled in the art will clearly understand that, for ease and brevity of description, the specific working processes of the devices and modules described above can refer to the corresponding process descriptions in the aforementioned method embodiments.

[0216] The exemplary embodiments of the present application further provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the methods of the various embodiments of the present application.

[0217] The methods according to the embodiments of the present invention described above can be implemented in hardware, firmware, or as software or computer code that can be stored on a recording medium (such as a CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or non-transitory machine-readable medium downloaded over a network and then stored on a local recording medium. Thus, the methods described herein can be processed by such software stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It will be understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods described herein are implemented. Furthermore, when a general-purpose computer accesses the code for implementing the methods described herein, the execution of the code transforms the general-purpose computer into a dedicated computer for performing the methods described herein.

[0218] Thus far, specific embodiments of the present invention have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired results. Additionally, the processes depicted in the accompanying drawings do not necessarily require the specific order shown, or sequential order, to achieve the desired results.

[0219] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, back, etc.) are only used to explain the relative position relationship between the components in a certain specific order (as shown in the accompanying drawings). If the specific order changes, the directional indication will also change accordingly.

[0220] In the description of the present invention, the terms "first" and "second" are used solely to facilitate description of different components or names and should not be construed as indicating or implying a sequential relationship, relative importance, or implicitly specifying the quantity of the technical features being described. Therefore, features specified as "first" or "second" may explicitly or implicitly include at least one of such features.

[0221] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0222] It should be noted that although the specific embodiments of the present invention are described in detail in conjunction with the accompanying drawings, this should not be construed as limiting the scope of protection of the present invention. Within the scope described by the claims, various modifications and variations that can be made by those skilled in the art without creative effort still fall within the scope of protection of the present invention.

[0223] The examples of the embodiments of the present invention are intended to briefly illustrate the technical features of the embodiments of the present invention so that those skilled in the art can intuitively understand the technical features of the embodiments of the present invention, and are not intended to improperly limit the embodiments of the present invention.

[0224] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A dual-mode fault-tolerant filtering fault diagnosis method for magnetic levitation vehicle sensor faults, characterized in that: include: Based on the electromagnet parameters of the maglev vehicle, a nonlinear coupling dynamic model of the multi-magnet suspension system of the high-speed maglev vehicle overlap structure is established. The nonlinear coupling dynamic model is expressed as: in, , is the voltage of the left and right electromagnets; is the resistance of the magnet coil; is the electromagnetic force coefficient, ; is the suspension gap between the left and right electromagnets; , is the first-order derivative of the left and right suspension gaps; , is the second-order derivative of the left and right suspension gaps; is the electromagnet excitation current; , is the first-order derivative of the excitation current; , is the electromagnetic force of the left and right electromagnets; is the laminated spring pressure; is the spring stiffness; is the displacement of the suspension support arm; is the first-order derivative of the suspension arm displacement; is the second-order derivative of the suspension arm displacement; is the mass of the left and right electromagnet modules, is the acceleration due to gravity, is the coupling stiffness of the left and right modules; is the mass of the support arm; According to the characteristic modeling theory and the projected gradient method, the nonlinear coupled dynamic model is reduced to a two-way linear low-order time-varying differential equation, which is expressed as: in, Represents the time step The characteristic model output of the gap value between the left and right suspension electromagnet modules; , The time steps The characteristic model output of the gap value between the left and right suspension electromagnet modules; , The time steps The characteristic model output of the gap value between the left and right suspension electromagnet modules; is the characteristic parameter, obtained through parameter identification; It is the input quantity of the left and right electromagnet modules; Based on the two-way linear low-order time-varying difference equation and the Kalman filtering method, gap information filtering estimation in the dual-modal fault-tolerant filter is performed to obtain a filtered value and a predicted value; Perform fault diagnosis based on the filtered value and the predicted value to obtain a fault diagnosis result; The information fusion weight coefficient is dynamically adjusted according to the fault diagnosis result to achieve gap information reconstruction in the event of a fault.

2. The method according to claim 1, characterized in that Also includes: Before performing the parameter identification, a closed-loop controller is designed using a full-coefficient adaptive control method based on a characteristic model, which is expressed as: in, It is The system input of the step; For the Step maintenance or tracking control quantity; For the The golden section adaptive control quantity of the step; For the The logical integral control quantity of the step; For the The logical differential control quantity of the step; The full-coefficient adaptive controller includes golden section adaptive control, maintenance or tracking control, logic differential control, and logic integral control.

3. The method according to claim 2, characterized in that The golden section adaptive control is expressed as: in, , is the golden ratio coefficient; Respectively The difference between the gap of the step feature output and the expected output gap value; Respectively The gap value of the expected output of the step; It is The characteristic parameter estimator of the step; Maintaining or tracking control is expressed as: in, is a positive number for adjustment; For the The gap value of the expected output.

4. The method according to claim 2, characterized in that The logical differential control is expressed as: in, is the parameter to be adjusted; is the differential coefficient; The logic integral control is expressed as: in, is the integration coefficient, is a small positive number, .

5. The method according to claim 1, wherein The gap information is estimated by a gap information prediction model; The gap information prediction model is expressed as: in, They are respectively , Prediction value vector of step gap information; is the introduced system process noise, which is independent of the sensor noise.

6. The method according to claim 1, characterized in that The fault diagnosis result is expressed as: in, is the fault detection function, which has 2 degrees of freedom. distribution, i.e. ; is a Gaussian random vector 、 A linear function of for The transposed amount of is the covariance matrix.

7. The method according to claim 6, characterized in that The judgment criteria of the fault diagnosis result are: in, is the fault judgment threshold.

8. The method according to claim 7, characterized in that Also includes: A fuzzy membership function is used to realize continuous quantification of the fault degree in the fault diagnosis result. The specific form of the fuzzy membership function is as follows: Among them, the lower limit The 90% quantile of the chi-square value under normal working conditions, the upper limit It is the 99% quantile of the chi-square value under normal operating conditions.

Images