Active early fault detection method for high-speed train suspension system

Abstract

The invention discloses an active early fault detection method for a high-speed train suspension system. The active early fault detection method comprises the following steps of 1, performing data modeling of an actual model and a nominal model of the high-speed train suspension system by utilizing T-S fuzzy; 2, constructing an input and output data matrix of data modeling; 3, designing a ToMFIR residual error according to the data model and the input and output data matrix; 4, constructing an auxiliary signal and an auxiliary signal distributor, and optimizing the ToMFIR residual error through the auxiliary signal and the auxiliary signal distributor; and 5, designing an evaluation function, and detecting a fault through the ToMFIR residual obtained in the step 4 and the evaluation function. By designing the auxiliary signal and the auxiliary signal distributor and injecting the auxiliary signal and the auxiliary signal distributor into the system, the ToMFIR residual error contains more fault information, so that the fault sensing capability of the system is improved.

Description

Technical Field

[0001] The invention relates to the technical field of high-speed train suspension system fault detection, and in particular to an active early fault detection method for a high-speed train suspension system. Background Art

[0002] The suspension system is an important component of high-speed trains. It not only supports the bogie and the train body, but also isolates the forces generated by track unevenness, ensuring the riding experience of passengers in the carriage. However, the train suspension system faces complex working conditions during operation, which induces early failures of various mechanical components and systems. These early failures are usually highly concealed and difficult to detect, so it is of great significance to study the early failure detection of the suspension system.

[0003] At present, although the existing analytical ToMFIR residuals, data-driven ToMFIR residuals and other methods and technologies have broken through the limitations of system structure and achieved the detection and diagnosis of early faults in control systems under certain restrictions, they still belong to the framework of passive fault detection and diagnosis. Compared with the active fault detection and diagnosis framework, the analytical ToMFIR residuals and data-driven ToMFIR residuals based on the passive framework only rely on the input and output data of the system, resulting in insufficient fault information extraction and fault feature mining, which greatly limits the existing technology for early fault detection, diagnosis and engineering application. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides an active early fault detection method for a high-speed train suspension system to solve the technical problems of insufficient fault information extraction and fault feature mining in the prior art.

[0005] The present invention provides an active early fault detection method for a high-speed train suspension system, comprising the following steps:

[0006] Step 1: Use TS fuzzy to perform data modeling of the actual model and nominal model of the high-speed train suspension system;

[0007] Step 2: Construct the input and output data matrix for data modeling;

[0008] Step 3: Design ToMFIR residual according to the data model and input and output data matrix;

[0009] Step 4: construct an auxiliary signal and an auxiliary signal distributor, and optimize the ToMFIR residual through the auxiliary signal and the auxiliary signal distributor;

[0010] Step 5: Design an evaluation function and detect faults using the ToMFIR residual obtained in step 4 and the evaluation function.

[0011] Furthermore, in step 3, an identification technique based on a combination of LQ decomposition and least squares method is used to design the ToMFIR residual.

[0012] Furthermore, in step 3, the designed ToMFIR residual is:

[0013] ToMFIR(k)=r y,l (k)-H(μ)r u,p,l (k)

[0014] In the formula, μ i (θ(k)) represents the membership function, satisfying μ i (θ(k))≥0 and

[0015] Furthermore, in step 4, the auxiliary signal and auxiliary signal distributor constructed are: Wherein, η is a constant; α1 and α2 are m×1 dimensional matrix vectors, and α1+α2=0.

[0016] Furthermore, the step 4 also includes: constructing a redundant actuator, the constructed redundant actuator is: B2, let B2=B1, wherein B1 and B2 are both n×m dimensions, and B1 is the input matrix coefficient of the normal system.

[0017] Furthermore, in step 4, the optimized ToMFIR residual is:

[0018]

[0019] In the formula,

[0020]

[0021] in,

[0022] Furthermore, in step 5, an evaluation function based on the Hellinger distance is designed.

[0023] Furthermore, the evaluation function is:

[0024]

[0025] In the formula, in,

[0026] in,

[0027] Beneficial effects of the present invention:

[0028] The present invention belongs to an active fault detection and diagnosis framework. The present invention designs auxiliary signals and auxiliary signal distributors and injects them into the system so that the ToMFIR residual contains more fault information, thereby improving the system's ability to perceive faults. At the same time, in order to reduce the impact of auxiliary signals on fault-free systems, the present invention designs redundant actuators for auxiliary signals to ensure that when the system is fault-free, the auxiliary signals have no impact on the normal operating system; and when a system fault occurs, the auxiliary signals can achieve a beneficial amplification of the ToMFIR residual of the faulty system, thereby improving the ability to detect faults. The present invention takes into account the part of early fault information that is compensated by the controller but not reflected in the system output, that is, in the process of residual design, the controller's adjustment amount before and after the early fault occurs is comprehensively considered, so the acquisition of early fault information is more comprehensive. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the present invention in any way. In the accompanying drawings:

[0030] Figure 1 It is a schematic diagram of a flow chart of a specific embodiment of the present invention;

[0031] Figure 2 It is a comparison diagram of the detection results with and without the auxiliary signal in the first fault of the specific embodiment of the present invention;

[0032] Figure 3 It is a comparison diagram of the detection results with and without the auxiliary signal in the second fault of the specific embodiment of the present invention;

[0033] Figure 4 It is a comparison diagram of the detection results with and without the auxiliary signal in the third fault of the specific embodiment of the present invention;

[0034] Figure 5 is a comparison diagram of the detection results with and without the auxiliary signal in the fourth fault of the specific embodiment of the present invention;

[0035] Figure 6 is a comparison diagram of the detection results with and without the auxiliary signal in the fifth fault of the specific embodiment of the present invention;

[0036] Figure 7 is a comparison diagram of detection results with and without auxiliary signals in the sixth fault of a specific embodiment of the present invention;

[0037] Figure 8 is a comparison diagram of the detection results with and without the auxiliary signal in the seventh fault in a specific embodiment of the present invention;

[0038] Fig. 9 is a comparison diagram of detection results with and without auxiliary signals in the eighth fault of a specific embodiment of the present invention;

[0039] Fig.10 It is a comparison diagram of detection results with and without auxiliary signals when the ninth fault occurs in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0041] The present invention is further illustrated below in conjunction with specific embodiments. Those skilled in the art should understand that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention, and modifications to various equivalent forms of the present invention fall within the scope defined by the appended claims of this application.

[0042] The present invention provides an active early fault detection method for a high-speed train suspension system, comprising the following steps:

[0043] Step 1: Use TS fuzzy to perform data modeling of the actual model and nominal model of the high-speed train suspension system. Under each fuzzy rule, the input-output relationship data model of the actual model and nominal model of the suspension system is obtained accordingly. The specific process is as follows:

[0044] Step 11: Considering faults and noise, the real system and the nominal system of the nonlinear suspension system are as follows:

[0045]

[0046] Where w(k) and v(k) represent process noise and measurement noise respectively; x0(k) represents the state variable of the nominal system; g(x0,k) represents the nonlinear term of the nominal system; u0(k) represents the input variable of the nominal system; y0(k) represents the output variable of the nominal system.

[0047] Step 12: Considering the fault and noise, the local linear model of the real system and its nominal system of the suspension system is as follows:

[0048]

[0049] Where θ(k)=[θ1(k) … θ p(k)] is a measurable antecedent variable; N j,i (j=1,2,…,p) is a fuzzy set; R i represents the i-th fuzzy inference rule, i = 1, 2,…, v, v is the number of TS fuzzy rules.

[0050] Through the local linear model, the real system of high-speed train suspension global fuzziness and its nominal system are expressed as follows:

[0051] R i :IFθ1(k)isN 1,i and...andθ p (k)isN p,i ,THEN

[0052]

[0053] In the formula, μ i (θ(k)) represents the membership function, satisfying: μ i (θ(k))≥0 and Step 13: The system output corresponding to the local TS fuzzy model of the real system of the suspension system is:

[0054] y l (k) = H i,u,p,l u p,l (k)+H i,d,p,l d p,l (k)+H i,w,p,l e p,l (k)

[0055] The derivation process is as follows:

[0056] Define S(k) as any data vector at time k, and define the following stack matrix:

[0057]

[0058] From formula (5), we can get:

[0059]

[0060] From formula (9), we can get:

[0061]

[0062] In the formula, Then from formula (5), we can get:

[0063] y l (k) = T i,l x(k)+H i,u,l u l (k)+Hi,d,l d l (k)+H i,w,l w l (k)+v l (k)(11)In formula,

[0064]

[0065]

[0066] when

[0067]

[0068] Then (11) is:

[0069]

[0070] That is, the system output corresponding to the local TS fuzzy model of the real system of the suspension system is obtained.

[0071] In the formula,

[0072] Step 14: Considering the time interval N, define the stack matrix Y in sequence k,l ,U k,p,l and D k,p,l for:

[0073]

[0074] From formula (7), we get:

[0075] Y k,l =H i,u,p,l U k,p,l +H i,d,p,l D k,p,l +H i,w,p,l E k,p,l (15)

[0076] The input and output data model of the real suspension system is shown in formula (15). The data model implies the relationship between the input and output of the suspension system. In practical applications, the data matrix Y can be obtained. k,l and U k,p,l specific information.

[0077] Step 2: Use displacement sensors and gyroscopes to obtain the vertical displacement and angular velocity signals of the center of mass of the train carriage and the vertical displacement signal of the center of mass of the bogie frame when the high-speed train is running. The signals of the sensors and gyroscopes are the system outputs. The obtained data are used to construct an input-output data matrix according to the data model.

[0078] Step 3: According to the data model and input and output data matrix, the ToMFIR residual is designed by using the identification technology based on LQ decomposition combined with the least squares method. The specific steps are as follows:

[0079] Step 31: According to formula (15), the local input and output data model of the real system is:

[0080] Y k,l =H i,u,p,l U k,p,l +H i,d,p,l D k,p,l +H i,w,p,l E k,p,l

[0081] Perform LQ decomposition on process data:

[0082]

[0083] We can get:

[0084]

[0085] Where, L i,33 Q i,3 =H i,e,p,l E k,p,l .

[0086] Step 32: By solving the following least squares method for [H i,u,p,l H i,d,p,l ] to make an estimate,

[0087]

[0088] We can get:

[0089]

[0090] Step 33: The system output under the local TS fuzzy model of the real system is:

[0091] y l (k) = H i,u,p,l u p,l (k)+H i,d,p,l d p,l (k)+H i,w,p,l e p,l (k)(21)

[0092] The system output under the local TS fuzzy model of the nominal system is:

[0093]

[0094] The system output of the local TS fuzzy model of the nominal system driven by the real system input is:

[0095] y l (k) * =H i,u,p,l u p,l (k) (23)

[0096] Define the output residual as: r y,l (k) = y l (k)-y l (k) 0 , the controller residual is: r u,p,l (k) = u p,l (k)-u p,l (k) 0 , then the residual error of the fault information of the local linear model is:

[0097]

[0098] Step 34: The matrix identified based on equation (20) is: H i,u,p,l , then ToMFIR under the local TS fuzzy model is expressed as:

[0099] R i :IFθ1(k)isN 1,i and...andθ p (k)isN p,i ,THEN

[0100]

[0101] In the formula,

[0102] ToMFIR under the global TS fuzzy model is expressed as:

[0103] ToMFIR(k)=r y,l (k)-H(μ)r u,p,l (k)(26)

[0104] In the formula,

[0105] Step 4: Construct redundant actuators, auxiliary signals, and auxiliary signal distributors, and optimize the ToMFIR residual through the auxiliary signals and auxiliary signal distributors. The specific process is as follows:

[0106] Step 41: Build Redundant Actuators:

[0107] The redundant actuator introduced in the design is: B2, let B2 = B1, then the input matrix coefficient of the real system is expanded to [B, B2][B, B2], where B1 and B2 are both n×m dimensions, B1 is the input matrix coefficient of the normal system, and B is the input matrix coefficient of the real system.

[0108] That is:

[0109]

[0110] The auxiliary signal and auxiliary signal distributor are constructed as follows: After the auxiliary signal is introduced, the system input matrix u(k) is expanded to:

[0111]

[0112] Where α1, α2, and u(k) are all m×1-dimensional matrix vectors; η is a constant; u * (k) is a 2m×1 dimensional matrix vector.

[0113] Step 42: Let α1+α2=0 in the auxiliary signal distributor. When the suspension system has no fault, the discrete state space description of the suspension system is as follows:

[0114] R i :IFθ1(k)isN 1,i and...andθ p (k)isN p,i ,THEN

[0115]

[0116] in,

[0117]

[0118] Step 43: Obtain the effect of the auxiliary signal on the ToMFIR residual:

[0119] After introducing redundant actuators, auxiliary signals and their distributors, the impact of auxiliary signals on system output is:

[0120]

[0121] in,

[0122] The process of optimizing the ToMFIR residual according to the generated impact is as follows:

[0123] It is known that after the introduction of auxiliary signals and redundant controllers:

[0124] [B i ,B2]u * (k) = Bi u(k)+(B i -B2)α1η(31)

[0125] Then the following formula (9) can be Rewritten as:

[0126]

[0127] Then we can get:

[0128]

[0129] y l (k)≈H i,u,p,l u p,l (k)+H i,d,p,l d p,l (k)+H i,e,p,l e p,l (k)+M i (36)

[0130] The influence of auxiliary signal on system output is:

[0131]

[0132] Therefore, the influence of the auxiliary signal on ToMFIR under the local TS fuzzy model can be expressed as:

[0133] R i :IFθ1(k)isN 1,i and…andθ p (k)isN p,i ,THEN

[0134]

[0135] In the formula,

[0136] The influence of the auxiliary signal on ToMFIR under the global TS fuzzy model can be expressed as:

[0137]

[0138] In the formula,

[0139] Step 5: Design an evaluation function based on the Hellinger distance, and detect the fault through the ToMFIR residual obtained in step 4 and the evaluation function. The process of designing the evaluation function based on the Hellinger distance is as follows:

[0140] Step 51: The global fuzzy system is:

[0141] ToMFIR new (k)=[t1,t2,…,t m ] T ∈R m

[0142] Where m = k y l

[0143] ToMFIR new (k) When it obeys or approximately obeys Gaussian distribution,

[0144] Then we have:

[0145]

[0146] In the formula,

[0147] Step 52: Introduce the fault-free residual signal:

[0148]

[0149] ToMFIR rf (k) is ToMFIR(k) at a certain fault-free moment, which has the same physical meaning as ToMFIR(k), then:

[0150]

[0151] In the formula,

[0152] Step 53: Based on the Hellinger distance, the evaluation function can be defined as:

[0153]

[0154] The following is a simulation verification of the method of the present invention:

[0155] Step A1: Set the characteristic information of the actuator fault in the suspension system, the failure rate is 10%, and the early actuator faults with fault evolution rates of 0.008, 0.04, and 0.1; set the value of the auxiliary input signal; the fault start time is the 60th second (the data of the first 20 seconds are discarded due to the large state fluctuations in the train start section), and the fault is injected into the software through the fault injection module of the Simpack-Matlab / Simulink joint simulation platform to establish a fault model;

[0156] Step A2: Set the characteristic information of the secondary spring in the suspension system, the failure rate is 10%, and the early fault of the secondary spring with the fault evolution rate of 0.01, 0.04, and 0.1; set the value of the auxiliary input signal; the fault start time is the 60th second (the data of the first 20 seconds are discarded due to the large state fluctuation of the train start section), and the fault is injected into the software through the fault injection module of the Simpack-Matlab / Simulink joint simulation platform to establish a fault model;

[0157] Step A3: Set the characteristic information of the secondary damper in the suspension system, the failure rate is 10%, and the early fault of the secondary damper with the fault evolution rate of 0.01, 0.04, and 0.1; set the value of the auxiliary input signal; the fault start time is the 60th second (the data of the first 20 seconds are discarded due to the large state fluctuation of the train start section), and the fault is injected by software through the fault injection module of the Simpack-Matlab / Simulink joint simulation platform to establish a fault model;

[0158] Step A4: Import the track-train coupling model established in SIMPACK into Matlab / Simulink, and establish a train vertical suspension system control simulation model in Simulink. Set the vehicle speed to 250 km / h, the simulation time to 120 seconds, and generate a total of 120,000 sets of sample data. The fault occurs at the 60th second, and the first 20 seconds of data are discarded due to the large state fluctuations in the train start section. The fault types include:

[0159] The first fault: the actuator on the right side of the front bogie has an early fault, θ = 0.1, β = 0.1;

[0160] The second fault: the actuator on the right side of the front bogie has an early fault, θ = 0.1, β = 0.04;

[0161] The third fault: the actuator on the right side of the front bogie has an early fault, θ = 0.1, β = 0.008;

[0162] The fourth fault: the secondary spring on the right side of the front bogie has an early fault, θ = 0.1, β = 0.1;

[0163] The fifth fault: the secondary spring on the right side of the front bogie has an early fault, θ = 0.1, β = 0.04;

[0164] The sixth fault: the secondary spring on the right side of the front bogie has an early fault, θ = 0.1, β = 0.01;

[0165] The seventh fault: the secondary damper on the right side of the front bogie has an early fault, θ = 0.1, β = 0.1;

[0166] The eighth fault: the secondary damper on the right side of the front bogie has an early fault, θ = 0.1, β = 0.04;

[0167] The ninth fault: An early fault occurred in the secondary damper on the right side of the front bogie, θ=0.1, β=0.01.

[0168] Figure 2-10 J _aux Corresponding to the auxiliary signal; J _without_aux Corresponding to no auxiliary signal; J th Corresponding threshold. Figure 2-10 It can be seen that compared with the existing passive detection framework method, the method of the present invention can more effectively realize the detection of early faults in the suspension system of high-speed trains, and more effectively solve the detection of early faults under the closed-loop control structure and its engineering practical problems, which is of great significance for the early warning and real-time monitoring of high-speed train suspension faults.

[0169] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. An active early fault detection method for a high-speed train suspension system, characterized in that: The steps include: Step 1: Use TS fuzzy to perform data modeling of the actual model and nominal model of the high-speed train suspension system; Step 2: Construct the input and output data matrix for data modeling; Step 3: Design ToMFIR residual according to the data model and input and output data matrix; Step 4: construct an auxiliary signal and an auxiliary signal distributor, and optimize the ToMFIR residual through the auxiliary signal and the auxiliary signal distributor; Step 5: Design an evaluation function and detect faults using the ToMFIR residual obtained in step 4 and the evaluation function.

2. The active early fault detection method for a high-speed train suspension system according to claim 1, characterized in that: In step 3, the ToMFIR residual is designed by using an identification technique based on a combination of LQ decomposition and least squares method.

3. The active early fault detection method for a high-speed train suspension system according to claim 1, characterized in that: In step 3, the designed ToMFIR residual is: ToMFIR(k)=r y,l (k)-H(μ)r u,p,l (k) In the formula, μ i (θ(k)) represents the membership function, satisfying μ i (θ(k))≥0 and 4. The active early fault detection method for a high-speed train suspension system according to any one of claims 1 to 3, characterized in that: In step 4, the auxiliary signal and auxiliary signal distributor constructed are: Wherein, η is a constant; α1 and α2 are m×1 dimensional matrix vectors, and α1+α2=0.

5. The active early fault detection method for a high-speed train suspension system according to claim 4, characterized in that: The step 4 also includes: constructing a redundant actuator, the constructed redundant actuator is: B2, let B2=B1, wherein B1 and B2 are both n×m dimensions, and B1 is the input matrix coefficient of the normal system.

6. The active early fault detection method for a high-speed train suspension system according to claim 5, characterized in that: In step 4, the optimized ToMFIR residual is: in, 7. The active early fault detection method for a high-speed train suspension system according to claim 1, characterized in that: In step 5, an evaluation function based on the Hellinger distance is designed.

8. The active early fault detection method for a high-speed train suspension system according to claim 7, characterized in that: The evaluation function is: In the formula, in, in,

Images