Multi-target fault detection method for aero-engine actuators

Abstract

This invention relates to a multi-target fault detection method for aero-engine actuators, belonging to the field of aero-engine fault diagnosis technology. The invention combines an adaptive event-triggered mechanism with a fault detection observer, while also considering network transmission delay effects, thus achieving a balance between saving network bandwidth resources and ensuring system fault detection performance. Design criteria for the multi-target fault detection observer are then presented, ensuring the robustness of the residual signal to external disturbances and its sensitivity to fault signals. Finally, an aero-engine control system model is used to verify the effectiveness of the proposed multi-target fault detection method based on the adaptive event-triggered mechanism.

Description

Technical Field

[0001] This invention belongs to the field of aero-engine fault diagnosis technology, and relates to fault diagnosis methods for aero-engine distributed control systems, particularly to a multi-target fault detection method for control system actuators based on an adaptive event triggering mechanism. Background Technology

[0002] As a future development direction for aero-engine control systems, distributed control has attracted widespread attention from researchers in recent years due to its advantages such as lightweight structure and ease of maintenance and upgrades. Considering the high risk of failure for intelligent sensors and actuators in distributed control systems under the harsh environment of high temperature and vibration in engines, research on fault detection in aero-engine distributed control systems has high practical value. Aero-engine distributed control is a type of networked control system, and theoretical research on fault detection in networked control systems has made some progress. For example, the paper "Event-Triggered Multiobjective Control and Fault Diagnosis: A Unified Framework" studied the multi-objective fault detection problem of actuators in linear discrete systems. However, research results on multi-objective fault detection of actuators in networked control systems are still relatively few, and existing methods in the literature do not consider the impact of network transmission delay on the fault detection results when performing multi-objective fault detection on actuators in networked control systems such as aero-engine distributed control, which differs somewhat from reality. Summary of the Invention

[0003] To conserve network bandwidth and reduce the impact of network bus transmission latency on the fault detection performance of actuators in aero-engine distributed control systems, this invention proposes a multi-objective fault detection method for aero-engine distributed control systems that considers network transmission latency. This method can adjust the threshold values ​​in the event triggering conditions according to the bandwidth occupancy in the communication network, thereby fully utilizing the network bandwidth. Furthermore, it enables fault residuals to be sensitive to fault signals while being robust to external disturbances.

[0004] The technical solution of this invention:

[0005] This invention combines an adaptive event-triggered mechanism with a fault detection observer, while considering network transmission delay effects, thus achieving a balance between saving network bandwidth resources and ensuring system fault detection performance. Then, design criteria for the multi-objective fault detection observer are presented, ensuring the robustness of the residual signal to external disturbances and its sensitivity to fault signals. Finally, a model of an aero-engine control system is used to verify the effectiveness of the proposed multi-objective fault detection method based on the adaptive event-triggered mechanism. The mathematical symbols used in this invention are defined as follows:

[0006] X T Let X represent the transpose of matrix X. -1 Let [X] denote the inverse of matrix X. s =X+X T The asterisk (*) is an abbreviated representation of the symmetric position in a symmetric matrix; diag{...} represents a block diagonal matrix, and col{...} represents a column vector composed of elements.

[0007] A multi-target fault detection method for aero-engine actuators, comprising the following steps:

[0008] S1. Establish a state-space model of the controlled object of the aero-engine actuator;

[0009] Consider the following continuous-time control system:

[0010]

[0011] Where x(t)∈R n Indicates the system status. The first derivative of the system state; f(t)∈R f Denotes an additive fault in the actuator, w(t)∈R w Let y(t) ∈ R represent external disturbances to the system. p Let A, B, C, and D represent the system's measured output, and let D represent the system's constant matrix. Assume that the system is observable and controllable.

[0012] S2. Establish adaptive event triggering conditions;

[0013] The adaptive event triggering mechanism used is as follows:

[0014]

[0015] Among them, e y (t)=y(ih)-y(t k h), ih=t k h+lh, i, l∈N, Ω is the event triggering weight matrix, and δ(t) is the event triggering parameter, the expression of which is as follows:

[0016]

[0017] Here, δ0 is a pre-selected constant threshold. It should be noted that compared to event triggering parameters using a fixed threshold, adaptive event triggering parameters can change with the triggering error and can adaptively reconstruct based on external disturbances, thus achieving a balance between ensuring system performance and conserving network resources. Furthermore, in adaptive event triggering conditions, since the event triggering interval is greater than or equal to the sampling period h, the Zeno phenomenon is fundamentally avoided.

[0018] Considering network communication delay and the characteristics of the zero-order hold, the input y(t) of the fault detection observer is y(t) = y(t) k h), t∈[t k h+η k , t k+1 h+η k+1 ), where η k Represents network transmission delay, k∈N, satisfying 0≤η k ≤η M , where η M This represents the maximum network transmission delay. The time interval [t] is used to represent this delay. k h+η k , t k+1 h+η k+1 The following division is performed:

[0019]

[0020] in, Define the piecewise time-varying delay τ(t) and the trigger error vector e. y (t) are respectively:

[0021]

[0022] Therefore, in the interval [t] k h+η k , t k+1 h+η k+1 On the fault detection observer, the input y(t) = y(t-τ(t)) + e y (t). Define variable τ M =h+η M , τ m =0, d=τ M -τ m Then the piecewise time-varying delay τ(t) satisfies 0≤τ(t)≤τ M .

[0023] S3. Design a fault detection observer based on aero-engine actuators and construct an augmented error dynamic system;

[0024] The fault detection observer based on aero-engine actuators is constructed as follows:

[0025]

[0026] in, This represents the state variables of the fault detection observer. This represents the output of the fault detection observer, where L and V are the gain matrices of the fault detection observer to be designed. It represents the first derivative of the observer's state variable.

[0027] Define variables ξ(t)=col{e x If x(t)}, d(t) = col{f(t), w(t)}, then the augmented error dynamic system can be expressed as:

[0028]

[0029] in

[0030]

[0031] S4. Obtain the fault detection observer gains L and V by solving the following multi-objective optimization problem;

[0032] (1) The residual system is robust to external disturbances:

[0033]

[0034] (2) Given the weight matrix Γ, the residual system is sensitive to fault signals:

[0035]

[0036] Where l2 represents the space of square-integrable vector functions in the non-negative range, and γ1 and γ2 represent H ∞ Performance metrics, H ∞ The design assumes that interference and noise signals have bounded energy in the time domain.

[0037] To detect and diagnose faults in aero-engine actuators, it is necessary to design a residual evaluation function z(t) and a corresponding fault threshold z. th When the residual evaluation function z(t) is greater than the fault threshold z th If a fault occurs, the system is considered to have a fault; otherwise, it is considered that no fault has occurred in the current system.

[0038] The expression for the residual evaluation function z(t) is shown below:

[0039]

[0040] The corresponding threshold is selected as follows:

[0041]

[0042] S5. First, consider the fault-free augmented residual system when f(t) = 0. Theorem 1 is given below to make the residual signal robust to external disturbances;

[0043] Theorem 1 Given parameters h, τ M γ1, δ0 are positive numbers. If there exist positive definite matrices P, Q, N, X1, X2, Ω, and matrices Z, V, X satisfying Make

[0044]

[0045] in, Ξ 13 =-Z, Ξ 14 =[-X;X], Ξ 15 = [0; -X1D + X2D], I1 = [-I, I], Ξ 23 =Z+N, Ξ 33 =-QN, Ξ 44 =-δ0Ω,Ξ 46 =τ M [-X;X] T Ξ 47 =V T , Ξ 56 =τ M [0;-X1D+X2D] T Ξ 66 =NPP T Ξ 77 =-I,

[0046] Then the error dynamic system (7) is asymptotically stable and satisfies constraint (8-a). Meanwhile, if the inequality in equation (10) has a feasible solution, then the fault detection observer gain matrix L is obtained from... To obtain.

[0047] Proof: The Lyapunov-Krasovskii function is constructed as follows:

[0048]

[0049] Differentiating the Lyapunov-Krasovskii function (11) with respect to time yields:

[0050]

[0051] in,

[0052]

[0053] For any matrix Z, the following holds: The following inequalities hold:

[0054]

[0055] Among them, λ1(t)=col{ξ(t-τ(t))-ξ(t), ξ(t-τ(t))-ξ(t-τ M )}.

[0056] The auxiliary function is constructed as follows:

[0057]

[0058] Define variables: ξ1(t)=col{ξ(t),ξ(t-τ(t)),ξ(t-τ M ), ey(t), w(t)}, combining (12)-(15), and applying Schur's complement lemma and congruence transformation, we obtain the following inequality:

[0059]

[0060] in,

[0061]

[0062] Ξ′ 12 =N+Z+PB f ,Ξ′ 13 =-Z,

[0063] Ξ′ 14 =PL f ,Ξ′ 15 =PD f ,

[0064] Ξ′ 23 =Z+N,

[0065] Ξ′ 33 =-QN,Ξ′ 44 =-δ0Ω, Ξ′ 47 =V T ,

[0066] Ξ′ 55 =-γ1, Ξ′ 66 =NPP T ,Ξ′ 77 =-I.

[0067] make X = X1L, thus obtaining Ξ < 0 in Theorem 1. When w(t) = 0, we obtain... This proves the asymptotic stability of the augmented system (7). Integrating both sides of inequality (16) from t = 0 to ∞, under zero initial conditions, we obtain (8-a).

[0068] S6. Next, consider the fault-free augmented residual system when w(t) = 0. Theorem 2 is given below to make the residual signal sensitive to actuator faults;

[0069] Theorem 2 Given parameters h, τ M γ2 and δ0 are positive numbers. If there exist positive definite matrices P1, Q1, N1, W1, W2, Ω, and matrices Z1, V, W satisfying Make

[0070]

[0071] in, Ψ 13 =-Z1, Ψ 14 =[-W;W],Ψ 15 = [0; -W1B + W2B], I1 = [-I, / ], Ψ 23 =Z1+N1, Ψ 33 =-Q1-N1,Ψ 44 =-δ0Ω,Ψ 46 =τ M [-W;W] T Ψ 47 =V T , Ψ 56 =τ M [0;-W1B+W2B] T Ψ 57 =-Γ T Ψ 66 =N1-P1-P1 T Ψ 77 =-I,

[0072] Then the error dynamic system (7) is asymptotically stable and satisfies constraint (8-b). Meanwhile, if the inequality in equation (17) has a feasible solution, then the fault detection observer gain matrix L is obtained from L = W1.-1 W is obtained.

[0073] Proof: The Lyapunov-Krasovskii function is constructed as follows:

[0074]

[0075] Differentiate the Lyapunov-Krasovskii function with respect to time.

[0076]

[0077]

[0078] For any matrix Z, the following holds: The following inequalities hold:

[0079]

[0080] Among them, λ1(t)=col{ξ(t-τ(t))-ξ(t), ξ(t-τ(t))-ξ(t-τ M )}.

[0081] The auxiliary function is constructed as follows:

[0082]

[0083] Define variables: ξ2(t)=col{ξ(t), ξ(t-τ(t)), ξ(t-τ M ), e y Combining (19) and (22), and applying Schur's complement lemma and congruent transformation, we obtain the following inequality:

[0084]

[0085] in,

[0086]

[0087] Ψ′ 12 =N1+Z1+P1B f Ψ′ 13 =-Z1,

[0088] Ψ′ 14 =P1L f Ψ′ 15 =P1D f ,

[0089] Ψ′ 23=Z1+N1,

[0090] Ψ′ 33 =-Q1-N1,Ψ′ 44 =-δ0Ω, Ψ′ 47 =V T ,

[0091] Ψ′ 55 =-γ2, Ψ′ 57 =-Γ T Ψ′ 66 =N1-P1-P1 T Ψ′ 77 =-I,

[0092] make W = W1L, thus obtaining Ψ < 0 in Theorem 2. When f(t) = 0, we obtain... This proves the asymptotic stability of the augmented system (7). Integrating both sides of inequality (23) from t = 0 to ∞, we obtain (8-b) under zero initial conditions.

[0093] S7. Based on Theorems 1 and 2 given above, Theorem 3 is given below to make the residual signal sensitive to actuator faults while being robust to external disturbances.

[0094] Theorem 3 Given parameters β1, β2, h, τ M Let γ1, γ2, δ0 be positive numbers. If there exist positive definite matrices Ω, P, P1, Q, Q1, N, N1, U1, and matrices Z1, Z2, V, W, X, U, satisfying

[0095] Make:

[0096]

[0097] in, and The augmented residual system (7) is asymptotically stable and has the characteristics of being robust to external disturbances and sensitive to actuator faults. The gain matrix of the fault detection observer is obtained by solving the following convex optimization problem.

[0098]

[0099] At this time, the fault detection observer gain

[0100] The beneficial effects of this invention are as follows: By introducing an adaptive event triggering mechanism, this invention designs a multi-target fault detection observer that takes into account network transmission delay; it can ensure that the fault residuals of the aero-engine control system actuators are robust to external disturbances while being sensitive to fault signals; at the same time, due to the adoption of the adaptive event triggering mechanism, network bandwidth resources can be utilized more effectively, further reducing network communication load. Attached Figure Description

[0101] Figure 1 This is a flowchart of a multi-target fault detection method based on an adaptive event triggering mechanism.

[0102] Figure 2 The image shows the effect of fault detection for the fuel valve actuator. As can be seen from the image, the evaluation function exceeds the fault threshold at 82 seconds. At this time, the fault is detected by the system and an alarm will be issued.

[0103] Figure 3 This is a graph showing how the adaptive event triggering parameters change over time when the fuel valve actuator malfunctions.

[0104] Figure 4 To determine the triggering time and interval under an adaptive event triggering mechanism when the fuel valve actuator malfunctions.

[0105] Figure 5 The diagram shows the effect of fault detection for the tail nozzle actuator. As can be seen from the diagram, the evaluation function exceeds the fault threshold at 98 seconds. At this time, the fault is detected by the system and an alarm will be issued.

[0106] Figure 6 This is a graph showing how the adaptive event triggering parameters change over time when the nozzle actuator malfunctions.

[0107] Figure 7 To determine the triggering time and interval under an adaptive event triggering mechanism when the tail nozzle actuator malfunctions. Detailed Implementation

[0108] The technical solution of the present invention will be further described below with reference to the technical solution and the accompanying drawings.

[0109] The flowchart of the multi-target fault detection method based on the actuators of the aero-engine control system is as follows: Figure 1 As shown. The experimental simulation verification is as follows:

[0110] This invention will use a model of an aero-engine control system to verify the effectiveness of the proposed multi-target fault detection method based on an adaptive event triggering mechanism. The output variables are: y = [ΔN², Δπ]. T ] T Where N2 is the high-voltage rotor speed, πT This represents the turbine outlet pressure ratio. The control input variables are selected as: u = [ΔWFM, ΔA8] T Where WFM is the main fuel flow rate and A8 is the tailpipe area.

[0111] When the fuel valve actuator experiences a drift failure, its system matrix is ​​as follows:

[0112]

[0113] Assume a sampling period of 0.025s, an upper bound of network-induced time delay of 0.04s, and δ0 = 20. According to Theorem 3, solve the following multi-objective optimization problem:

[0114] minλ1γ1+λ2γ2

[0115] st constraint(8)

[0116] Where, λ i (i = 1, 2) represent the weighting coefficients. Choosing λ1 = 3, λ2 = 1, γ1 = 1.5, and γ2 = 1.2, the fault detection observer gain and event triggering weight matrix can be obtained as follows:

[0117] V = [0.1563 0.1414],

[0118] Consider the following external disturbances to the aircraft engine control system:

[0119]

[0120] Consider the offset fault signal f(t) occurring in the fuel valve actuator of the aircraft engine as shown below:

[0121]

[0122] When the aforementioned drift fault occurs in the fuel valve actuator of an aircraft engine, the multi-target fault detection method based on the adaptive event triggering mechanism described in this invention achieves the following results: Figure 2 As shown; the corresponding adaptive event triggering parameters change over time as follows: Figure 3 As shown; Figure 4 This reflects the trigger interval and trigger time under the adaptive event triggering mechanism when an offset failure occurs in the fuel valve actuator of an aircraft engine. At this time, the number of packets sent within 200 seconds is 329; while if the traditional periodic event triggering mechanism is used, the number of packets sent within 200 seconds is 414.

[0123] When the nozzle actuator experiences a drift failure, its system matrix is ​​as follows:

[0124]

[0125] Similarly, the fault detection observer gain and event triggering weight matrix for the corresponding cases can be obtained:

[0126] V = [0.1593 0.1543],

[0127] The drift fault signal f(t) of the aero-engine exhaust nozzle actuator is considered as follows:

[0128]

[0129] When the aero-engine exhaust nozzle actuator experiences the aforementioned drift fault, the multi-target fault detection method based on the adaptive event triggering mechanism described in this invention achieves the following results: Figure 5 As shown; the corresponding adaptive event triggering parameters change over time as follows: Figure 6 As shown; Figure 7 This reflects the trigger interval and trigger time under the adaptive event triggering mechanism when a drift fault occurs in the aero-engine exhaust nozzle actuator. In this case, the number of packets sent within 200 seconds is 111; if the traditional periodic event triggering mechanism is used, the number of packets sent within 200 seconds is 178. Therefore, it can be seen that compared with the traditional periodic event triggering mechanism, the multi-target fault detection method based on the actuator of the aero-engine distributed control system proposed in this invention can further reduce the load on network communication resources while ensuring the fault detection effect.

Claims

1. A method for multi-objective fault detection of an aeroengine actuator, characterized in that, The steps are as follows: S1. Establish a state-space model of the controlled object of the aero-engine actuator; Consider the following continuous-time control system: (1) in, Indicates the system status. The first derivative represents the system state; This indicates an additive failure in the actuator. This indicates external disturbances to the system. Indicates the system measurement output, A , B , C , D Let represent the constant matrix of the system; assume the system is observable and controllable; S2. Establish adaptive event triggering conditions; The adaptive event triggering mechanism used is as follows: ， (2) in, , , , For the event triggering weight matrix, The event trigger parameter has the following expression: (3) in, It is a pre-selected constant threshold; Considering network communication delay and the characteristics of the zero-order hold, the input of the fault detection observer... ,in Indicates network transmission latency. ,satisfy ,in Indicates the maximum network transmission delay; divides the time interval Perform the following segmentation: (4) in, Define piecewise time-varying delays and trigger error vector They are respectively: (5) Therefore, in the interval Above, the fault detection observer input Define variables , , Then the segmented time-varying time delay satisfy ; S3. Design a fault detection observer based on aero-engine actuators and construct an augmented error dynamic system; The fault detection observer based on aero-engine actuators is constructed as follows: (6) in, This represents the state variables of the fault detection observer. This represents the output of the fault detection observer. L, V The gain matrix of the fault detection observer to be designed, The first derivative of the observer's state variable; Define variables , , The augmented error dynamic system is then expressed as: (7) in ， ， ， S4. The gain of the fault detection observer is obtained by solving the following multi-objective optimization problem. L and V ; (1) The residual system is robust to external disturbances: (2) The residual system is sensitive to fault signals: To detect and diagnose faults in aero-engine actuators, it is necessary to design a residual evaluation function and a corresponding fault threshold. When the residual evaluation function is greater than the fault threshold, a fault is considered to have occurred; otherwise, the current system is considered to be without fault. S5. First consider when The fault-free augmented residual system is described below; Theorem 1 is given below to make the residual signal robust to external disturbances. Theorem 1 Given parameters , , , It is a positive number; if a positive definite matrix exists. , There exists a matrix satisfy , making (8) in, The error dynamic system (7) is asymptotically stable and satisfies the condition that the residual system is robust to external disturbances; meanwhile, if the inequality in equation (8) has a feasible solution, then the fault detection observer gain matrix... L Depend on Seek; S6. Then consider when The fault-free augmented residual system is described below; Theorem 2 is given below to make the residual signal sensitive to actuator faults; Theorem 2 Given parameters , , , It is a positive number; if a positive definite matrix exists. , There exists a matrix satisfy , making (9) in, The error dynamic system (7) is asymptotically stable and satisfies the requirement that the residual system is sensitive to fault signals; meanwhile, if the inequality in equation (9) has a feasible solution, then the gain matrix of the fault detection observer is... L Depend on Seek; S7. Based on Theorems 1 and 2 given above, Theorem 3 is given below to make the residual signal sensitive to actuator faults while being robust to external disturbances. Theorem 3 Given parameters It is a positive number; if a positive definite matrix exists. , , There exists a matrix , , satisfy , Make: (10) in, and From Theorem 1 and Theorem 2 respectively , By replacement for , for Thus, the augmented residual system (7) is asymptotically stable and simultaneously possesses the characteristics of being robust to external disturbances and sensitive to actuator faults. The gain matrix of the fault detection observer is obtained by solving the following convex optimization problem. (11) At this time, the fault detection observer gain .

Images