CN114771866A - A long-endurance UAV fault detection method triggered by dynamic events - Google Patents

Abstract

The invention discloses a long-duration unmanned aerial vehicle fault detection method triggered by a dynamic event, and belongs to the technical field of unmanned aerial vehicle fault detection. The method of the invention considers the unmanned aerial vehicle system with wind interference and fault, adopts the dynamic event trigger mechanism to reduce the occupation of the communication resources of the unmanned aerial vehicle in the network environment, and the method can realize the complete decoupling of the residual error and the transmission error triggered by the dynamic event , to eliminate the error caused by event triggering, and the designed fault filter can calculate the optimal solution online.

Description

A long-endurance UAV fault detection method triggered by dynamic events

technical field

The invention belongs to the technical field of unmanned aerial vehicle fault detection, and in particular relates to a long-duration unmanned aerial vehicle fault detection method triggered by dynamic events.

Background technique

UAVs are increasingly used in military, transportation and wireless communication fields due to their unique advantages. With the wide application of UAVs, it is particularly important to ensure the safety and reliability of UAV flight control systems. Rapid detection of faults is an important prerequisite for ensuring UAV system safety and reducing economic losses.

For long-endurance flying UAVs, it is often necessary to exchange data with the ground station through the communication network, so as to implement the fault detection algorithm in the ground station computer, which is a typical networked control system. A communication network is required to realize data transmission between the ground station and the UAV, and continuous communication is bound to waste limited network resources.

In the event-triggered fault detection process, there is an error between the data at the non-triggered moment and the actual system data, that is, the event transmission error, which will inevitably affect the fault detection performance. Under the dynamic event-triggered mechanism, it is crucial to avoid the influence of event-triggered transmission errors on the residual signal of the faulty filter.

SUMMARY OF THE INVENTION

Aiming at the above technical problems existing in the prior art, the present invention proposes a long-endurance UAV fault detection method triggered by a dynamic event, which has a reasonable design, overcomes the deficiencies of the prior art, and has good effects.

In order to achieve the above object, the present invention adopts the following technical solutions:

A long-duration UAV fault detection method triggered by a dynamic event, comprising the following steps:

Step 1: Establish a non-uniform sampling period model of the UAV under the dynamic event trigger mechanism;

The nonlinear attitude discrete system model of the UAV considering the actuator fault is:

In the formula, k represents the sampling time of the discrete system, x=[ω _x ,ω _y ,ω _z ] ^T , y=[γ,ψ,θ] ^T , u=[δ _x ,δ _y ,δ _z ] ^T ,d (k)=[w _xg , w _yg , w _zg ], ω _x , ω _y , ω _z are the roll angular velocity, yaw angular velocity and pitch angular velocity of the drone in the body coordinate system, respectively; γ, ψ, θ are respectively are the roll, yaw and pitch angles of the aircraft; δ _x is the aileron deflection angle, δ _y is the rudder deflection angle, and δ _z is the elevator deflection angle; w _xg , w _yg , and w _zg are the wind disturbances along the body axis gradient;

u _f =l ₁ u+l ₂ , u is the control input, u _f is the actual input, l ₁ is the diagonal matrix, l ₂ is the rudder surface deviation fault, f, g _u , g _d are nonlinear functions, C and D _d is the matrix of the corresponding dimension;

For discrete UAV nonlinear systems, in the time series [k ₀ , k ₁ , ..., k _i , ...], the dynamic event trigger decides whether to apply the sampled UAV control input u(k ) and the measurement output y(k) are transmitted to the ground station through the wireless network, and save the most recently transmitted data packets; the ground station uses these data to complete fault detection, and the sampling time sequence is triggered by the dynamic event trigger mechanism (2) of the following formula:

为参数；In the formula, _ki is the latest event trigger time, Ω∈R ^q×q is the dynamic event trigger weight matrix, σ>0 is the event trigger threshold,

is a parameter;

η(k) is a positive definite internal dynamic variable that satisfies the following differential equation:

where φ is the local Lipchitz continuous κ _∞ function, and η ₀ ∈ R is the parameter to be designed;

y(k _i ) is the latest transmission measurement output, and y(k) is the current sampling data; if equation (2) is established, y(k) is y(k _i+1 ), and it is sent to the fault detection module; the fault detection module The transmission input data y(k _i ) of is updated through dynamic event trigger condition (2);

处进行泰勒级数展开，并略去其高阶项，则模型(1)写为：Put the UAV nonlinear attitude system model (1) in

Perform Taylor series expansion at , and omit its higher-order terms, then model (1) is written as:

in,

According to the UAV linear attitude system model (4), the system state at the event trigger time k _i+1 is expressed as the following formula (5):

Restate equation (5) as:

υ(k_i)＝[υ^T(k_i) υ^T(k_i+1) ... υ^T(k_i+1-1)]^T，υ代表w，u_f和d，D _d(k)＝[D_d 0 ... 0]^T；In the formula,

υ ( _ki )=[υ ^T ( _ki ) υ ^T ( _ki +1) ... υ ^T (ki ₊₁ -1)] ^T , υ stands for w, u _f and d, D _d (k )=[D _d 0 ... 0] ^T ;

Step 2: Design a dynamic event-triggered H _i /H _∞ fault detection filter;

The system state estimation of the design event trigger time k _i+1 is shown in formula (7):

为状态x(k_i)的估计向量，

为输出y(k_i)的估计向量，r(k_i)为生成的残差信号，L(k_i)∈R^n×q为观测器增益矩阵，W(k_i)∈R^q×q为后置滤波器加权矩阵，动态事件触发的y(k_i)驱动残差发生器工作；In the formula, u ( _ki )=[u ^T ( _ki ) u ^T ( _ki ) ... u ^T ( _ki )] ^T ,

is the estimated vector of state x(k _i ),

is the estimated vector of the output y(k _i ), r(k _i ) is the generated residual signal, L(k _i )∈R ^n×q is the observer gain matrix, W(k _i )∈R ^q×q is Post filter weighting matrix, y(k _i ) triggered by dynamic events drives the residual generator to work;

The observer gain matrix L(k _i ) and the post-filter weight matrix W(k _i ) are designed according to the following formulas:

Step 3: Perform data processing on the residual r(k _i ) generated in the previous step to obtain a fault detection result;

Define the residual evaluation function as formula (12):

In the formula, _ki is the current event trigger time, and N is the moving time window;

The residual threshold is determined according to formula (13):

的均值和均方差：In the formula, i=1,2,...,M, M is the triggering times of the dynamic event triggering condition (2), which is calculated according to the following formula

The mean and mean squared deviation of :

The threshold value J _th is determined according to formula (14):

According to formulas (12) and (14), the residual evaluation logic is expressed as the following formula (15):

When the residual evaluation function J _N (r(k _i )) is greater than the threshold value J _th , it is judged that the UAV has malfunctioned during flight and a signal indication is given; when the residual evaluation function J _N (r(k _i )) is less than or equal to When the threshold value is J _th , it is judged that the drone is flying normally.

Beneficial technical effects brought by the present invention:

The method of the invention takes into account the unmanned aerial vehicle system with wind interference and faults, adopts the dynamic event trigger mechanism to reduce the occupation of the communication resources of the unmanned aerial vehicle in the network environment, and the method can realize the complete decoupling of the residual error and the transmission error triggered by the dynamic event , to eliminate the error caused by event triggering, and the designed fault filter can calculate the optimal solution online.

Description of drawings

Figure 1 is a schematic structural diagram of a dynamic event-triggered UAV fault detection method;

Fig. 2 is a method flow chart of a dynamic event-triggered UAV fault detection method;

Fig. 3 is the dynamic event trigger sequence diagram of the unmanned aerial vehicle system in the example of the present invention;

FIG. 4 is a failure detection diagram of an unmanned aerial vehicle system in an embodiment of the present invention.

Detailed ways

The present invention is described in further detail below in conjunction with the accompanying drawings and specific embodiments:

As shown in Figure 1, a dynamic event-triggered long-endurance UAV fault detection method adopts a controller, an actuator, an UAV, a sensor, a dynamic event trigger module, a wireless communication network and a fault detection module; control The actuator, the actuator, the drone, the sensor, and the dynamic event trigger module are connected in sequence through the line; the dynamic event trigger module is connected with the fault detection module through the wireless communication network; the controller, the actuator, the drone, and the sensor form a closed loop; One end of the dynamic event triggering module is connected to the common end of the controller and the actuator; the fault detection module is set at the ground station;

The control input signal u(k) output by the controller and the measurement output signal y(k) output by the sensor are screened by the dynamic event trigger module and transmitted to the fault detection module of the ground station through the wireless communication network.

A dynamic event triggering mechanism decides whether to transmit the sampled control input and measurement output to the ground station, and saves the most recently transmitted data packets. The ground station uses this data to complete fault detection. Since the construction of the dynamic event-triggered fault detection filter requires the relevant information of the system control input, the control input u(k) and the measurement output y(k) of the UAV attitude control system are packaged and transmitted to the ground station through the dynamic event-triggered module. Fault detection module. The dynamic event-triggered fault detection method proposed in this paper is to use the available u(k _i ) and y(k _i ) to construct a residual generator and a residual evaluation function to eliminate the influence of event-triggered transmission errors on the residual signal. as shown in picture 2.

Step 1: Establish a non-uniform sampling period model of the UAV under the dynamic event trigger mechanism;

The nonlinear attitude discrete system model of the UAV considering the actuator fault is:

In the formula, k represents the sampling time of the discrete system, x=[ω _x ,ω _y ,ω _z ] ^T , y=[γ,ψ,θ] ^T , u=[δ _x ,δ _y ,δ _z ] ^T ,d (k)=[w _xg , w _yg , w _zg ], ω _x , ω _y , ω _z are the roll angular velocity, yaw angular velocity and pitch angular velocity of the drone in the body coordinate system, respectively; γ, ψ, θ are respectively are the roll, yaw and pitch angles of the aircraft; δ _x is the aileron deflection angle, δ _y is the rudder deflection angle, and δ _z is the elevator deflection angle; w _xg , w _yg , and w _zg are the wind disturbances along the body axis gradient.

u _f =l ₁ u+l ₂ , u is the control input, u _f is the actual input, l ₁ is the diagonal matrix, l ₁ =I means no multiplicative fault occurs, if l ₁ ≠I and the corresponding diagonal If the element is between (0, 1), it means that the corresponding rudder surface has a control effect loss fault, which is a multiplicative fault, and l ₂ represents a rudder surface deviation fault, which is an additive fault.

f, g _u , g _d are nonlinear functions, and C and D _d are matrices of corresponding dimensions.

The sensor passes the drone output y(k) to the dynamic event trigger. For discrete UAV systems, in the time series [k ₀ , k ₁ , ..., k _i , ...], the dynamic event triggers decide whether to combine the sampled UAV control inputs u(k) and The measurement output y(k) is transmitted to the ground station through the wireless network, and saves the most recently transmitted data packets; the ground station uses these data to complete fault detection, and the sampling time sequence is triggered by the dynamic event trigger mechanism (2) of the following formula:

为参数；In the formula, _ki is the latest event trigger time, Ω∈R ^q×q is the dynamic event trigger weight matrix, σ>0 is the event trigger threshold,

is a parameter;

η(k) is a positive definite internal dynamic variable that satisfies the following differential equation:

where φ is the local Lipchitz continuous κ _∞ function, and η ₀ ∈ R is the parameter to be designed;

y(k _i ) is the latest transmission measurement output, and y(k) is the current sampling data; if equation (2) is established, y(k) is y(k _i+1 ), and it is sent to the fault detection module; the fault detection module The transmission input data y(k _i ) of is updated through dynamic event trigger condition (2);

Since some data will be lost under dynamic event triggering, there is a difference between the non-triggered time data of the fault detection module and the actual system data. This data difference is defined as the event transmission error e _y (k):

处进行泰勒级数展开，并略去其高阶项：Put the UAV nonlinear attitude system model (1) in

Taylor series expansion is performed at , and its higher-order terms are omitted:

make

Then formula (1) can be reformulated as:

According to the UAV linear attitude system model (4), the system state at the event trigger time k _i+1 is expressed as the following formula (5):

Restate equation (5) as:

υ(k_i)＝[υ^T(k_i) υ^T(k_i+1) ... υ^T(k_i+1-1)]^T，υ代表w，u_f和d，D _d(k)＝[D_d 0 ... 0]^T；In the formula,

υ ( _ki )=[υ ^T ( _ki ) υ ^T ( _ki +1) ... υ ^T (ki ₊₁ -1)] ^T , υ stands for w, u _f and d, D _d (k )=[D _d 0 ... 0] ^T ;

Formula (6) represents the system model of the UAV fault model at the time of dynamic event triggering [ _ki , ki ₊₁ ).

Step 2: Design a dynamic event-triggered H _i /H _∞ fault detection filter;

The state estimation of the UAV system at the design event trigger time k _i+1 is shown in formula (7):

为状态x(k_i)的估计向量，

为输出y(k_i)的估计向量，r(k_i)为生成的残差信号，L(k_i)∈R^n×q为观测器增益矩阵，W(k_i)∈R^q×q为后置滤波器加权矩阵，动态事件触发的y(k_i)驱动残差发生器工作；In the formula, u ( _ki )=[u ^T ( _ki ) u ^T ( _ki ) ... u ^T ( _ki )] ^T ,

is the estimated vector of state x(k _i ),

is the estimated vector of the output y(k _i ), r(k _i ) is the generated residual signal, L(k _i )∈R ^n×q is the observer gain matrix, W(k _i )∈R ^q×q is Post filter weighting matrix, y(k _i ) triggered by dynamic events drives the residual generator to work;

公式(7)减去公式(6)得到状态估计误差方程如下：Define the estimated error vector

Formula (7) subtracts formula (6) to obtain the state estimation error equation as follows:

In the formula, F _LC ( _ki ) = F ( _ki )-L( _ki )C,

It can be seen that the fault detection filter achieves a complete decoupling of the residual r(k _i ) from the dynamic event-triggered transmission error e _y (k).

For i∈N, the key parameters L(k _i ) and W(k _i ) of the fault detection filter are calculated by the following formulas respectively, and the calculation process is as follows:

In the formula, P(k _i )>0, which is calculated recursively by the Riccati equation;

Based on the UAV system equation (7) and the definitions of F ( _ki ) and G _d ( _ki ), the key parameters of the fault detection filter, the observer gain matrix L( _ki ) and the post filter, are designed according to the following formulas Weighting matrix W(k _i ):

Step 3: Perform data processing on the residual r(k _i ) generated in the previous step to obtain a fault detection result;

Define the residual evaluation function as formula (12):

In the formula, _ki is the current event trigger time, and N is the moving time window;

The residual threshold is determined according to formula (13):

的均值和均方差：In the formula, i=1,2,...,M, M is the triggering times of the dynamic event triggering condition (2), which is calculated according to the following formula

The mean and mean squared deviation of :

The threshold value J _th is determined according to formula (14):

According to formulas (12) and (14), the residual evaluation logic is expressed as the following formula (15):

When the residual evaluation function J _N (r(k _i )) is greater than the threshold value J _th , it is judged that the UAV has malfunctioned during flight and a signal indication is given; when the residual evaluation function J _N (r(k _i )) is less than or equal to When the threshold value is J _th , it is judged that the drone is flying normally.

Wherein step 1, step 2 and step 3 are all completed at the ground station in the UAV fault detection device.

In step 1, the ground station uses the information received by the wireless communication network to construct a non-uniform sampling model of the UAV;

On this basis, steps 2 and 3 complete the fault detection work in the fault detection module.

It is easy to obtain from the above process, and the designed dynamic event-triggered fault detection method can achieve better detection performance in the network-controlled long-endurance UAV system.

The following will describe the long-endurance UAV fault detection method triggered by the dynamic event proposed by the present invention in combination with the experiments to verify the effectiveness of the method proposed by the present invention.

In the experiment process: take the experimental step size as 90, and simulate in Matlab software through the dynamic event trigger mechanism to transmit the control input and measurement output of the UAV to the fault detection module of the ground station.

Using the fault detection method proposed by the present invention, using Matlab software to generate dynamic event trigger sequence and residual evaluation function. Figures 3 to 4 show the failure situation of the UAV elevator with 10% loss of control effectiveness when k=60. The fault detection method generates a dynamic event trigger sequence and a residual evaluation function.

Figure 3 shows the dynamic event-triggered time sequence of the UAS. Among them, the stem-and-leaf diagram with circles represents the trigger moment, and the height of the stem-leaf represents the sampling period interval between two adjacent triggers. As can be seen from Figure 3, compared with equal period sampling, the dynamic event trigger mechanism can effectively reduce the communication burden and reduce data transmission by about 42.2%.

Figure 4 shows the residual evaluation function under this fault condition. When k=60, the elevator has a 10% loss of control effectiveness. The dynamic event-triggered fault detection method can effectively detect the occurrence of faults when k=62.

To sum up, the present invention transmits the information of the UAV system through the dynamic event triggering mechanism, and completes the fault detection, which can effectively reduce the occupation of network communication resources and achieve a better fault detection effect.

Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the essential scope of the present invention should also belong to the present invention. the scope of protection of the invention.

Claims (1)

1. A method for detecting faults of a long-endurance unmanned aerial vehicle triggered by dynamic events is characterized by comprising the following steps: the method comprises the following steps:

step 1: establishing an unmanned aerial vehicle non-uniform sampling period model under a dynamic event trigger mechanism;

the model of the unmanned aerial vehicle nonlinear attitude discrete system considering the faults of the actuating mechanism is as follows:

where k denotes a discrete system sampling time, and x ═ ω_x,ω_y,ω_z]^T，y＝[γ,ψ,θ]^T，u＝[δ_x,δ_y,δ_z]^T，d(k)＝[w_xg,w_yg,w_zg]，ω_x，ω_y，ω_zRespectively the rolling angular velocity, the yaw angular velocity and the pitch angular velocity of the unmanned aerial vehicle in a body coordinate system; gamma, psi and theta are respectively a rolling angle, a yaw angle and a pitch angle of the airplane; delta_xFor aileron deflection angle, delta_yIs rudder deflection angle, delta_zIs the elevator deflection angle; w is a_xg，w_yg，w_zgThe gradient of wind disturbance along the axis of the body;

u_f＝l₁u+l₂u is a control input, u_fFor actual input, /)₁Is a diagonal matrix,/₂Indicating control plane deviation fault, f, g_u，g_dAs a non-linear function, C and D_dA matrix of corresponding dimensions;

for discrete drone nonlinear systems, in time series k₀，k₁，...，k_i，…]The dynamic event trigger is used for determining whether to transmit the sampled unmanned aerial vehicle control input u (k) and the measurement output y (k) to the ground station through the wireless network and storing a recently transmitted data packet; the ground station uses the data to complete fault detection, and the sampling time sequence is triggered by the following formula dynamic event triggering mechanism (2):

in the formula, k_iFor the most recent event trigger time, Ω ∈ R^q×qFor dynamic event-triggered weighting matrices, σ>0 is the threshold value of the event trigger,

is a parameter;

η (k) is a positive internal dynamic variable satisfying the following differential equation:

wherein phi is local Lipchitz continuous kappa_∞Function η₀E, taking R as a parameter to be designed;

y(k_i) For the most recently transmitted measurement output, y (k) is the current sample data; if formula (2) holds, y (k) is y (k)_i+1) And transmitting to the fault detection module; transmission input data y (k) of fault detection module_i) Updating through a dynamic event trigger condition (2);

the nonlinear attitude system model (1) of the unmanned aerial vehicle is arranged

Where a taylor series expansion is performed and its higher order terms are omitted, the model (1) is written as:

wherein,

according to the linear attitude system model (4) of the unmanned aerial vehicle, at the event triggering moment k_i+1Is expressed as the following formula (5):

restated equation (5) as:

in the formula,

υ(k_i)＝[υ^T(k_i)υ^T(k_i+1)…υ^T(k_i+1-1)]^Tv represents w, u_fAnd (d) a second step of,D _d(k)＝[D_d 0 … 0]^T；

step 2: design dynamic event trigger H_i/H_∞A fault detection filter;

design event trigger time k_i+1Is shown in equation (7):

in the formula,u(k_i)＝[u^T(k_i) u^T(k_i) … u^T(k_i)]^T，

is in a state x (k)_i) The estimated vector of (a) is calculated,

is output y (k)_i) Estimated vector of r (k)_i) To generate a residual signal, L (k)_i)∈R^n×qIs an observer gain matrix, W (k)_i)∈R^q×qY (k) of dynamic event trigger for post-filter weighting matrix_i) Driving the residual generator to work;

an observer gain matrix L (k) is designed according to the following formula_i) And a postfilter weighting matrix W (k)_i)：

And 3, step 3: for the residual r (k) generated in the previous step_i) Performing data processing to obtain a fault detection result;

defining a residual evaluation function as formula (12):

in the formula, k_iTriggering time for the current event, wherein N is a moving time window;

determining a residual threshold value according to equation (13):

where i is 1,2, a, M, and M is the number of times of triggering of the dynamic event trigger condition (2), and is calculated according to the following formula

Mean and mean square error of (d):

threshold J_thDetermined according to equation (14):

according to the formulas (12) and (14), the residual evaluation logic is expressed as the following formula (15):

when residual evaluation function J_N(r(k_i) Greater than a threshold J)_thJudging that the unmanned aerial vehicle breaks down during flying and giving a signal indication; when residual evaluation function J_N(r(k_i) ) is less than or equal to the threshold value J_thAnd judging that the unmanned aerial vehicle normally flies.

Images