CN103984233B - A kind of quadrotor dual-granularity method for diagnosing faults based on mixed model - Google Patents

Abstract

The present invention relates to a dual granularity fault diagnosis method for a quadrotor aircraft based on a hybrid model, which includes analyzing various physical effects affecting quadrotor flight to determine the modeling category of prior models and non-parametric models; for non-parametric models, Measure the degree of nonlinearity; choose the appropriate parameter identification method according to the degree of nonlinearity, and establish the hybrid model of the quadrotor aircraft; divide the granularity level of the process variable according to the mixed model; judge the channel of the fault according to the coarse-grained level and determine the occurrence of the fault by the fine-grained level Components, thereby achieving double granularity fault diagnosis. The invention makes full use of the structural characteristics of the hybrid model, and is suitable for the feasibility verification of the process detection and diagnosis of the structural failure of the quadrotor helicopter.

Description

A Dual-Granularity Fault Diagnosis Method for Quadrotor Aircraft Based on Hybrid Model

technical field

The invention relates to the technical field of fault diagnosis of aviation aircraft, in particular to a dual granularity fault diagnosis method for quadrotor aircraft based on a hybrid model.

Background technique

Quadrotor aircraft is a typical modern complex system. Compared with fixed-wing aircraft, it has more complex aerodynamic characteristics and more special flight states. It requires higher-precision mathematical models and more robust control laws to ensure flight quality. and flight safety.

The traditional modeling methods of quadrotor aircraft are roughly divided into mechanism modeling and data-driven modeling. Mechanism modeling parameters have strong interpretability and model extension, but for modern complex systems such as multivariable, nonlinear, and strong coupling, modeling is difficult; data-driven modeling does not require prior knowledge of process objects, but Model accuracy and generalization are highly dependent on the modeling data. Therefore, it is difficult to obtain a high-quality target model only by a single modeling method.

The quadrotor aircraft outputs three attitude angle signals through four actuators. It belongs to the overdrive system, which can effectively improve the structural load capacity and response speed. It will inevitably receive external disturbances or malfunction during operation; Fault refers to a large deviation of at least one characteristic or parameter of the drive system, which is beyond the acceptable range. At this time, the performance of the system is obviously lower than its normal level; the classification of this fault can be carried out from different aspects, and from the point of view of the location of the fault, it can be divided into actuator fault, sensor fault and structural fault.

The fault diagnosis method for quadrotor aircraft mainly has the following problems:

(1) In terms of diagnosis methods, the model-based fault diagnosis method is the main method, and the diagnosis process ignores the data characteristics of a large number of process variables and some fault information;

(2) In terms of fault types, most literature considers the fault diagnosis of quadrotor aircraft under actuator faults, a small amount of literature considers the fault diagnosis of sensors, and structural faults are rare due to complex fault models and various fault types Studied.

To sum up, how to overcome the shortcomings of the existing technology has become one of the key problems to be solved urgently in the field of aviation aircraft fault diagnosis technology.

Contents of the invention

The purpose of the present invention is to provide a dual granularity fault diagnosis method for quadrotor aircraft based on a hybrid model in order to overcome the existing deficiencies in the prior art. The established hybrid model can improve the accuracy of fault diagnosis by refining the granularity of data, and is suitable for the feasibility verification of process detection and diagnosis of quadrotor helicopter structural faults.

A kind of quadrotor aircraft double granularity fault diagnosis method based on hybrid model proposed according to the present invention is characterized in that it comprises following specific steps:

Step A: Apply various physical effects to analyze the degree of influence of the quadrotor aircraft, divide it into main influencing factors and secondary influencing factors according to the degree of influence, and determine the prior model and non-parametric model of the quadrotor aircraft hybrid model. Modeling categories and building prior models based on the main influencing factors;

Step B: Analyze various nonlinear items and coupling items in the non-parametric model for the secondary influencing factors, and measure the nonlinear degree of the quadrotor aircraft;

Step C: According to the degree of influence of physical effects and various non-linear degrees, use the fuzzy reasoning machine to select appropriate parameter identification and linearization methods to establish a hybrid model of the quadrotor aircraft;

Step D: According to the source of the data and its own physical meaning, the granularity level of the quadrotor aircraft is divided;

Step E: Use the process monitoring method based on principal component analysis to first determine the channel where the fault occurred for the coarse-grained data, and then locate the component where the fault occurred for the fine-grained data, thereby realizing double-grained fault diagnosis .

Compared with the prior art, the present invention has the following remarkable advantages: firstly, the hybrid model established by the present invention fully considers the degree of influence of various physical effects on the quadrotor aircraft, and simultaneously avoids model analysis caused by nonlinear terms and coupling terms , processing difficulties; second, the dual granularity fault diagnosis method makes full use of the data information of each granularity level, which improves the accuracy and reliability of diagnosis; The modification of the fault-tolerant control law of different channels provides convenience; the fourth is to use fine-grained data to accurately diagnose the faults of the components, effectively avoiding the complex fault modeling and multiple problems in the traditional fault diagnosis methods, which is better Expanded the scope of fault diagnosis.

Description of drawings

Fig. 1 is a schematic block diagram of steps of a dual-granularity fault diagnosis method for a quadrotor aircraft based on a hybrid model proposed by the present invention.

Fig. 2 is a schematic diagram of a Z=X*Y three-dimensional image.

Figure 3 is Schematic diagram of a 3D image.

Figure 4 is a schematic diagram of the general process of hybrid modeling.

Fig. 5 is a schematic diagram of fault detection (attitude angle) for a coarse-grained level.

Figure 6 is a schematic diagram of fault detection (aerodynamic moments) for a coarse-grained level.

Figure 7 is a schematic diagram of the contribution rate of the attitude angle to the fault.

Figure 8 is a schematic diagram of the total contribution rate of the attitude angle to the fault.

Fig. 9 is a schematic diagram of the contribution rate of the aerodynamic moment to the fault.

Figure 10 is a schematic diagram of the total contribution rate of the aerodynamic moment to the fault.

Figure 11 is a schematic diagram of the contribution rate of the fuselage gyro torque to the fault.

Figure 12 is a schematic diagram of the total contribution rate of the fuselage gyro torque to the fault.

Figure 13 is a schematic diagram of fault detection (voltage) for a fine-grained level.

Figure 14 is a schematic diagram of fault detection (rotor gyro torque) for a fine-grained level.

Figure 15 is a schematic diagram of the contribution rate of the rotor gyro torque to the fault.

Figure 16 is a schematic diagram of the total contribution rate of the rotor gyro torque to the fault.

detailed description

The specific implementation manners of the present invention will be further described in detail below in conjunction with the accompanying drawings.

In conjunction with Fig. 1, a kind of dual granularity fault diagnosis method of quadrotor aircraft based on hybrid model proposed by the present invention includes analyzing various physical effects affecting quadrotor flight to determine the modeling category of prior model and non-parametric model; For the non-parametric model, measure the degree of nonlinearity; choose the appropriate parameter identification method according to the degree of nonlinearity, and establish the mixed model of the quadrotor aircraft; divide the granularity level of the process variable according to the mixed model; The level of granularity determines the component where the fault occurs; the specific implementation steps are as follows:

Step A: Apply various physical effects to analyze the degree of influence of the quadrotor aircraft, divide it into main influencing factors and secondary influencing factors according to the degree of influence, and determine the prior model and non-parametric model of the quadrotor aircraft hybrid model. Modeling category, and establish a priori model according to the main influencing factors; wherein: the various physical effects described are mainly in the form of aerodynamic moment, which is caused by the pulling force and resistance generated by the rotation of the rotor of the quadrotor aircraft, It is the most important type of moment suffered by the aircraft, including rolling moment, pitching moment and yaw moment, which belongs to the modeling category of non-parametric model; taking the aerodynamic moment as the main influencing factor, a priori model can be established as follows :

In the above formula, φ, θ, Respectively represent the pitch angle, roll angle and yaw angle; are the aerodynamic moments on the x, y, and z axes respectively; l is the distance from the motor center point to the coordinate origin; K _f , K _t,c are the constant coefficients between the motor voltage and torque; V ₁ , V ₃ , V _2. V ₄ is the voltage generated by the front, rear, left and right motors; Indicate the moments of inertia of the roll channel, pitch channel, and yaw channel about the x, y, and z axes respectively; J _xx , J _yy , and J _zz respectively indicate the roll channel, pitch channel, and yaw channel about the x, y, and z axes moment of inertia.

The selected state quantity is Control variable u=[V ₁ V ₃ V ₂ V ₄ ] ^T , then the prior model of quadrotor aircraft described in the form of state equation is:

y=Cx

In the above formula:

D=0;

In addition to the main influencing factor of the aerodynamic moment, the gyroscopic moment on the system is relatively weak but cannot be ignored. Its existence determines the local approximation performance of the system. It is a secondary influencing factor and belongs to the modeling category of non-parametric models. It mainly includes the gyro torque of the fuselage and the gyro torque of the rotor. Based on the above analysis, a more accurate mechanism model of the quadrotor aircraft can be established. The formula is as follows:

After considering the secondary influencing factors, nonlinear terms and cross-coupling terms appeared in the model formula.

Step B: Analyze various nonlinear items and coupling items in the non-parametric model for the secondary influencing factors, and measure the nonlinear degree of the quadrotor aircraft; wherein: the nonlinear degree refers to: for the input signal A stable causal system N: U _a → Y, its degree of nonlinearity Defined as the following non-negative equation:

In the above formula: G:U _a →Y is a linear operator, and the norm ||·|| _VT is defined as T is the period; inf and sup represent the infimum and supremum respectively; select a series of sinusoidal signals to form the input set U _LB :

U _LB ＝<u|u(t)＝A sin(ωt),A∈A,ω∈Ω>

A＝<A∈R ⁺ |A _min ≤A≤A _max >

Ω＝<ω∈R ⁺ |ω _min ≤ω≤ω _max >

Then the steady-state output can be expressed as the following formula:

According to the definition of the degree of nonlinearity, the calculation formula of the lower bound of the degree of nonlinearity is deduced as:

In the modeling process of quadrotor aircraft, the secondary influencing factors such as gyro effect belong to the modeling category of non-parametric model; from the above analysis, it can be seen that the degree of nonlinearity of each part should be measured first.

Now take the rolling channel as an example to illustrate the calculation process of the degree of nonlinearity. For the airframe gyroscopic effect, its original expression is is the coupling term of the two state quantities. Therefore, this type of gyro effect can be abstracted as x×y type; for the rotor gyro torque, its original expression is Ω _i is the rotational speed of the rotor, which is proportional to the square of the control voltage V _i , namely Therefore, the rotor gyro torque is essentially the coupling item of the state quantity and the control quantity, and this kind of gyro effect can be abstracted as type. Figure 2 and Figure 3 describe the three-dimensional images of two types of abstract models, which can intuitively reflect their respective nonlinear degrees; x and y are respectively represented by sinusoidal signals Asin(αt) and Bsin(βt) with different amplitudes and phase angles , then the x×y type is further written as the A sin(αt)×B sin(βt) type, type further writing By selecting enough representative amplitudes and phase angles for the two types of abstract models, the respective nonlinear degrees can be calculated. Table 1 shows the calculation results of the nonlinear degree of the fuselage gyro moment and the rotor gyro moment taking the roll channel as an example. Expand to the degree of nonlinearity of the i-th first and second terms.

Table 1 Calculation of nonlinear degree of gyro effect

In order to determine the appropriate linearization method, after obtaining the nonlinear degree of the two types of gyroscopic effects, it is necessary to comprehensively consider the accuracy of the model and the complexity of the algorithm. Four representative linearization methods are: least squares linear fitting (linearfit, LF), equilibrium point Taylor series expansion (series expansion, SE), T-S fuzzy inference model (Takagi-sugeno, TS) and subspace System identification (parameter identification, PI).

Although the least squares method linear fitting algorithm has high identification efficiency, but the identification error is large, it is a rough data processing method, which is suitable for the situation with weak nonlinearity; although the subspace system identification algorithm has high accuracy, the identification The efficiency is low, the implementation process is complicated, and it is suitable for situations with strong nonlinearity; according to the nonlinearity of the variables and the applicable conditions of the linearization method, fuzzy rules can be established between the nonlinearity and the linearization method, as shown in Table 2 Shown; select the non-linear degree and the mean square error of the output quantity as the input quantity, the appropriate linearization method as the output quantity, and establish the mamdani type fuzzy inference model.

Table 2 Fuzzy rule base

For the gyro effect on each channel of the quadrotor aircraft, by analyzing the variable relationship, the influence area of the two types of gyro effect on the state equation is obtained; after fuzzy reasoning, the respective linearization methods are output from the fuzzy rule observation window as shown in Table 3 Show.

Table 3 Optimal linearization methods for nonparametric models

According to Table 3, each part of the non-parametric model is linearized, and combined with the prior model, the state equation of the mixed model is obtained. To sum up, Figure 4 shows the general process of hybrid modeling.

Step C: According to the degree of influence of physical effects and various non-linear degrees, use the fuzzy reasoning machine to select the appropriate parameter identification and linearization methods to establish the hybrid model of the quadrotor aircraft; The mapping between parameter identification and linearization methods is established, and finally the quadrotor hybrid model formula based on physical effect analysis and nonlinear measurement is obtained as follows:

In the above formula, f _SE <·>, f _TS <·>, f _PI <·> represent the three parameter identification methods of SE, TS, and PI to deal with various gyro effect expressions in "<>", specifically as Table 6 shows.

Taking the working condition where the desired attitude angle is 1° and the reference voltage is U _bias = 2V as an example, the state equation matrix of the hybrid model is as follows:

Step D: According to the source of the data and its own physical definition, the granularity level of the quadrotor aircraft is divided; the specific implementation methods include:

Compared with the single model, the hybrid model can obtain some unknown parameters from the non-parametric model through intelligent algorithms such as parameter identification and fuzzy reasoning, in addition to obtaining conventional information such as the output quantity and state quantity in the prior model. The physical meaning of these unknown parameters can be obtained through the modeling process of the experimental model and the mechanism knowledge of the controlled object; only from the perspective of information volume, the more data output by the hybrid model, the more comprehensive the fault information can be obtained; however, from the fault From the perspective of diagnosis, efficient fault diagnosis needs to divide the granularity level of data.

In the hybrid model, the prior model grasps the global characteristics of the system, and the degree of refinement of the hybrid model is low, from which coarse-grained data information, such as system output and state quantities, are obtained; the non-parametric model has good local approximation performance, The degree of refinement of the mixed model is high, from which fine-grained data information is obtained, such as unknown parameters in the prior model. For quadrotor aircraft, the attitude angle and aerodynamic moment of the three channels of roll, pitch and yaw can be selected as coarse-grained data, and the fuselage gyro torque, rotor gyro torque and motor voltage can be selected as fine-grained data.

Step E: Use the process monitoring method based on principal component analysis to first determine the channel where the fault occurred for the coarse-grained data, and then locate the component where the fault occurred for the fine-grained data, thereby realizing double-grained fault diagnosis ; Specific implementation methods include:

Using the process monitoring method based on principal component analysis, the fault diagnosis method based on the traditional contribution graph is adopted, and the square prediction error statistic (SPE) is used as the evaluation index; Coarse-grained data such as angular velocity determine the approximate range of the fault, that is, determine the channel where the fault occurred.

Under the working condition of the reference voltage 3.5V and the expected attitude angle of 0.3°, collect 15 sets of data under normal working conditions, each with 155 sampling points, to establish the principal component analysis model; collect 12 sets of data under fault conditions , each group of 155 sampling points.

Taking one type of fault as an example, the fault detection is first carried out with coarse-grained data; as can be seen from Figure 5 and Figure 6, the three attitude angles and aerodynamic moments all exceed the control limits to varying degrees, so the fault is detected;

Secondly, using the fault diagnosis method based on the traditional contribution graph, for the three types of coarse-grained data of attitude angle, aerodynamic moment and fuselage gyro moment, Fig. 7, Fig. 8 and Fig. The variation trend of the sampling point variables to the failure contribution rate; at the same time, Figure 10, Figure 11 and Figure 12 more intuitively count the total contribution rate of each variable failure in the form of a histogram.

Based on the diagnostic results in Figures 7 to 12, the roll angle, roll moment, and roll channel fuselage gyro torque have the largest contribution to the fault; therefore, according to the fault diagnosis results at the coarse-grained level, it can be inferred that the roll channel has occurred Fault.

For the roll channel, repeat the above fault diagnosis steps for fine-grained physical quantities (rotor gyro torque and motor voltage); it can be seen from Figure 13 that the voltage does not exceed the control limit, indicating that the motor is not faulty. When performing fault detection on the rotor gyro torque, the SPE statistic exceeds the control limit, as shown in Figure 14, indicating that the fault occurred in the rotor. To sum up, according to the contribution rate analysis shown in Figure 15 and Figure 16, the components that failed are the front and back rotors.

On this basis, according to the fine-grained data such as the fuselage gyro effect, the faulty parts are finally determined, and finally a dual-grained fault diagnosis method for quadrotor aircraft based on the hybrid model is realized.

The invention has been verified through repeated tests and has achieved satisfactory application effects.

Claims (6)

1. a dual granularity fault diagnosis method for quadrotor aircraft based on hybrid model, is characterized in that the method comprises the following specific steps: Step A: Applying various physical effects to analyze the degree of influence of the quadrotor aircraft, the various physical effects are divided into main influencing factors and secondary influencing factors according to their degree of influence, and determine the prior model in the mixed model of the quadrotor aircraft and the modeling category of non-parametric models, and establish prior models according to the main influencing factors; wherein: the various physical effects refer to the forms of aerodynamic effects, fuselage gyroscopic effects and rotor gyroscopic effects; the main influencing factors Refers to the aerodynamic moment, which is caused by the pulling force and resistance generated by the rotor rotation of the quadrotor aircraft. The aerodynamic moment is the most important type of moment that the aircraft bears, including the form of rolling moment, pitching moment and yaw moment. The modeling category of the non-parametric model; the secondary influencing factor refers to the influence of the gyro moment on the system in addition to the aerodynamic moment, mainly including the fuselage gyro moment and the rotor gyro moment, which is the modeling category of the non-parametric model; Step B: Analyze various nonlinear items and coupling items in the non-parametric model for the secondary influencing factors, and measure the nonlinear degree of the quadrotor aircraft; Step C: According to the degree of influence of physical effects and various non-linear degrees, use the fuzzy inference machine to select parameter identification and linearization methods to establish a hybrid model of the quadrotor aircraft; Step D: According to the source of the data and its own physical definition, the granularity level of the quadrotor aircraft is divided; Step E: Use the process monitoring method based on principal component analysis to first determine the channel where the fault occurred for the coarse-grained data, and then locate the component where the fault occurred for the fine-grained data, thereby realizing double-grained fault diagnosis . 2. a kind of dual granularity fault diagnosis method of quadrotor aircraft based on hybrid model according to claim 1, is characterized in that described aerodynamic moment is main influencing factor, establishes prior model formula as follows: In the above formula, φ, θ, Respectively represent the pitch angle, roll angle and yaw angle; are the aerodynamic moments on the x, y, and z axes respectively; l is the distance from the center point of the motor to the coordinate origin; K _f is the constant coefficient of the rotor moment; K _{t, c} are the constants between the thrust and torque when the rotor rotates in reverse Coefficients; V ₁ , V ₃ , V ₂ , V ₄ are the voltages generated by the front, rear, left, and right four motors; J _xx , J _yy , and J _zz respectively represent the roll channel, pitch channel, and yaw channel about x ,Moments of inertia of the y,z axes. 3. a kind of dual granularity fault diagnosis method of quadrotor aircraft based on hybrid model according to claim 1, it is characterized in that the degree of nonlinearity described in step B refers to: for input signal A stable causal system N: U _a → Y, the degree of nonlinearity of the causal system Defined as the following non-negative equation: ${\mathrm{<span\; class="notranslate">\; \&phi;</span>}}_{\mathrm{<span\; class="notranslate">\; N</span>}}^{\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; Y</span>}}{\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; no</span>}\mathrm{<span\; class="notranslate">\; f</span>}}\underset{\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; u</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; p</span>}}\frac{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; |</span>{<span\; class="notranslate">\; |</span>}_{\mathrm{<span\; class="notranslate">\; V</span>}\mathrm{<span\; class="notranslate">\; T</span>}}}{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; \&rsqb;</span><span\; class="notranslate">\; |</span>{<span\; class="notranslate">\; |</span>}_{\mathrm{<span\; class="notranslate">\; V</span>}\mathrm{<span\; class="notranslate">\; T</span>}}}$ G in the above formula is a linear operator in the causal system N; G[u] represents the function of the linear part G in the causal system with respect to the input signal u; N[u] represents the function of the stable causal system N with respect to the input signal u; The norm ||·|| _VT is defined as T is the period; inf and sup represent the infimum and supremum respectively; after measuring the degree of nonlinearity of the nonlinear term and the coupling term in the non-parametric model, select the method of linearization and parameter identification to measure the nonlinear term and coupling term Coupling terms are linearized. 4. a kind of dual granularity fault diagnosis method based on hybrid model of quadrotor aircraft according to claim 2, is characterized in that the hybrid model described in step C refers to: utilize the method of fuzzy reasoning, in nonlinear degree and parameter identification , linearization method to establish a mapping, and finally obtain the quadrotor hybrid model formula based on physical effect analysis and nonlinear measurement as follows: In the above formula, Ω _i represents the rotational speed of the four rotors, i=1, 2, 3, 4; F<g(x)> represents the corresponding parameter identification or linearization method F to process the The result obtained after the expression g(x); F chooses one of the three methods of SE, TS, and PI, SE, TS, and PI respectively represent the equilibrium point Taylor series expansion, TS fuzzy reasoning model and subspace system identification, g( x) is various expressions representing the gyro effect, and J _r represents the moment of inertia of the rotor motor. 5. a kind of dual granularity fault diagnosis method of quadrotor aircraft based on hybrid model according to claim 1, it is characterized in that the granularity level described in step D divides and refers to: in the hybrid model, the prior model grasps the Global characteristics, compared with the non-parametric model, the a priori model has a lower degree of refinement to the mixed model, from which the coarse-grained data information including the system output and state quantity is obtained; the non-parametric model has good local approximation performance, and Compared with the prior model, the non-parametric model has a higher degree of refinement for the mixed model, from which the fine-grained data information including the unknown parameters in the prior model can be obtained. 6. a kind of dual granularity fault diagnosis method based on hybrid model of quadrotor aircraft according to claim 1 is characterized in that the double granularity fault diagnosis described in step E refers to: utilize the process monitoring method based on principal component analysis, with The square prediction error statistic is used as the evaluation index for fault detection, and at the same time, the method based on the traditional contribution graph is used for fault diagnosis, and the variable with a large contribution rate is the causal variable that causes the fault, including: Step E1: Using the coarse-grained data of the angular velocity of the three attitude angles of the quadrotor aircraft roll, pitch and yaw to determine the scope of the fault, so as to determine the channel where the fault occurred; Step E2: On the basis of determining the fault channel, finally determine the component where the fault occurs according to the fine-grained data of the fuselage gyroscopic effect.