CN111240304B - Machine learning sample generation method for online identification of thrust fault of aircraft - Google Patents
Machine learning sample generation method for online identification of thrust fault of aircraft Download PDFInfo
- Publication number
- CN111240304B CN111240304B CN202010076010.9A CN202010076010A CN111240304B CN 111240304 B CN111240304 B CN 111240304B CN 202010076010 A CN202010076010 A CN 202010076010A CN 111240304 B CN111240304 B CN 111240304B
- Authority
- CN
- China
- Prior art keywords
- fault
- data
- training
- aircraft
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0256—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults injecting test signals and analyzing monitored process response, e.g. injecting the test signal while interrupting the normal operation of the monitored system; superimposing the test signal onto a control signal during normal operation of the monitored system
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/20—Pc systems
- G05B2219/24—Pc safety
- G05B2219/24065—Real time diagnostics
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a machine learning sample generation method for online identification of thrust faults of an aircraft, which is suitable for the field of online identification of faults of a typical power system in the flight process of the aircraft. Data fusion is carried out on the flight motion information (such as flight position, speed, acceleration, attitude angle, angular velocity and the like) of the control system, and corresponding data are intercepted according to the design method of the invention and are used as machine learning training and testing samples. The method considers factors such as mass center motion, disturbance center motion, structural disturbance, aerodynamic force and moment of the aircraft, generates data by introducing deviation combination circulation in the simulation model, is more real and credible, and is beneficial to improving the identification precision of actual faults. The invention refines the fault mode, generates the related data with finer granularity of the fault mode, and is beneficial to improving the identification precision.
Description
Technical Field
The invention relates to a machine learning sample generation method for online identification of thrust faults of an aircraft, which is suitable for the field of online identification of faults in the flight process of a carrier rocket.
Background
Because real data is limited, the aircraft fault identification technology based on machine learning mostly depends on an aircraft simulation model to generate a large number of machine learning samples, and if the model is inaccurate or has a large difference with the real model, the actual flight identification precision is greatly influenced. The existing machine learning fault identification can achieve higher precision in a simulation test, the effect is often not good in practical application, and the difference between machine learning sample data and real data is large. And the current machine learning sample data has large scale, occupies large computing resources, influences computing efficiency, and is not suitable for computing hardware with limited computing power.
Disclosure of Invention
The technical problem solved by the invention is as follows: the method overcomes the defects of the prior art, provides a machine learning sample generation method for the online identification of the thrust fault of the aircraft, and establishes a simulation model by considering factors such as mass center motion, disturbance center motion, structural disturbance, aerodynamic force and moment of the aircraft. By introducing deviation combination cycle into the simulation model to generate data, the data is more real and credible, and the improvement of the fault identification precision in actual flight is facilitated. In addition, the invention refines the fault mode, generates related data with finer granularity of the fault mode, and is beneficial to improving the identification precision. The invention avoids the situation of large quantity of repeated positive samples by designing a corresponding algorithm. The machine learning training and test sample data obtained by intercepting corresponding data according to the design method of the invention has smaller scale, can save computing resources and improve identification efficiency. The method provided by the invention is adopted to generate the machine learning sample, so that the fault identification speed can be improved, the fault can be quickly identified, and the identification precision can be ensured. And moreover, a mode of manually marking a large number of data labels can be omitted, and batch data are automatically marked.
The invention is realized by the following technical scheme: as shown in FIG. 1, a method for generating machine learning samples for online identification of thrust faults of an aircraft includes the following steps:
(1) Constructing a six-degree-of-freedom dynamic simulation model of the aircraft according to the real aircraft and the environment where the real aircraft is located;
(2) Setting each simulation deviation combination, and inputting the set simulation deviation combination into the aircraft six-degree-of-freedom dynamic simulation model constructed in the step (1); carrying out the step (3);
(3) Setting the occurrence time and the degree of each fault, and inputting the set occurrence time and the set degree of the faults into the aircraft six-degree-of-freedom dynamic simulation model constructed in the step (1);
(4) And (3) carrying out permutation and combination on the simulation deviation combination set in the step (2) and the fault occurrence time and the fault degree set in the step (3) to obtain simulation data of the six-degree-of-freedom dynamic simulation model of the aircraft under different conditions, wherein the simulation data comprises: acceleration, attitude angle deviation; storing the simulation data under different conditions;
(5) Intercepting the simulation data under different conditions to generate a data sample; and designing a data label according to the number of the fault engine and the fault degree, and labeling each data sample.
Preferably, the method further comprises the steps (6) to (9);
(6) Randomly taking more than most of the labeled data in the step (5) to divide the data into a training set, randomly taking half of the rest part to divide a verification set, and dividing the other half into a test set; building a BP neural network, wherein the structure comprises a single hidden layer and an output layer; the number of single hidden layers is 10, and the activation function is a Sigmoid function; the activation function is a softmax function, and the number of the neurons in the output layer is 11; the layers are all connected;
(7) Training adopts cross entropy as a loss function, adopts a gradient descent method to update parameters of the neural network, and updates the weight and the bias of the neural network; inputting the data samples in the training set in the step (5) into the BP neural network built in the step (6) for training; training and updating network parameters by adopting a gradient descent method; testing the training process by adopting the data samples in the verification set, and finishing the training when the error of the samples on the verification set is not reduced for N times continuously by the neural network to obtain a training result;
(8) Testing the training result by adopting the data sample in the test set, if the training result meets the requirement, storing the trained BP neural network, and performing the step (9); if not, adjusting the BP neural network built in the step (6), and returning to the step (6);
(9) And (4) embedding the BP neural network stored in the step (8) into an aircraft control computer, and performing online fault identification by using the trained BP neural network.
Preferably, the step (6) randomly takes more than most of the labeled data in the step (5) to be divided into training sets, specifically: and (4) randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set.
Preferably, a total of 11 neurons of the layer are output in step (6), and the number of the neurons is the total number of fault categories.
Preferably, the total number of failure categories is the total number of tags in step (5).
Preferably, the method further comprises the steps (6) to (9);
(6) Randomly dividing more than most of the labeled data in the step (5) into a training set, and dividing the rest into a test set; training by using a training set data sample and using a CART algorithm to generate a decision tree, and finishing training to obtain a training result when the decision tree is generated;
(7) Judging the training result, if the training result meets the requirement, storing the generated decision tree, and performing the step (8); if not, returning to the step (4);
(8) Verifying the decision tree in the step (7) by adopting a data sample in the test set, and if the verification accuracy meets the requirement, extracting and storing a judgment rule in the decision tree; if the requirement is not met, returning to the step (4);
(9) And using the decision tree to perform online fault identification.
Preferably, the step (6) randomly takes more than most of the labeled data in the step (5) to be divided into training sets, specifically: and (4) randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set.
Preferably, the method further comprises the steps (6) to (9);
(6) Randomly taking more than most of the labeled data in the step (5) to divide the data into a training set, randomly taking half of the rest part to divide a verification set, and dividing the other half into a test set; building an LSTM neural network, wherein the structure comprises an input layer, an LSTM layer and an output layer; LSTM layer has 50 neurons in total; the number of the neurons of the output layer is the total number of the fault categories, namely the total number of the labels in the step (5), and the activation function is a softmax function; the layers are all connected;
(7) Training adopts cross entropy as a loss function, adopts a gradient descent method to update parameters of the neural network, and updates the weight and the bias of the neural network; inputting the data samples in the training set in the step (5) into the LSTM neural network built in the step (6) for training; training and updating network parameters by adopting a gradient descent method; testing the training process by adopting the data samples in the verification set, finishing the training when the error of the neural network continuously iterates the samples on the verification set for N times is not reduced to obtain a training result,
(8) Testing the training result by adopting the data sample in the test set, if the training result meets the requirement, storing the trained LSTM neural network, and performing the step (9); if not, adjusting the LSTM neural network built in the step (6), and returning to the step (6);
(9) And (4) embedding the LSTM neural network stored in the step (8) into an aircraft control computer, and performing online fault identification by using the trained LSTM neural network.
Preferably, (6) randomly taking most of the labeled data in (5) to divide into a training set, specifically: and (4) randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set.
Preferably, the input layer, i.e. the input data, is a 9 x 20 vector.
Compared with the prior art, the invention has the following advantages:
(1) The method takes the factors of mass center motion, disturbance center motion, structural disturbance, aerodynamic force, moment and the like of the aircraft into consideration, and generates data by introducing deviation combination cycle in the simulation model, so that the data is more real and credible, and the method is favorable for improving the fault identification precision in actual flight.
(2) According to the invention, by designing the corresponding algorithm, the situation that the positive samples are repeated in a large quantity is avoided, the scale of the data samples is reduced, the computing resources are saved, and the fault identification precision is ensured.
(3) The invention refines the fault mode, generates the related data with finer granularity of the fault mode, and is beneficial to improving the identification precision.
(4) The method provided by the invention is adopted to generate the machine learning sample, so that the fault identification speed can be improved, the fault can be quickly identified, and the identification precision can be ensured.
(5) The invention can save the mode of manually marking data labels in batches and automatically mark the batch data.
Drawings
FIG. 1 is a schematic view of a scheme;
FIG. 2 is a schematic flow chart of a training sample interception method;
FIG. 3 is a schematic flow chart of a test sample capture method;
FIG. 4 is a schematic diagram of an algorithm flow for avoiding excessive positive samples;
FIG. 5 is a block diagram of the control system components;
FIG. 6 is a schematic view of a turbine engine yaw angle.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments.
The invention relates to a machine learning sample generation method for online identification of thrust faults of an aircraft, which is suitable for the field of online identification of faults of a typical power system in the flight process of the aircraft. Data fusion is carried out on the flight motion information (such as flight position, speed, acceleration, attitude angle, angular velocity and the like) of the control system, and corresponding data are intercepted according to the design method of the invention and are used as machine learning training and testing samples. The method considers factors such as mass center motion, disturbance center motion, structural disturbance, aerodynamic force and moment of the aircraft, generates data by introducing deviation combination circulation in the simulation model, is more real and credible, and is beneficial to improving the identification precision of actual faults. In addition, the invention refines the fault mode, generates related data with finer granularity of the fault mode, and is beneficial to improving the identification precision. The invention avoids the situation of large quantity of repeated positive samples by designing a corresponding algorithm. The machine learning training and test sample data obtained by intercepting corresponding data according to the design method of the invention has smaller scale, can save computing resources and improve identification efficiency. The invention can save the mode of manually marking a large number of data labels and automatically mark batch data.
The invention discloses a machine learning sample generation method for online identification of thrust faults of an aircraft. And (4) establishing a simulation model by considering factors such as mass center motion, disturbance center motion, structural disturbance, aerodynamic force and moment of the aircraft. By introducing proper deviation combination cycle generation data into the simulation model, the data can be more real and credible, the improvement of the fault identification precision in actual flight is facilitated, the data amount is reduced, and the calculation resources are saved. In addition, the invention refines the fault mode, generates related data with finer granularity of the fault mode, and is beneficial to improving the identification precision. The method provided by the invention is adopted to generate the machine learning sample, so that the fault identification speed can be improved, the fault can be quickly identified, and the identification precision can be ensured. And moreover, a mode of manually marking a large number of data labels can be omitted, and batch data are automatically marked. In addition, the invention avoids the situation that the positive samples are repeated in large quantity by designing the corresponding algorithm. The machine learning training and test sample data obtained by intercepting corresponding data according to the design method of the invention has smaller scale, can save computing resources and improve identification efficiency.
The aircraft is provided with 2 engines, the formed resultant thrust force is the total thrust, the environment is a real low-altitude environment, and the influence of wind disturbance and aerodynamics is considered;
the six-degree-of-freedom dynamic simulation model of the aircraft is preferably as follows: an aircraft thrust fault online identification six-degree-of-freedom dynamics simulation model based on machine learning.
The vehicle is preferably a rocket with a mass of about 90 kg. The control system is shown in a block diagram in fig. 5. And guidance and attitude control are designed according to a PID control law.
Since the aircraft is flying for a short time, at low altitude, and at low speed, the gravitational acceleration is calculated as a constant value, taking into account a flat, stationary ground.
Coefficient of aerodynamic force C x ,
Axial force coefficient, normal force coefficient derivative, and lateral force coefficient derivative, respectively.
The direction of the gravitational acceleration is vertical downward, and the magnitude is preferably:
g 0 ＝9.80665m/s 2
decomposed into g under the target system x 、g y 、g z 。
The resultant force here includes a thrust force, a pneumatic force, and no attractive force.
Initial position: x =0m, y =0m, z =0m
Initial speed: v. of x ＝0m/s,v y ＝0m/s,v z ＝0m/s
Initial attitude angle:
ψ＝0,γ＝0
the initial attitude angular velocity is 0.
The thrust regulating characteristic of the turbojet engine is preferably as follows:
maximum swing angle of the engine: 10 degrees.
The dynamic characteristics of the electric steering engine are preferably as follows:
inertial group dynamics, preferably:
other preferred embodiments of the model are described in the detailed preferred embodiment.
The deviations contained in the deviation combinations include: mass deviation, centroid deviation, rotational inertia deviation, wind speed deviation, wind direction deviation and engine flow deviation. According to the deviation combination in the invention, the data scale can be reduced, and meanwhile, the real model can be fitted as much as possible, so that the actual identification precision is ensured.
The two engine numbers are 1 and 2, respectively. From the start of takeoff, the minimum value of 2s after the fault occurs and the flight ending time is preferably taken as the end of simulation, and the fault occurring time is set once every equal interval of 0.5 s. And (4) finishing the uniform injection of the fault in 2s, so that the thrust of the single engine is reduced to the corresponding fault degree at the fault starting injection moment.
The failure degree specifically refers to the degree of reduction of the thrust of a single engine. One scheme of the fault degree is that a single fault engine is reduced by 30-90%, and the interval is preferably 10%; the other scheme is that a single fault engine is reduced by 10% -30%, and the interval is 5%.
X in simulation data generated through model simulation under target system with acceleration T 、Y T 、Z T Acceleration in three directions; the attitude angle is a pitch angle, a yaw angle and a rolling angle under a target system; the attitude angle deviation is the target system pitch angle deviation, yaw angle deviation and roll angle deviation.
Step (5) designing a data label according to the existence of thrust faults, the number of a fault engine and the fault degree of the aircraft, wherein the first scheme is as follows: label 0 indicates thrust normal, i.e. no fault condition; 1-7 are respectively 30% -90% lower than the No. 1 engine; 8-14 are respectively 30% -90% of the reduction of the No. 2 engine, and the interval is 10%.
Label settings table below 1:
TABLE 1
Designing a data label according to the existence of thrust faults, the number of a fault engine and the fault degree of the aircraft, wherein the second specific preferred scheme is as follows:
label 0 indicates thrust normal, i.e. no fault condition; 1-5 are respectively 10% -30% lower than the No. 1 engine; 6-10 are respectively 10% -30% reduction of No. 2 engine, and the interval is 5%.
The tag settings table is as follows:
intercepting the simulation data under different conditions to generate a data sample, wherein the specific scheme is as follows: the intercepting process of the training sample is shown in figure 2, the intercepting process of the testing sample is shown in figure 3, the abscissa t in figures 2 and 3 is the flight time, the ordinate y represents the acquired simulation data, which can be the acceleration, the attitude angle or the attitude angle deviation, L is the length of the sample, I is the overlapping size of two adjacent samples, t is the overlapping size of two adjacent samples fault For fault injection time, t end Is the simulation end time.
Preferably, 9-dimensional information including the acceleration, attitude angle and attitude angle deviation of 20 time points in the period is taken as training data every 50ms in every 1s in the flight complete data. The number of each group is 9 × 20, i.e. 180 dimensional state quantities.
If n samples are total in the time period from t1 to t2, the length of each sample is L, and the overlapping length between samples is I, then:
for the kth sample: let k =1,2, \ 8230;, n,
start and end lines of the kth sample:
if a positive sample is taken, then in the equation:
t 1 ＝0
t 2 ＝t fault
if a positive sample is intercepted, then in the equation:
t 1 ＝t fault
t 2 ＝t end
wherein, t fault For fault injection time, t end Is the simulation end time.
In this test, the values of the parameters are as follows:
TABLE 2
| Type of parameter | Value of | Unit of |
| L | 1 | s |
| I | 0.8 | s |
| t fault | Injection every 0.5s for 1-20s is a single simulation | s |
| t end | Actual flight end time and t of ascent fault Minimum value of +2 | s |
Step (5) intercepting the simulation data under different conditions to generate a data sample, wherein the preferable scheme is as follows: and setting an algorithm for avoiding excessive positive samples, wherein the preferred algorithm flow is shown in figure 4, and intercepting by adopting the algorithm to generate data samples. And judging whether the fault injection time is the last time under the current deviation combination, wherein the fault degree is the last gear under the current deviation combination. If yes, intercepting a positive sample and a negative sample; otherwise, only negative samples are intercepted.
The invention relates to a machine learning sample generation method for online identification of thrust faults of an aircraft, which has the following specific preferred scheme:
(1) According to a real aircraft and the environment where the real aircraft is located, a six-degree-of-freedom dynamic simulation model of the aircraft is constructed, and the method specifically comprises the following steps:
the aircraft is provided with 2 engines, the formed resultant thrust force is the total thrust, the environment is a real atmospheric environment, and the influences of wind, structural interference and pneumatics are considered;
the simulation model fault injection mode is that from the start of takeoff, faults are injected once every 0.5s, and once simulation is carried out. And after the uniform injection of the fault is finished within 2s, the thrust of a single engine is reduced to the corresponding fault degree at the fault starting injection moment.
The six-degree-of-freedom dynamic simulation model of the aircraft is preferably as follows: the method comprises the steps of identifying a six-degree-of-freedom dynamics simulation model on line based on the thrust fault of an aircraft based on machine learning, wherein the model is specifically as follows.
The vehicle is a rocket with a mass of about 90 kg. The control system is shown in a block diagram in fig. 5. And guidance and attitude control are designed according to a PID control law.
Maximum swing angle of the engine: 10 degrees.
The overall aircraft data values are as follows in table 3:
TABLE 3 aircraft Overall data value-taking Table (10 kg Fuel)
The coordinate system is defined as follows:
target relative coordinate system (T series)
To the target point O T Is the origin of coordinates, O T Y T Opposite to the local gravity direction of the target point, O T X T Axis and OY T The axis being perpendicular and directed in the direction of the flying point, O T Z T And O T X T Shaft, OX T Y T The axes forming a right-hand coordinate system, the target relative coordinate system OX T Y T Z T Rotating with the earth's rotation.
Arrow coordinate system (b series)
The origin of coordinates O is the center of mass of the rocket, OX b The axis pointing along the rocket longitudinal axis towards the head, OY b In the longitudinal plane of symmetry of the rocket, perpendicular to the longitudinal axis, OZ b Shaft and OX b 、OY b The axes constitute a right-hand coordinate system.
The coordinate system transfer matrix is as follows:
the variable symbols used by the simulation model are defined as follows:
t: time of flight, takeoff time t =0s;
τ: guidance control period, τ =10ms.
V x 、V y 、V z : object system X T 、Y T 、Z T Speed in three directions, unit: m/s; (ii) a
x, y, z: object system X T 、Y T 、Z T Position in three directions, unit: m; (ii) a
g x 、g y 、g z : object system X T 、Y T 、Z T Acceleration of gravity in three directions, unit: m/s 2 ；
jx, jx: object system X T 、Y T 、Z T Moment of inertia in three directions, unit: kg x m 2
The aircraft control system is shown in a block diagram. And guidance and attitude control are designed according to a PID control law.
Since the aircraft is flying for a short time, at low altitude, and at low speed, the gravitational acceleration is calculated as a constant value, taking into account a flat, stationary ground. At OX T Y T Z T Mass center equation of motion under coordinate system:
the arrow system is according to the acceleration equation:
in the above formula, W x 、W y 、W z Is X under the target system T 、Y T 、Z T Apparent acceleration in three directions; w x1 、W y1 、W z1 For X under arrow coordinate system b 、Y b 、Z b Apparent acceleration in three directions. F x1 、F y1 、F z1 For arrow coordinate systemX of b 、Y b 、Z b Main thrust in three directions.
Mass change equation:
in the formula, R F The fuel consumption rate.
The specific fuel consumption may be linearly interpolated according to the thrust command using the following table 4 relationship, and linearly extrapolated beyond the ranges of the table.
TABLE 4
Equation of motion around the center:
in the above formula, M x1 、M y1 、M z1 The resultant moment omega in the three directions of Xb, yb and Zz under the arrow coordinate system x1 、ω y1 、ω z1 The angular velocities in the Xb, yb and Zz directions under the arrow coordinate system.
And (3) calculating pneumatic force and moment:
in the above formula, in the formula,
for a three-axis relative velocity of the centre of mass of the projectile system relative to the air stream>
Is the airspeed.
Are respectively X under arrow system b 、Y b 、Z z The sum of the wind speed in three directions and the flight speed of the aircraft. Wherein the target is X T 、Y T 、Z T Wind speeds f in three directions x 、f y 、f z The calculation formula is as follows:
f x ＝-V wind cos(A wind -π)
f y ＝0
f z ＝-V wind sin(A wind -π)
in the above formula, A wind Is the wind direction. In batch simulation, 8 directions of wind are often set. For ease of labeling, contract:
wind direction 0: a. The w Is 0 degree
Wind direction 1: a. The w Is at 45 degrees
……
The wind direction is 7: a. The w Is 315 degrees
V wind The wind speed is adopted, and the simulation is 0-5m/s.
Angle of attack α, sideslip angle β:
dynamic pressure q:
ρ is the atmospheric density at the current altitude.
Aerodynamic forces include axial forces R xv Normal force R yv Lateral force R zv Calculated as follows:
R xv ＝-C x qS M
S M is the reference area, given by the aircraft transverse plane. Coefficient of aerodynamic force C x ,
Axial force coefficient, normal force coefficient derivative, and lateral force coefficient derivative, respectively.
The specific aerodynamic parameters are calculated and given according to the actual aircraft.
The direction of the gravitational acceleration is vertical downward, and the magnitude is preferably:
g 0 ＝9.80665m/s 2
decomposed into g under the target system x 、g y 、g z 。
The resultant force here includes a thrust force, a pneumatic force, and no attractive force.
F x1 ＝R x1 +P x1
F y1 ＝R y1 +P y1
F z1 ＝R z1 +P z1
Initial position: x =0m, y =0m, and z =0m
Initial speed: v. of x ＝0m/s,v y ＝0m/s,v z ＝0m/s
Initial attitude angle:
ψ＝0,γ＝0
the initial attitude angular velocity is 0.
Turbojet thrust modulation characteristics
Maximum swing angle of the engine: 10 degrees.
Dynamic characteristics of electric steering engine
Dynamic characteristic of inertial measurement unit
The swing angle decomposition formula of the turbojet engine is as follows:
engine mounting angle
Actual angle A of engine in four quadrants of FIG. 6 1y 、A 1z 、A 2y 、A 2z ：
Two engines thrust P in x1 direction 1x1 、P 2x1 Respectively as follows:
P 1x1 ＝P 1 cos(A 1y )cos(A 1z )
P 2x1 ＝P 2 cos(A 2y )cos(A 2z )
in the above formula, P 1 、P 2 Respectively, thrust commands of the two engines.
Two engines thrust P in y1 direction 1y1 、P 2y1 Respectively as follows:
P 1y1 ＝-P 1 sin(A 1y )cos(A 1z )
P 2y1 ＝-P 2 sin(A 2y )cos(A 2z )
z1 direction thrust P of two engines 1z1 、P 2z1 Respectively as follows:
P 1z1 ＝-P 1 sin(A 1z )
P 2z1 ＝-P 2 sin(A 2z )
x under arrow system of two engines b 、Y b 、Z z Resultant force P in three directions x1 、P y1 、P z1 Respectively as follows:
x under arrow system b 、Y b 、Z b Control moments in three directions:
and (3) calculating an attitude angle:
(2) Setting various simulation deviation combinations, inputting the set simulation deviation combinations into the aircraft six-degree-of-freedom dynamic simulation model constructed in the step (1), and performing the step (3), wherein the preferable scheme is as follows:
the bias combination includes: mass, center of mass, moment of inertia, wind speed, wind direction, thrust line deflection, engine flow deviation, and the like.
The deviation combination values in the test are as follows:
TABLE 5
According to the deviation combination, the data scale can be reduced, and meanwhile, a real model is fitted as much as possible, so that the actual identification precision is ensured.
(3) Setting each fault occurrence time and fault degree, and inputting the set fault occurrence time and fault degree into the aircraft six-degree-of-freedom dynamic simulation model constructed in the step (1), wherein the preferred scheme is as follows:
the two engine numbers are 1 and 2, respectively. From the start of takeoff, the minimum value of 2s after the fault occurs and the flight ending time is taken as the end of simulation, and the fault occurring time is set once every equal interval of 0.5 s. One scheme of the fault degree is that a single fault engine is reduced by 30-90%, and the interval is 10%; the other scheme is that a single fault engine is reduced by 10% -30%, and the interval is 5%.
(4) Arranging the simulation deviation combination set in the step (2) and the fault occurrence time and the fault degree set in the step (3)Combining to obtain simulation data of the six-degree-of-freedom dynamic simulation model of the aircraft under different conditions, wherein the simulation data comprises: acceleration, attitude angle deviation; acceleration of X in the target system of (1) T 、Y T 、Z T Acceleration in three directions; the attitude angle is a pitch angle, a yaw angle and a rolling angle under a target system; the attitude angle deviation is the target system pitch angle deviation, yaw angle deviation and roll angle deviation. Storing the simulation data under different conditions;
(5) Intercepting the simulation data under different conditions to generate a data sample; designing a data label according to the number of the fault engine and the fault degree, and labeling each data sample, wherein the preferable scheme is as follows:
the intercepting process of the training sample is shown in figure 2, the intercepting process of the testing sample is shown in figure 3, the abscissa t in figures 2 and 3 is the flight time, the ordinate represents the acquired simulation data, which can be the acceleration, the attitude angle or the attitude angle deviation, L is the length of the sample, I is the overlapping size of two adjacent samples, t is the overlapping size of two adjacent samples fault For fault injection time, t end Is the simulation end time.
And taking 9-dimensional information including the acceleration, attitude angle and attitude angle deviation of 20 time points in the period as training data every 50ms in every 1s in the flight complete data. The number of each group is 9 × 20, i.e. 180 dimensional state quantities.
Assuming that n samples are total in the time from t1 to t2, the length of each sample is L, and the overlapping length between the samples is I, then:
for the kth sample: let k =1,2, \ 8230;, n,
start and end lines of the kth sample:
if a positive sample is intercepted, then in the equation:
t 1 ＝0
t 2 ＝t fault
if a positive sample is intercepted, then in the equation:
t 1 ＝t fault
t 2 ＝t end
wherein, t fault For fault injection time, t end Is the simulation end time.
In this test, the values of the parameters are as follows:
TABLE 6
| Type of parameter | Value of | Unit of |
| L | 1 | s |
| I | 0.8 | s |
| t fault | Injecting every 0.5s within 1-20s for single simulation | s |
| t end | Actual flight end time and t of ascent fault Minimum value of +2 | s |
Step (5) intercepting the simulation data under different conditions to generate a data sample, wherein the preferable scheme is as follows: and setting an algorithm for avoiding excessive positive samples, wherein the algorithm flow refers to the attached figure 4, and intercepting by adopting the algorithm to generate data samples. And judging whether the fault injection time is the last time under the current deviation combination, wherein the fault degree is the last gear under the current deviation combination. If yes, intercepting a positive sample and a negative sample; otherwise, only negative samples are intercepted.
Label 0 indicates thrust normal, i.e. no fault condition; 1-7 are respectively 30% -90% lower than the No. 1 engine; 8-14 are respectively 30% -90% of the reduction of the No. 2 engine, and the interval is 10%.
The label settings table is shown in table 7 below:
TABLE 7
Label 0 indicates thrust normal, i.e. no fault condition; 1-5 are respectively 10% -30% lower than the No. 1 engine; 6-10 are respectively 10% -30% reduction of No. 2 engine, and the interval is 5%.
The label settings table is shown in table 8 below:
TABLE 8
The aircraft thrust fault machine learning online identification sample generated according to the steps (1) to (5), preferably scheme 1 is online fault identification based on a BP neural network, and preferably further comprises the following steps (6) to (9); and (5) further solving the problem of online fault identification by using the labeled sample data obtained in the step (5).
(6) Randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set, randomly taking half of the rest part to divide a verification set, and dividing the other half into a test set; building a BP neural network, wherein the structure comprises a single hidden layer and an output layer; the number of the single hidden layer is 10, and the activation function is a Sigmoid function; the number of the neurons of the output layer is the total number of the fault categories, namely the total number of the labels in the step (5), and the activation function is a softmax function; the layers are all connected, and the preferred scheme is as follows:
the neural network structure is two layers.
The linear relationship between the layers is preferably:
and (3) neuron output:
aσ(z)
wherein:
m is the input signal dimension of the layer, i.e. the output dimension of the previous layer
x i For the input signal, i =1,2, \ 8230;, m
w i Is a weight value
b is an offset
z being the result of a linear relationship, i.e. the input to the activation function
σ is an activation function
a is neuron output
The neural network comprises a single hidden layer, 10 neurons in total, and the activation function is a Sigmoid function, preferably:
x is the linear result of the neurons in the layer, using the parameter α =1.
The output layer has 11 neurons, and the activation function is a softmax function:
the calculation formula of the jth neuron output:
k is the number of neurons in the output layer and p is the total number of classes.
The layers are all connected.
(7) Training adopts cross entropy as a loss function, adopts a gradient descent method to update parameters of the neural network, and updates the weight and the bias of the neural network; inputting the data samples in the training set in the step (5) into the BP neural network built in the step (6) for training; the training adopts a gradient descent method to update the network parameters; adopting a data sample in the verification set to test the training process, and finishing training when the error of the neural network continuously iterates the sample on the verification set for N times is not reduced to obtain a training result, wherein the method specifically comprises the following steps:
the cross-entropy expression for multi-classification is preferably as follows:
in the above formula, M is the total number of categories; yc is an indicating variable, if the output type is the same as the sample type, 1 is obtained, otherwise 0 is obtained; pc is the predicted probability that the output class belongs to c.
(8) Testing the training result by adopting the data sample in the test set, if the training result meets the requirement, storing the trained BP neural network, and performing the step (9); if not, adjusting the BP neural network built in the step (6), and returning to the step (6); the method comprises the following specific steps:
if the test accuracy is greater than or equal to 90%, the trained neural network is stored, and if the test accuracy is not greater than 90%, the number of hidden layers of the neural network in the step (6) or the number of neurons in the hidden layers is adjusted, and the training is carried out again.
(9) Embedding the BP neural network stored in the step (8) into an aircraft control computer, and performing fault online identification by using the trained BP neural network, wherein the fault online identification method specifically comprises the following steps:
in the actual flight process of the aircraft, the acceleration, attitude angle and attitude angle deviation in every three directions within every 1s are input into a BP neural network embedded in an aircraft control computer in a sliding window mode at intervals of 0.2s, and the neural network outputs an identification result in real time to guide flight control decision.
The aircraft thrust fault machine learning online identification sample generated according to the steps (1) to (5), preferably scheme 2 is online fault identification based on a decision tree, and comprises the following steps (6) to (9); and (5) further solving the problem of online fault identification by using the labeled sample data obtained in the step (5).
(6) Randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set, and dividing the rest of the labeled data into a test set; training by using a training set data sample and using a CART algorithm to generate a decision tree, and finishing training to obtain a training result when the decision tree is generated, wherein the preferable scheme is as follows:
decision trees can be considered as a set of decision rules, or as conditional probability distributions defined over a feature space and a class space. The number of discrimination rules generated in this test is 178, that is, 178 determinations of the input features.
(7) Judging the training result, if the training result meets the requirement, storing the generated decision tree, and performing the step (8); if not, returning to the step (4), wherein the preferable scheme is as follows:
and (3) testing the decision tree by adopting the training set as a training result to generate the decision tree and the accuracy of the decision tree on the training set, wherein the preferable judgment standard is as follows: if the accuracy is more than 90%, the requirement is met, and if the accuracy is less than 90%, the requirement is not met.
(8) Verifying the decision tree in the step (7) by adopting a data sample in the test set, and if the verification accuracy meets the requirement, extracting and storing a judgment rule in the decision tree; if the requirement is not met, returning to the step (4), wherein the preferable scheme is as follows:
and testing the decision tree by adopting a test set, wherein the preferable judgment standard is as follows: if the accuracy is more than 90%, the requirement is met, and if the accuracy is less than 90%, the requirement is not met.
(9) Using a decision tree to perform fault online identification, wherein the preferred scheme is as follows:
and (4) using the decision tree to perform fault online identification, and embedding the judgment rule of the decision tree into an aircraft control computer. In the actual flight process of the aircraft, the acceleration, attitude angle and attitude angle deviation in every three directions within every 1s are input into a decision tree discrimination rule embedded in an aircraft control computer in a sliding window mode at intervals of 0.2s, and an identification result is output in real time and used for guiding flight control decision.
The aircraft thrust fault machine learning online identification sample generated according to the steps (1) to (5), wherein the preferred scheme 3 is online fault identification based on an LSTM neural network, and comprises the following steps (6) to (9); and (5) further solving the problem of online fault identification by using the labeled sample data obtained in the step (5).
(6) Randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set, randomly taking half of the rest part to divide a verification set, and dividing the other half into a test set; building an LSTM neural network, wherein the structure comprises an input layer, an LSTM layer and an output layer; LSTM layer of 50 neurons in total; the input layer is input data, and is a vector of 9 × 20; the number of the neurons of the output layer is the total number of the fault categories, namely the total number of the labels in the step (5), and the activation function is a softmax function; the full-connection mode is adopted among the layers, and the full-connection mode specifically comprises the following steps:
the neural network structure is two layers, as shown in fig. 2:
the linear relationship between the layers is preferably:
and (3) neuron output:
a＝σ(z)
wherein:
m is the input signal dimension of the layer, i.e. the output dimension of the previous layer
x i For the input signal, i =1,2, \ 8230;, m
w i Is a weight value
b is an offset
z being the result of a linear relationship, i.e. the input to the activation function
σ is an activation function
a is neuron output
The output layer has 11 neurons, and the activation function is a softmax function:
the calculation formula of the jth neuron output is preferably:
k is the number of neurons in the output layer and p is the total number of classes.
The layers are all connected.
(7) Training adopts cross entropy as a loss function, adopts a gradient descent method to update parameters of the neural network, and updates the weight and the bias of the neural network; inputting the data samples in the training set in the step (5) into the LSTM neural network built in the step (6) for training; the training adopts a gradient descent method to update the network parameters; adopting a data sample in the verification set to test the training process, and finishing training when the error of the neural network continuously iterates the sample on the verification set for N times is not reduced to obtain a training result, wherein the method specifically comprises the following steps:
the cross-entropy expression for multi-classification is preferably as follows:
in the above formula, M is the total number of categories; yc is an indicating variable, if the output type is the same as the sample type, 1 is obtained, otherwise 0 is obtained; pc is the predicted probability that the output class belongs to c.
(8) Testing the training result by adopting the data sample in the test set, if the training result meets the requirement, storing the trained LSTM neural network, and performing the step (9); if not, adjusting the LSTM neural network built in the step (6), and returning to the step (6); the method comprises the following specific steps:
if the test accuracy is greater than or equal to 90%, the trained neural network is stored, and if the test accuracy is not greater than 90%, the number of hidden layers of the neural network in the step (6) or the number of neurons in the hidden layers is adjusted, and the training is carried out again.
The test precision in the test is 90.5%.
(9) Embedding the LSTM neural network stored in the step (8) into an aircraft control computer, and using the trained LSTM neural network to perform fault online identification, wherein the method specifically comprises the following steps:
in the actual flight process of the aircraft, the acceleration, attitude angle and attitude angle deviation in every three directions in every 1s are input into an LSTM neural network embedded into an aircraft control computer in a sliding window mode at intervals of 0.2s, and the neural network outputs an identification result in real time for guiding flight control decision.
The invention discloses a machine learning sample generation method for online identification of thrust faults of an aircraft, which further preferably comprises the following implementation steps:
(1) According to a real aircraft and the environment where the real aircraft is located, a six-degree-of-freedom dynamic simulation model of the aircraft is constructed, and the method specifically comprises the following steps:
the aircraft is provided with 2 engines, the resultant force of formed thrust is the total thrust, the environment is a real atmospheric environment, and the influences of wind disturbance, structural disturbance and pneumatics are considered;
the six-degree-of-freedom dynamic simulation model of the aircraft is preferably as follows: the method is characterized in that a six-degree-of-freedom dynamic simulation model is identified on line based on the thrust fault of the aircraft based on machine learning, and the specific situation of the model is as described in the above specific scheme (1).
(2) Setting each simulation deviation combination, inputting the set simulation deviation combination into the aircraft six-degree-of-freedom dynamic simulation model constructed in the step (1), and performing the step (3), wherein the specific steps are as follows:
the bias combination includes: mass, center of mass, moment of inertia, wind speed, wind direction, thrust line deflection, engine flow deviation, and the like.
The combined values of the deviations in this test are as follows:
according to the deviation combination, the data scale can be reduced, and meanwhile, a real model is fitted as much as possible, so that the actual identification precision is ensured.
(3) Setting each fault occurrence time and fault degree, and inputting the set fault occurrence time and fault degree into the aircraft six-degree-of-freedom dynamic simulation model constructed in the step (1), wherein the specific steps are as follows:
the two engine numbers are 1 and 2, respectively. From the start of takeoff, the minimum value of 2s after the fault occurs and the flight ending time is taken as the end of simulation, and the fault occurring time is set once every equal interval of 0.5 s. One scheme of the fault degree is that a single fault engine is reduced by 30-90%, and the interval is 10%; the other scheme is that a single fault engine is reduced by 10% -30%, and the interval is 5%.
(4) And (3) carrying out permutation and combination on the simulation deviation combination set in the step (2) and the fault occurrence time and the fault degree set in the step (3) to obtain simulation data of the six-degree-of-freedom dynamic simulation model of the aircraft under different conditions, wherein the simulation data comprises: acceleration, attitude angle deviation; acceleration of X in the target system of (1) T 、Y T 、Z T Acceleration in three directions; the attitude angle is a pitch angle, a yaw angle and a rolling angle under a target system; the attitude angle deviation is the target system pitch angle deviation, yaw angle deviation and roll angle deviation. Storing the simulation data under different conditions; the preferred scheme is as follows:
the characteristic quantity attributes are as follows:
(5) Intercepting the simulation data under different conditions to generate a data sample; designing a data label according to the number and the fault degree of the fault engine, and labeling each data sample, wherein the preferable scheme is as follows:
the intercepting process of the training sample is shown in figure 2, the intercepting process of the testing sample is shown in figure 3, the abscissa t in figures 2 and 3 is the flight time, the ordinate y represents the acquired simulation data, which can be the acceleration, the attitude angle or the attitude angle deviation, L is the length of the sample, I is the overlapping size of two adjacent samples, t is the overlapping size of two adjacent samples fault For fault injection time, t end Is the simulation end time.
And taking 9-dimensional information including the acceleration, attitude angle and attitude angle deviation of 20 time points in the period as training data every 50ms in every 1s in the flight complete data. The number of each group is 9 × 20, i.e. 180 dimensional state quantities.
Assuming that n samples are total in the time from t1 to t2, the length of each sample is L, and the overlapping length between the samples is I, then:
for the kth sample: let k =1,2, \ 8230;, n,
start and end lines of the kth sample:
if a positive sample is intercepted, then in the equation:
t 1 ＝0
t 2 ＝t fault
if a positive sample is intercepted, then in the equation:
t 1 ＝t fault
t 2 ＝t end
wherein, t fault For fault injection time, t end Is the simulation end time.
In this test, the values of the parameters are as follows:
step (5) intercepting the simulation data under different conditions to generate a data sample, wherein the preferable scheme is as follows: and setting an algorithm for avoiding excessive positive samples, wherein the algorithm flow refers to the attached figure 4, and intercepting by adopting the algorithm to generate data samples. And judging whether the fault injection time is the last time under the current deviation combination, wherein the fault degree is the last gear under the current deviation combination. If yes, intercepting a positive sample and a negative sample; otherwise, only negative samples are intercepted.
Label 0 indicates thrust normal, i.e. no fault condition; 1-7 are respectively 30% -90% lower than the No. 1 engine; 8-14 are respectively 30% -90% of the reduction of the No. 2 engine, and the interval is 10%.
The tag settings table is as follows:
label 0 indicates thrust normal, i.e. no fault condition; 1-5 are respectively 10% -30% lower than the No. 1 engine; 6-10 are respectively 10% -30% reduction of No. 2 engine, and the interval is 5%.
The tag settings table is as follows:
by using the obtained data sample, a preferred scheme is selected for fault identification according to the preferred scheme based on machine learning algorithm online fault identification described in embodiment 3.
Thrust fault online identification based on a BP neural network is adopted, and the method specifically comprises the following steps:
(6) Randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set, randomly taking half of the rest part to divide a verification set, and dividing the other half into a test set; building a BP neural network, wherein the structure comprises a single hidden layer and an output layer; the number of single hidden layers is 10, and the activation function is a Sigmoid function; the number of the neurons of the output layer is the total number of the fault categories, namely the total number of the labels in the step (5), and the activation function is a softmax function; the full-connection mode is adopted among the layers, and the full-connection mode specifically comprises the following steps:
the neural network structure is two layers, and the linear relationship between the layers is as follows:
and (3) neuron output:
a＝σ(z)
wherein:
m is the input signal dimension of the layer, i.e. the output dimension of the previous layer
x i For the input signal, i =1,2, \ 8230;, m
w i Is a weight value
b is an offset
z being the result of a linear relationship, i.e. the input to the activation function
σ is an activation function
a is neuron output
The neural network comprises a single hidden layer, 10 neurons in total, and the activation function is a Sigmoid function:
x is the linear result of the neurons in the layer, using the parameter α =1.
The output layer has 11 neurons, and the activation function is a softmax function:
the calculation formula of the jth neuron output:
k is the number of neurons in the output layer and p is the total number of classes.
The layers are all connected.
(7) Training adopts cross entropy as a loss function, adopts a gradient descent method to update parameters of the neural network, and updates the weight and the bias of the neural network; inputting the data samples in the training set in the step (5) into the BP neural network built in the step (6) for training; the training adopts a gradient descent method to update the network parameters; adopting a data sample in the verification set to test the training process, and finishing training when the error of the neural network continuously iterates the sample on the verification set for N times is not reduced to obtain a training result, wherein the method specifically comprises the following steps:
the cross-entropy expression for multi-classification is as follows:
in the above formula, M is the total number of categories; yc is an indicating variable, if the output type is the same as the sample type, 1 is obtained, otherwise 0 is obtained; pc is the predicted probability that the output class belongs to c.
(8) Testing the training result by adopting the data sample in the test set, if the training result meets the requirement, storing the trained BP neural network, and performing the step (9); if not, adjusting the BP neural network built in the step (6), and returning to the step (6); the method comprises the following specific steps:
if the test accuracy is greater than or equal to 90%, the trained neural network is saved, if the test accuracy is not met, the number of hidden layers of the neural network in the step (6) or the number of neurons in the hidden layers is adjusted, and the training is carried out again.
(9) Embedding the BP neural network stored in the step (8) into an aircraft control computer, and performing fault online identification by using the trained BP neural network, wherein the fault online identification method specifically comprises the following steps:
in the actual flight process of the aircraft, the acceleration, attitude angle and attitude angle deviation in every three directions in every 1s are input into a BP neural network embedded into an aircraft control computer in a sliding window mode at intervals of 0.2s, and the neural network outputs an identification result in real time for guiding flight control decision.
Through flight demonstration verification, through data analysis, whether a fault exists or not and a single engine with reduced thrust can be identified by 100%, the real-time identification is within 2s after the fault is stable, and the identification error of the thrust phase difference value of the two engines is within 10%.
Claims (10)
1. A machine learning sample generation method for online identification of thrust faults of an aircraft is characterized by comprising the following steps:
(1) Constructing a six-degree-of-freedom dynamic simulation model of the aircraft according to the real aircraft and the environment where the real aircraft is located;
(2) Setting each simulation deviation combination, wherein the deviation combination comprises: mass, mass center, rotational inertia, wind speed, wind direction, thrust line deflection and engine flow deviation; inputting the set simulation deviation combination into the aircraft six-degree-of-freedom dynamic simulation model constructed in the step (1); carrying out the step (3);
(3) Setting each fault occurrence time and fault degree, wherein the fault occurrence time is set as follows: from the start of takeoff, taking the minimum value of Ns after the fault and the flight ending time as the end of simulation, and setting the fault occurring time once every equal interval (N/4) s; the degree of failure is set to: the single fault engine is decreased by a first preset percentage to a second preset percentage, and the fault degree is divided at equal intervals; inputting the set fault occurrence time and fault degree into the aircraft six-degree-of-freedom dynamic simulation model constructed in the step (1);
(4) And (3) carrying out permutation and combination on the simulation deviation combination set in the step (2) and the fault occurrence time and the fault degree set in the step (3) to obtain simulation data of the six-degree-of-freedom dynamic simulation model of the aircraft under different combination conditions, wherein the simulation data comprises: acceleration, attitude angle deviation; storing the simulation data under different combination conditions;
(5) Intercepting the simulation data under different combination conditions to generate a data sample; the method specifically comprises the following steps: judging whether the fault injection time is the last time under the current deviation combination, and the fault degree is the last gear under the current deviation combination; if so, intercepting the positive sample and the negative sample, otherwise, only intercepting the negative sample; and designing a data label according to the number of the fault engine and the fault degree, and labeling each data sample.
2. The machine learning sample generation method for the online identification of the thrust fault of the aircraft according to claim 1, characterized in that: further comprising steps (6) to (9);
(6) Randomly taking more than most of the labeled data in the step (5) to divide the data into a training set, randomly taking half of the rest part to divide a verification set, and dividing the other half into a test set; building a BP neural network, wherein the structure comprises a single hidden layer and an output layer; the number of single hidden layers is 10, and the activation function is a Sigmoid function; the number of neurons in the output layer is 11, and the activation function is a softmax function; the layers are all connected;
(7) Training adopts cross entropy as a loss function, adopts a gradient descent method to update parameters of the neural network, and updates the weight and the bias of the neural network; inputting the data samples in the training set in the step (5) into the BP neural network built in the step (6) for training; the training adopts a gradient descent method to update the network parameters; testing the training process by adopting the data samples in the verification set, and finishing the training when the error of the samples on the verification set is not reduced for N times continuously by the neural network to obtain a training result;
(8) Testing the training result by adopting the data sample in the test set, if the training result meets the requirement, storing the trained BP neural network, and performing the step (9); if not, adjusting the BP neural network built in the step (6), and returning to the step (6);
(9) And (4) embedding the BP neural network stored in the step (8) into an aircraft control computer, and performing online fault identification by using the trained BP neural network.
3. The machine learning sample generation method for the online identification of the thrust fault of the aircraft according to claim 2, characterized in that: and (6) randomly taking more than most of the labeled data in the step (5) and dividing the data into training sets, specifically: and (4) randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set.
4. The machine learning sample generation method for the online identification of the thrust fault of the aircraft according to claim 2, characterized in that: and (4) outputting 11 neurons in the layer in the step (6), wherein the number of the neurons is the total number of the fault categories.
5. The machine learning sample generation method for the online identification of the thrust fault of the aircraft according to claim 4, wherein: the total number of fault categories is the total number of labels in step (5).
6. The machine learning sample generation method for the online identification of the thrust fault of the aircraft according to claim 1, characterized in that: further comprising steps (6) to (9);
(6) Randomly taking more than most of the labeled data in the step (5) to be divided into a training set, and dividing the rest into a test set; training by using a training set data sample and a CART algorithm to generate a decision tree, and when the decision tree is generated, finishing training to obtain a training result;
(7) Judging the training result, if the training result meets the requirement, storing the generated decision tree, and performing the step (8); if not, returning to the step (4);
(8) Verifying the decision tree in the step (7) by adopting a data sample in the test set, and if the verification accuracy meets the requirement, extracting and storing a judgment rule in the decision tree; if the requirement is not met, returning to the step (4);
(9) And using a decision tree to perform online fault identification.
7. The machine learning sample generation method for aircraft thrust fault online identification according to claim 6, wherein: and (6) randomly taking more than most of the labeled data in the step (5) and dividing the data into training sets, specifically: and (4) randomly taking more than 2/3 of the labeled data in the step (5) to be divided into a training set.
8. The machine learning sample generation method for the online identification of the thrust fault of the aircraft according to claim 1, characterized in that: further comprising steps (6) to (9);
(6) Randomly taking more than most of the labeled data in the step (5) to divide the data into a training set, randomly taking half of the rest part to divide a verification set, and dividing the other half into a test set; building an LSTM neural network, wherein the structure comprises an input layer, an LSTM layer and an output layer; LSTM layer has 50 neurons in total; the number of the neurons of the output layer is the total number of the fault categories, namely the total number of the labels in the step (5), and the activation function is a softmax function; the layers are all connected;
(7) Training adopts cross entropy as a loss function, adopts a gradient descent method to update parameters of the neural network, and updates the weight and the bias of the neural network; inputting the data samples in the training set in the step (5) into the LSTM neural network built in the step (6) for training; the training adopts a gradient descent method to update the network parameters; testing the training process by adopting the data samples in the verification set, finishing the training when the error of the neural network continuously iterates the samples on the verification set for N times is not reduced to obtain a training result,
(8) Testing the training result by adopting the data sample in the test set, if the training result meets the requirement, storing the trained LSTM neural network, and performing the step (9); if not, adjusting the LSTM neural network built in the step (6), and returning to the step (6);
(9) And (4) embedding the LSTM neural network stored in the step (8) into an aircraft control computer, and performing online fault identification by using the trained LSTM neural network.
9. The machine learning sample generation method for aircraft thrust fault online identification according to claim 8, wherein: (6) Randomly taking more than most of the labeled data in the step (5) and dividing the data into training sets, specifically: and (4) randomly taking more than 2/3 of the labeled data in the step (5) to divide the labeled data into a training set.
10. The machine learning sample generation method for aircraft thrust fault online identification according to claim 8, wherein: the input layers, i.e. the input data, are vectors of 9 x 20.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010076010.9A CN111240304B (en) | 2020-01-23 | 2020-01-23 | Machine learning sample generation method for online identification of thrust fault of aircraft |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010076010.9A CN111240304B (en) | 2020-01-23 | 2020-01-23 | Machine learning sample generation method for online identification of thrust fault of aircraft |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111240304A CN111240304A (en) | 2020-06-05 |
| CN111240304B true CN111240304B (en) | 2023-03-24 |
Family
ID=70869934
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010076010.9A Active CN111240304B (en) | 2020-01-23 | 2020-01-23 | Machine learning sample generation method for online identification of thrust fault of aircraft |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111240304B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112987695A (en) * | 2021-03-12 | 2021-06-18 | 北京航天自动控制研究所 | Aircraft health prediction method |
| CN114237269B (en) * | 2021-11-29 | 2024-06-07 | 北京航天自动控制研究所 | Method for constructing polar fault mode recognition model of attitude control system |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101625571B (en) * | 2009-07-25 | 2010-12-29 | 大连理工大学 | Method for simulating six degrees of freedom movement of spinning aircraft |
| CN102880182B (en) * | 2012-09-12 | 2015-01-14 | 北京航空航天大学 | Microminiature unmanned aerial vehicle controlling method having network random delay problem |
| CN103116280B (en) * | 2013-01-16 | 2016-03-02 | 北京航空航天大学 | A kind of exist the longitudinal control method of the microminiature unmanned vehicle becoming distributed network random delay |
| CN103065022B (en) * | 2013-01-24 | 2015-07-15 | 无锡华航电子科技有限责任公司 | Incentive model simulation platform for aerocraft electronic system |
| CN103412557A (en) * | 2013-07-17 | 2013-11-27 | 浙江工业大学 | Industrial fault detection and diagnostic method suitable for nonlinear process on-line monitoring |
| CN103587723B (en) * | 2013-11-07 | 2015-12-02 | 北京临近空间飞行器系统工程研究所 | One reenters initial segment analytic expression longitudinally online Trajectory Design and tracking |
| CN104122885B (en) * | 2014-07-15 | 2017-02-01 | 北京航空航天大学 | 422 bus solid-liquid power aircraft flying control testing system |
| RU2601367C1 (en) * | 2015-06-18 | 2016-11-10 | Открытое акционерное общество Московский научно-производственный комплекс "Авионика" имени О.В. Успенского (ОАО МНПК "Авионика") | Method of aircraft engine thrust determining at operating mode change and device for its implementation |
-
2020
- 2020-01-23 CN CN202010076010.9A patent/CN111240304B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111240304A (en) | 2020-06-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111176263B (en) | Online identification method for thrust fault of aircraft based on BP neural network | |
| CN111258302B (en) | Aircraft thrust fault online identification method based on LSTM neural network | |
| CN111221345B (en) | Decision tree-based aircraft power system fault online identification method | |
| CN109614633A (en) | A kind of composite rotor craft non-linear modeling method and Calculate Ways | |
| CN106842926B (en) | A kind of aerial vehicle trajectory optimization method based on positive real B-spline | |
| CN109270947A (en) | Tilting rotor wing unmanned aerial vehicle flight control system | |
| CN111240304B (en) | Machine learning sample generation method for online identification of thrust fault of aircraft | |
| CN112348124B (en) | Data-driven micro fault diagnosis method and device | |
| CN109635494B (en) | Flight test and ground simulation aerodynamic force data comprehensive modeling method | |
| Lu et al. | Real-time simulation system for UAV based on Matlab/Simulink | |
| CN103995540A (en) | Method for rapidly generating finite time track of hypersonic aircraft | |
| CN107103164A (en) | Unmanned plane performs the distribution method and device of multitask | |
| CN107065901A (en) | A kind of rotor wing unmanned aerial vehicle attitude control method, device and unmanned plane | |
| CN106896722B (en) | The hypersonic vehicle composite control method of adoption status feedback and neural network | |
| CN107367941B (en) | Method for observing attack angle of hypersonic aircraft | |
| CN111813146A (en) | Reentry prediction-correction guidance method based on BP neural network prediction voyage | |
| CN111638654A (en) | Fault-adaptive intelligent control semi-physical simulation method for carrier rocket | |
| Li et al. | Anti-disturbance control for attitude and altitude systems of the helicopter under random disturbances | |
| CN111830848A (en) | Unmanned aerial vehicle super-maneuvering flight performance simulation training system and method | |
| CN106096091A (en) | A kind of airplane motion analogy method | |
| CN116300992A (en) | L-based 1 Adaptive dynamic inverse variant aircraft control method | |
| CN106774385A (en) | A kind of dirigible spot hover control method of use adaptive variable structure | |
| CN115407294B (en) | Aircraft dynamic radar scattering cross section simulation method for double-station radar system | |
| CN116414153A (en) | Unmanned aerial vehicle take-off and landing correction method based on artificial intelligence | |
| Wang et al. | Deep learning based missile trajectory prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |