CN114313307A - Unmanned aerial vehicle control plane fault alarm method and device based on machine learning - Google Patents
Unmanned aerial vehicle control plane fault alarm method and device based on machine learning Download PDFInfo
- Publication number
- CN114313307A CN114313307A CN202210092099.7A CN202210092099A CN114313307A CN 114313307 A CN114313307 A CN 114313307A CN 202210092099 A CN202210092099 A CN 202210092099A CN 114313307 A CN114313307 A CN 114313307A
- Authority
- CN
- China
- Prior art keywords
- control surface
- oscillation
- data set
- fault diagnosis
- flight parameter
- 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.)
- Granted
Links
Images
Landscapes
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention provides a control surface fault alarm method and device of an unmanned aerial vehicle based on machine learning, wherein the method comprises the following steps: acquiring a flight parameter data set, and preprocessing the flight parameter data set to obtain a control surface prior data set; dividing a control surface prior data set into a training set and a test set according to a preset proportion; learning the training set by adopting a machine learning method to obtain a control surface fault diagnosis model after training; testing the diagnostic precision of the model through the test set, and gradually iterating the model algorithm; acquiring a flight parameter data set to be diagnosed, inputting the flight parameter data set to be diagnosed into a trained control surface fault diagnosis model to perform fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition. The method can effectively establish an accurate control surface fault diagnosis model, can input the flight parameter data set to be diagnosed in real time, has high fault diagnosis accuracy, has stronger characteristic prediction capability of the trained control surface fault diagnosis model, and can accurately and timely carry out fault alarm.
Description
Technical Field
The invention relates to the field of unmanned aerial vehicle fault diagnosis and alarm, in particular to a control plane fault alarm method and device of an unmanned aerial vehicle based on machine learning.
Background
The unmanned aerial vehicle control surface system is an important component of an unmanned aerial vehicle, the reliability of the unmanned aerial vehicle control surface system is an important guarantee of safe flight, particularly, the unmanned aerial vehicle control system can be directly influenced by the ailerons, the elevators and the rudders which are used as the three-large control surface system, once the ailerons, the elevators and the rudders break down, the stability of the unmanned aerial vehicle flight control system is reduced, and the airplane crash is directly caused. Therefore, the method has important significance for state detection and fault early warning research of the unmanned aerial vehicle control surface system.
At present, when the control surface limit ring of an unmanned aerial vehicle vibrates, the traditional fault diagnosis algorithm cannot effectively judge fault characteristics, and due to the fact that a complex coupling relation exists inside a flight control system, a physical model and a fault model which are accurate in system are difficult to effectively establish, and the problems that the accuracy of fault diagnosis of the flight control system is low and the prediction capability is high in uncertainty are caused.
Therefore, how to provide a control plane fault alarm method of an unmanned aerial vehicle, which is suitable for real-time data input, has high accuracy and characteristic prediction capability, becomes a problem to be solved urgently.
Disclosure of Invention
The present invention is directed to solving one of the problems set forth above.
The invention mainly aims to provide a control surface fault alarm method of an unmanned aerial vehicle based on machine learning.
The invention further aims to provide a control surface fault alarm device of the unmanned aerial vehicle based on machine learning.
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
the invention provides a control surface fault alarm method of an unmanned aerial vehicle based on machine learning, which comprises the following steps:
acquiring a flight parameter data set, wherein the flight parameter data set comprises data influencing control surface oscillation;
preprocessing the flight parameter data set to obtain a control surface prior data set, wherein the control surface prior data set comprises: the flight parameter data set and limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the flight parameter data set, wherein the limit ring oscillation data identification comprises a numerical value for identifying control surface oscillation and a numerical value for identifying non-oscillation of the control surface;
dividing the control surface prior data set into a training set and a test set according to a preset proportion;
learning the training set by adopting a machine learning method to obtain a trained control surface fault diagnosis model;
acquiring a flight parameter data set to be diagnosed, inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model to perform fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition.
Optionally, the data affecting the oscillation of the control surface includes: time mark parameters, rudder position parameters and rudder deflection instruction parameters; the preprocessing is performed on the flight parameter data set to obtain a control surface prior data set, and the method comprises the following steps: dividing the flight parameter data set into n groups of flight parameter data sets, wherein each group of flight parameter data sets comprises m pieces of data influencing control surface oscillation, and m and n are positive integers; performing the following operations for each set of flight parameter data: sequentially acquiring m pieces of data influencing the oscillation of the control surface, calculating a deviation value of the rudder deflection instruction parameter and the rudder position parameter in the m pieces of data influencing the oscillation of the control surface, judging whether the absolute value of the deviation value is greater than a first preset threshold value or not and whether the absolute value of the rudder deflection instruction parameter is greater than a second preset threshold value or not, and if so, adding 1 to a counter; otherwise, the counter is increased by 0; judging whether the count value of the counter is larger than a third preset threshold value, if so, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation as a numerical value for identifying the control surface oscillation; otherwise, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation as a numerical value for identifying that the control surface does not oscillate.
Optionally, the learning of the training set by using a machine learning method to obtain a trained control surface fault diagnosis model includes: and learning the training set by adopting various machine learning methods to obtain a plurality of to-be-evaluated fault diagnosis models, and selecting an optimal to-be-evaluated fault diagnosis model as the trained control surface fault diagnosis model according to the accuracy of each to-be-evaluated fault diagnosis model and the ROC curve value of the operating characteristics of the testees.
Optionally, the method of machine learning includes at least one of: linear algorithms, proximity algorithms, support vector machine algorithms, and neural network algorithms.
Optionally, before learning the training set by using the machine learning method to obtain the trained control surface fault diagnosis model, the method further includes: and carrying out normalization processing on the control surface prior data set.
Optionally, the method further includes: and inputting the test set into the trained control surface fault diagnosis model for testing, evaluating the performance of the trained control surface fault diagnosis model according to the accuracy, the ROC curve value, the precision ratio and the recall ratio of the test result, and adjusting the parameters of the trained control surface fault diagnosis model.
Optionally, inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model to perform fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition, including:
and inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model in groups according to a preset mode to obtain a plurality of groups of analysis results, and performing control surface oscillation alarm when the oscillation times in the plurality of groups of analysis results exceed the preset times.
The invention provides a control surface fault alarm device of an unmanned aerial vehicle based on machine learning, which comprises the following components:
the flight parameter data set comprises data influencing the oscillation of the control surface;
the pre-processing module is used for pre-processing the flight parameter data set to obtain a control surface prior data set, the control surface prior data set comprises the flight parameter data set and a limit ring oscillation data identifier obtained by carrying out control surface oscillation evaluation on the flight parameter data set, and the limit ring oscillation data identifier comprises a numerical value used for identifying control surface oscillation and a numerical value used for identifying that the control surface does not oscillate;
the model construction module is used for dividing the control surface prior data set into a training set and a test set according to a preset proportion; learning the training set by adopting a machine learning method to obtain a trained control surface fault diagnosis model;
and the fault diagnosis alarm module is used for acquiring a flight parameter data set to be diagnosed, inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model for fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition.
Optionally, the data affecting the oscillation of the control surface includes: time mark parameters, rudder position parameters and rudder deflection instruction parameters; the preprocessing module preprocesses the flight parameter data set in the following mode to obtain a control surface prior data set: dividing the flight parameter data set into n groups of flight parameter data sets, wherein each group of flight parameter data sets comprises m pieces of data influencing control surface oscillation, and m and n are positive integers; performing the following operations for each set of flight parameter data: sequentially acquiring m pieces of data influencing the oscillation of the control surface, calculating a deviation value of the rudder deflection instruction parameter and the rudder position parameter in the m pieces of data influencing the oscillation of the control surface, judging whether the absolute value of the deviation value is greater than a first preset threshold value or not and whether the absolute value of the rudder deflection instruction parameter is greater than a second preset threshold value or not, and if so, adding 1 to a counter; otherwise, the counter is increased by 0; judging whether the count value of the counter is larger than a third preset threshold value, if so, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation as a numerical value for identifying the control surface oscillation; otherwise, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation as a numerical value for identifying that the control surface does not oscillate.
Optionally, the model building module learns the training set by using a machine learning method in the following manner to obtain a trained control surface fault diagnosis model: and learning the training set by adopting various machine learning methods to obtain a plurality of to-be-evaluated fault diagnosis models, and selecting an optimal to-be-evaluated fault diagnosis model as the trained control surface fault diagnosis model according to the accuracy of each to-be-evaluated fault diagnosis model and the ROC curve value of the operating characteristics of the testees.
According to the technical scheme provided by the invention, the method and the device for alarming the fault of the control surface of the unmanned aerial vehicle based on machine learning can effectively establish a systematically accurate control surface fault diagnosis model, can input a flight parameter data set to be diagnosed in real time through the diagnosis precision of a test set test model and gradual iteration of a model algorithm, have high fault diagnosis accuracy, have strong characteristic prediction capability of the trained control surface fault diagnosis model, and can accurately and timely alarm the fault.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
Fig. 1 is a flowchart of a control plane fault alarm method of an unmanned aerial vehicle based on machine learning according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a control surface fault alarm device of an unmanned aerial vehicle based on machine learning according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
In the description of the present invention, it is to be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", and the like, indicate orientations or positional relationships based on those shown in the drawings, and are used only for convenience in describing the present invention and for simplicity in description, and do not indicate or imply that the referenced devices or elements must have a particular orientation, be constructed and operated in a particular orientation, and thus, are not to be construed as limiting the present invention. Furthermore, the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or quantity or location.
In the description of the present invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
Embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.
Examples
Machine learning is a data-driven algorithm, a system model is not required to be established, and only historical data of system operation is required to be collected, so that the optimal characteristic representation of the system can be obtained, and tasks such as fault classification and fault prediction are completed.
Fig. 1 shows a flowchart of a control surface fault alarm method of a drone based on machine learning according to the present embodiment.
As shown in fig. 1, the control surface fault alarm method of the unmanned aerial vehicle based on machine learning includes (steps S101-S105):
s101, acquiring a flight parameter data set, wherein the flight parameter data set comprises: data affecting control surface oscillation.
In this embodiment, the data affecting the control surface oscillation may be different control surface data, such as a rudder position parameter, a flap position parameter, an aileron position parameter, a slat position parameter, and the like. Different control surface data can be adopted according to different control surface faults. Furthermore, the data affecting the control surface oscillations may also be other types of flight parameter data sets affecting the control surface oscillations, such as the vibration coefficient of the engine, aerodynamic parameters, the material of the control surface, the steering engine stiffness, the clearance of the steering engine, the control surface mass, etc. Therefore, the prediction capability of fault diagnosis caused by different reasons can be improved according to different data influencing the control surface oscillation.
The unmanned aerial vehicle transmits a flight parameter data set back to the ground command station, and can contain thousands or even tens of thousands of pieces of flight parameter data. Specifically, table 1 exemplarily presents partial data of the flight parameter data set, specifically including a time scale parameter, a left rudder position parameter, a right rudder position parameter, and a rudder deflection command parameter, specifically see table 1.
Table 1 partial flight parameter data set example
S102, preprocessing a flight parameter data set to obtain a control surface prior data set, wherein the control surface prior data set comprises the flight parameter data set and a limit ring oscillation data identifier obtained by carrying out control surface oscillation evaluation on the flight parameter data set, and the limit ring oscillation data identifier comprises a numerical value for identifying control surface oscillation and a numerical value for identifying that a control surface does not oscillate;
in this embodiment, the flight parameter data set is evaluated, and if the control surface oscillation is evaluated, the limit cycle oscillation data flag is recorded as a value for identifying the control surface oscillation, for example, 1, and if the control surface oscillation is evaluated as not oscillating, the limit cycle oscillation data flag is recorded as a value for identifying the control surface oscillation, for example, 0.
In step S102, the preprocessing the flight parameter data set includes: and carrying out control surface oscillation evaluation on the flight parameter data set, recording the limit ring oscillation data identifier obtained by evaluation as a numerical value for identifying control surface oscillation if the control surface oscillation is evaluated, and recording the limit ring oscillation data identifier obtained by evaluation as a numerical value for identifying control surface oscillation if the control surface oscillation is not evaluated, thereby obtaining control surface prior data.
As an optional implementation manner in the embodiment of the present invention, the data affecting the oscillation of the control surface includes: under the condition of time mark parameter, rudder position parameter and rudder skewness instruction parameter, the flight parameter data set is preprocessed to obtain a rudder surface prior data set, which comprises the following steps: dividing the flight parameter data set into n groups of flight parameter data sets, wherein each group of flight parameter data sets comprises m data influencing the oscillation of the control surface, and m and n are positive integers; performing the following operations for each set of flight parameter data: sequentially acquiring m data influencing the oscillation of the control surface, calculating a deviation value of a rudder deflection instruction parameter and a rudder position parameter in the m data influencing the oscillation of the control surface, judging whether the absolute value of the deviation value is greater than a first preset threshold value or not and whether the absolute value of the rudder deflection instruction parameter is greater than a second preset threshold value or not, and if so, adding 1 to a counter; otherwise, the counter is increased by 0; judging whether the count value of the counter is larger than a third preset threshold value, if so, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on m pieces of data influencing control surface oscillation as a numerical value for identifying control surface oscillation; otherwise, the limit ring oscillation data obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation are recorded as a numerical value for identifying that the control surface does not oscillate.
Specifically, in an application example, assuming that the flight parameter data set has 20000 pieces of flight parameter data, which are divided into 200 groups, and each group has 100 pieces of flight parameter data, taking 100 pieces of flight parameter data of the 1 st group as an example, the control surface oscillation evaluation is performed starting from the 1 st flight parameter data of the 1 st group, which specifically includes: calculating a deviation value of a rudder deflection instruction parameter and a rudder position parameter in the 1 st flight parameter data to obtain a 1 st deviation value, judging whether the absolute value of the deviation value is greater than a first preset threshold value and whether the absolute value of the rudder deflection instruction parameter is greater than a second preset threshold value, and if so, adding 1 to a counter; otherwise, the counter is increased by 0; the initial value of the counter in each group is 0, the first preset threshold may be set to 0.66 ° and the second preset threshold may be set to 1 ° according to experience of a technician, and of course, the first preset threshold and the second preset threshold may be set as needed, which is not limited in this embodiment. Subsequently, the subsequent flight parameter data in the 1 st group are sequentially subjected to control surface oscillation evaluation in the same manner as described above. During the control surface oscillation evaluation of the flight parameter data of the 1 st group, if the count value of the counter is greater than a third preset threshold value, the limit ring oscillation data flag of the group is recorded as a numerical value for flag of control surface oscillation, for example, 1. If the count value of the counter is less than or equal to a third preset threshold, the limit cycle oscillation data flag of the set is recorded as a value for identifying that the control surface is not oscillating, such as 0. And sequentially carrying out control surface oscillation evaluation on the subsequent 199 groups of flight parameter data in the same manner of carrying out control surface oscillation evaluation on the 100 flight parameter data of the 1 st group, thereby finishing the control surface oscillation evaluation on the 200 groups of flight parameter data and obtaining a control surface prior data set.
Specifically, the control surface prior data set may include n rows, that is, n sets of divided flight parameter data, where each row includes m columns of flight parameter data and 1 column of limit ring oscillation data identifier, that is, each set of flight parameter data set includes m pieces of data affecting control surface oscillation and a limit ring oscillation data identifier obtained by performing control surface oscillation evaluation on the m pieces of data affecting control surface oscillation. Table 2 gives an exemplary partial control surface prior data set. Referring to table 2, the control surface prior data set is composed of a flight parameter data set and a limit ring oscillation data identifier obtained by performing control surface oscillation evaluation on the flight parameter data set. The last column is a limit ring oscillation data identifier obtained by carrying out control surface oscillation evaluation on the flight parameter data set, and the data before the last column is the flight parameter data set.
The specific value of the limit ring oscillation data identifier is 0 or 1, in a 3 rd row example, the value of the last column is 0, and the control surface oscillation of the flight parameter data of the row is evaluated as the control surface does not oscillate. For another example, each 100 flight parameter data sets may be subjected to control surface oscillation evaluation, where in table 2, 101 columns of data are shown, the first 100 columns are flight parameter data, and the 101 th column is identified by limit ring oscillation data.
Table 2 partial control surface prior data set example
S103, dividing a control surface prior data set into a training set and a test set according to a preset proportion;
specifically, the preset ratio may be set empirically, for example, according to a training set: and the test set is a preset ratio of 3:1, and the control plane prior data set is divided into a training set and a test set. Any of the 3/4 data in Table 2 was used as a training set, leaving 1/4 data set as a test set. During division, after each group of data in the control surface prior data set is disordered randomly, the training set and the test set can be divided according to a preset proportion. For example, the predetermined ratio may also be 8:2, which is not limited in the present invention. The control surface fault diagnosis model after training can be obtained by learning the training set, and the control surface fault diagnosis model after training can be verified and parameter-adjusted by the test set.
S104, learning the training set by adopting a machine learning method to obtain a trained control surface fault diagnosis model;
as an optional implementation manner in the embodiment of the present invention, learning a training set by using a machine learning method to obtain a trained control surface fault diagnosis model includes: and learning the training set by adopting various machine learning methods to obtain a plurality of to-be-evaluated fault diagnosis models, and selecting an optimal to-be-evaluated fault diagnosis model as the trained control surface fault diagnosis model according to the accuracy of each to-be-evaluated fault diagnosis model and the curve value of an ROC curve (receiver operating characteristic curve).
The machine learning method comprises one of the following steps: linear algorithms, proximity algorithms (k Nearest Neighbor, abbreviated as KNN), Support Vector Machine algorithms (SVM) and Neural Network (ANN) algorithms.
The accuracy and the ROC curve are both an evaluation mode for judging whether the prediction capability of the fault diagnosis model to be evaluated is accurate. Specifically, as an optional application example, table 3 exemplarily shows the accuracy and ROC curve values of four types of fault diagnosis models to be evaluated (linear models, MLP models, KNN models, and SVM models) obtained by using the above 4 types of learning algorithms.
| Model name | Value of ROC curve | Rate of accuracy |
| Linear model | 0.90 | 0.8592 |
| ANN model | 0.93 | 0.8930 |
| KNN model | 0.92 | 0.8704 |
| SVM model | 0.82 | 0.7944 |
TABLE 3 accuracy and ROC curve values for each to-be-evaluated fault diagnosis model
Referring to table 3, it can be seen that, in the 4 to-be-evaluated fault diagnosis models, the accuracy of the neural network model obtained by learning the training set by the neural network algorithm is 0.8930, the ROC curve value is 0.93, and the value is optimal, so in this application example, the trained neural network model is selected as the trained control surface fault diagnosis model.
As an optional implementation manner in this embodiment, before learning a training set by using a machine learning method to obtain a trained control surface fault diagnosis model, the control surface fault alarm method for an unmanned aerial vehicle based on machine learning provided in this embodiment further includes: and carrying out normalization processing on the control surface prior data set.
Specifically, the control plane prior data set may be normalized by a Principal Component Analysis (PCA), and particularly, when the flight parameter data set includes a plurality of different types of data (such as different control plane data, or vibration coefficients of an engine, aerodynamic parameters, and the like), on the premise of less information loss, the data in the high dimension is converted into the data in the low dimension, and the values are normalized to 0 to 1, so as to reduce the calculation amount.
As an optional implementation manner in this embodiment, the method for alarming a control plane fault of an unmanned aerial vehicle based on machine learning provided in this embodiment further includes: and inputting the test set into the trained control surface fault diagnosis model for testing, evaluating the performance of the trained control surface fault diagnosis model according to the accuracy of the test result, the ROC curve value, the precision ratio and the recall ratio, and adjusting the parameters of the trained control surface fault diagnosis model. Particularly, the error judgment can be reduced through the accuracy, the ROC curve value and the precision ratio. The misjudgment means that the control surface is evaluated as oscillating when there is no oscillation of the control surface. According to the method, the diagnosis precision of the test set test model is tested, and the model algorithm is iterated step by step, so that the optimal performance of the trained control surface fault diagnosis model is ensured, and the fault diagnosis accuracy is high.
And S105, acquiring a flight parameter data set to be diagnosed, inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model to perform fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition.
As an optional implementation manner in this embodiment, inputting a flight parameter data set to be diagnosed into a trained control surface fault diagnosis model to perform fault diagnosis to obtain an analysis result, and giving an alarm when the analysis result meets a preset condition, where the method includes: and inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model in groups according to a preset mode to obtain a plurality of groups of analysis results, and alarming when the oscillation frequency in each group of analysis results exceeds the preset frequency. For example, the left rudder position parameter data sets acquired in real time are 1 group every 100, 6 groups are used as one rudder surface oscillation detection, 6 groups of data are sequentially input into the trained rudder surface fault diagnosis model, and if the diagnosis results of 3 groups of 6 groups are that the rudder surface oscillates, the rudder surface oscillation alarm is performed.
By the control surface fault alarm method of the unmanned aerial vehicle based on machine learning, a systematically accurate control surface fault diagnosis model can be effectively established, a flight parameter data set to be diagnosed can be input in real time through the diagnosis precision of a test set test model and gradual iteration of a model algorithm, the fault diagnosis accuracy is improved, the trained control surface fault diagnosis model has strong characteristic prediction capability, and fault alarm can be accurately and timely carried out.
The embodiment also provides a control surface fault alarm device of the unmanned aerial vehicle based on machine learning. Fig. 2 shows a schematic structural diagram of the control surface fault alarm device of the unmanned aerial vehicle based on machine learning provided by the embodiment. The following only briefly describes the structure of the control surface fault alarm device of the unmanned aerial vehicle based on machine learning, and please refer to the above detailed description of the control surface fault alarm method of the unmanned aerial vehicle based on machine learning for other relevant matters.
Referring to fig. 2, the control surface fault alarm device of the unmanned aerial vehicle based on machine learning includes:
the flight parameter data set comprises data influencing the oscillation of the control surface;
the control system comprises a preprocessing module, a control plane priori data set and a control plane oscillation data identification module, wherein the preprocessing module is used for preprocessing the flight parameter data set to obtain the control plane priori data set, the control plane priori data set comprises the flight parameter data set and a limit ring oscillation data identification obtained by carrying out control plane oscillation evaluation on the flight parameter data set, and the limit ring oscillation data identification comprises a numerical value used for identifying control plane oscillation and a numerical value used for identifying that the control plane does not oscillate;
the model construction module is used for dividing the control surface prior data set into a training set and a test set according to a preset proportion; learning the training set by adopting a machine learning method to obtain a control surface fault diagnosis model after training;
and the fault diagnosis alarm module is used for acquiring a flight parameter data set to be diagnosed, inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model for fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition.
As an optional implementation manner in this embodiment, the data affecting the oscillation of the control surface includes: time mark parameters, rudder position parameters and rudder deflection instruction parameters;
the preprocessing module preprocesses the flight parameter data set in the following mode to obtain a control surface prior data set: dividing the flight parameter data set into n groups of flight parameter data sets, wherein each group of flight parameter data sets comprises m data influencing the oscillation of the control surface, and m and n are positive integers; performing the following operations for each set of flight parameter data: sequentially acquiring m data influencing the oscillation of the control surface, calculating a deviation value of a rudder deflection instruction parameter and a rudder position parameter in the m data influencing the oscillation of the control surface, judging whether the absolute value of the deviation value is greater than a first preset threshold value or not and whether the absolute value of the rudder deflection instruction parameter is greater than a second preset threshold value or not, and if so, adding 1 to a counter; otherwise, the counter is increased by 0; judging whether the count value of the counter is larger than a third preset threshold value, if so, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on m pieces of data influencing control surface oscillation as a numerical value for identifying control surface oscillation; otherwise, the limit ring oscillation data obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation are recorded as a numerical value for identifying that the control surface does not oscillate.
As an optional implementation manner in this embodiment, the model building module learns the training set by using a machine learning method in the following manner to obtain a trained control surface fault diagnosis model: and learning the training set by adopting various machine learning methods to obtain a plurality of to-be-evaluated fault diagnosis models, and selecting an optimal to-be-evaluated fault diagnosis model as a trained control surface fault diagnosis model according to the accuracy of each to-be-evaluated fault diagnosis model and the ROC curve value of the operating characteristics of the testees.
Through the control surface fault alarm device of the unmanned aerial vehicle based on machine learning provided by the embodiment, a systematically accurate control surface fault diagnosis model can be effectively established, a flight parameter data set to be diagnosed can be input in real time through the diagnosis precision of a test set test model and a model algorithm is iterated step by step, the fault diagnosis accuracy is improved, and the trained control surface fault diagnosis model has strong characteristic prediction capability and can accurately and timely give a fault alarm.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: discrete logic circuits with logic gates for implementing logic functions on data signals, application specific integrated circuits with appropriate combinational logic gates, Programmable Gate Arrays (PGAs), Field Programmable Gate Arrays (FPGAs), Tx2, raspberry pies, etc.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made in the above embodiments by those of ordinary skill in the art without departing from the principle and spirit of the present invention. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (10)
1. The utility model provides a control plane fault alarm method of unmanned aerial vehicle based on machine learning which characterized in that includes:
acquiring a flight parameter data set, wherein the flight parameter data set comprises data influencing control surface oscillation;
preprocessing the flight parameter data set to obtain a control surface prior data set, wherein the control surface prior data set comprises: the flight parameter data set and limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the flight parameter data set, wherein the limit ring oscillation data identification comprises a numerical value for identifying control surface oscillation and a numerical value for identifying non-oscillation of the control surface;
dividing the control surface prior data set into a training set and a test set according to a preset proportion;
learning the training set by adopting a machine learning method to obtain a trained control surface fault diagnosis model;
acquiring a flight parameter data set to be diagnosed, inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model to perform fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition.
2. The method of claim 1,
the data affecting control surface oscillation include: time mark parameters, rudder position parameters and rudder deflection instruction parameters;
the preprocessing is performed on the flight parameter data set to obtain a control surface prior data set, and the method comprises the following steps:
dividing the flight parameter data set into n groups of flight parameter data sets, wherein each group of flight parameter data sets comprises m pieces of data influencing control surface oscillation, and m and n are positive integers;
performing the following operations for each set of flight parameter data: sequentially acquiring m pieces of data influencing the oscillation of the control surface, calculating a deviation value of the rudder deflection instruction parameter and the rudder position parameter in the m pieces of data influencing the oscillation of the control surface, judging whether the absolute value of the deviation value is greater than a first preset threshold value or not and whether the absolute value of the rudder deflection instruction parameter is greater than a second preset threshold value or not, and if so, adding 1 to a counter; otherwise, the counter is increased by 0; judging whether the count value of the counter is larger than a third preset threshold value, if so, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation as a numerical value for identifying the control surface oscillation; otherwise, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation as a numerical value for identifying that the control surface does not oscillate.
3. The method of claim 1,
the learning of the training set by adopting a machine learning method to obtain the trained control surface fault diagnosis model comprises the following steps:
and learning the training set by adopting various machine learning methods to obtain a plurality of to-be-evaluated fault diagnosis models, and selecting an optimal to-be-evaluated fault diagnosis model as the trained control surface fault diagnosis model according to the accuracy of each to-be-evaluated fault diagnosis model and the ROC curve value of the operating characteristics of the testees.
4. The method of claim 3,
the method of machine learning includes at least one of: linear algorithms, proximity algorithms, support vector machine algorithms, and neural network algorithms.
5. The method according to claim 1, before learning the training set by the machine learning method to obtain a trained control surface fault diagnosis model, further comprising:
and carrying out normalization processing on the control surface prior data set.
6. The method of claim 1, further comprising:
and inputting the test set into the trained control surface fault diagnosis model for testing, evaluating the performance of the trained control surface fault diagnosis model according to the accuracy, the ROC curve value, the precision ratio and the recall ratio of the test result, and adjusting the parameters of the trained control surface fault diagnosis model.
7. The method of claim 1,
inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model to perform fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition, wherein the method comprises the following steps:
and inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model in groups according to a preset mode to obtain a plurality of groups of analysis results, and performing control surface oscillation alarm when the oscillation times in the plurality of groups of analysis results exceed the preset times.
8. The utility model provides an unmanned aerial vehicle's control surface fault alarm device based on machine learning which characterized in that includes:
the flight parameter data set comprises data influencing the oscillation of the control surface;
the pre-processing module is used for pre-processing the flight parameter data set to obtain a control surface prior data set, the control surface prior data set comprises the flight parameter data set and a limit ring oscillation data identifier obtained by carrying out control surface oscillation evaluation on the flight parameter data set, and the limit ring oscillation data identifier comprises a numerical value used for identifying control surface oscillation and a numerical value used for identifying that the control surface does not oscillate;
the model construction module is used for dividing the control surface prior data set into a training set and a test set according to a preset proportion; learning the training set by adopting a machine learning method to obtain a trained control surface fault diagnosis model;
and the fault diagnosis alarm module is used for acquiring a flight parameter data set to be diagnosed, inputting the flight parameter data set to be diagnosed into the trained control surface fault diagnosis model for fault diagnosis to obtain an analysis result, and performing control surface oscillation alarm when the analysis result meets a preset condition.
9. The apparatus of claim 8,
the data affecting control surface oscillation include: time mark parameters, rudder position parameters and rudder deflection instruction parameters;
the preprocessing module preprocesses the flight parameter data set in the following mode to obtain a control surface prior data set: dividing the flight parameter data set into n groups of flight parameter data sets, wherein each group of flight parameter data sets comprises m pieces of data influencing control surface oscillation, and m and n are positive integers; performing the following operations for each set of flight parameter data: sequentially acquiring m pieces of data influencing the oscillation of the control surface, calculating a deviation value of the rudder deflection instruction parameter and the rudder position parameter in the m pieces of data influencing the oscillation of the control surface, judging whether the absolute value of the deviation value is greater than a first preset threshold value or not and whether the absolute value of the rudder deflection instruction parameter is greater than a second preset threshold value or not, and if so, adding 1 to a counter; otherwise, the counter is increased by 0; judging whether the count value of the counter is larger than a third preset threshold value, if so, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation as a numerical value for identifying the control surface oscillation; otherwise, recording limit ring oscillation data identification obtained by carrying out control surface oscillation evaluation on the m pieces of data influencing the control surface oscillation as a numerical value for identifying that the control surface does not oscillate.
10. The apparatus of claim 8,
the model construction module learns the training set by adopting a machine learning method in the following way to obtain a trained control surface fault diagnosis model: and learning the training set by adopting various machine learning methods to obtain a plurality of to-be-evaluated fault diagnosis models, and selecting an optimal to-be-evaluated fault diagnosis model as the trained control surface fault diagnosis model according to the accuracy of each to-be-evaluated fault diagnosis model and the ROC curve value of the operating characteristics of the testees.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210092099.7A CN114313307B (en) | 2022-01-26 | 2022-01-26 | Unmanned aerial vehicle control plane fault alarm method and device based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210092099.7A CN114313307B (en) | 2022-01-26 | 2022-01-26 | Unmanned aerial vehicle control plane fault alarm method and device based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114313307A true CN114313307A (en) | 2022-04-12 |
| CN114313307B CN114313307B (en) | 2022-07-01 |
Family
ID=81027796
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210092099.7A Active CN114313307B (en) | 2022-01-26 | 2022-01-26 | Unmanned aerial vehicle control plane fault alarm method and device based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114313307B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103968940A (en) * | 2014-05-27 | 2014-08-06 | 中国航天科技集团公司烽火机械厂 | Method and system for diagnosing oscillation fault of electric steering engine |
| CN106240839A (en) * | 2015-06-04 | 2016-12-21 | 波音公司 | For analyzing the system and method for flutter test data |
| CN107830996A (en) * | 2017-10-10 | 2018-03-23 | 南京航空航天大学 | A kind of vehicle rudder diagnosis method for system fault |
| US20190092495A1 (en) * | 2017-09-22 | 2019-03-28 | The Boeing Company | System and method for aircraft failure prediction |
| CN110618699A (en) * | 2019-10-24 | 2019-12-27 | 四川航天烽火伺服控制技术有限公司 | Unmanned aerial vehicle control surface control method, system and device computer readable storage medium |
| CN112884051A (en) * | 2021-02-26 | 2021-06-01 | 哈尔滨工业大学 | Data-driven multi-component online complex fault diagnosis method for lightweight unmanned aerial vehicle |
| CN113485304A (en) * | 2021-07-22 | 2021-10-08 | 北京航空航天大学 | Aircraft hierarchical fault-tolerant control method based on deep learning fault diagnosis |
| CN113514109A (en) * | 2021-08-18 | 2021-10-19 | 广东新创华科环保股份有限公司 | Unmanned aerial vehicle flight fault detection method and system |
-
2022
- 2022-01-26 CN CN202210092099.7A patent/CN114313307B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103968940A (en) * | 2014-05-27 | 2014-08-06 | 中国航天科技集团公司烽火机械厂 | Method and system for diagnosing oscillation fault of electric steering engine |
| CN106240839A (en) * | 2015-06-04 | 2016-12-21 | 波音公司 | For analyzing the system and method for flutter test data |
| US20190092495A1 (en) * | 2017-09-22 | 2019-03-28 | The Boeing Company | System and method for aircraft failure prediction |
| CN107830996A (en) * | 2017-10-10 | 2018-03-23 | 南京航空航天大学 | A kind of vehicle rudder diagnosis method for system fault |
| CN110618699A (en) * | 2019-10-24 | 2019-12-27 | 四川航天烽火伺服控制技术有限公司 | Unmanned aerial vehicle control surface control method, system and device computer readable storage medium |
| CN112884051A (en) * | 2021-02-26 | 2021-06-01 | 哈尔滨工业大学 | Data-driven multi-component online complex fault diagnosis method for lightweight unmanned aerial vehicle |
| CN113485304A (en) * | 2021-07-22 | 2021-10-08 | 北京航空航天大学 | Aircraft hierarchical fault-tolerant control method based on deep learning fault diagnosis |
| CN113514109A (en) * | 2021-08-18 | 2021-10-19 | 广东新创华科环保股份有限公司 | Unmanned aerial vehicle flight fault detection method and system |
Non-Patent Citations (2)
| Title |
|---|
| 杨建忠: "作动器故障对阵风减缓系统的影响研究", 《机械设计与制造》 * |
| 杨建忠: "作动器故障对阵风减缓系统的影响研究", 《机械设计与制造》, 30 November 2020 (2020-11-30), pages 94 - 97 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114313307B (en) | 2022-07-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107329088B (en) | Battery state of health diagnostic device and method | |
| CN114585983B (en) | Method, device and system for detecting abnormal operation state of equipment | |
| CN109000940B (en) | Abnormal axle temperature diagnosis method and system for rolling stock | |
| CN106528975B (en) | A kind of prognostic and health management method applied to Circuits and Systems | |
| EP1960853B1 (en) | Evaluating anomaly for one-class classifiers in machine condition monitoring | |
| CN109115501A (en) | A kind of Civil Aviation Engine Gas path fault diagnosis method based on CNN and SVM | |
| CN107490964B (en) | Rotating machinery fault feature reduction method based on feature evidence discretization | |
| US8560279B2 (en) | Method of determining the influence of a variable in a phenomenon | |
| CN111665066B (en) | Equipment fault self-adaptive upper and lower early warning boundary generation method based on convolutional neural network | |
| Al-Haddad et al. | Improved UAV blade unbalance prediction based on machine learning and ReliefF supreme feature ranking method | |
| US10565343B2 (en) | Circuit configuration optimization apparatus and machine learning device | |
| EP3669244A1 (en) | Turbine diagnostic feature selection system | |
| CN112712305A (en) | Aircraft system and health assessment method and device thereof | |
| CN114313307B (en) | Unmanned aerial vehicle control plane fault alarm method and device based on machine learning | |
| CN110457731A (en) | A method of it is analyzed applied to Helicopter Dynamic Components vibration trend | |
| CN117828510A (en) | Elevator door abnormality detection method and device, electronic equipment and storage medium | |
| TWI400619B (en) | Product quality fault detection method and real metrolgy data evaluation method | |
| CN115272166A (en) | Model training method, defect target detection method, device and storage medium | |
| CN113505039A (en) | Communication fault analysis method, device and system | |
| CN116736173B (en) | Energy storage battery model construction and energy storage battery state judgment method and device | |
| CN115345210B (en) | Bearing signal fault diagnosis method, device, equipment and readable storage medium | |
| KR102697177B1 (en) | A Drone Flight Control and Diagnosing System and Method therefor | |
| Chinta | Review on Various Supervised Machine Learning Algorithms used in Data Science | |
| CN116343131B (en) | Intelligent monitoring and early warning method, device, equipment and storage medium for camera set | |
| US20230375636A1 (en) | Lebesgue sampling-based deep belief network for lithium-ion battery diagnosis and prognosis |
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 |