CN110390177B - Method and device for determining outlier flying object - Google Patents
Method and device for determining outlier flying object Download PDFInfo
- Publication number
- CN110390177B CN110390177B CN201910700625.1A CN201910700625A CN110390177B CN 110390177 B CN110390177 B CN 110390177B CN 201910700625 A CN201910700625 A CN 201910700625A CN 110390177 B CN110390177 B CN 110390177B
- Authority
- CN
- China
- Prior art keywords
- flying object
- outlier
- detection point
- ground
- parameters
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
The disclosure relates to an outlier flying object determination method and device, wherein the method comprises the following steps: acquiring operation parameters of at least one flying object when the flying object reaches a preset detection point in a preset take-off and landing stage; and determining the outlier flying object in the at least one flying object by using the operation parameter and an outlier flying object determination model corresponding to the preset take-off and landing stage. Through the method, the outlier flying object in at least one flying object can be rapidly and accurately determined, and the outlier flying object is comprehensively judged based on the QAR data and the parameters in the whole industry, so that dependence on subjective experience is avoided.
Description
Technical Field
The disclosure relates to the technical field of data processing, and in particular relates to an outlier flyer determining method and device.
Background
According to the statistics report in the boeing 2017, the take-off stage only accounts for 2% of the whole flight stage, the landing stage only accounts for 4% of the whole flight stage, but during the period from 2008 to 2017, the proportion of the fatal accidents in the take-off stage accounts for 14% of all the fatal accidents, the proportion of the fatal accidents in the landing stage accounts for 49% of all the fatal accidents, and the take-off and landing are the most risky stages in the whole flight process. With the development of technology in recent years, the aviation accident rate is reduced year by year, by the 2 months of 2019, the transportation aviation in China has realized continuous safe flight for 102 months, and the record is also growing continuously. Although there is no fatal accident, unsafe accidents possibly causing the fatal accident occur from time to time, and the tail wiping event 24 is recorded to occur in the take-off stage of civil aviation in China between 1 month in 1990 and 4 months in 2008. Safety is always the theme of civil aviation, how to further improve the safety level, especially the incident of taking off and landing is prevented in advance and is an important problem that needs to be constantly considered.
In order to improve the aviation safety level, the civil aviation bureau of China specifies that transport planes registered and operated in China should be equipped with a quick access recorder (quick access recorder, QAR) or equivalent equipment since 1 month 1 1998 and that QAR data should be applied directly to the civil aviation flight quality monitoring (Flight Operations Quality Assurance, FOQA) from 2000 to improve the flight safety management level. In 2017, the national civil aviation quality monitoring base station is formally put into operation, and receives QAR data of more than 3100 airplanes and 16000 flights of each national airline company every day, so as to monitor flight conditions and discover operation risks in time. The QAR data encompasses very rich aircraft flight data including time, speed, altitude, attitude, position, aircraft engines, APU auxiliary power units, flight control systems, fuel systems, air traffic control information, landing gear, inertial navigation systems, gearboxes, and the like. The QAR data records various parameters of the aircraft in the flight process, can continuously and completely reflect the actual state and various symptoms of the aircraft system in the operation, and is an important basis for application of data science in the civil aviation safety and operation fields.
The appearance of QAR data provides a new application direction for aviation security, but current application is mainly focused on warning of overrun events, for example, related technology generally presets risk thresholds of certain single variables according to aircraft suppliers or experiences, then compares each variable with the threshold one by one, and judges that flight of the flight has security risk after exceeding the preset threshold. The related technology is too dependent on subjective experience, so that the artificial influence is large, a given threshold value is usually loose, and the safety early warning capability is reduced in a certain sense; secondly, the method mainly comprises the steps of comparing single variables, and lacks consideration of relevance among the variables; in addition, most of the existing applications are based on data of a single airline company or a single airport, do not cover the application level of the whole industry, and lack the capability of coping with the whole deviation.
Disclosure of Invention
In view of this, the present disclosure proposes a method of determining an outlier flying object, the method comprising: acquiring operation parameters of at least one flying object when the flying object reaches a preset detection point in a preset take-off and landing stage; and determining the outlier flying object in the at least one flying object by using the operation parameter and an outlier flying object determination model corresponding to the preset take-off and landing stage.
In a possible implementation manner, the determining the outlier flying object in the at least one flying object by using the operation parameter and the outlier flying object determination model corresponding to the predetermined take-off and landing stage includes: determining estimated outlier flying objects under N preset detection points in the preset take-off and landing stage; and determining the flying object with the number of times of being determined to be not less than M as the outlier flying object, wherein M, N is a natural number and M is less than or equal to N.
In one possible embodiment, the method further comprises: acquiring historical operation parameters of a plurality of flying objects reaching a preset detection point in a preset take-off and landing stage; carrying out data analysis on the historical operation parameters to obtain data characteristics, wherein the data characteristics comprise density functions of all the historical operation parameters, states among all the historical operation parameters and correlation; and establishing an outlier flying object determination model according to the data characteristics.
In a possible implementation manner, the data analysis on the historical operation parameters to obtain data features includes: establishing a scatter diagram matrix between every two parameters by utilizing the historical operation parameters; and determining the data characteristics according to the scatter diagram matrix.
In one possible embodiment, building an outlier aircraft determination model from the data features includes: determining that the outlier flyer determination model is a multi-class cluster model under the condition that the existing historical operation parameters are in discrete states or the density function of the existing historical operation parameters is in multi-peak characteristics; or under the condition that all the historical operation parameters are in a converging state and the density functions of all the historical operation parameters are in a unimodal characteristic, determining that the outlier flyer determination model is a single-class clustering model.
In one possible embodiment, in the case where there is a correlation between the plurality of historical operating parameters, one of the plurality of historical operating parameters having the correlation and the other historical operating parameters having no correlation are utilized as the training data.
In a possible implementation manner, the building an outlier flyer determination model according to the data features further includes: training the outlier flyer determination model by utilizing the training data on the premise of avoiding over-fitting and under-fitting; and adjusting model parameters of the outlier flying object determining model, determining final model parameters of the outlier flying object determining model when the number of the outlier flying objects determined in training accounts for the number of all flying objects to reach a first proportion, and obtaining the trained outlier flying object determining model according to the final model parameters.
In one possible embodiment, the first proportion is between 5% and 10%.
In one possible implementation manner, in the case that the predetermined take-off and landing stage is a take-off stage, the predetermined take-off and landing stage includes a first detection point, where the first detection point is a position of a flying object at a take-off moment, the operation parameters include any one or more of a take-off distance, a ground elevation angle, a mean value of a change rate of the ground elevation angle, a standard deviation of the change rate of the ground elevation angle, a vertical speed of the ground elevation, a mean value of a change rate of the ground elevation angle, and a standard deviation of the change rate of the ground elevation angle, where the take-off distance represents a horizontal distance when the flying object reaches the predetermined safe height from a start to a point where the flying object is in a moment, the ground elevation angle represents an elevation angle of the flying object at the moment of the flying object, the ground elevation angle of the flying object is in the moment, the ground elevation angle represents a mean value of each second of change in the flying object in the moment of the flying object, the ground elevation angle of the ground elevation angle represents a mean value of each second of the flying object in front and back n seconds, the ground elevation angle of the change rate of the flying object is in the moment, the ground elevation angle of the flying object is represented by n seconds, the standard deviation of the ground elevation angle of the flying object from the ground elevation angle of the flying object, and the ground elevation angle of the flying object is represented by n seconds of the ground elevation angle of the flying object, and the ground elevation angle of the flying object from the ground elevation angle of the ground moment is represented by the ground n seconds, and the ground elevation angle of the ground plane from the ground plane.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is the approach stage, the preset detection points include a second detection point, a third detection point and a fourth detection point, where the second detection point is a position where the flying object reaches a stable approach detection height under the Metro weather condition IMC, the third detection point is a position where the flying object reaches a stable approach detection height under the visual weather condition VMC, and the fourth detection point is a position where the flying object is at a stable approach detection height at five sides of the airport.
In one possible embodiment, the second detection point has a height of 1000 feet, the third detection point has a height of 500 feet, and the fourth detection point has a height of 300 feet.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is the approach stage, the operation parameters include any one or more of a pitch angle mean, a heading mean, a relative speed mean, a vertical speed mean, a standard deviation of low-pressure rotors of each engine, and a standard deviation of high-pressure rotors of each engine, where the pitch angle mean represents a mean value of a pitch angle of each second m seconds before and after the altitude of the flying object reaches the preset detection point, the heading mean represents a mean value of a heading change of each second m seconds before and after the altitude of the flying object reaches the preset detection point,
The relative speed mean value represents the average value of the airspeed per second of m seconds before and after the flying object height reaches the preset detection point minus the reference speed, the vertical speed mean value represents the average value of the inertial vertical speed per second of m seconds before and after the flying object height reaches the preset detection point, the standard deviation of the low-pressure rotor of the engine represents the standard deviation of the speed change of the low-pressure rotor of the engine per second of m seconds before and after the flying object height reaches the preset detection point, and the standard deviation of the high-pressure rotor of the engine represents the standard deviation of the speed change of the low-pressure rotor of the engine per second of m seconds before and after the flying object height reaches the preset detection point, wherein m is more than 0.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is a ground-level floating stage, the preset detection points include a fifth detection point, a sixth detection point and a seventh detection point, where the fifth detection point is a position where the flying object is modified from a pitch rule to a roll-up rule, the sixth detection point is a position where the flying object reaches a roll-up height specified under a stable condition, and the seventh detection point is a position where the flying object receives an automatic call reminding pilot to withdraw the thrust handle.
In one possible embodiment, the fifth detection point has a height of 50 feet, the sixth detection point has a height of 30 feet, and the seventh detection point has a height of 20 feet.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is a ground-level drift stage, the running parameters include any one or more of an inertial vertical velocity mean, a pitch angle mean, an airspeed mean, a vertical overload mean, a ground distance and a ground time, where the inertial vertical velocity mean represents an average value of the inertial vertical velocity every second k seconds before and after the flying object height reaches a preset detection point, the pitch angle mean represents an average value of the pitch angle every second k seconds before and after the flying object height reaches the preset detection point, the airspeed mean represents an average value of airspeed (IAS) every second k seconds before and after the flying object height reaches the preset detection point, the vertical overload mean represents an average value of vertical overload VRTG every second k seconds before and after the flying object height reaches the preset detection point, the ground distance is a horizontal distance from the preset ground point to the ground point, and the ground time represents a time from the preset ground point to the ground point, where k > 0.
According to another aspect of the present disclosure, there is provided an outlier flying object determining apparatus, the apparatus comprising: the first acquisition module is used for acquiring operation parameters when at least one flying object reaches a preset detection point in a preset take-off and landing stage; and the determining module is connected with the acquiring module and is used for determining the outlier flying object in the at least one flying object by using the operation parameters and the outlier flying object determining model corresponding to the preset take-off and landing stage.
According to another aspect of the present disclosure, there is provided an outlier flying object determining apparatus, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to perform the above method.
According to another aspect of the present disclosure, there is provided a non-transitory computer readable storage medium having stored thereon computer program instructions, wherein the computer program instructions, when executed by a processor, implement the above-described method.
Through the method and the device, the operation parameters of at least one flying object when reaching the preset detection point in the preset take-off and landing stage can be obtained, and the outlier flying object in the at least one flying object is determined by utilizing the operation parameters and the outlier flying object determination model corresponding to the preset take-off and landing stage. The method and the device can rapidly and accurately determine the outlier flying object in at least one flying object, and comprehensively judge the outlier flying object based on the QAR data and a plurality of parameters in the whole industry, so that the dependence on subjective experience is avoided.
Other features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments, which proceeds with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments, features and aspects of the present disclosure and together with the description, serve to explain the principles of the disclosure.
FIG. 1 illustrates a flow chart of a method of outlier flying object determination according to an embodiment of the present disclosure.
Fig. 2 shows a flow chart of an outlier flying object determination method according to an embodiment of the present disclosure.
Fig. 3a shows a schematic representation of historical operating parameters of the same model of aircraft acquired during the take-off phase.
Fig. 3b shows a schematic diagram of a scatter plot matrix created during the takeoff phase using historical operating parameters.
Fig. 3c shows a schematic representation of the adjustment of parameters of an outlier determination model of the takeoff phase.
Fig. 3d shows a schematic diagram of an outlier determination using an outlier determination model of a flight phase.
Fig. 4a shows a schematic diagram of historical operating parameters of the same model of aircraft acquired at the approach stage.
FIG. 4b shows a schematic of a scatter plot matrix created at the near stage using historical operating parameters.
Fig. 4c shows a schematic of the adjustment of parameters of the outlier determination model at the approach stage.
Fig. 4d shows a schematic representation of an outlier determination at a second detection point using an outlier determination model at an approach stage.
Fig. 4e shows a schematic diagram of an outlier determination at a third detection point using an outlier determination model at the approach stage.
Fig. 4f shows a schematic representation of an outlier determination at a fourth detection point using an outlier determination model at the approach stage.
Fig. 4g shows a schematic diagram of determining outlier flights in combination with a plurality of predetermined detection points at the approach stage.
Fig. 5a shows a schematic diagram of historical operating parameters of the same model of aircraft acquired during the ground level drift phase.
Fig. 5b shows a schematic diagram of a scatter plot matrix created during the horizon drift phase using historical operating parameters.
Fig. 5c shows a schematic of the adjustment of parameters of an outlier fly determination model at the ground level drift stage.
Fig. 5d shows a schematic diagram of an outlier determination at a fifth detection point using an outlier determination model at a ground level drift stage.
Fig. 5e shows a schematic diagram of an outlier determination at a sixth detection point using an outlier determination model at a ground level drift stage.
Fig. 5f shows a schematic diagram of an outlier determination at a seventh detection point using an outlier determination model at a ground level drift stage.
Fig. 5g shows a schematic diagram of an outlier determination at an eighth detection point using an outlier determination model at a ground level drift stage.
Fig. 5h shows a schematic diagram of determining an outlier flight in combination with a plurality of predetermined detection points of the ground level drift phase.
Fig. 6 shows a block diagram of an outlier flying object determination device according to an embodiment of the disclosure.
Fig. 7 shows a block diagram of an outlier flying object determination device according to an embodiment of the disclosure.
Detailed Description
Various exemplary embodiments, features and aspects of the disclosure will be described in detail below with reference to the drawings. In the drawings, like reference numbers indicate identical or functionally similar elements. Although various aspects of the embodiments are illustrated in the accompanying drawings, the drawings are not necessarily drawn to scale unless specifically indicated.
The word "exemplary" is used herein to mean "serving as an example, embodiment, or illustration. Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments. In addition, numerous specific details are set forth in the following detailed description in order to provide a better understanding of the present disclosure. It will be understood by those skilled in the art that the present disclosure may be practiced without some of these specific details. In some instances, methods, means, elements, and circuits well known to those skilled in the art have not been described in detail in order not to obscure the present disclosure.
Referring to fig. 1, fig. 1 shows a flow chart of an outlier flyer determination method according to an embodiment of the disclosure.
The method can be applied to the terminal and/or the server.
As shown in fig. 1, the method includes:
step S110, obtaining operation parameters when at least one flying object reaches a preset detection point in a preset take-off and landing stage;
and step S120, determining the outlier flying object in the at least one flying object by using the operation parameters and an outlier flying object determination model corresponding to the preset take-off and landing stage.
By the method, the operation parameters of at least one flying object when reaching the preset detection point in the preset take-off and landing stage can be obtained, and the outlier flying object in the at least one flying object is determined by utilizing the operation parameters and the outlier flying object determination model corresponding to the preset take-off and landing stage. The method and the device can rapidly and accurately determine the outlier flying object in at least one flying object, and comprehensively judge the outlier flying object based on the QAR data and a plurality of parameters in the whole industry, so that the dependence on subjective experience is avoided.
Under the condition that the outlier flying object in at least one flying object is determined by using the outlier flying object determining method disclosed by the disclosure, the deviation between the taking-off and landing performance of the flying object near a preset detection point and the taking-off and landing performance of a group near the preset detection point is large, and adjustment is needed. After the results are obtained, the results may be sent to a control center for reference, or a warning may be issued to alert personnel to the precautionary risk.
In one possible implementation, the present disclosure may obtain, in real time, operational parameters of one or more flyers to determine whether the one or more flyers are outlier flyers. One or more operational parameters of the flyer may also be obtained over a period of time to make the determination. In short, the outlier flying object determining method provided by the disclosure can be applied to a strong real-time scene and a weak real-time scene.
In one possible implementation, the predetermined take-off and landing phases may include a take-off phase, a approach phase, and a ground level drift phase.
In one possible embodiment, the preset detection point may be determined with reference to a technical description of take-off and landing in an airline flight crew technical manual.
For different preset take-off and landing stages, different preset detection points can be determined, the number of the preset detection points can be one or a plurality of the preset detection points, the specific number of the preset detection points can be determined according to analysis requirements, the specific preset detection points are not limited, and the number of the preset detection points is not limited.
In one possible embodiment, the measure with height as the preset detection point may be selected.
For example, the approach stage can select the air pressure height as the height measurement, avoiding the use of radar height, as radar height fluctuates greatly due to the influence of terrain and buildings; the radar height can be used as the height measurement in the ground level drifting stage, and the radar height after approach is influenced by the obstacle and is stable in fluctuation and can be used as a height reference.
In one example, the pre-set detection point for the takeoff phase may be selected to be at ground level.
In one example, the preset detection point of the approach phase may include a stable approach detection altitude under meter weather conditions (IMC), a stable approach detection altitude under visual weather conditions (VMC), a stable approach detection altitude when the aircraft is on the five sides of the airport, and the like.
In one example, the preset detection points of the ground level drift phase may include the height at which the pitch law is modified to the roll-up law, the roll-up height specified under stable conditions, the height at which the pilot is automatically alerted to retract the thrust handles, and so forth.
In one possible embodiment, the operating parameters of the different take-off and landing phases can be determined as a function of the actual situation. In one example, the operating parameters can be categorized into direct measurement parameters, such as speed, pitch angle, rate of rise, etc., provided by the onboard QAR data, and calculation parameters; the calculated parameter may be a calculated value calculated based on a directly measured parameter, such as an average value, a standard deviation, etc.
The above has been a general introduction to preset detection points, operation parameters, etc., and will be described in detail later.
It should be noted that, those skilled in the art may select the preset detection point, the operation parameter, etc. according to actual needs.
In one possible embodiment, the flyer may be various types of aircraft. Such as flights currently in operation, including various types of aircraft. Of course, in other embodiments, other unmanned aerial vehicles are possible, for example, commercial or non-commercial use.
In a possible implementation manner, the determining, in step S120, the outlier flying object in the at least one flying object by using the operation parameter and the outlier flying object determining model corresponding to the predetermined take-off and landing stage may include: determining estimated outlier flying objects under N preset detection points in the preset take-off and landing stage; and determining the flying object with the number of times of being determined to be not less than M as the outlier flying object, wherein M, N is a natural number and M is less than or equal to N.
The present disclosure may set a minimum detection number (e.g., M) of continuous outlier flyers according to actual usage requirements, and if the number of outliers detected by some flyers is not less than the minimum detection number in all preset detection points in a predetermined take-off and landing stage, the flyers are marked as continuous outlier flyers, which indicates that the flyers do not recover the take-off and landing performance adjustment of the flyers through flight control, and the flyers determined as outlier flyers have a higher risk of occurrence of problems than the normal flyer population, and need to pay attention to.
Referring to fig. 2, fig. 2 shows a flow chart of an outlier flyer determination method according to an embodiment of the disclosure.
In one possible embodiment, the method may further comprise:
step S130, acquiring historical operation parameters of a plurality of flying objects reaching a preset detection point in a preset take-off and landing stage;
step S140, carrying out data analysis on the historical operation parameters to obtain data characteristics, wherein the data characteristics comprise density functions of all the historical operation parameters, states and correlations among all the historical operation parameters;
and step S150, establishing an outlier flying object determination model according to the data characteristics.
By the method, the historical operation parameters of the plurality of flying objects reaching the preset detection points in the preset take-off and landing stage can be obtained, the historical operation parameters are subjected to data analysis to obtain data characteristics, the data characteristics comprise density functions of the historical operation parameters, states and correlations among the historical operation parameters, and an outlier flying object determination model is built according to the data characteristics. According to the outlier flying object determining model determined by the method, outlier flying objects in the detected flying objects can be obtained rapidly and accurately.
In one possible implementation, the historical operating parameter may be historical QAR data for a month, week, or day, or any other period of time.
In a possible implementation manner, the step S140 of performing data analysis on the historical operating parameters to obtain data features may include: establishing a scatter diagram matrix between every two parameters by utilizing the historical operation parameters; and determining the data characteristics according to the scatter diagram matrix.
In one possible implementation, step S150 of building an outlier aircraft determination model according to the data features may include:
determining that the outlier flyer determination model is a multi-class cluster model under the condition that the existing historical operation parameters are in discrete states or the density function of the existing historical operation parameters is in multi-peak characteristics; or under the condition that all the historical operation parameters are in a converging state and the density functions of all the historical operation parameters are in a unimodal characteristic, determining that the outlier flyer determination model is a single-class clustering model.
In One possible implementation, the single class cluster model may include, for example, one-class SVM, robust covariance, and the like. In one possible implementation, the multi-class cluster model may include, for example, DBSCAN, GMM, and the like. Of course, the single-class cluster model and the multi-class cluster model are not limited thereto, and the above is merely exemplary description thereof and examples, and those skilled in the art may select other single-class cluster models and multi-class cluster models, which are not limited thereto.
Through the method, under the condition that the data characteristics of a plurality of historical operating parameters meet different conditions, different types of clustering models can be determined to serve as the outlier flight non-determination model, so that the adaptability of different environments is improved, and the outlier flight object determination method can be applied to various different conditions.
In one possible implementation, where there is a correlation between a plurality of historical operating parameters, one of the plurality of historical operating parameters having the correlation and the other historical operating parameters having no correlation may be utilized as training data. In one possible embodiment, the correlation may refer to a pearson correlation coefficient between parameters of not less than 0.8.
By determining the correlation between parameters and using one of a plurality of historical operating parameters with correlation and other historical operating parameters without correlation as training data, the number of model dimensions can be effectively reduced, so that operation resources are saved, and operation speed is improved.
In one possible implementation, step S150 may further include establishing an outlier flyer determination model according to the data characteristic:
Training the outlier flyer determination model by utilizing the training data on the premise of avoiding over-fitting and under-fitting; and adjusting model parameters of the outlier flying object determining model, determining final model parameters of the outlier flying object determining model when the number of the outlier flying objects determined in training accounts for the number of all flying objects to reach a first proportion, and obtaining the trained outlier flying object determining model according to the final model parameters.
By training the outlier determination model by using the training data on the premise of avoiding over-fitting and under-fitting, the accurate outlier determination model can be obtained, and therefore, an accurate result can be obtained when the outlier determination is performed.
In one possible embodiment, the first proportion is between 5% and 10%. Through setting the first proportion between 5% and 10%, the requirement of engineering application can be met, and the judgment of the outlier flying object is accurate.
The different phases of the predetermined take-off and landing phase will be described separately.
During the takeoff phase:
in a possible implementation manner, in the case that the predetermined take-off and landing stage is a take-off stage, the preset detection point may include a first detection point, where the first detection point is the position of the take-off moment of the flying object. During the takeoff phase of the flyer, the flyer is in contact with the ground and has a ground clearance of 0 feet.
In one possible embodiment, the operating parameters may include any one or more of take-off distance (dist_take off), ground speed (ias_liftoff), ground elevation angle (pitch_liftoff), ground elevation angle change RATE average (pitch_rate_ave), ground elevation angle change RATE standard deviation (pitch_rate_std), ground vertical speed (IVV _liftoff), ground speed change RATE average (IVV _rate_ave), ground speed change RATE standard deviation (IVV _rate_std), wherein,
the take-off distance represents the horizontal distance from the beginning of flying to the moment when the flying object reaches a preset safety height, the ground-leaving speed represents the horizontal speed of the flying object at the ground-leaving moment of the flying object main wheel, the ground-leaving elevation angle represents the elevation angle of the flying object at the ground-leaving moment of the flying object main wheel, the mean value of the ground-leaving elevation angle change rate represents the mean value of every second elevation angle change in n seconds before and after the ground-leaving of the flying object main wheel, the standard deviation of the ground-leaving elevation angle change rate represents the standard deviation of every second elevation angle change in n seconds before and after the ground-leaving of the flying object main wheel, the ground-leaving vertical speed represents the vertical speed of the flying object at the ground-leaving moment of the flying object main wheel, the mean value of the ground-leaving speed change rate represents the standard deviation of every second vertical speed change in n seconds before and after the ground-leaving of the flying object main wheel, and n is more than 0.
In one possible embodiment, the specific size of n may be determined according to practical situations, and in one example, n may be 5s.
The above gives examples of operating parameters during the takeoff phase, it should be understood that those skilled in the art may increase and decrease the operating parameters according to actual conditions, and the present disclosure is not limited thereto.
Model training during the take-off phase and determination of outlier flights are described in exemplary fashion below.
Referring to fig. 3a, fig. 3a shows a schematic diagram of historical operating parameters of the same model of aircraft acquired during the takeoff phase.
According to step S130, the present disclosure may obtain historical operating parameters of the same model of aircraft acquired during the takeoff phase as shown in fig. 3 a.
It should be appreciated that the same type of flyer is described herein as an example, however, the present disclosure is not so limited and applies equally to the determination of outlier flyers for different types of flyers.
As shown in fig. 3a, in one example, the present disclosure selects historical operating parameters for 28 flights (numbered No-X0001-X0028), which include: take-off distance (dist_take off), ground speed (ias_liftoff), ground elevation angle (pitch_liftoff), ground elevation angle change RATE mean (pitch_rate_ave), ground elevation angle change RATE standard deviation (pitch_rate_std), ground vertical speed (IVV _liftoff), ground speed change RATE mean (IVV _rate_ave), ground speed change RATE standard deviation (IVV _rate_std).
Here and hereinafter, the present disclosure does not describe historical operating parameters, units of operating parameters, it being understood that each unit of parameter may be determined in accordance with actual circumstances.
After obtaining the historical operating parameters, according to step S140, the present disclosure may perform data analysis on the historical operating parameters to obtain data features, where the data features include density functions of each historical operating parameter, states and correlations between each historical operating parameter.
For example, a scatter plot matrix between every two parameters may be established using the historical operating parameters; and determining the data characteristics according to the scatter diagram matrix.
Referring to FIG. 3b, FIG. 3b shows a schematic diagram of a scatter plot matrix created during the takeoff phase using historical operating parameters. As shown in fig. 3b, on the diagonal of the scatter plot matrix, the density function of each parameter has a unimodal characteristic (i.e., only one peak), approximating a normal distribution; in addition, obvious data convergence characteristics exist between parameters, and no discrete parameters exist.
According to step S150, an outlier determination model may be built from the data features.
In One example, if there are no discrete parameters in the historical operation parameters, the outlier flying object detection problem can be regarded as a single type of clustering problem, namely, a normal flying object and an outlier flying object, and a corresponding single type of clustering algorithm can be selected for outlier flying object detection, for example, one-class SVM can be selected.
Referring to fig. 3c, fig. 3c shows a schematic diagram of the adjustment of parameters of the model for determining the outlier flying object during the take-off phase.
Here, for convenience of description, two parameters of the TAKEOFF distance dist_take off and the ground leaving speed ias_liftoff will be selected for analysis, and it should be noted that the model and the training method are equally applicable when multiple parameters (including parameters other than the TAKEOFF distance dist_take off and the ground leaving speed ias_liftoff) are included.
Based on the data characteristics of each historical operating parameter in the take-off stage, the present disclosure illustratively selects One-Class SVM algorithm for cluster learning, wherein the kernel function illustratively selects Radial Basis Function (RBF), and performs parameter adjustment on two parameters nu (for defining the training error upper boundary of the SVM algorithm) and gamma (for adjusting the scale parameters of the radial basis function) of the algorithm. On the premise of avoiding over-fitting and under-fitting, the parameters nu and gamma are adjusted until the quantity of the outlier flying objects is adjusted to be optimal in the range of 5-10%.
As shown in fig. 3c, the black border (ellipse, etc.) is a classification boundary obtained by algorithm learning, and outside the classification boundary, the outlier flyers marked in the training process are normal flyers marked in the training process. It can be seen that the larger the parameter nu, the more like a central collection the classification boundary; the parameter gamma can be used to adjust the shape of the classification boundary, the larger the parameter, the more curved the classification boundary, but the larger the parameter, the more likely the overfitting occurs (as shown in the lower right corner of fig. 3 c). Both the over-fit and under-fit affect the results of anomaly detection, so this disclosure is avoided in parameter adjustment. As can be seen from fig. 3c, the outlier determination model of the take-off phase performs better when the parameter nu is 0.1 and the parameter gamma is 5.
After training the outlier flying object determining model, the outlier flying object determining model can be utilized to judge the flying object needing outlier judgment.
Referring to fig. 3d, fig. 3d shows a schematic diagram of an outlier determination using an outlier determination model of a flight phase.
As shown in fig. 3d, the parameter nu of the outlier determination model is set to 0.1, the parameter gamma is set to 5, the operation parameter (one or more of the TAKEOFF distance (dist_take off), the ground clearance velocity (ias_liftoff), the ground clearance elevation angle (pitch_liftoff), the ground clearance elevation angle change RATE mean (pitch_rate_ave), the ground clearance elevation angle change RATE standard deviation (pitch_rate_std), the ground clearance vertical velocity (IVV _liftoff), the ground clearance velocity change RATE mean (IVV _rate_ave), and the ground clearance velocity change RATE standard deviation (IVV _rate_std) of at least one of the flying objects (for example, 480) are input into the outlier determination model, and it can be determined that the outlier flying object is about 43 (dark color point in fig. 3 d) accounting for about 9.03% of all training data.
In the above example, for the take-off phase, the preset detection points selected by the present disclosure are only 1, so for the above example, the determined 43 outlier flyers are the final result.
It should be noted that, in this and the following examples, other parameters (except dist_take off, ias_liftoff) cannot be represented in the diagram due to the visualization limitation (visualization can only be in 3 dimensions), but whether the flyer is an outlier flyer or not may be represented in the diagram. For example, 8 parameters are input into the model, each flyer will be marked as normal or outlier, and normal flights (light) are represented in the three-dimensional map; an outlier flight (dark color).
For the approach phase:
according to the technical manual of the airline crewmember, it is considered to be stable in the near phase when all the following criteria are met:
1) The aircraft is in the correct flight path; 2) Only slight changes in pitch and heading are required to maintain the correct flight path; 3) The aircraft is at approach speed; 4) The aircraft is in a correct landing configuration; 5) The sinking rate is not more than a certain set value; 6) The thrust setting is adapted to the shape of the aircraft; 7) All profiles and checklists have been executed.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is the approach stage, the preset detection points may include a second detection point, a third detection point and a fourth detection point, where the second detection point may be a position where the flying object reaches a stable approach detection height under the meter meteorological condition IMC, the third detection point may be a position where the flying object reaches a stable approach detection height under the visual meteorological condition VMC, and the fourth detection point may be a position where the flying object is at a stable approach detection height on five sides of the airport.
In one possible embodiment, the second detection point may have a height of 1000 feet, the third detection point may have a height of 500 feet, and the fourth detection point may have a height of 300 feet.
In one possible implementation manner, in the case that the predetermined take-off and landing stage is the approach stage, the operation parameters may include any one or more of a PITCH angle average (pitch_ave), a heading average (head_diff), a relative speed average (iassinusvref), a vertical speed average (IVV _ave), a standard deviation of low-pressure rotors of each engine, and a standard deviation of high-pressure rotors of each engine, where the PITCH angle average represents an average value of PITCH angle of each second m seconds before and after the altitude reaches the preset detection point, the heading average represents an average value of changes of heading of each second m seconds before and after the altitude reaches the preset detection point, the relative speed average represents an average value of airspeed of each second after the altitude reaches the preset detection point, the vertical speed average value represents an average value of inertial vertical speeds of each second m seconds before and after the altitude reaches the preset detection point, the standard deviation of low-pressure rotors of each second represents a low-pressure rotor speed change of each engine before and after the altitude reaches the preset detection point, and the standard deviation of low-pressure rotors of each second represents a standard deviation of m seconds before and after the altitude reaches the preset detection point, and the standard deviation of low-pressure rotors of each engine is less than 0 seconds.
In one possible embodiment, the specific size of m may be determined according to practical situations, and m may be 5s in one example.
The above gives examples of operating parameters at the approach stage, and it should be understood that those skilled in the art may increase and decrease the operating parameters according to actual situations, and the disclosure is not limited thereto.
Model training at the approach stage and determination of outlier flights are described in exemplary fashion below.
Referring to fig. 4a, fig. 4a shows a schematic diagram of historical operation parameters of the same model of flying object acquired in the approach phase.
According to step S130, the present disclosure may obtain historical operating parameters of the same model of aircraft acquired at the approach stage as shown in fig. 4 a.
As shown in FIG. 4a, in one example, the present disclosure selects historical operating parameters for 27 flights (numbered No-X0001-X0027).
In one possible embodiment, where the predetermined take-off and landing phase is a near phase, the operating parameters may include PITCH angle average (pitch_ave), heading average (head_diff), relative speed average (iassinusvref), vertical speed average (IVV _ave), low pressure rotor standard deviation per engine, high pressure rotor standard deviation per engine. In this example, as shown in fig. 4a, the present disclosure obtains a standard deviation (n11_std) of the low-pressure rotor speed variation of the engine 1 m seconds before and after the altitude reaches the detection point, a standard deviation (n12_std) of the low-pressure rotor speed variation of the engine 2 m seconds before and after the altitude reaches the detection point, a standard deviation (n21_std) of the high-pressure rotor speed variation of the engine 1 m seconds before and after the altitude reaches the detection point, and a standard deviation (n22_std) of the high-pressure rotor speed variation of the engine 2 m seconds before and after the altitude reaches the detection point. Of course, it should be understood that the number of engines for the flyer may be other, for example the flyer may comprise 4 engines, in which case the parameters of the other 2 engines may be introduced.
After obtaining the historical operating parameters, according to step S140, the present disclosure may perform data analysis on the historical operating parameters to obtain data features, where the data features include density functions of each historical operating parameter, states and correlations between each historical operating parameter.
For example, a scatter plot matrix between every two parameters may be established using the historical operating parameters; and determining the data characteristics according to the scatter diagram matrix.
Referring to FIG. 4b, FIG. 4b shows a schematic diagram of a scatter plot matrix created during the approach phase using historical operating parameters.
From fig. 4b, it can be derived that: 1) N11_std, n12_std, n21_std, and n22_std are highly correlated, so model training using n11_std to represent n12_std, n21_std, and n22_std may be selected (of course, any other parameter may be selected as a representative); 2) Head_diff is a discrete parameter, and outlier flyer detection will be a multi-class clustering problem; 3) The density functions of the other operating parameters, except head_diff, are unimodal, approximately normal distribution, and each scatter plot has distinct data convergence properties.
According to step S150, an outlier determination model may be built from the data features.
In one example, due to the presence of the discrete parameter head_diff in the historical operating parameters, the outlier detection problem is a multi-class clustering problem, i.e., a plurality of normal flight species and outlier flight species, based on which a corresponding multi-class clustering algorithm may be selected for outlier detection, e.g., a DBSCAN algorithm may be selected.
Referring to fig. 4c, fig. 4c shows a schematic diagram of the adjustment of parameters of the model for determining an outlier fly in the approach phase.
Here, for convenience of description, three historical operation parameters of head_diff, IVV _ave and IASminusVREF are selected for analysis, and it should be noted that the model and training method are equally applicable in the case of multiple parameters. There are two parameters of the DBSCAN algorithm to be adjusted, namely a parameter eps and a parameter min_samples, wherein eps represents a neighborhood radius specified in the DBSCAN algorithm, and min_samples represents the minimum number of samples contained in a core object neighborhood specified in the DBSCAN algorithm.
And on the premise of avoiding over-fitting and under-fitting, the parameters eps and the parameters min_samples are adjusted until the quantity of the outlier flying objects is adjusted to be optimal between 5% and 10%.
Similarly, when adjusting parameters, one parameter may be fixed first to adjust another parameter, for example, please refer to fig. 4c, in which the first graph is in over-fit, the third graph is in under-fit, and the second graph is normal, by fixing the min_samples first to adjust the min_samples. The parameter eps is then adjusted by fixing the parameter min sample of the second graph, as shown in the second row. And repeating the iteration until the quantity of the outlier flying objects is adjusted to be 5-10%.
As shown in fig. 4c, dark dots (oval circled parts) are outlier flyers marked during training, and light dots are normal flyers marked during training.
As shown in fig. 4c, in this example, the outlier determination model of the approach phase performs better when the parameter eps is 0.18 or 0.22, parameter min_samples bit 5.
After training the outlier flying object determining model, the outlier flying object determining model can be utilized to judge the flying object needing outlier judgment.
Referring to fig. 4d, fig. 4d shows a schematic diagram of an outlier determination at a second detection point using an outlier determination model at a near stage. Wherein setting eps to 0.18 and min_samples to 5, inputting the operational parameters (pitch_ave, head_diff, IASminusVREF, IVV _ave, and n11_std) of at least one flyer (e.g., 480) into the outlier flyer determination model, it can be determined that the outlier flyers are approximately 43 (dark dots in fig. 4 d), accounting for about 9.03% of all training data.
Referring to fig. 4e, fig. 4e shows a schematic diagram of an outlier determination at a third detection point using an outlier determination model at a near stage. Wherein setting eps to 0.22 and min_samples to 5, inputting the operational parameters (pitch_ave, head_diff, IASminusVREF, IVV _ave, and n11_std) of at least one flyer (e.g., 480) into the outlier flyer determination model, it can be determined that the outlier flyers are approximately 45 (dark dots in fig. 4 e), accounting for about 9.45% of all training data.
Referring to fig. 4f, fig. 4f shows a schematic diagram of an outlier determination at a fourth detection point using an outlier determination model at a near stage. Wherein setting eps to 0.18 and min_samples to 5, inputting the operational parameters (pitch_ave, head_diff, IASminusVREF, IVV _ave, and n11_std) of at least one flyer (e.g., 480) into the outlier flyer determination model, it can be determined that the outlier flyer is approximately 40 (dark dots in fig. 4 f), accounting for about 8.40% of all training data.
For the approach stage, the number of the preset detection points selected for detecting the outlier flying object is 3, so that when the result of the outlier flying object at the second detection point, the third detection point and the fourth detection point is obtained, the estimated outlier flying object under N preset detection points in the preset take-off and landing stage can be determined according to the following steps; determining a flying object determined to be estimated as an outlier flying object whose number of times is not less than M as the outlier flying object "further determines an outlier flying object. For example, the continuous outlier flying objects detected by the second detection point (1000 feet), the third detection point (500 feet) and the fourth detection point (300 feet) can be combined to further screen out the outlier flying objects.
Referring to fig. 4g, fig. 4g shows a schematic diagram of determining outlier flyers in combination with a plurality of predetermined detection points at the approach stage.
When the minimum detection number of persistent outliers is set to 3 (M), a total of 12 persistent outliers are marked, which is marked with dark dots in fig. 4 g. These flyers do not recover the landing performance adjustment of the aircraft through flight control, and although they do not necessarily trigger an alarm, the risk of problems occurring with continuous outlier flyers is high relative to group performance, requiring major attention.
For the ground-horizon drift stage:
in a possible implementation manner, in the case that the predetermined take-off and landing stage is a ground-level floating stage, the preset detection points include a fifth detection point, a sixth detection point and a seventh detection point, where the fifth detection point is a position where the flying object is modified from a pitch rule to a roll-up rule, the sixth detection point is a position where the flying object reaches a roll-up height specified under a stable condition, and the seventh detection point is a position where the flying object receives an automatic call reminding pilot to withdraw the thrust handle.
In one possible embodiment, the fifth detection point has a height of 50 feet, the sixth detection point has a height of 30 feet, and the seventh detection point has a height of 20 feet.
In one possible implementation manner, in the case that the predetermined take-off and landing stage is a landing stage, the operation parameters may include any one or more of an inertial vertical velocity average (IVV _ave), a PITCH angle average (pitch_ave), an airspeed average (ias_ave), a vertical overload average (vrtg_ave), a ground distance (dist_ld), and a ground TIME (time_ld), wherein the inertial vertical velocity average represents an average value of an Inertial Vertical Velocity (IVV) of k seconds before and after the flying object height reaches the preset detection point, the PITCH angle average represents an average value of a PITCH angle of k seconds before and after the flying object height reaches the preset detection point, the airspeed average represents an average value of an airspeed (IAS) of k seconds before and after the flying object height reaches the preset detection point, the vertical overload average represents an average value of vertical overload tg of k seconds before and after the flying object height reaches the preset detection point, and the ground distance is a horizontal distance of the preset ground point to the ground point, and the ground TIME is greater than 0.
In one possible embodiment, the specific size of k may be determined according to the actual situation, and in one example, k may be 5s.
The above gives examples of the operation parameters in the ground-level drift stage, and it should be understood that those skilled in the art may increase and decrease the operation parameters according to actual situations, which is not limited in this disclosure.
The model training at the ground level drift stage and determination of outlier flights are described in exemplary fashion below.
Referring to fig. 5a, fig. 5a shows a schematic diagram of historical operating parameters of the same model of aircraft obtained during the ground level drift stage.
According to step S130, the present disclosure may obtain the historical operating parameters of the same model of flying object acquired in the ground level drift stage as shown in fig. 5 a.
As shown in fig. 5a, in one example, the present disclosure selects historical operating parameters for 28 flights (numbered No-X0001-X0028), including, in one possible implementation, inertial vertical velocity average (IVV _ave), PITCH angle average (pitch_ave), airspeed average (ias_ave), vertical overload average (vrtg_ave), ground distance (dist_ld), and ground TIME (time_ld) with the predetermined take-off and landing phase being a ground-plane phase.
After obtaining the historical operating parameters, according to step S140, the present disclosure may perform data analysis on the historical operating parameters to obtain data features, where the data features include density functions of each historical operating parameter, states and correlations between each historical operating parameter.
For example, a scatter plot matrix between every two parameters may be established using the historical operating parameters;
and determining the data characteristics according to the scatter diagram matrix.
Referring to fig. 5b, fig. 5b shows a schematic diagram of a scatter plot matrix created during the ground level drift phase using historical operating parameters.
As can be seen from fig. 5 b: 1) In fig. 5b, the density functions on the diagonal have unimodal characteristics, approximating a normal distribution; 2) Obvious data convergence characteristics exist between parameters, and no discrete parameters exist; 3) DIST_LD and TIME_LD are highly correlated, so DIST_LD may be selected to represent TIME_LD for training (DIST_LD may also be used to represent TIME_LD in subsequent outlier fly determination).
According to step S150, an outlier determination model may be built from the data features.
In One example, if there are no discrete parameters in the historical operating parameters, the outlier flying object detection problem can be regarded as a single class clustering problem, namely a normal flying object and an outlier flying object, based on which a corresponding single class clustering algorithm can be selected for outlier flying object detection, for example, an One-class SVM algorithm can be selected.
Referring to fig. 5c, fig. 5c shows a schematic diagram of the adjustment of parameters of an outlier fly determination model at the ground level drift stage.
Here, for convenience of description, two historical operation parameters, IVV _ave and dist_ld, are selected for performing anomaly detection analysis of the flying object, and it should be noted that the model and training method are equally applicable in the case of multiple parameters. The present disclosure exemplarily selects One-Class SVM algorithm for cluster learning, where a kernel function selects a Radial Basis Function (RBF), and performs parameter adjustment on two core parameters nu and gamma of the algorithm, where nu is used to define an upper boundary of a training error of the SVM algorithm, gamma is used to adjust a scale parameter of the radial basis function, and on the premise of avoiding over-fitting and under-fitting, the result is shown in fig. 5c, until the number of outlier flyers is adjusted to be optimal between 5% and 10%.
As shown in fig. 5c, the boundary line is internally marked by the outlier flyer determination model, the boundary line is externally marked by the normal point marked by the outlier flyer determination model, and the boundary line is the classification boundary learned by the outlier flyer determination model. It can be seen that the larger the parameter nu, the more like the center converges the classification boundary; the parameter gamma can be used to adjust the shape of the classification boundary, the larger the parameter, the more curved the classification boundary, but the larger the parameter, the more likely the overfitting occurs (as shown in the bottom right hand corner legend in fig. 5 c). Both overfitting and underfilling affect the results of anomaly detection, and this disclosure is avoided in parameter adjustment.
As can be seen from fig. 5c, when the parameter nu is 0.05 and the parameter gamma is 5, the model for determining the outlier flying object in the ground level drift stage has better performance.
After training the outlier flying object determining model, the outlier flying object determining model can be utilized to judge the flying object needing outlier judgment.
Referring to fig. 5d, fig. 5d shows a schematic diagram of an outlier determination at a fifth detection point using an outlier determination model at a ground level drift stage. In this example, the parameter nu is set to 0.05, the parameter gamma is set to 5, and the operation parameters (IVV _ave, pitch_ave, ias_ave, vrtg_ave and dist_ld) of at least one flying object at the fifth detection point (50 feet) are selected and input into the outlier flying object determination model, so that it can be determined that there are about 27 outlier flying objects (dark points in fig. 5 d), which account for about 5.67% of all training data.
Referring to fig. 5e, fig. 5e shows a schematic diagram of an outlier determination at a sixth detection point using an outlier determination model at a ground level drift stage. In this example, the parameter nu is set to 0.05, the parameter gamma is set to 5, and the operation parameters (IVV _ave, pitch_ave, ias_ave, vrtg_ave and dist_ld) of at least one flying object at the sixth detection point (30 feet) are selected and input into the outlier flying object determination model, so that it can be determined that there are about 25 outlier flying objects (dark points in fig. 5 e), which account for about 5.25% of all training data.
Referring to fig. 5f, fig. 5f shows a schematic diagram of an outlier determination at a seventh detection point using an outlier determination model at a ground level drift stage. In this example, the parameter nu is set to 0.05, the parameter gamma is set to 5, and the operation parameters (IVV _ave, pitch_ave, ias_ave, vrtg_ave and dist_ld) of at least one flying object at the seventh detection point (20 feet) are selected and input into the outlier flying object determination model, so that it can be determined that there are about 27 outlier flying objects (dark points in fig. 5 f), which account for about 5.67% of all training data.
It should be noted that in the determination of an outlier flight at each of the take-off and landing phases, a predetermined detection point other than that exemplified by the present disclosure may be selected, an example of which is given below.
Referring to fig. 5g, fig. 5g shows a schematic diagram of an outlier determination at an eighth detection point using an outlier determination model at a ground level drift stage. In this example, the parameter nu is set to 0.05, the parameter gamma is set to 5, and the operation parameters (IVV _ave, pitch_ave, ias_ave, vrtg_ave and dist_ld) of at least one flying object at the eighth detection point (10 feet) are selected and input into the outlier flying object determination model, so that it can be determined that the number of outlier flying objects is about 24 (dark points in fig. 5 g), which is about 5.04% of all training data.
For the ground-level drift stage, the number of the preset detection points selected for detecting the outlier flying object is 4, so that when the result of the outlier flying object of the fifth detection point, the sixth detection point, the seventh detection point and the eighth detection point is obtained, the estimated outlier flying object under the N preset detection points in the preset take-off and landing stage can be determined according to the following; determining a flying object determined to be estimated as an outlier flying object whose number of times is not less than M as the outlier flying object "further determines an outlier flying object. For example, the continuous outlier flying objects can be further screened by integrating outlier flying objects detected by the fifth detection point (50 feet), the sixth detection point (30 feet), the seventh detection point (20 feet) and the eighth detection point (10 feet).
Referring to fig. 5h, fig. 5h shows a schematic diagram of determining outlier flying objects in combination with a plurality of predetermined detection points of the ground level drift stage.
When the minimum detection number of persistent outliers is set to 4 (M), a total of 8 persistent outliers are marked, which is marked with dark dots in fig. 5 h. These flyers do not recover the landing performance adjustment of the aircraft through flight control, and although they do not necessarily trigger an alarm, the risk of problems occurring with continuous outlier flyers is high relative to group performance, requiring major attention.
According to the above description, the method for determining the outlier flying object is completely data-driven, is full-industry data, and is used for multi-parameter simultaneous analysis, so that the dependence of subjective experience is avoided, the data characteristics of the full industry are covered, and the correlation characteristics among the parameters are included. The method is realized without prior threshold value, the outlier flying object and the judgment boundary are completely searched from the data, the influence of subjective experience on management can be reduced, the flying object with the taking-off and landing performance outlier can be objectively extracted, the airlines can concentrate on the outlier flying object to improve the safety management service level, and pilots can utilize the information to improve the flying technology of the pilots.
Referring to fig. 6, fig. 6 shows a block diagram of an outlier flying object determination device according to an embodiment of the disclosure. As shown in fig. 6, the apparatus includes: a first obtaining module 10, configured to obtain an operation parameter when at least one flying object reaches a preset detection point in a preset take-off and landing stage; a determining module 20, coupled to the acquiring module 10, for determining an outlier flying object of the at least one flying object by using the operation parameter and an outlier flying object determining model corresponding to the predetermined take-off and landing stage.
Through the device, the operation parameters of at least one flying object when reaching the preset detection point in the preset take-off and landing stage can be obtained, and the outlier flying object in the at least one flying object is determined by utilizing the operation parameters and the outlier flying object determination model corresponding to the preset take-off and landing stage. The method and the device can rapidly and accurately determine the outlier flying object in at least one flying object, and comprehensively judge the outlier flying object based on the QAR data and a plurality of parameters in the whole industry, so that the dependence on subjective experience is avoided.
Under the condition that the device disclosed by the disclosure is used for determining the outlier flying object in at least one flying object, the deviation between the taking-off and landing performance of the flying object near a preset detection point and the taking-off and landing performance of a group near the preset detection point is large, and adjustment is needed. After the results are obtained, the results may be sent to a control center for reference, or a warning may be issued to alert personnel to the precautionary risk.
In one possible implementation, the determining module includes: a first determining sub-module for determining estimated outlier flyers at N preset detection points in the predetermined take-off and landing phase; and the second determining sub-module is connected with the first determining sub-module and is used for determining the flying object with the number of times being not less than M, which is determined to estimate the outlier flying object, as the outlier flying object, wherein M, N is a natural number, and M is less than or equal to N.
Referring to fig. 7, fig. 7 shows a block diagram of an outlier flying object determination device according to an embodiment of the disclosure. In one possible embodiment, as shown in fig. 7, the apparatus may further include: a second obtaining module 30, configured to obtain historical operation parameters of the plurality of flying objects reaching the preset detection point in the preset take-off and landing stage; the analysis module 40 is connected to the second obtaining module 30, and is configured to perform data analysis on the historical operating parameters to obtain data features, where the data features include a density function of each historical operating parameter, and states and correlations between each historical operating parameter; the establishing module 50 is connected to the analyzing module 40 and the first acquiring module 10, and is configured to establish an outlier flyer determining model according to the data characteristics.
In a possible implementation manner, the data analysis on the historical operation parameters to obtain data features includes: establishing a scatter diagram matrix between every two parameters by utilizing the historical operation parameters; and determining the data characteristics according to the scatter diagram matrix. In one possible embodiment, building an outlier aircraft determination model from the data features includes: determining that the outlier flyer determination model is a multi-class cluster model under the condition that the existing historical operation parameters are in discrete states or the density function of the existing historical operation parameters is in multi-peak characteristics; or under the condition that all the historical operation parameters are in a converging state and the density functions of all the historical operation parameters are in a unimodal characteristic, determining that the outlier flyer determination model is a single-class clustering model.
In one possible embodiment, in the case where there is a correlation between the plurality of historical operating parameters, one of the plurality of historical operating parameters having the correlation and the other historical operating parameters having no correlation are utilized as the training data.
In a possible implementation manner, the building an outlier flyer determination model according to the data features further includes: training the outlier flyer determination model by utilizing the training data on the premise of avoiding over-fitting and under-fitting; and adjusting model parameters of the outlier flying object determining model, determining final model parameters of the outlier flying object determining model when the number of the outlier flying objects determined in training accounts for the number of all flying objects to reach a first proportion, and obtaining the trained outlier flying object determining model according to the final model parameters.
In one possible embodiment, the first proportion is between 5% and 10%.
In one possible implementation manner, in the case that the predetermined take-off and landing stage is a take-off stage, the predetermined take-off and landing stage includes a first detection point, where the first detection point is a position of a flying object at a take-off moment, the operation parameters include any one or more of a take-off distance, a ground elevation angle, a mean value of a change rate of the ground elevation angle, a standard deviation of the change rate of the ground elevation angle, a vertical speed of the ground elevation, a mean value of a change rate of the ground elevation angle, and a standard deviation of the change rate of the ground elevation angle, where the take-off distance represents a horizontal distance when the flying object reaches the predetermined safe height from a start to a point where the flying object is in a moment, the ground elevation angle represents an elevation angle of the flying object at the moment of the flying object, the ground elevation angle of the flying object is in the moment, the ground elevation angle represents a mean value of each second of change in the flying object in the moment of the flying object, the ground elevation angle of the ground elevation angle represents a mean value of each second of the flying object in front and back n seconds, the ground elevation angle of the change rate of the flying object is in the moment, the ground elevation angle of the flying object is represented by n seconds, the standard deviation of the ground elevation angle of the flying object from the ground elevation angle of the flying object, and the ground elevation angle of the flying object is represented by n seconds of the ground elevation angle of the flying object, and the ground elevation angle of the flying object from the ground elevation angle of the ground moment is represented by the ground n seconds, and the ground elevation angle of the ground plane from the ground plane.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is the approach stage, the preset detection points include a second detection point, a third detection point and a fourth detection point, where the second detection point is a position where the flying object reaches a stable approach detection height under the Metro weather condition IMC, the third detection point is a position where the flying object reaches a stable approach detection height under the visual weather condition VMC, and the fourth detection point is a position where the flying object is at a stable approach detection height at five sides of the airport.
In one possible embodiment, the second detection point has a height of 1000 feet, the third detection point has a height of 500 feet, and the fourth detection point has a height of 300 feet.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is the approach stage, the operation parameters include any one or more of a pitch angle mean, a heading mean, a relative speed mean, a vertical speed mean, a low-pressure rotor standard deviation of each engine, and a high-pressure rotor standard deviation of each engine, where the pitch angle mean represents a pitch angle mean of m seconds before and after the altitude reaches the preset detection point, the heading mean represents a heading change mean of m seconds before and after the altitude reaches the preset detection point, the relative speed mean represents an airspeed of m seconds before and after the altitude reaches the preset detection point minus a reference speed mean of m seconds, the vertical speed mean represents an inertial vertical speed mean of m seconds before and after the altitude reaches the preset detection point, and the low-pressure rotor standard deviation of each engine represents a standard deviation of low-pressure rotor speed change of each engine of m seconds before and after the altitude reaches the preset detection point, and the high-pressure rotor standard deviation of each engine represents a low-pressure rotor speed change of m seconds of each engine, where m > 0.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is a ground-level floating stage, the preset detection points include a fifth detection point, a sixth detection point and a seventh detection point, where the fifth detection point is a position where the flying object is modified from a pitch rule to a roll-up rule, the sixth detection point is a position where the flying object reaches a roll-up height specified under a stable condition, and the seventh detection point is a position where the flying object receives an automatic call reminding pilot to withdraw the thrust handle.
In one possible embodiment, the fifth detection point has a height of 50 feet, the sixth detection point has a height of 30 feet, and the seventh detection point has a height of 20 feet.
In a possible implementation manner, in the case that the predetermined take-off and landing stage is a ground-level drift stage, the running parameters include any one or more of an inertial vertical velocity mean, a pitch angle mean, an airspeed mean, a vertical overload mean, a ground distance and a ground time, where the inertial vertical velocity mean represents an average value of the inertial vertical velocity every second k seconds before and after the flying object height reaches a preset detection point, the pitch angle mean represents an average value of the pitch angle every second k seconds before and after the flying object height reaches the preset detection point, the airspeed mean represents an average value of airspeed (IAS) every second k seconds before and after the flying object height reaches the preset detection point, the vertical overload mean represents an average value of vertical overload VRTG every second k seconds before and after the flying object height reaches the preset detection point, the ground distance is a horizontal distance from the preset ground point to the ground point, and the ground time represents a time from the preset ground point to the ground point, where k > 0.
According to the above description, the outlier flyer determining device is completely data-driven, is full-industry data, performs multi-parameter simultaneous analysis, avoids the dependence of subjective experience, covers the data characteristics of the full industry, and comprises the correlation characteristics among the parameters. The method is realized without prior threshold value, the outlier flying object and the judgment boundary are completely searched from the data, the influence of subjective experience on management can be reduced, the flying object with the taking-off and landing performance outlier can be objectively extracted, the airlines can concentrate on the outlier flying object to improve the safety management service level, and pilots can utilize the information to improve the flying technology of the pilots.
The foregoing description of the embodiments of the present disclosure has been presented for purposes of illustration and description, and is not intended to be exhaustive or limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the various embodiments described. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or the improvement of technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (11)
1. A method of determining an outlier flying object, the method comprising:
acquiring operation parameters of at least one flying object when the flying object reaches a preset detection point in a preset take-off and landing stage;
determining an outlier flying object in the at least one flying object by using the operation parameter and an outlier flying object determination model corresponding to the preset take-off and landing stage;
the method further comprises the steps of:
acquiring historical operation parameters of a plurality of flying objects reaching a preset detection point in a preset take-off and landing stage;
performing data analysis on the historical operating parameters to obtain data features, wherein the data features comprise density functions of each historical operating parameter, states and correlations among each historical operating parameter, and the data analysis on the historical operating parameters to obtain the data features comprises the following steps: establishing a scatter diagram matrix between every two parameters by utilizing the historical operation parameters; determining the data features according to the scatter diagram matrix;
and establishing an outlier flying object determination model according to the data characteristics.
2. The method of claim 1, wherein determining an outlier of the at least one outlier using the operating parameter and an outlier determination model corresponding to the predetermined take-off and landing phase comprises:
Determining estimated outlier flying objects under N preset detection points in the preset take-off and landing stage;
a flying object determined to estimate an outlier flying object a number of times not less than M is determined as the outlier flying object,
wherein M, N is natural number, M is less than or equal to N.
3. The method of claim 1, wherein building an outlier determination model from the data features comprises:
determining that the outlier flyer determination model is a multi-class cluster model under the condition that the existing historical operation parameters are in discrete states or the density function of the existing historical operation parameters is in multi-peak characteristics; or (b)
And under the condition that all the historical operation parameters are in a converging state and the density functions of all the historical operation parameters are in a unimodal characteristic, determining that the outlier flyer determination model is a single-class clustering model.
4. A method according to claim 3, characterized in that in case there is a correlation between a plurality of historical operating parameters, one of the plurality of historical operating parameters with correlation and the other historical operating parameters without correlation are utilized as training data.
5. The method of claim 4, wherein said building an outlier fly determination model from said data characteristics further comprises:
Training the outlier flyer determination model by utilizing the training data on the premise of avoiding over-fitting and under-fitting;
and adjusting model parameters of the outlier flying object determining model, determining final model parameters of the outlier flying object determining model when the number of the outlier flying objects determined in training accounts for the number of all flying objects to reach a first proportion, and obtaining the trained outlier flying object determining model according to the final model parameters.
6. The method of claim 1, wherein the predetermined detection point comprises a first detection point, the first detection point being a location of a flying object at a take-off instant, the operating parameter comprises any one or more of a take-off distance, a ground clearance elevation angle change rate mean, a ground clearance elevation angle change rate standard deviation, a ground clearance vertical velocity, a ground clearance velocity change rate mean, a ground clearance velocity change rate standard deviation, wherein the take-off distance represents a horizontal distance from a start of the flying object to a time when the flying object reaches the predetermined safe height, the ground clearance velocity represents a horizontal velocity of the flying object at the ground clearance instant of a main wheel of the flying object, the elevation angle of the ground leaving represents the elevation angle of the flying object at the moment of the ground leaving of the main wheel of the flying object, the mean value of the elevation angle change of the ground leaving elevation angle represents the mean value of elevation angle change of each second in n seconds before and after the main wheel of the flying object leaves the ground, the standard deviation of the elevation angle change of the ground leaving elevation angle represents the standard deviation of elevation angle change of each second in n seconds before and after the main wheel of the flying object leaves the ground, the vertical velocity of the flying object at the moment of the ground leaving of the main wheel of the flying object represents the vertical velocity change of each second in n seconds before and after the main wheel of the flying object leaves the ground, and the standard deviation of the vertical velocity change of the ground leaving elevation angle represents the standard deviation of vertical velocity change of each second in n seconds before and after the main wheel of the flying object leaves the ground, wherein n is more than 0.
7. The method according to claim 1, wherein in the case where the predetermined take-off and landing stage is the approach stage, the predetermined detection points include a second detection point, a third detection point and a fourth detection point, the second detection point being a position where the flying object reaches a stable approach detection altitude under the meter meteorological condition IMC, the third detection point being a position where the flying object reaches a stable approach detection altitude under the visual meteorological condition VMC, and the fourth detection point being a position where the flying object is at a stable approach detection altitude on the fifth side of the airport.
8. The method of claim 1, wherein, in the case where the predetermined take-off and landing phase is a near phase, the operating parameters include any one or more of pitch angle mean, heading mean, relative velocity mean, vertical velocity mean, low pressure rotor standard deviation per engine, high pressure rotor standard deviation per engine, wherein the pitch angle mean represents the mean of pitch angle per second m seconds before and after the altitude reaches the predetermined detection point, the heading mean represents the mean of heading change per second m seconds before and after the altitude reaches the predetermined detection point, the relative velocity mean represents the mean of airspeed per second m seconds after the altitude reaches the predetermined detection point minus a reference velocity, the vertical velocity mean represents the mean of inertial vertical velocity per second m seconds before and after the altitude reaches the predetermined detection point, the low pressure rotor standard deviation per engine represents the standard deviation of low pressure rotor velocity change per second of the engine m seconds before and after the altitude reaches the predetermined detection point, and the high pressure rotor standard deviation per engine represents the low pressure rotor speed change per second m seconds before and after the altitude reaches the predetermined detection point, wherein m > 0.
9. The method of claim 1, wherein, in the case where the predetermined take-off and landing stage is a ground-level drift stage, the predetermined detection points include a fifth detection point, a sixth detection point, and a seventh detection point, wherein the fifth detection point is a position where the flying object is modified from a pitch law to a roll-up law, the sixth detection point is a position where the flying object reaches a roll-up height specified under a stable condition, and the seventh detection point is a position where the flying object receives an automatic shout alert pilot to retract a thrust handle.
10. The method of claim 1, wherein, in the case where the predetermined take-off and landing phase is a landing drift phase, the operation parameters include any one or more of an inertial vertical velocity average representing an average value of inertial vertical velocity every second k seconds before and after the flying object height reaches a predetermined detection point, a pitch angle average representing an average value of pitch angle every second k seconds before and after the flying object height reaches the predetermined detection point, an airspeed average representing an average value of airspeed (IAS) every second k seconds before and after the flying object height reaches the predetermined detection point, a vertical overload average representing an average value of vertical overload VRTG every second before and after the flying object height reaches the predetermined detection point, a ground distance being a horizontal distance from the predetermined ground point to the ground point, and a ground time representing a time from the predetermined ground point to the ground point, wherein k > 0.
11. An outlier flying object determination apparatus, the apparatus comprising:
the first acquisition module is used for acquiring operation parameters when at least one flying object reaches a preset detection point in a preset take-off and landing stage;
the determining module is connected with the acquiring module and is used for determining the outlier flying object in the at least one flying object by utilizing the operation parameters and the outlier flying object determining model corresponding to the preset take-off and landing stage;
the apparatus further comprises:
the second acquisition module is used for acquiring historical operation parameters of a plurality of flying objects reaching a preset detection point in a preset take-off and landing stage;
the analysis module is connected to the second acquisition module, and is used for performing data analysis on the historical operation parameters to obtain data characteristics, wherein the data characteristics comprise density functions of all the historical operation parameters, states and correlations among all the historical operation parameters, and the data analysis on the historical operation parameters to obtain the data characteristics comprises the following steps: establishing a scatter diagram matrix between every two parameters by utilizing the historical operation parameters; determining the data features according to the scatter diagram matrix;
the establishing module is connected with the analyzing module and the first obtaining module and is used for establishing an outlier flying object determining model according to the data characteristics.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910700625.1A CN110390177B (en) | 2019-07-31 | 2019-07-31 | Method and device for determining outlier flying object |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910700625.1A CN110390177B (en) | 2019-07-31 | 2019-07-31 | Method and device for determining outlier flying object |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110390177A CN110390177A (en) | 2019-10-29 |
CN110390177B true CN110390177B (en) | 2023-06-16 |
Family
ID=68288108
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910700625.1A Active CN110390177B (en) | 2019-07-31 | 2019-07-31 | Method and device for determining outlier flying object |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110390177B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111477036B (en) * | 2020-04-08 | 2021-01-29 | 中国电子科技集团公司第二十八研究所 | Method for detecting aircraft height abnormity of air traffic control automation system |
CN112200052B (en) * | 2020-09-30 | 2023-06-09 | 驭势(上海)汽车科技有限公司 | Track deviation recognition and vehicle running analysis method, device, equipment and medium |
CN114004292B (en) * | 2021-10-29 | 2023-02-03 | 重庆大学 | Pilot flat-floating ejector rod behavior analysis method based on flight parameter data unsupervised clustering |
CN117314173A (en) * | 2023-11-08 | 2023-12-29 | 重庆大学 | Method and system for identifying and analyzing tail wiping risks based on radio altitude auxiliary measurement and calculation |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10643480B2 (en) * | 2016-04-19 | 2020-05-05 | George Mason University | Method and apparatus for probabilistic alerting of aircraft unstabilized approaches using big data |
CN107045638B (en) * | 2016-12-30 | 2021-03-30 | 中国民航管理干部学院 | Flight safety event analysis method based on situational awareness model |
CN110033194B (en) * | 2019-04-17 | 2021-02-26 | 中国民航科学技术研究院 | Method for detecting and evaluating unstable approach of airplane and electronic equipment |
-
2019
- 2019-07-31 CN CN201910700625.1A patent/CN110390177B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN110390177A (en) | 2019-10-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110390177B (en) | Method and device for determining outlier flying object | |
US10643480B2 (en) | Method and apparatus for probabilistic alerting of aircraft unstabilized approaches using big data | |
US8798898B2 (en) | Methods and systems for inferring aircraft parameters | |
US9177479B2 (en) | System and method for determining aircraft operational parameters and enhancing aircraft operation | |
FR2939558A1 (en) | METEOROLOGICAL MODELING METHOD FOR CALCULATING AN AIRCRAFT FLIGHT PLAN | |
EP2743739B1 (en) | Using aircraft trajectory data to infer atmospheric conditions | |
Olive et al. | Quantitative assessments of runway excursion precursors using Mode S data | |
US20220139233A1 (en) | Method and system for updating a flight plan | |
CN111627259A (en) | Unmanned aerial vehicle intrusion classification early warning method in airport clearance protection area and storage medium | |
Das et al. | Comparison of algorithms for anomaly detection in flight recorder data of airline operations | |
CN115064009A (en) | Method for grading risk of unmanned aerial vehicle and manned conflict in terminal area | |
Milano et al. | Air risk maps for unmanned aircraft in urban environments | |
CN113487914A (en) | Low landing operation standard identification and alarm method based on airport meteorological messages | |
CN116415818B (en) | Method and system for confirming risk points in aircraft approach stage based on clustering algorithm | |
US20180079495A1 (en) | Systems and methods for forecasting and reducing the occurrence of tire overspeed events during aircraft takeoff and landing | |
Payan et al. | Improvement of rotorcraft safety metrics using performance models and data integration | |
Jiang et al. | Research on the flight anomaly detection during take-off phase based on FOQA data | |
Jiang et al. | A method for the outlier flights detection of the final approach based on FOQA data | |
Payan et al. | Review of proactive safety metrics for rotorcraft operations and improvements using model-based parameter synthesis and data fusion | |
De Visscher et al. | Data mining and Machine Learning techniques supporting Time-Based Separation concept deployment | |
Hernández et al. | Time in Advance Go-Around Predictions for Decision Support in Air Traffic Management | |
Distefano et al. | Investigation of the causes of runway excursions | |
He et al. | Data-driven method for estimating aircraft mass from quick access recorder using aircraft dynamics and multilayer perceptron neural network | |
Johnson et al. | Investigating the impacts of a separation standard for UAS operations in enroute and transition airspace | |
Venturelli Cavalheiro et al. | Characterizing Effective Navigation Performance of Legacy Aircraft to Enable Reduced Procedural Separation in Support of Advanced Air Mobility |
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 |