A method for classifying aircraft maneuvers in a two-dimensional plane

Abstract

The present invention discloses a method for classifying aircraft maneuvers in a two-dimensional plane. Aiming at the drawbacks that most existing maneuver classifications are based on existing standard maneuvers, an automatic maneuver classification method based on the projections of flight trajectories in the horizontal and vertical planes is proposed in combination with the Sequential Important Points (SIP). This method extracts maneuver segments from flight data according to the trends of two-dimensional plane trajectory data, combines and refines the maneuver segments with the idea of maneuver splicing, and conducts classification experiments with complete aircraft landing and takeoff data. It is proved by experiments that the present invention can improve the recognition efficiency of flight maneuvers on the premise of ensuring the classification efficiency.

Description

Technical Field

[0001] The present invention relates to the field of aviation, and particularly to a method for dividing aircraft maneuvers in a two-dimensional plane. Background Art

[0002] The division of aircraft maneuvering actions is an important basis for research work such as flight action evaluation, engine load spectrum research, and flight simulation. Under different flight training requirements, aircraft flights have different laws. It is difficult to compare and summarize the laws of different flights by directly observing and studying flight takeoff and landing data. In fact, flight training is a combination of a series of maneuvering actions. By dividing the complex flight takeoff and landing data of the entire section into multiple basic, regular, and representative maneuvering action segments through aircraft maneuver division, the laws of different flights can be compared and summarized. In addition, the idea of maneuvering action division can also be used for flight action evaluation, aircraft load spectrum research, engine load spectrum research, and risk assessment.

[0003] Currently, domestic and foreign maneuver divisions mainly include knowledge base matching method, pattern matching method, neural network, trend recognition method, etc. The division and recognition of maneuvering actions have been achieved through various methods, but there is a basic problem: the acquisition of standard maneuvering action templates. The key to the knowledge base matching method lies in the establishment of the knowledge base, which often requires experienced personnel to interpret a large amount of flight parameter data and summarize through experience, and each time it can only be for a specific type of aircraft. Although other methods do not require manual reading and judgment of parameters, they require standard action templates for comparative analysis, and there are certain operation errors when different pilots fly the same action, and it is also difficult to determine the standard template. Summary of the Invention

[0004] Aiming at the above deficiencies in the prior art, the aircraft maneuver division method in a two-dimensional plane provided by the present invention solves the problem of difficult determination of the standard maneuver template and can quickly, automatically, and accurately complete the task segment division.

[0005] To achieve the above invention purpose, the technical solution adopted by the present invention is: an aircraft maneuver division method in a two-dimensional plane, the method comprising the following steps:

[0006] S1: Divide the flight parameter data into multiple sequences according to a fixed length, project the sequence trajectory onto the horizontal plane, use the PLR_SIP algorithm to segment the horizontal trajectory projection, identify the horizontal sequence trend, and use the merging rule to merge adjacent horizontal trend sequences with the same trend;

[0007] S2: Project the merged horizontal trend sequence onto the vertical plane, use the PLR_SIP algorithm to segment the vertical trajectory projection, identify the vertical sequence trend, and use the splitting rule to split adjacent vertical trend sequences with different trends;

[0008] S3: Superimpose the merged horizontal trend sequence and the split vertical trend sequence to obtain the basic maneuvering actions;

[0009] S4: Refine the basic maneuvering actions using the subdivision rules to achieve the aircraft maneuvering division in the two-dimensional plane.

[0010] The beneficial effects of the above solution are as follows: Through the above technical solution, the maneuvering segments in the flight data are extracted according to the trends of the two-dimensional plane trajectory data. With the idea of maneuvering action splicing, the maneuvering segments are combined and refined, and the division experiment is carried out using the complete aircraft landing and takeoff data. This division method can improve the recognition efficiency of flight actions while ensuring the division efficiency, and solves the problem that it is difficult to determine the standard maneuvering template.

[0011] Furthermore, the PLR_SIP algorithm in S1 and S2 includes the following sub-steps:

[0012] A1: Set the time series X = (x1, x2, …, x n ), the set of important point sequences IPs, and add the first point x1 and the x n of the time series X = (x1, x2, …, x n ) to the set of important point sequences IPs;

[0013] A2: Take two adjacent points in the time series X = (x1, x2, …, x n ) as a subsequence, calculate the SIP points in the time series X = (x1, x2, …, x n ), and add them to the set of important point sequences IPs;

[0014] A3: Form a SIP subsequence with two adjacent SIP points, perform linear interpolation on the SIP subsequence, and calculate the root mean square error with the original sequence of relative positions;

[0015] A4: Determine whether the root mean square error meets the set fitting error threshold condition. If so, the algorithm ends; if not, return to step A2 and continue to execute.

[0016] The beneficial effects of the above further solution are as follows: Through the above technical solution, the horizontal trajectory and the vertical trajectory are segmented, and the PLR_SIP algorithm is used to identify the horizontal sequence trend and the vertical sequence trend.

[0017] Furthermore, the distance between two SIP points in the PLR_SIP algorithm in S1 and S2 uses the VD distance as the metric value, and the formula is as follows:

[0018]

[0019] Among them, both (x1, y1) and (x2, y2) are the coordinate points of the interval endpoints, and (x s , y s ) is the coordinate point of the data point.

[0020] The beneficial effect of the above further solution is that the distance between two SIP points can be expressed by the vertical distance (PD), the plumb distance (VD), and the Euclidean distance (ED). For the three distances, the number of calculations required is the same. The graphs fitted using PD and VD are the same, but using the VD distance minimizes the amount of calculation. Therefore, the VD distance is used as the metric value in analyzing flight parameters in this solution.

[0021] Further, the identification of the horizontal sequence trend in S1 specifically includes the following situations:

[0022] B1: When the maximum value d of the VD distance value of the SIP point max ≤ the specified distance threshold d SIP , it is determined that the horizontal sequence is a straight primitive;

[0023] B2: When the maximum value d of the VD distance value of the SIP point max > the specified distance threshold d SIP , it is determined that the horizontal sequence is a curved primitive.

[0024] The beneficial effect of the above further solution is that by comparing the maximum value of the VD distance value with the specified distance threshold, the identification of the straight primitive and the curved primitive of the horizontal trend sequence is completed.

[0025] Further, the identification of the plumb sequence trend in S2 specifically includes the following situations:

[0026] B3: When the slope k of two adjacent SIP points i ≤ the negative value of the specified slope threshold k SIP , it is determined that the plumb sequence is a descending primitive;

[0027] B4: When the negative value of the specified slope threshold k SIP < the slope k of two adjacent SIP points i < the specified slope threshold k SIP , it is determined that the plumb sequence is a horizontal primitive;

[0028] B5: When the specified slope threshold k SIP ≤ the slope k of two adjacent SIP points i , it is determined that the plumb sequence is an ascending primitive.

[0029] The beneficial effect of the above further solution is that when the sequence is in the plumb coordinate system, the slope of two adjacent SIP points needs to be calculated and compared with the specified slope threshold to complete the identification of the plumb sequence trend. Description of the Drawings

[0030] Figure 1 It is a flow chart of a method for dividing aircraft maneuvers in a two-dimensional plane.

[0031] Figure 2 It is a schematic diagram for calculating three distances.

[0032] Figure 3 It is a process diagram of PLR_SIP calculation.

[0033] Figure 4 It is a diagram for merging trend sequences.

[0034] Figure 5 It is a diagram for splitting trend sequences.

[0035] Figure 6 It is a diagram of the horizontal plane trend recognition result.

[0036] Figure 7 It is a diagram for dividing simple maneuver actions and mission segments.

[0037] Figure 8 It is a detailed diagram of simple maneuver actions. Specific implementation manner

[0038] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0039] As Figure 1 shown, a method for dividing aircraft maneuvers in a two-dimensional plane, the method includes the following steps:

[0040] S1: Divide the flight parameter data into multiple segments of sequences according to a fixed length, project the sequence trajectory onto the horizontal plane, use the PLR_SIP algorithm to segment the horizontal trajectory projection, identify the horizontal sequence trend, and merge adjacent identical horizontal trend sequences using the merging rule;

[0041] S2: Project the merged horizontal trend sequence onto the vertical plane, use the PLR_SIP algorithm to segment the vertical trajectory projection, identify the vertical sequence trend, and split adjacent different vertical trend sequences using the splitting rule;

[0042] S3: Superimpose the merged horizontal trend sequence and the split vertical trend sequence to obtain the basic maneuver actions;

[0043] S4: Refine the basic maneuver actions using the refinement rule to achieve the division of aircraft maneuvers in the two-dimensional plane.

[0044] The PLR_SIP algorithm in S1 and S2 includes the following sub-steps:

[0045] A1: Set the time series X = (x1, x2,..., xn ), the set of important point sequences IPs, takes the first points x1 and x of the time series X = (x1, x2, …, x n ) and adds them to the set of important point sequences IPs; n

[0046] A2: Using adjacent two points in the time series X = (x1, x2, …, x n ) as a subsequence, calculates the SIP points in the time series X = (x1, x2, …, x n ) and adds them to the set of important point sequences IPs;

[0047] A3: Forms a SIP subsequence with two adjacent SIP points, performs linear interpolation on the SIP subsequence, and calculates the root mean square error with the original sequence of relative positions;

[0048] A4: Judges whether the root mean square error meets the set fitting error threshold condition. If so, the algorithm ends; if not, returns to step A2 to continue execution.

[0049] The distance between two SIP points of the PLR_SIP algorithm in S1 and S2 uses the VD distance as the metric value, and the formula is as follows:

[0050]

[0051] Among them, (x1, y1) and (x2, y2) are the coordinates of the interval endpoints, and (x s , y s ) are the coordinates of the data points.

[0052] The basic idea of the PLR_SIP algorithm is to select the sequence points in the sequence that are decisive for the overall shape of the sequence, and other points with less decisiveness can be ignored. The selected points are called sequence important points (SIP). Solving for SIP points mainly involves calculating the distances from each sequence point in the region formed by two adjacent SIP points to these two SIP points. Currently, 3 distance representation methods are often used: perpendicular distance (PD), vertical distance (VD), and Euclidean distance (ED). As Figure 2 shown, it gives 3 distance calculation methods for the data point P S to the interval endpoints P1 and P2:

[0053]

[0054]

[0055] ​

[0056] For the three distances, the number of calculations required is the same. The graphs fitted using PD and VD are the same, but the computational amount is the smallest when using the VD distance. Therefore, the VD distance is used as the metric value in analyzing flight parameters in this paper.

[0057] In an embodiment of the present invention, taking the altitude data of a certain flight as an example, perform PLR_SIP calculation on it, and set the fitting error threshold Δ SIP to be 0.0001. The calculation process is demonstrated as Figure 3 shown. The dotted line part is the state of the original sequence, and the straight line part is the form described using SIP points. It can be observed that as the number of SIP points increases, Δ SIP decreases, but at the same time the data compression efficiency is lower. Therefore, it needs to be set according to the actual situation during the calculation.

[0058] The trend recognition technology is a technology for extracting trend information from noisy process data. Flight parameter data has the characteristics of being massive, complex, trending, and often accompanied by noise effects. And the flight parameter data can be regarded as a time series form. This solution uses the commonly used PLR_SIP algorithm in time series patterns to describe the trend of the parameter sequence. When identifying, two linear primitives, curved and straight, are selected on the horizontal plane, and three linear primitives, rising, falling, and straight, are selected on the vertical plane. Finally, the state of the parameter sequence is obtained by superimposing the primitive information in the two planes. The specific process is as follows:

[0059] Divide the original flight parameter data sequence into multiple sequences with the same length of n. Suppose a sequence contains n sequence points. The horizontal axis coordinates X = (x1, x2,..., x n ) and the vertical axis coordinates Y = (y1, y2,..., y n ) of this sequence in the horizontal plane coordinate system. Calculate the SIP point sequence of this sequence segment. Denote the horizontal coordinates of the SIP point sequence as T = (t1, t2,..., t m ), the vertical coordinates as V = (v1, v2,..., v m ), the VD distance values of the SIP points as D = (d1, d2,..., d m-2 ), and stipulate the distance threshold d SIP . The identification of the horizontal sequence trend in S1 specifically includes the following situations:

[0060] B1: When the maximum value d max of the VD distance value of the SIP points ≤ the stipulated distance threshold d SIP , determine that the horizontal sequence is a straight primitive;

[0061] B2: When the maximum value d max of the VD distance value of the SIP points > the stipulated distance threshold d SIP, it is determined that the horizontal sequence is a bending element.

[0062] When the sequence is in the vertical coordinate system, after calculating the horizontal and vertical coordinates of the SIP point sequence of the sequence segment, calculate the slope of two adjacent SIP points to form a slope sequence k = (k1, k2, …, k l-2 ), and a slope threshold k SIP is specified. The identification of the vertical sequence trend in S2 specifically includes the following situations:

[0063] B3: When the slope k i of two adjacent SIP points ≤ the negative value of the specified slope threshold k SIP , it is determined that the vertical sequence is a descending element;

[0064] B4: When the negative value of the specified slope threshold k SIP < the slope k i of two adjacent SIP points < the specified slope threshold k SIP , it is determined that the vertical sequence is a horizontal element;

[0065] B5: When the specified slope threshold k SIP ≤ the slope k i of two adjacent SIP points, it is determined that the vertical sequence is an ascending element.

[0066] The setting of the three parameters has a great influence on the trend recognition accuracy. d SIP and k SIP mainly affect the ability of the algorithm to recognize trend turning points. d SIP and k SIP The smaller the preset value, the stronger the ability of the algorithm to recognize trend turning points. However, too small a value will limit the trend recognition to local features, resulting in too many classification quantities and losing the meaning of trend recognition. n reflects the length of the initial trend recognition segment of the algorithm. The larger n is, the less attention the algorithm pays to the local trend of the overall initial segment, and the less affected by noise. However, too small a value will make it impossible to recognize important trends in the segment.

[0067] In an embodiment of the present invention, when splitting and merging trend sequences, when initially defining n, a basic maneuver segment may be split into multiple adjacent trend sequences with the same trend. Therefore, it is necessary to merge these trend sequences into one trend sequence. The merged sequence may have a situation where the overall trend changes. For example, when two straight primitives are merged together, it may become a curved primitive. At this time, it is necessary to re-identify the trend of the sequence. Repeat the process of sequence recognition and merging until there are no adjacent trend sequences with the same trend, and the merging of trend sequences is completed. When projecting the "horizontal trend sequence" onto the vertical plane for trend recognition, there may be multiple primitives in a single sequence, such as an ascending primitive - a horizontal primitive. At this time, it is necessary to use the SIP point to check the trend of the sequence. If the trends of the sequences are inconsistent, it is necessary to split each primitive and use the idea of merging to check for adjacent trend sequences with the same trend. At this time, the basic maneuver segment has been formed. The specific situation is as Figure 4 and Figure 5 shown.

[0068] In an embodiment of the present invention, the simulation experiment and analysis of this solution are as follows:

[0069] (1) Analysis of trajectory data in the horizontal plane

[0070] Taking a complete takeoff and landing data as an example, integrate the flight speed and some attitude angles to reproduce the aircraft's motion trajectory. Project the trajectory onto the horizontal plane (xoy plane), and set the trend recognition parameters as shown in Table 1. Cut the trajectory data into multiple sequences with a length of n, perform trend recognition on each segment of the sequence, and merge adjacent sequences with the same trend. At this time, the division result of the aircraft's flight trajectory on the horizontal plane can be obtained, that is, multiple groups of horizontal trend sequences are obtained, such as Figure 6 shown.

[0071] Table 1 Setting of trend recognition parameters in the horizontal plane

[0072]

[0073] (2) Analysis of trajectory data in the vertical plane

[0074] Project the 86 horizontal trend sequences generated in the horizontal trend onto the vertical plane to obtain the corresponding vertical trend sequences. According to Δ SIP Take out the SIP points of the vertical trend sequences, calculate the slopes between each SIP point, and compare k SIP to judge whether there is a situation where the primitive type changes. If so, split the sequence at the change position. The trend recognition parameters are shown in Table 2.

[0075] Table 2 Setting of trend recognition parameters in the vertical plane

[0076]

[0077] (3) Maneuvering segment division

[0078] The trend states of the superimposed sequences in the horizontal and vertical planes. According to flight dynamics, 6 sequence types can be superimposed, including level flight, climb, glide, turn, climbing turn, and descending turn. By flying and examining each sequence type in the overall sequence, adjacent identical sequence types are merged. The 3 types of superimposed results are shown in Table 3.

[0079] Table 3 Summary of type superposition

[0080]

[0081] Since in practice, more attention is paid to the maneuver types of the aircraft between the takeoff and landing segments, the following rules are added to identify the takeoff and landing segments: First, determine the time when the barometric altitude first reaches near the maximum value and there is no obvious altitude increase action afterwards, and define this time value as the separation point between the takeoff segment and the intermediate maneuvering flight segment. Similarly, determine the moment when the barometric altitude starts to make a descending action and there is no obvious altitude increase action afterwards, and define this moment as the separation point between the landing segment and the intermediate maneuvering flight segment. Using the barometric altitude value, the takeoff and landing segments can be further divided into ground roll segments. The specific division is as Figure 7 shown.

[0082] (4) Maneuvering segment subdivision

[0083] After obtaining the maneuvering action division result of this flight through (3), different subdivision criteria can be set according to different requirements. According to the flight training syllabus and flight mode, using the change patterns of flight parameters such as speed, heading angle, and roll angle within the segment, this application summarizes the following 14 subdivision criteria for reference, as shown in Table 4.

[0084] Table 4 Summary of flight action subdivision

[0085]

[0086] As Figure 8 shown, it shows the basic maneuvering action division results after refining some intermediate maneuvering actions. In addition, complex maneuvering actions can also be formed according to the composition order of the basic actions and the cumulative change values of the parameters. For example, if the cumulative change value of the horizontal turn heading angle exceeds 360°, it can be considered a hovering action. If in a continuous action, first make a horizontal roll of half a turn and then dive until the pitch angle is 0°, it can be considered that these two actions form a half-roll inverted action. Using this method, the complex maneuvering actions in the figure can be further analyzed to obtain horizontal turns of 90°, 180°, descending turns of 180°, etc.

[0087] (5) Validation of the effectiveness of action segments

[0088] According to the division points in (4), corresponding segments are intercepted under the three-dimensional trajectory, and the maneuver action categories of each segment are determined by domain expert personnel, and compared with the action categories divided by this method. Some comparison results are shown in Table 5.

[0089] Table 5 Comparison of the detailed results of flight actions

[0090]

[0091] The above division results show that this solution realizes the division of basic maneuver actions, that is, there are no data points in the takeoff and landing process that are not recognized, and the divided categories of maneuver actions are very consistent, proving that the action recognition method established by this solution is effective.

[0092] The present invention proposes a method for dividing aircraft maneuver actions based on the trend of trajectory projection features under two two-dimensional planes, fully utilizes the characteristics of maneuver actions themselves to realize the recognition and division of basic maneuver actions, deeply analyzes the characteristics of maneuver action flights in daily aircraft training subjects, combines the change types of basic action parameters and the combination rules of complex actions, and realizes the refinement of maneuver actions and the combination recognition of complex actions. The division algorithm proposed by this solution has the advantages of accurate and effective division method, clear and comprehensive division results, and fast division speed, and can be used as an effective solution for maneuver action division.

[0093] Those of ordinary skill in the art will realize that the embodiments described herein are for helping readers understand the principles of the present invention, and it should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations that do not depart from the essence of the present invention according to these technical revelations disclosed by the present invention, and these deformations and combinations are still within the protection scope of the invention.

Claims

1. A method for classifying aircraft maneuvers in a two-dimensional plane, characterized in that, The method includes the following steps: S1: Divide the flight parameter data into multiple sequences of fixed length, project the sequence trajectories onto the horizontal plane, use the PLR_SIP algorithm to segment the horizontal trajectory projections, identify the horizontal sequence trends, and merge adjacent horizontal trend sequences with the same trend using the merging rule; S2: Project the merged horizontal trend sequences onto the vertical plane, use the PLR_SIP algorithm to segment the vertical trajectory projections, identify the vertical sequence trends, and split adjacent vertical trend sequences with different trends using the splitting rule; The PLR_SIP algorithm includes the following sub-steps: A1: Set the time series X = (x1, x2, …, x n ), the set of important point sequences IPs, and add the first points x1 and x n of the time series X = (x1, x2, …, x n ) to the set of important point sequences IPs; A2: Take two adjacent points in the time series X = (x1, x2, …, x n ) as a subsequence, calculate the SIP points in the time series X = (x1, x2, …, x n ), and add them to the important point sequence set IPs; A3: Form a SIP subsequence with two adjacent SIP points, perform linear interpolation on the SIP subsequence, and calculate the root mean square error with the original sequence of relative positions; A4: Determine whether the root mean square error meets the set fitting error threshold condition. If so, the algorithm ends. If not, return to step A2 and continue to execute; S3: Superimpose the merged horizontal trend sequences and the split vertical trend sequences to obtain the basic maneuvering actions; S4: Refine the basic maneuvering actions using the refinement rule to achieve the maneuvering division of the aircraft in the two-dimensional plane.

2. The aircraft maneuver division method under a two-dimensional plane according to claim 1, wherein In S1 and S2, the VD distance is used as the metric value for the distance between two SIP points in the PLR_SIP algorithm. The formula is as follows: Among them, (x1, y1) and (x2, y2) are both the coordinates of the interval endpoints, and (x s , y s ) is the coordinate of the data point.

3. The aircraft maneuver division method under a two-dimensional plane according to claim 2, wherein The identification of the horizontal sequence trend in S1 specifically includes the following situations: B1: When the maximum value d of the VD distance value at the SIP point max ≤ the specified distance threshold d SIP , it is determined that the horizontal sequence is a flat primitive; B2: When the maximum value d of the VD distance value of the SIP point max > the specified distance threshold d SIP , it is determined that the horizontal sequence is a bending primitive.

4. The aircraft maneuver division method under a two-dimensional plane according to claim 3, wherein The identification of the vertical sequence trend in S2 specifically includes the following situations: B3: When the slope k between two adjacent SIP points i ≤ the specified slope threshold k SIP is negative, determine the vertical sequence as a descending primitive; B4: When the specified slope threshold k SIP is negative < the slope k of adjacent two SIP points i < the specified slope threshold k SIP , it is determined that the vertical sequence is a horizontal primitive; B5: When the specified slope threshold k SIP ≤ the slope k of two adjacent SIP points i , it is determined that the plumb sequence is a rising primitive.

Images