A Method for Quantifying and Evaluating the Collision Risk of Aircraft Operations at Hub Airports

Abstract

The present invention discloses a method for quantifying and evaluating the collision risk of aircraft operations at hub airports. It obtains the airport flight operation information within a certain period of time, integrates and analyzes the data to obtain the airport apron operation network diagram, and establishes an airport operation Petri net for the airport apron operation network; predicts the probability of an aircraft arriving at different positions for the aircraft taxiing on the airport surface and converts it into the probability of different places in the Petri net; calculates the collision risk of different aircraft in different regions using the arrival probabilities of the places corresponding to different aircraft; classifies the collision risk using the XGBoost algorithm, thereby generating an evaluation of the airport operation collision risk. The present invention solves the problems of inaccurate quantification of the aircraft collision risk during the traditional apron taxiing process, inaccurate definition of the collision area, and inaccurate risk classification and evaluation; it is beneficial for the airport to accurately quantify the collision risk when undertaking most flight operation tasks, highlight the risk areas, and provide a reference for the daily management of the apron.

Description

Technical Field

[0001] The present invention belongs to the field of civil aviation airport operation safety, and particularly relates to a method for quantifying and evaluating the collision risk of aircraft operations at hub airports. Background Art

[0002] The airport apron is the core area of airport operation and transportation, where aircraft perform operations such as takeoff, landing, and taxiing. With the increasing demand in the civil aviation market, the number of flights undertaken by airports daily is gradually increasing, and hub airports bear greater flight pressure compared to regional airports. In the case of large-scale flight transportation, there will be multiple aircraft taxiing on the airport apron simultaneously, significantly increasing the collision risk between multiple aircraft. Since the collision risk is an abstract indicator, there will be a large deviation in visually judging the degree of danger of the collision risk by apron controllers alone, and controllers with different work experiences also have different judgment criteria for the collision risk of aircraft. Therefore, accurately quantifying the collision risk between different aircraft and classifying and evaluating different degrees of collision risk are particularly important, and can also provide relevant references for airport operation management.

[0003] Currently, there are mainly two types of quantification methods for aircraft collision risk. One type of quantification method mostly involves establishing a collision risk index system, which mainly includes: ranking of safety influencing factors, analytic hierarchy process, etc.; the other is to conduct physical motion modeling for aircraft in operation, which mainly includes: lateral and longitudinal collision models of aircraft in the air, following collision models, etc. The former can classify and evaluate the collision risk by indicators, but lacks a corresponding accurate quantification process. Although the latter can quantify the collision risk from a physical perspective, the model lacks consideration of the possible collision areas and research on classifying and evaluating the collision risk, and is obviously not applicable to the current requirements of hub airports for quantifying and evaluating aircraft collision risk. Summary of the Invention

[0004] Object of the Invention: The present invention proposes a method for quantifying and evaluating the collision risk of aircraft operations at hub airports, aiming to solve the problems of inaccurate quantification of aircraft collision risk during traditional apron taxiing, inaccurate definition of collision areas, and inaccurate risk classification and evaluation.

[0005] Technical Solution: A method for quantifying and evaluating the collision risk of aircraft operations at hub airports according to the present invention includes the following steps:

[0006] (1) Obtain the airport flight operation information within a certain period of time, integrate and analyze the data, and select the flight operation information during the peak hour of the peak day;

[0007] (2) Obtain the airport apron operation network diagram, and establish an airport operation Petri net for the airport apron operation network;

[0008] (3) Predict the probability of an aircraft on the airport surface reaching different positions during taxiing, and convert it into the probability of different places in the Petri net;

[0009] (4) Calculate the collision risk of different aircraft in different areas using the arrival probabilities of the places corresponding to different aircraft;

[0010] (5) Classify the collision risk using the XGBoost algorithm to generate an evaluation of the collision risk of airport operations.

[0011] Further, the airport operation flight information in step (1) includes the takeoff and landing times of flights, the types of flight arrivals and departures, the corresponding parking positions, and the taxiing speed of the aircraft.

[0012] Further, the airport apron operation network diagram in step (2) includes the specific location distribution of the airport taxiways, the operation conditions of the airport taxiway areas, and the length conditions of the airport taxiway areas.

[0013] Further, the implementation process of establishing an airport operation Petri net for the airport apron operation network in step (2) is as follows:

[0014] Discretize the surface movement area into multiple taxiway sub-areas as places p i , where the area length satisfies that only one aircraft is allowed to operate in it, the operation state of the surface flight corresponds to the state identifier m, and the transition t i is the event that the flight enters the next taxiway sub-area p i from the taxiway sub-area p i+1 ; establish a Petri net in the following form:

[0015] N = {P, T, Pre, Post, m, K}

[0016] Among them, the place set P represents the sub-areas where the aircraft operates, the transition set T represents the boundaries between the aircraft operation sub-areas and the next ones; Pre and Post represent the forward and backward incidence matrices of P and T; m is the surface state identifier; the mapping K controls the aircraft that cannot enter the active area.

[0017] Further, the implementation process of step (3) is as follows:

[0018] For the operation state of the aircraft, calculate the potential movement space of the taxiing aircraft at different times, regard the taxiing movement of the aircraft from the starting point to the end point as a Brownian motion to calculate the probability of the aircraft reaching different positions; to ensure that the aircraft arrives on time, calculate the reachable interval of the aircraft:

[0019]

[0020] Among them, L x (t) is the lower bound of the operating position of the aircraft at time t, and U x (t) is the upper bound of the operating position of the aircraft at time t, and t i is the time when the aircraft starts taxiing from the starting point, and t j is the time when the aircraft is expected to taxi to the end point, and D ij is the distance from the starting point to the end point of the aircraft, is the average operating speed of the aircraft, and V m is the maximum taxiing speed of the aircraft;

[0021] The probability that the aircraft may reach different positions on the taxiing route is obtained by calculating the upper and lower bounds of the aircraft position at different times. The calculation formula is:

[0022]

[0023] Among them, h(x) is the corrected probability density function, f(x) is the probability density function of the standard unbounded Brownian motion, F(x) is the cumulative distribution function, and σ′ 2 is the corrected variance; thus, the probability that the aircraft inside the interval reaches any position is obtained;

[0024] Assume that the designated length of a certain place p k is m meters. The sum of the arrival probabilities corresponding to different positions in the place is used as the probability that the aircraft reaches this place at the current time:

[0025]

[0026] Among them, h(x h ) is the arrival probability of the aircraft at the position x k in the place p h .

[0027] Furthermore, the step (4) is implemented through the following formula:

[0028]

[0029] Among them, n represents the number of aircraft with a non-zero arrival probability for the place k on the airfield.

[0030] Furthermore, the implementation process of classifying the collision risk by using the XGBoost algorithm in step (5) is as follows:

[0031]

[0032] Among them, (x j , y j ) is the observation value; l is the loss function; is the estimated value of the (k - 1)-th iteration calculation; Ω(f k (x)) is a penalty term that penalizes the complexity of the tree; is to sum up the results of each tree; taking different regions as the x vector, the regional collision risk value as the corresponding value, and the comprehensive classification level as the y vector, and using the XGBoost algorithm for training and classification.

[0033] Further, the collision risk described in step (5) has six levels; if the risk value is 0, the risk level is 0 and no warning is required; if the risk value is 0 - 0.01, the risk level is 1 and a mild warning is given; if the risk value is 0.01 - 0.1, the risk level is 2 and a moderate warning is given; if the risk value is 0.1 - 0.3, the risk level is 3 and a high warning is given; if the risk value is 0.3 - 0.5, the risk level is 4 and a severe warning is given; if the risk value is above 0.5, the risk level is 5 and an emergency warning is given.

[0034] Beneficial effects: Compared with the prior art, the beneficial effects of the present invention are as follows: By calculating the potential movement space of the aircraft on the taxiing path, the present invention models the aircraft activities as Brownian motion under constrained conditions, calculates the arrival probabilities of the aircraft at different positions at different times, maps them to the established apron Petri net to calculate the collision risks of different regions, and uses the XGBoost algorithm to classify and evaluate the collision risks to obtain the final evaluation result; it solves the problems of inaccurate quantification of aircraft collision risks during the traditional apron taxiing process, inaccurate definition of collision regions, and inaccurate risk classification and evaluation; it is beneficial for the airport to accurately quantify the collision risks when undertaking most flight operation tasks, highlight the risk regions, and provide a reference for the daily management of the apron. Brief Description of the Drawings

[0035] Figure 1 is the flow chart of the present invention;

[0036] Figure 2 is the schematic diagram of the collision risk quantification of the example;

[0037] Figure 3 is the simplified schematic diagram of the Petri net collision risk classification and evaluation of the example. Detailed Embodiment

[0038] The present invention will be further described in detail below with reference to the drawings.

[0039] The present invention provides a method for quantitatively evaluating the collision risk of aircraft operations at hub airports, including establishing an apron operation Petri net, calculating the collision risk of taxiing aircraft at different places at different times on the apron, and using the XGBoost algorithm to classify and evaluate the collision risk. The present invention is applicable to accurately quantitatively classifying and evaluating the collision risk generated by operating aircraft at hub airports, and can highlight the possible collision risk areas, with good practicability, providing a reference for the daily management of the apron. As Figure 1 shown, the specific steps are as follows:

[0040] Step 1: Obtain relevant flight information of the airport, which mainly includes flight departure and arrival times, flight in-and-out types, and corresponding docking positions, etc., providing a basis for subsequent simulations.

[0041] Through on-site investigations, online inquiries, etc., obtain the information of passenger flights operating at the airport within a week, mainly including flight departure and arrival times, flight in-and-out types, corresponding docking positions, aircraft taxiing speeds, and other main information on aircraft surface operations. And integrate and analyze the obtained flight data, and select the flight operation information during the peak hour of the peak day as the support for simulation data.

[0042] Step 2: Obtain relevant information such as the distribution of taxiways at the airport, and establish a corresponding apron operation Petri net according to its distribution.

[0043] The detailed information obtained mainly includes the specific location distribution of the airport taxiways, the operation conditions of the airport taxiway areas (whether there are abnormal conditions such as maintenance), and the length conditions of the airport taxiway areas.

[0044] Establish an airport operation Petri net for the airport apron operation network. The principle is: discretely divide the surface activity area into multiple taxiway sub-areas as places p i , the area length is such that only one aircraft is allowed to operate in it, the operation state of the surface flight corresponds to the state flag m, and the transition t i is the event that the flight enters the next taxiway sub-area p i from the taxiway sub-area p i+1 . Thus, a Petri net in the following form is established:

[0045] N = {P, T, Pre, Post, m, K}

[0046] Among them, the place set P represents the sub-areas where the aircraft operates; the transition set T represents the boundary between the aircraft operation sub-areas and the next one; Pre and Post represent the forward (backward) incidence matrices of P and T; m is the surface state flag; the mapping K controls the aircraft that cannot enter the activity area.

[0047] Step 3: Establish an aircraft trajectory prediction model for the aircrafts operating on the scene, so as to obtain the probabilities of the aircraft arriving at different positions at different times, and calculate the probabilities of different places in the Petri net.

[0048] During the taxiing process of the aircraft, it is restricted by various factors such as time, speed, and position. It cannot taxi smoothly at a stable speed, which results in a deviation between its actual position and the expected position (the expected position is the position reached at each moment when operating at the average speed). However, in order to ensure arriving on time, the aircraft always keeps this deviation within an interval, and this deviation interval is called the aircraft reachable interval.

[0049] First, it is necessary to calculate the potential movement space of the taxiing aircraft. In the state of uniform motion, the position space that an object can reach at any moment is represented by a circle, and the radius of the circle is the product of its maximum operating speed and the corresponding time. Therefore, under the condition that the starting and ending points are determined, there are two corresponding reachable areas, one is the starting point reachable area, and the other is the ending point reachable area. The two areas will change with time, and the intersection of the areas determines the calculation of the upper and lower bounds of the aircraft at this moment:

[0050]

[0051] Among them, t i is the moment when the aircraft starts to taxi from the starting point, t j is the moment when the aircraft is expected to taxi to the end point, D ij is the distance from the starting point to the end point of the aircraft, is the average operating speed (expected speed) of the aircraft, V m is the maximum taxiing speed of the aircraft. U x U(t) is the upper bound of the operating position of the aircraft at time t, and L x L(t) is the lower bound of the operating position of the aircraft at time t. Thus, the movement of the aircraft during taxiing can be regarded as a Brownian motion with boundaries. The standard Brownian motion formula without boundaries is:

[0052]

[0053] Among them, x i and x j are the coordinates of the relative positions of the starting and ending points respectively. When considering the upper and lower bounds, the position probability distribution of the aircraft within the upper and lower bound intervals still satisfies the corresponding formula, but the position probability distribution outside the interval is regarded as 0. In the Petri network, calculate the path of the aircraft operation and the places passed through. By using the calculated upper and lower bounds of the aircraft position at different times, the probabilities of the aircraft possibly reaching different positions on the taxiing route can be obtained. The calculation formula is:

[0054]

[0055]

[0056] Among them, h(x) is the corrected probability density function, f(x) is the probability density function of the standard boundaryless Brownian motion, F(x) is the cumulative distribution function, and σ′ 2 is the corrected variance. Thus, the probability that the aircraft inside the interval reaches any position can be obtained.

[0057] In the Petri net, assume that the defined length of a certain place (p i ) is k meters. The sum of the arrival probabilities corresponding to different positions in the place (accumulated by meter) is used as the probability that the aircraft reaches this place at the current moment:

[0058]

[0059] Among them, h(x h ) is the arrival probability of the aircraft at the position x k in the place p h .

[0060] Step 4: Calculate the corresponding collision risk generated by the place for the arrival probability of the aircraft at different moments.

[0061] Arrange the arrival probabilities of each aircraft at different places in different time periods vertically in sequence and horizontally in chronological order to obtain the arrival probability matrix of different aircraft. Assume there are m aircraft, and perform a product operation on the arrival probabilities (non-zero) of all aircraft corresponding to the place p k to obtain the collision risk of this place at the current moment:

[0062]

[0063] Among them, n represents the number of aircraft with non-zero arrival probability for the place k on the field. According to the above formula, the collision risks of different places in the Petri net at different moments can be calculated, as shown in Figure 2 .

[0064] After that, classification can be carried out according to the value of the collision risk. The risk is divided into 6 levels: if the risk value is 0, the risk level is 0, and no warning is required; if the risk value is 0 - 0.01, the risk level is 1, and a mild warning is given; if the risk value is 0.01 - 0.1, the risk level is 2, and a moderate warning is given; if the risk value is 0.1 - 0.3, the risk level is 3, and a high warning is given; if the risk value is 0.3 - 0.5, the risk level is 4, and a severe warning is given; if the risk value is above 0.5, the risk level is 5, and an emergency warning is given.

[0065] Step 5: Classify and evaluate the collision risks for different regions, and use the XGBoost algorithm for classification verification.

[0066] The XGBoost algorithm is an algorithm based on the accumulation of multiple decision trees. In each iterative training, the decision tree is obtained by minimizing an objective function:

[0067]

[0068] where (x j , y j ) are the observed values; l is the loss function; is the estimated value calculated in the (k - 1)-th iteration; Ω(f k (x)) is the penalty term, which penalizes the complexity of the tree; is to sum the results of each tree; obtain the regional collision risks and corresponding classification levels at different times. Take different regions as the x vector, the regional collision risk values as the corresponding values, and the comprehensive classification level as the y vector. Use the XGBoost algorithm for training and classification, and take the classification accuracy as the accuracy of this classification method.

[0069] After classifying and evaluating the collision risks for different regions, the color mapping method can be used to perform hierarchical mapping on the levels of different regions to highlight the differences in risk levels between apron regions, as Figure 3 shown. This processing effect can be better visualized, making the risk areas more prominent.

[0070] The above-disclosed are only the preferred embodiments of the present invention, and the scope of the rights of the present invention cannot be limited thereby. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.

Claims

1. A method for quantifying and evaluating the collision risk of aircraft operations at a hub airport, characterized in that The following steps are involved: (1) Obtain airport flight operation information within a certain period of time, integrate and analyze the data, and select flight operation information during peak hours on peak days; (2) Obtain the airport apron operation network diagram and establish the airport operation Petri net for the airport apron operation network; (3) For taxiing aircraft on the airport surface, predict the probability of the aircraft arriving at different locations and convert it into the probability of different locations in the Petri net; (4) Calculate the collision risk of different aircraft in different areas using the arrival probability of different aircraft at corresponding locations; (5) Using the XGBoost algorithm to classify collision risks, thus generating an evaluation of the collision risk of airport operations; The implementation process of step (3) is as follows: Based on the aircraft's operational status, the potential motion space of the taxiing aircraft at different times is calculated. The taxiing motion of the aircraft from the starting point to the end point is regarded as Brownian motion to calculate the probability of the aircraft arriving at different locations. To ensure that the aircraft arrives on time, the reachable range of the aircraft is calculated: Among them, L x (t) is the lower bound of the operating position of the aircraft at time t, U x (t) is the upper bound of the operating position of the aircraft at time t, t i is the time when the aircraft starts taxiing from the starting point, t j is the time when the aircraft is expected to taxi to the end point, D ij is the distance from the starting point to the end point of the aircraft, is the average operating speed of the aircraft, V m is the maximum taxiing speed of the aircraft; By calculating the upper and lower bounds of the aircraft's position at different times, we can obtain the probability that the aircraft may reach different positions along the taxi route. The calculation formula is: where h(x) is the corrected probability density function, f(x) is the probability density function of the standard boundaryless Brownian motion, F(x) is the cumulative distribution function, and σ′ 2 is the corrected variance; thus, the probability that the aircraft within the interval reaches any position is obtained. Assume a certain place p k with a defined length of m meters. The sum of the arrival probabilities corresponding to different positions within the place is used as the probability of the aircraft arriving at this place at the current moment: Among them, h(x h ) is the arrival probability of the aircraft at the position x k in the place p h .

2. The method for quantifying and evaluating collision risk of aircraft operations at a hub airport according to claim 1, characterized in that: The airport flight information in step (1) includes flight take-off and landing times, flight entry and exit types, corresponding parking spaces, and aircraft taxiing speeds.

3. The method for quantifying and evaluating collision risk of aircraft operations at a hub airport according to claim 1, characterized in that: The airport apron operation network diagram in step (2) includes the specific location distribution of the airport taxiways, the operation status of the airport taxiway area, and the length of the airport taxiway area.

4. A method for quantifying and evaluating the collision risk of aircraft operations at a hub airport according to claim 1, characterized in that, The implementation process of establishing the airport operation Petri net for the airport apron operation network in step (2) is as follows: The scene activity area is discretely divided into multiple taxiway sub-areas as places p i , the area length is such that only one aircraft is allowed to operate in it, and the operation state of the scene flight corresponds to the state identifier m, and the transition t i is the event that the flight enters from the taxiway sub-area p i to the next taxiway sub-area p i+1 ; A Petri net in the following form is established: N={P,T,Pre,Post,m,K} Among them, the place set P represents the sub-area of aircraft operation, the transition set T represents the boundary between the aircraft operation sub-area and the next one; Pre and Post represent the forward and backward association matrices of P and T; m is the surface state identifier; the mapping K controls the aircraft that cannot enter the activity area.

5. The method for quantifying and evaluating collision risk of aircraft operations at a hub airport according to claim 1, characterized in that: The step (4) is achieved by the following formula: Where n is the number of aircraft on the scene whose arrival probability at place k is not zero.

6. The method for quantifying and evaluating the collision risk of aircraft operations at a hub airport according to claim 1, wherein The process of implementing the collision risk classification using the XGBoost algorithm in step (5) is as follows: Among them, (x j , y j ) are the observed values; l is the loss function; is the estimated value calculated in the (k-1)-th iteration; Ω(f k (x)) is the penalty term, which penalizes the complexity of the tree; is to sum up the results of each tree; taking different regions as the x vector, the regional collision risk value as the corresponding value, and the comprehensive classification level as the y vector, and using the XGBoost algorithm for training and classification.

7. A method for quantifying and evaluating the collision risk of aircraft operations at a hub airport according to claim 1, characterized in that, The collision risk in step (5) is divided into six levels; If the risk value is 0, the risk level is 0 and no alarm is required; if the risk value is 0-0.01, the risk level is 1 and a slight alarm is issued; If the risk value is 0.01-0.1, the risk level is 2, a moderate warning; if the risk value is 0.1-0.3, the risk level is 3, a high warning; if the risk value is 0.3-0.5, the risk level is 4, a serious warning; if the risk value is above 0.5, the risk level is 5, an emergency warning.

Images