Method for estimating the detachment position of an aircraft wreckage based on a double-loop optimization model

Abstract

The application discloses an aircraft wreckage separation position estimation method based on a double-loop optimization model, which comprises the following steps: obtaining wreckage data of an aircraft wreckage and flight data of an aircraft corresponding to the aircraft wreckage; constructing a double-loop optimization model comprising a wreckage movement module and a likelihood function module; the wreckage movement module is used for constructing a wreckage movement equation for predicting a falling point based on a ballistic coefficient; the likelihood function module takes a probability density function of a predicted falling point output by the wreckage movement module and an actual falling point coordinate as a likelihood function; the double-loop optimization model sets a time step and a drag coefficient step, performs a likelihood value loop iteration, and outputs an optimal separation time and an optimal drag coefficient; and the wreckage position of the aircraft flight data at the optimal separation time is obtained as the wreckage separation position. The application realizes accurate estimation of the wreckage separation position and the separation time, and improves the aircraft accident investigation efficiency and the conclusion reliability.

Description

Technical Field

[0001] This invention relates to the field of aircraft accident investigation, and in particular to a method for estimating the detachment location of aircraft debris based on a dual cyclic optimization model. Background Technology

[0002] During flight, aircraft may experience airborne debris detachment due to factors such as structural failure, sudden aerodynamic load changes, external impacts, or disintegration. This debris scatters over a wide and dispersed area, posing significant challenges to accident investigation, debris search, and accident reconstruction. Current methods for estimating aircraft debris detachment locations generally face numerous technical difficulties. First, reverse engineering is extremely difficult: relying solely on the actual impact point of the debris to deduce the airborne detachment time and location is a typical technical challenge, as key parameters such as detachment time, drag coefficient, and initial attitude are unknown, making it difficult to uniquely determine the solution space with a single observation. Second, aerodynamic and environmental uncertainties are significant: the irregular shape and random attitude of debris make it difficult to accurately obtain the drag coefficient; high-altitude wind fields and air density dynamically change with altitude and time, leading to large errors in traditional fixed-parameter ballistic models and significant deviations in impact point prediction. Third, estimation accuracy and robustness are insufficient: existing methods often employ simplified motion equations or fixed drag models without joint optimization of detachment time and drag coefficient, making them susceptible to initial value deviations, measurement noise, and model errors, thus failing to achieve high-precision and stable three-dimensional detachment location inversion. Fourth, there is a lack of quantitative evaluation mechanisms: most methods only provide single-point estimation results and do not establish probability matching criteria between predicted and actual landing points. They cannot quantitatively determine the credibility of the estimation results and cannot meet the rigorous requirements of accident investigation.

[0003] These technical challenges make it difficult to accurately pinpoint the location and time of debris separation in mid-air using traditional techniques, severely hindering accident cause analysis, debris search and rescue efficiency, and the complete reconstruction of the accident process. The ability to deduce the location and altitude at which debris separated from the aircraft based on its weight and size information is crucial for reconstructing the accident and analyzing its causes. Summary of the Invention

[0004] The purpose of this invention is to provide a method for estimating the separation location of aircraft debris based on a dual cyclic optimization model. The dual cyclic optimization model accurately estimates the separation location and time of debris based on debris motion calculation and likelihood value calculation, providing key technical support for accident reconstruction.

[0005] The objective of this invention is achieved through the following technical solution: A method for estimating the separation location of aircraft debris based on a dual-cycle optimization model, the method comprising: S1. Obtain debris data of the aircraft, including area S, mass m, and actual impact point coordinates. ; Obtain flight data of the aircraft corresponding to the aircraft debris, including trajectory data, speed data, attitude data, air density data, and wind field data; S2. Construct a dual-loop optimization model comprising a fragment motion module and a likelihood function module. The fragment motion module constructs a fragment motion equation based on ballistic coefficients to predict the impact point. The likelihood function module uses the predicted impact point output by the fragment motion module. Coordinates of the actual landing point The probability density function is used as the likelihood function, and the parameters of the likelihood function are the initial state data. With drag coefficient ; S3. The dual-cycle optimization model selects a specific time recorded in the aircraft's flight data and assumes it to be the takeoff time. And collect detachment time Flight data as initial state data And set the time step Set the range of the drag coefficient and the drag coefficient step size. Iterative calculation of different detachment times Likelihood values ​​under different drag coefficients Output the takeoff time at the maximum likelihood value. The drag coefficient is used as the optimal time for disengagement. With the optimal drag coefficient Obtain the optimal takeoff time from the aircraft flight data. The location of the remaining fragments is used as the location where the wreckage detached.

[0006] To better implement this invention, in method S2, the ballistic coefficient includes the drag coefficient, the area of ​​the fragment, and the weight of the area. The dual-loop optimization model internally constructs a three-dimensional coordinate system with the north-south direction as the X-axis, the east-west direction as the Y-axis, and the vertical direction as the Z-axis. The fragment motion equation in the fragment motion module includes the following acceleration equation: , , ,in The acceleration in the X-axis direction is... The acceleration in the Y-axis direction, The acceleration in the Z-axis direction is... γ is the vacuum velocity in the flight data, and γ is the flight trajectory angle in the flight data. The heading angle in the flight data. The drag coefficient, The area of ​​the fragment in the fragment data. This is the acceleration due to gravity.

[0007] Preferably, the fragment motion module further includes position and velocity equations in a three-dimensional coordinate system. Both the position and velocity equations are based on acceleration, and the expression for the position equation is as follows: , , , , These represent the velocity components of the fragment along the X, Y, and Z axes, respectively. , , These are the position coordinates of the fragment along the X, Y, and Z axes, respectively. , , These are the initial position coordinates at the instant of departure; the velocity equation is expressed as follows: , , ,in These are the initial velocity components at the instant of departure from time. These are the acceleration components of the fragment in three directions, and t is the fragment's motion time from the separation time.

[0008] Preferably, the debris motion module selects a specific time recorded in the aircraft's flight data as the separation time. And collect detachment time Flight data as initial state data Output the predicted landing point according to the position equation. Initial state data Including the velocity components of the fragment in the X, Y, and Z axes, the initial position coordinates at the moment of separation, and the predicted impact point. This includes the position coordinates of the predicted location of the fragment along the X, Y, and Z axes.

[0009] Preferably, in method S3, the likelihood function module uses the predicted landing point coordinates output by the fragment motion module. Coordinates of the actual landing point The constructed likelihood function expression is as follows: , Let the likelihood value be... Let be the covariance of the propagation of the equation of motion of the fragment.

[0010] Preferably, the distribution of the predicted landing points by the fragment motion module conforms to a normal distribution, and the three-dimensional normal distribution probability density expression of the predicted landing point constraint is as follows: ,in To predict the landing point The probability density distribution, The mean of all predicted landing points. For covariance.

[0011] Preferably, the dual-loop optimization model includes an outer loop processing mechanism module and an inner loop processing mechanism module. The outer loop processing mechanism module uses the initial resistance coefficient and adjusts the steps according to the resistance coefficient. The process iterates through the drag coefficient range; the inner loop processing module calls the fragment motion equation and uses numerical integration methods such as the Runge-Kutta method until the fragment height drops to zero to obtain the predicted landing point. .

[0012] Preferably, the maximum likelihood value is obtained iteratively as follows: first, iteratively calculate and record different exit times. First, the likelihood value is calculated for different drag coefficients. Then, the local maximum likelihood value is calculated. If the current likelihood value obtained from the iterative calculation is greater than the local maximum likelihood value, then the current likelihood value is used as the local maximum likelihood value; otherwise... Then update to the global maximum likelihood value, and output the optimal escape time under the global maximum likelihood value. With the optimal drag coefficient .

[0013] Preferably, in method S3, the optimal takeoff time is obtained from the aircraft flight data. The optimal escape time is calculated based on the aircraft's position and the location of the debris within the aircraft. The location of the remaining fragments is used as the location where the wreckage was separated.

[0014] Compared with the prior art, the present invention has the following advantages and beneficial effects: (1) The dual-loop optimization model of this invention realizes the joint optimal inversion optimization process of the departure time and the drag coefficient. The outer loop traverses the drag coefficient, the inner loop traverses the candidate departure time, and combines the maximum likelihood estimation for global optimization, and outputs the optimal departure time synchronously. With the optimal drag coefficient This enables precise estimation of the location and timing of fragment detachment, providing crucial technical support for accident reconstruction.

[0015] (2) This invention constructs a debris motion module that couples air resistance, gravity, flight attitude and wind field. Based on aerodynamics and particle ballistics theory, it performs high-precision prediction of three-dimensional debris motion trajectory, accurately decomposes parameters such as position and velocity in the north-south, east-west and vertical directions, and fully describes the motion law of the debris from separation to landing. It adopts a three-dimensional normal distribution likelihood function to match the probability density of the predicted landing point with the actual landing point, and gives a quantitative likelihood value for each set of parameters, so as to achieve an objective evaluation of the reliability of the estimation results.

[0016] (3) When applied to accident investigation engineering, this invention provides key technical support for centralized search of debris, time sequence analysis of aerial disintegration, and reconstruction of accident chain, significantly improving the efficiency and reliability of aircraft accident investigations. By integrating multi-source data such as wind field data, air density data, and flight data, it can be extended to meet the debris location needs of civil aircraft, general aviation aircraft, and other scenarios. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method for estimating the location of aircraft debris detachment according to the present invention; Figure 2 This is a schematic diagram illustrating the force decomposition of an example fragment in mid-air, as shown in the embodiment. Figure 3 This is a geometric diagram illustrating the relationship between resistance decomposition and direction of motion in an example embodiment. Figure 4 This is a schematic diagram illustrating the principle of the outer and inner loops of the dual-loop optimization model in the embodiment. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to embodiments: Example like Figure 1 As shown, the method for estimating the detachment location of aircraft debris based on a dual-cycle optimization model includes the following steps: S1. Obtain debris data of the aircraft, including area S, mass m, and actual impact point coordinates. Obtain flight data of the aircraft corresponding to the aircraft debris. Flight data includes trajectory data, speed data (including vacuum speed, etc.), attitude data (including flight trajectory angles, heading, etc.), and air density data (including air density). Data such as wind direction and wind speed.

[0019] S2. Construct a dual-loop optimization model comprising a fragment motion module and a likelihood function module. The fragment motion module constructs the fragment motion equations based on ballistic coefficients to predict the impact point. The likelihood function module uses the predicted impact point output by the fragment motion module. Coordinates of the actual landing point The probability density function is used as the likelihood function, and the parameters of the likelihood function are the initial state data. With drag coefficient In some embodiments, the ballistic coefficient includes the drag coefficient. The area S of the fragment, the weight of that area m, and the air drag coefficient. (abbreviated as drag coefficient) The value can be obtained through wind tunnel testing, or a range of values ​​can be set and the values ​​can be taken in actual situations.

[0020] The dual-loop optimization model internally constructs a three-dimensional coordinate system with the north-south direction as the X-axis (north of the X-axis is positive), the east-west direction as the Y-axis (east of the Y-axis is positive), and the direction perpendicular to the ground as the Z-axis (Z-axis is positive when facing the ground). The fragment motion module includes the following mechanical equations: ; ; ; in Vacuum speed, vacuum ; The drag coefficient, This refers to the cross-sectional area (i.e., the area of ​​the fragment in the fragment data). The flight trajectory angle. That is, the angle between the velocity vector and the horizontal plane; For heading angle, heading angle That is, the angle between the projection of the velocity vector onto the horizontal plane and the X-axis (north direction); For the quality of the fragment, This refers to the density of air. For example... Figure 2 As shown, Figure 2 This diagram illustrates the force decomposition of the debris in mid-air, with gravity W... t Along the positive Z-axis (vertically downward), the air resistance F is opposite to the direction of the debris's motion, and its magnitude is determined by the air density ρ and the vacuum velocity V. true Drag coefficient C D The drag vector is determined by the reference area S; the projections of the drag vector onto the X, Y, and Z axes are determined by the heading angle Ψ and the flight trajectory angle γ, respectively. Figure 3 The geometric relationship between resistance decomposition and motion direction is further explained. The X and Y axes are the horizontal and vertical axes, respectively. γ is set as the angle between the velocity vector and the horizontal plane, and Ψ is set as the angle between the projection of the velocity vector on the horizontal plane and the X axis (north direction).

[0021] The equations of motion for the fragments in the fragment motion module include the following acceleration equations (considering the combined effects of air resistance and gravity): ; ; ; in The acceleration in the X-axis direction is... The acceleration in the Y-axis direction, The acceleration in the Z-axis direction is... γ is the vacuum velocity in the flight data, and γ is the flight trajectory angle in the flight data. The heading angle in the flight data. The drag coefficient, The area of ​​the fragment in the fragment data. It is the acceleration due to gravity; The ballistic coefficient is a key parameter (reflecting the relationship between the aerodynamic characteristics and weight of the debris).

[0022] The fragment motion module of this invention also includes position and velocity equations in a three-dimensional coordinate system. Both the position and velocity equations are based on acceleration. The expression for the position equation is as follows: ; ; ; in , These represent the velocity components of the fragment along the X, Y, and Z axes, respectively. , , These are the position coordinates of the fragment along the X, Y, and Z axes, respectively. , , These are the initial position coordinates at the instant of departure. The velocity equation is expressed as follows: ; ; ; in These are the initial velocity components at the instant of departure from time. These are the acceleration components of the fragment in the three directions, and t is the fragment's motion time starting from the separation time (i.e., the motion time starting from the separation moment).

[0023] In some embodiments, the debris motion module of the present invention selects a certain time recorded in the aircraft's flight data as the separation time. And collect detachment time The flight data (i.e., the position and velocity at the moment of debris separation) is used to extract the position (S) at that moment from the flight trajectory. X0 ,S Y0 ,S Z0 ) and velocity (V X0 V Y0 V Z0 )) as initial state data Output the predicted landing point according to the position equation. Initial state data Including the velocity components of the fragment in the X, Y, and Z axes, the initial position coordinates at the moment of separation, and the predicted impact point. This includes the position coordinates of the predicted location of the fragment along the X, Y, and Z axes. Preferably, the distribution of the predicted landing points by the fragment motion module of the present invention conforms to a normal distribution (the predicted landing point may be at the actual landing point s). obs Nearby, which aligns with the actual landing point s obs The probability density expression of the three-dimensional normal distribution for predicting the landing point constraint is as follows: (The distribution is centered on a normal distribution). ,in To predict the landing point The probability density distribution, The mean of all predicted landing points. For covariance.

[0024] S3. The dual-cycle optimization model selects a specific time recorded in the aircraft's flight data and assumes it to be the takeoff time. And collect detachment time Flight data as initial state data And set the time step Set the range of the drag coefficient and the drag coefficient step size. Iterative calculation of different detachment times Likelihood values ​​under different drag coefficients Preferably, the likelihood function module uses the predicted landing point coordinates output by the fragment motion module. Coordinates of the actual landing point The constructed likelihood function expression is as follows: , Let the likelihood value be... Let be the covariance of the propagation of the debris's motion equations. During the debris's motion in the air, the uncertainty of the initial state gradually accumulates and amplifies over time; in each calculation step, the uncertainty of the previous moment... The error at time k propagates through the equations of motion to the current time (affected by the coefficients of the equations of motion, for example, it is amplified proportionally), resulting in uncertainty. (i.e., the error at time k+1); at the same time, the new process noise generated in this step Adding the two together gives the total uncertainty at the current moment. Assuming a given drag coefficient CD and a takeoff time t0, the dual-cycle optimization model at takeoff time... drag coefficient (Its corresponding actual landing point coordinates) The landing point (and its corresponding actual landing point coordinates) The predicted likelihood value The expression is as follows: ; in In order to break away from time drag coefficient The predicted landing point coordinates are below. These are the actual landing point coordinates. In order to break away from time drag coefficient The variance below.

[0025] like Figure 4 As shown, the dual-loop optimization model of this invention includes an outer loop processing mechanism module and an inner loop processing mechanism module. The outer loop processing mechanism module uses the initial resistance coefficient according to the resistance coefficient step size. The process iterates through the drag coefficient range. The inner loop processing module calls the fragment's motion equations and uses numerical integration methods such as the Runge-Kutta method until the fragment's height drops to zero (i.e., it hits the ground, S). Z =0) to obtain the predicted landing point The dual-loop optimization model utilizes an outer loop processing mechanism module and an inner loop processing mechanism module for collaborative iterative computation.

[0026] The off-time of the dual-loop optimization model of this invention outputs the maximum likelihood value. The drag coefficient is used as the optimal time for disengagement. With the optimal drag coefficient Preferably, the maximum likelihood value is obtained iteratively as follows: first, iteratively calculate and record different exit times. First, the likelihood value is calculated for different drag coefficients. Then, the local maximum likelihood value is calculated. If the current likelihood value obtained from the iterative calculation is greater than the local maximum likelihood value, then the current likelihood value is used as the local maximum likelihood value; otherwise... Then update to the global maximum likelihood value, and output the optimal escape time under the global maximum likelihood value. With the optimal drag coefficient .

[0027] Obtain the optimal takeoff time from the aircraft flight data The location of the debris is taken as the wreckage separation location. Preferably, the optimal separation time is obtained from the aircraft flight data. The aircraft position corresponds to the position (S) X0 ,S Y0 ,S Z0 The optimal separation time was calculated based on the location of the debris within the aircraft. The location of the debris (for example, if the debris happens to be the location of the corresponding sensor in the aircraft's flight data, then the debris location is the flight data recorded by the aircraft. For example, the relative distance L between the debris and the recorded location in the aircraft's flight data can be used to obtain the specific location data of the debris. Of course, the relative distance L can also be defined as the distance from the debris to the center of gravity of the aircraft. For example, if the debris is a winglet, then the relative distance L is the distance from the winglet to the center of gravity of the aircraft, which is approximately equal to the wingspan) and is used as the location of the debris separation (i.e., the location is calculated based on the location of the debris on the aircraft).

[0028] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for estimating the separation location of aircraft debris based on a dual-cycle optimization model, characterized in that: The methods include: S1. Obtain debris data of the aircraft, including area S, mass m, and actual impact point coordinates. ; Obtain flight data of the aircraft corresponding to the aircraft debris, including trajectory data, speed data, attitude data, air density data, and wind field data; S2. Construct a dual-loop optimization model comprising a fragment motion module and a likelihood function module. The fragment motion module constructs a fragment motion equation based on ballistic coefficients to predict the impact point. The likelihood function module uses the predicted impact point output by the fragment motion module. Coordinates of the actual landing point The probability density function is used as the likelihood function, and the parameters of the likelihood function are the initial state data. With drag coefficient ; S3. The dual-cycle optimization model selects a specific time recorded in the aircraft's flight data and assumes it to be the takeoff time. And collect detachment time Flight data as initial state data And set the time step Set the range of the drag coefficient and the drag coefficient step size. Iterative calculation of different detachment times Likelihood values ​​under different drag coefficients Output the takeoff time at the maximum likelihood value. The drag coefficient is used as the optimal time for disengagement. With the optimal drag coefficient Obtain the optimal takeoff time from the aircraft flight data. The location of the remaining fragments is used as the location where the wreckage detached.

2. The method for estimating the wreckage separation location based on a dual-cycle optimization model according to claim 1, characterized in that: In method S2, the ballistic coefficient includes the drag coefficient, the area of ​​the fragment, and the weight of the area. The dual-loop optimization model internally constructs a three-dimensional coordinate system with the north-south direction as the X-axis, the east-west direction as the Y-axis, and the vertical direction as the Z-axis. The fragment motion equation in the fragment motion module includes the following acceleration equation: , , ,in The acceleration in the X-axis direction is... The acceleration in the Y-axis direction, The acceleration in the Z-axis direction is... γ is the vacuum velocity in the flight data, and γ is the flight trajectory angle in the flight data. The heading angle in the flight data. The drag coefficient, The area of ​​the fragment in the fragment data. This is the acceleration due to gravity.

3. The method for estimating the wreckage separation location based on a dual-cycle optimization model according to claim 2, characterized in that: The fragment motion module also includes position and velocity equations in a three-dimensional coordinate system. Both the position and velocity equations are based on acceleration. The expression for the position equation is as follows: , , , , These represent the velocity components of the fragment along the X, Y, and Z axes, respectively. , , These are the position coordinates of the fragment along the X, Y, and Z axes, respectively. , , These are the initial position coordinates at the instant of departure; the velocity equation is expressed as follows: , , ,in These are the initial velocity components at the instant of departure from time. These are the acceleration components of the fragment in three directions, and t is the fragment's motion time from the separation time.

4. The method for estimating the wreckage separation location based on a dual-cycle optimization model according to claim 3, characterized in that: The debris motion module selects a specific time recorded in the aircraft's flight data and assumes it to be the separation time. And collect detachment time Flight data as initial state data Output the predicted landing point according to the position equation. Initial state data Including the velocity components of the fragment in the X, Y, and Z axes, the initial position coordinates at the moment of separation, and the predicted impact point. This includes the position coordinates of the predicted location of the fragment along the X, Y, and Z axes.

5. The method for estimating the wreckage separation location based on a dual-cycle optimization model according to claim 1, characterized in that: In method S3, the likelihood function module uses the predicted landing point coordinates output by the fragment motion module. Coordinates of the actual landing point The constructed likelihood function expression is as follows: , Let the likelihood value be... Let be the covariance of the propagation of the equation of motion of the fragment.

6. The method for estimating the separation location of aircraft debris based on a dual-cycle optimization model according to claim 1, characterized in that: The predicted impact point distribution of the fragment motion module conforms to a normal distribution, and the three-dimensional normal distribution probability density expression of the predicted impact point constraint is as follows: ,in To predict the landing point The probability density distribution, The mean of all predicted landing points. For covariance.

7. The method for estimating the wreckage separation location based on a dual-cycle optimization model according to claim 1, characterized in that: The dual-loop optimization model includes an outer loop processing mechanism module and an inner loop processing mechanism module. The outer loop processing mechanism module uses the initial resistance coefficient and adjusts the step size according to the resistance coefficient. The process iterates through the drag coefficient range; the inner loop processing module calls the fragment motion equation and uses numerical integration methods such as the Runge-Kutta method until the fragment height drops to zero to obtain the predicted landing point. .

8. The method for estimating the wreckage separation location based on a dual cyclic optimization model according to claim 1 or 7, characterized in that: The maximum likelihood value is obtained iteratively as follows: First, calculate and record different exit times iteratively. First, the likelihood value is calculated for different drag coefficients. Then, the local maximum likelihood value is calculated. If the current likelihood value obtained from the iterative calculation is greater than the local maximum likelihood value, then the current likelihood value is used as the local maximum likelihood value; otherwise... Then update to the global maximum likelihood value, and output the optimal escape time under the global maximum likelihood value. With the optimal drag coefficient .

9. The method for estimating the wreckage separation location based on a dual-cycle optimization model according to claim 1, characterized in that: In method S3, the optimal takeoff time is obtained from the aircraft flight data. The optimal escape time is calculated based on the aircraft's position and the location of the debris within the aircraft. The location of the remaining fragments is used as the location where the wreckage was separated.

Images