A Markov chain-based method for compiling probabilistic flight load spectra
Through the Markov chain-based probabilistic flight load spectrum preparation method, the problem of failure to consider fleet flight training correlation and future flight load uncertainty in the single-aircraft fatigue life prediction of aircraft structures is solved, and accurate load prediction for future rise and fall and missing rise and fall is achieved, supporting structural probability fatigue damage assessment and residual life prediction.
Patent Information
- Application Number
- CN202211713405.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-12-29
- Publication Date
- 2025-08-19
- Estimated Expiration
- 2042-12-29
AI Technical Summary
The prior art fails to effectively consider the correlation of fleet flight training and the uncertainty of future flight loads in the prediction of single-air fatigue life of aircraft structures, resulting in residual life prediction deviations and safety hazards. Especially in the absence of take-off and landing data or unknown mission-type, structural fatigue damage cannot be accurately evaluated.
The probability flight load spectrum composition method based on Markov chain is adopted, and the probability flight load prediction of future rise and fall and missing rise and fall are achieved through classification and statistical mission-type take-off and landing ratio, extraction and identification of maneuvering flight parametric process, maneuver transfer probability matrix construction, random arrangement of maneuvering actions and take-off and landing sequence, combined with the flight parametric-load model, the probability flight load prediction of future rise and fall and missing rise and fall are achieved.
The probability flight load prediction of future rise and fall and missing rise and fall is achieved, supporting structural probability fatigue damage assessment and residual life prediction, reducing the deviation of residual life prediction and improving safety.
Smart Images
Figure CN115935130B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of aircraft load spectrum compilation, and in particular relates to a probabilistic flight load spectrum compilation method based on a Markov chain. Background Art
[0002] Single-machine tracking of aircraft structures (or single-machine life monitoring) requires obtaining the single-machine load spectrum based on actual field flight data and evaluating the fatigue life consumption of each key structure. In order to further predict the remaining life of the structure, it is also necessary to probabilistically predict and compile future flight load spectra.
[0003] A simple approach currently used in China assumes that future flight loads will be comparable to current flight loads. This means that the fatigue damage accumulation rate is constant and fixed. This allows predictions of the remaining life when damage reaches a threshold or crack size reaches a critical value based on current damage characterization parameters. This approach, on the one hand, considers only the historical flight history of a single aircraft and ignores the correlation between flight training and overall fleet performance. This is clearly unrealistic, especially for aircraft with relatively few flight loads, which can lead to significant deviations in the remaining life estimate. Furthermore, due to the uncertainty of future usage, treating future flight loads as deterministic variables may pose safety risks.
[0004] At the same time, there are cases in field use where some takeoff and landing data are missing and the mission type (also called mission profile or subject) is unknown. If it is believed that the flight severity of the missing takeoff and landing is equivalent to that of other takeoffs and landings, this may not be the true situation. The probabilistic method should also be used to predict the flight load of the missing takeoff and landing. Summary of the Invention
[0005] The purpose of this invention is to propose a probabilistic flight load spectrum compilation method based on Markov chains, which can realize the probabilistic flight load prediction of future takeoffs and landings and missing takeoffs and landings, thereby supporting structural probabilistic fatigue damage assessment, probabilistic remaining life prediction and fracture risk analysis.
[0006] The technical solution of the present invention:
[0007] A Markov chain-based probabilistic flight load spectrum compilation method comprises the following steps:
[0008] Step 1: Based on the actual flight takeoffs and landings in the field and the mission type information of each takeoff and landing, classify and count each mission type to obtain the takeoff and landing ratio of each mission type in the fleet;
[0009] Step 2: Based on the actual field flight data and the maneuver feature library, the flight parameter history of all takeoffs and landings is extracted and maneuver action recognition is performed. The number of maneuvers and the maneuver sequence for each mission type are obtained by combining the mission type information of each takeoff and landing.
[0010] Step 3: Based on the maneuver sequence of each task type, use the Markov chain model to obtain the maneuver transition probability matrix of each task type;
[0011] Step 4: Based on the number of single flight maneuvers of each mission type and the maneuver transfer probability matrix of each mission type, the maneuvers of each mission type are randomly arranged using the probability sampling method; combined with the extracted maneuver flight parameter history, the probability flight parameter history of each mission type single flight is constructed;
[0012] Step 5: For future takeoffs and landings or missing takeoffs and landings, combine the takeoff and landing ratios of each mission type and randomly arrange each mission type to obtain the takeoff and landing order of each mission type;
[0013] Step 6: Based on the probabilistic flight parameter history of each mission type and the take-off and landing sequence of each mission type, a complete probabilistic flight parameter history is formed, and the complete probabilistic flight parameter history is substituted into the flight parameter-load model to obtain the probabilistic flight load spectrum of future take-off and landing or missing take-off and landing.
[0014] Furthermore, in step 1, the number of takeoffs and landings of each mission type is m i , then the take-off and landing ratio of each mission type is p i =m i / Σm i , i=1:n, n is the number of task categories.
[0015] Furthermore, in step 2, the maneuver feature library refers to a feature parameter template library of various maneuver actions, and the feature parameters include flight altitude, Mach number, center of gravity normal overload, roll angle, pitch angle, heading angle, roll angular velocity, pitch angular velocity, and heading angular velocity;
[0016] Maneuvers include: turns / circles, pitches, diving turns, leaping turns, half-rolls, loops, half-loops, rolls, and ground attacks.
[0017] Furthermore, in step 2, the method for extracting the maneuver flight parameter history is: at the peak point of the center of gravity normal overload, search forward and backward along the time axis for the first time when the center of gravity normal overload is 1g±0.25g, the roll angle is 0±20°, and the pitch angular velocity and roll angular velocity are 0±10° / second, as the start and end times of the maneuver.
[0018] Furthermore, in step 2, maneuvering action recognition refers to using a clustering algorithm to compare and match the extracted maneuvering flight parameter history with the maneuvering feature library to identify the maneuvering action.
[0019] Furthermore, in step 3, the maneuver transfer probability matrix of each task type is obtained through the following process:
[0020] Assuming the number of maneuver categories is k, the dimension of the maneuver transition probability matrix is k×k, and the sum of the probabilities in each row is 1.0; the form is as follows:
[0021]
[0022] Among them, p ij represents the probability of transferring from maneuver i to maneuver j. Maneuver transfers need to be counted and converted into probabilities based on the actual maneuver sequence of each task type. The sum of the elements in each row of the matrix is 1. The values of i and j range from 1 to k.
[0023] Furthermore, in step 4, the random arrangement process of the maneuvers of each task type is performed using a probability sampling method as follows:
[0024] For a specific mission type, the initial maneuver is arbitrarily selected, and the next maneuver is randomly drawn according to the data of the row where the initial maneuver is located in the maneuver transfer probability matrix. Then the next maneuver is randomly drawn according to the data of the row where the newly drawn maneuver is located in the maneuver transfer probability matrix, and so on, until the number of maneuvers reaches the number of maneuvers of a single flight of the mission type.
[0025] Furthermore, in step 5, the probability random arrangement process of each task type is as follows: let the number of future take-offs and landings or the number of missing take-offs and landings be n, and the take-off and landing ratio of each task type be p i , then the number of take-offs and landings of each mission type is n i =n×p i ; Then use the "mixed multiplication and congruential method" to randomly arrange each task type.
[0026] Beneficial effects of the present invention:
[0027] The present invention proposes a Markov chain-based probabilistic flight load spectrum compilation method, which can realize the probabilistic flight load prediction of future takeoffs and landings and missing takeoffs and landings, thereby supporting structural probabilistic fatigue damage assessment, probabilistic remaining life prediction and fracture risk analysis. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 A flow chart of the method for compiling a probabilistic flight load spectrum based on a Markov chain;
[0029] Figure 2 This is a schematic diagram of the maneuver transfer probability matrix for mission type 1#;
[0030] Figure 3 The diagram below compares the probability overload spectrum exceedance curve for the fleet's future 500 takeoffs and landings with the overload exceedance curve for the No. 1 aircraft that has already flown. DETAILED DESCRIPTION
[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.
[0032] The core of the present invention is to use the Markov chain model to construct the transfer probability matrix of different maneuvers for each mission type, and to make a probability prediction of the flight load of future take-offs and landings and missing take-offs and landings based on the take-off and landing ratio of each mission type.
[0033] The solution of the present invention comprises the following steps:
[0034] S1: Based on the actual flight takeoffs and landings in the field and the mission type information of each takeoff and landing, classify and count each mission type to obtain the takeoff and landing ratio of each mission type in the fleet.
[0035] In a possible embodiment, in step S1, the mission type is also called mission profile, including stunt, instrument, formation, air combat, ground attack and other mission types.
[0036] In a possible embodiment, in step S1, the calculation method of the take-off and landing ratio of each mission type is as follows: Assuming that the number of take-offs and landings of each mission type is m i , then the take-off and landing ratio of each mission type is p i =m i / Σm i
[0037] S2: Based on actual field flight data and a maneuver feature library, we extract the flight parameters and identify the maneuvers for all takeoffs and landings. Combined with the mission information for each takeoff and landing, we calculate the number and sequence of maneuvers for each mission type in a single flight.
[0038] In one possible embodiment, in step S2, the maneuver signature library refers to a library of characteristic parameter templates for various maneuvers, including parameters such as flight altitude, Mach number, normal center of gravity overload, roll angle, pitch angle, heading angle, roll angular velocity, pitch angular velocity, and heading angular velocity. The parameter variation patterns of various maneuvers are generally obtained based on a large amount of flight simulation data and measured flight data. Maneuvers are derived from the military flight training syllabus and include turns / circles, pitches, dive turns, jump turns, half rolls, dogfights, dogfights, rolls, and ground attacks.
[0039] In one possible embodiment, in step S2, the maneuver history is extracted by searching forward and backward along the time axis from the peak point of the normal overload at the center of gravity to the time point at which the normal overload at the center of gravity is near 1g, and the aircraft state parameters such as the roll rate, pitch rate, and roll rate are all near zero. These points are used as the start and end times of the maneuver. That is, a complete maneuver starts and ends at a 1g level flight state.
[0040] In a possible embodiment, in step S2, maneuvering action recognition refers to using a clustering algorithm to compare and match the extracted maneuvering flight parameter history with a maneuvering feature library, thereby identifying the maneuvering action.
[0041] S3: Based on the maneuver sequence of each task type obtained in step S2, a Markov chain model is used to obtain a maneuver transition probability matrix for each task type;
[0042] In a possible embodiment, in step S3 , a Markov chain model is used to define the probability of transitioning from one maneuver to another, thereby constructing a maneuver transition probability matrix.
[0043] Assuming the number of maneuver categories is k, the dimension of the maneuver transition probability matrix is k×k, and the sum of the probabilities in each row is 1.0; the form is as follows:
[0044]
[0045] Among them, p ij represents the probability of transitioning from maneuver i to maneuver j. Maneuver transitions must be counted and converted to probabilities based on the actual maneuver sequence of each mission type. The sum of the elements in each row of the matrix is 1. The values of i and j range from 1 to k. The i-th row of the matrix represents the probability of transitioning from maneuver i to another maneuver. Maneuver transitions must be counted and converted to probabilities based on the actual maneuver sequence of each mission type.
[0046] For example, assuming that the maneuver identifiers are A, B, and C, and the actual maneuver sequence for a specific task type is: AABCBABCBACACAB, then the first row of the transition probability matrix is [1 / 6, 1 / 2, 2 / 6], because the number of transitions from A to A is 1, the number of transitions from A to B is 3, and the number of transitions from A to C is 2; the second row of the transition probability matrix is [1 / 3, 0, 2 / 3], because the number of transitions from B to A is 1, the number of transitions from B to B is 0, and the number of transitions from B to C is 2; the third row of the transition probability matrix is [1 / 2, 1 / 2, 0], because the number of transitions from C to A is 2, the number of transitions from C to B is 2, and the number of transitions from C to C is 0.
[0047] S4: Based on the number of maneuvers for each mission type single flight obtained in step S2 and the maneuver transfer probability matrix for each mission type obtained in step S3, a probability sampling method is used to randomly arrange the maneuvers for each mission type, and the maneuvers are combined with the flight parameter history of each maneuver obtained in step S2 to form a probabilistic flight parameter history for each mission type single flight;
[0048] In one possible embodiment, in step S4, a probability sampling method is used to probabilistically arrange the maneuvers of each task type. The specific method is as follows: for a specific task type, an initial maneuver is arbitrarily selected, and the next maneuver is randomly selected based on the row of the transition probability matrix corresponding to the initial maneuver. Then, the next maneuver is randomly selected based on the row of the transition probability matrix corresponding to the maneuver, and so on, until the number of maneuvers reaches the number of maneuvers of a single flight of the task type.
[0049] S5: Based on the future take-off and landing number or the missing take-off and landing number, and the take-off and landing ratio of each task type obtained in step S1, randomly arrange each task type by probability to form a take-off and landing sequence for each task type;
[0050] In a possible embodiment, in step S5, each mission type is randomly arranged by probability. The specific method is: assuming that the number of future take-offs and landings or the number of missing take-offs and landings is n, and the take-off and landing ratio of each mission type is p i , then the number of take-offs and landings of each mission type is n i =n×p i ,Then the “mixed multiplication and congruence method” is used to randomly arrange each task type.
[0051] S6: Based on the probabilistic flight parameter history of each mission type obtained in step S4 and the take-off and landing sequence of each mission type obtained in step S5, a complete probabilistic flight parameter history is combined. The flight parameter history is substituted into the "flight parameter-load" model to obtain the probabilistic flight load spectrum of future take-off and landing or missing take-off and landing.
[0052] In a possible embodiment, in step S6, the "flight parameter-load" model refers to a mapping model from flight parameters to component loads constructed using machine learning algorithms such as neural networks and multiple regression analysis.
[0053] Example 1
[0054] Taking a certain type of aircraft as an example, the flight load spectrum of the next 500 takeoffs and landings is predicted based on the actual flight data of a fleet.
[0055] 1) According to step S1, there is a fleet of 10 aircraft with a total of approximately 2000 takeoff and landing flight data and mission type information for each takeoff and landing. The mission types are classified into 5 categories, and the takeoff and landing ratios of each mission type are [0.1, 0.3, 0.3, 0.2, 0.1];
[0056] 2) According to step S2, based on the maneuver feature library (the aircraft has 100 types of maneuvers), maneuver history extraction and maneuver recognition are performed on the flight parameter history of all takeoffs and landings, and the number of maneuvers (approximately 20) and the maneuver sequence for each mission type single flight are counted.
[0057] 3) According to step S3, the Markov chain model is used to obtain the maneuver transfer probability matrix of each task type, such as Figure 2 , is the maneuver transfer probability matrix diagram of task type 1#. The darker the color, the greater the probability of maneuver transfer.
[0058] 4) According to step S4, the probability of each maneuver is arranged according to the maneuver transfer probability matrix of each mission type, and the probabilistic flight parameter history of each mission type single flight is combined;
[0059] 5) According to step S5, based on the take-off and landing ratio of each mission type [0.1, 0.3, 0.3, 0.2, 0.1], each mission type is randomly sampled probabilistically to form the mission type arrangement for the next 500 take-offs and landings;
[0060] 6) According to step S6, a complete probabilistic flight parameter history is assembled and substituted into the "flight parameter-load" model to obtain the probabilistic load spectrum for the next 500 flights.
[0061] Repeat steps 4) to 6) 20 times to obtain 20 future probabilistic overload spectra. Draw the exceedance curve and compare it with the actual overload exceedance curve of aircraft 1. Figure 3 .
[0062] The above description is merely a detailed description of specific embodiments of the present invention. Any unspecified portions are conventional techniques. However, the scope of the present invention is not limited thereto. Any changes or substitutions that can be readily conceived by a person skilled in the art within the technical scope disclosed in the present invention are intended to be encompassed within the scope of the present invention. The scope of the present invention shall be determined by the scope of the claims.
Claims
1. A Markov chain-based probabilistic flight load spectrum compilation method, characterized by: The method comprises the following steps: Step 1: Based on the actual flight takeoffs and landings in the field and the mission type information of each takeoff and landing, classify and count each mission type to obtain the takeoff and landing ratio of each mission type in the fleet; Step 2: Based on the actual field flight data and the maneuver feature library, the flight parameter history of all takeoffs and landings is extracted and maneuver action recognition is performed. The number of maneuvers and the maneuver sequence for each mission type are obtained by combining the mission type information of each takeoff and landing. Step 3: Based on the maneuver sequence of each task type, use the Markov chain model to obtain the maneuver transition probability matrix of each task type; Step 4: Based on the number of single flight maneuvers of each mission type and the maneuver transfer probability matrix of each mission type, the maneuvers of each mission type are randomly arranged using the probability sampling method; combined with the extracted maneuver flight parameter history, the probability flight parameter history of each mission type single flight is constructed; Step 5: For future takeoffs and landings or missing takeoffs and landings, combine the takeoff and landing ratios of each mission type and randomly arrange each mission type to obtain the takeoff and landing order of each mission type; Step 6: Based on the probabilistic flight parameter history of each mission type and the take-off and landing sequence of each mission type, a complete probabilistic flight parameter history is formed, and the complete probabilistic flight parameter history is substituted into the flight parameter-load model to obtain the probabilistic flight load spectrum of future take-off and landing or missing take-off and landing.
2. The method according to claim 1, wherein: In step 1, the number of takeoffs and landings of each mission type is m i , then the take-off and landing ratio of each mission type is p i =m i / Σm i , i=1:n, n is the number of task types.
3. The method according to claim 2, wherein: In step 2, the maneuver feature library refers to a feature parameter template library of various maneuver actions, and the feature parameters include flight altitude, Mach number, center of gravity normal overload, roll angle, pitch angle, heading angle, roll angular velocity, pitch angular velocity, and heading angular velocity; Maneuvers include: turning, pitching, diving turning, jumping turning, half-roll, bucket, half-buoy flip, roll, and ground attack.
4. The method according to claim 3, wherein: In the step 2, the method for extracting the maneuvering flight parameter history is: at the peak point of the center of gravity normal overload, search forward and backward along the time axis for the first time when the center of gravity normal overload is 1g±0.25g, the roll angle is 0±20°, and the pitch angular velocity and roll angular velocity are 0±10° / second, as the start time and end time of the maneuvering action.
5. The method according to claim 4, characterized in that: In the step 2, maneuvering action recognition refers to using a clustering algorithm to compare and match the extracted maneuvering flight parameter history with the maneuvering feature library, thereby identifying the maneuvering action.
6. The method according to claim 5, characterized in that: In step 3, the maneuver transfer probability matrix of each task type is obtained through the following process: Assuming the number of maneuver categories is k, the dimension of the maneuver transition probability matrix is k×k, and the sum of the probabilities in each row is 1.0; the form is as follows: Among them, p ij represents the probability of transferring from maneuver i to maneuver j. Maneuver transfers need to be counted and converted into probabilities based on the actual maneuver sequence of each task type. The sum of the elements in each row of the matrix is 1. The values of i and j range from 1 to k.
7. The method according to claim 6, characterized in that: In step 4, the random arrangement process of the maneuvers of each task type is performed using a probability sampling method as follows: For a specific mission type, the initial maneuver is arbitrarily selected, and the next maneuver is randomly drawn according to the data of the row where the initial maneuver is located in the maneuver transfer probability matrix. Then the next maneuver is randomly drawn according to the data of the row where the newly drawn maneuver is located in the maneuver transfer probability matrix, and so on, until the number of maneuvers reaches the number of maneuvers of a single flight of the mission type.
8. The method according to claim 7, wherein: In step 5, the probability random arrangement process of each task type is as follows: let the number of future take-offs and landings or the number of missing take-offs and landings be N, and the take-off and landing ratio of each task type be p i , then the number of take-offs and landings of each mission type is n i =N×p i ; Then use the "mixed multiplication and congruential method" to randomly arrange each task type.
Citation Information
Patent Citations
Compilation method for reliability test load spectrum of electric drive assembly mechanical system
CN111581893A
Method and device for predicting crack damage of train component
WO2019201176A1