Reactive wind shear alarm reliability evaluation method based on flight parameter data

Abstract

The invention belongs to the technical field of aircraft avionics airworthiness, and particularly discloses a reaction type wind shear alarm reliability evaluation method based on flight parameter data, which comprises the following steps: extracting flight wind shear operation data; calibrating a wind shear alarm event; calculating a wind shear alarm reliability evaluation index; building and training based on a VAE-LSTM model; and performing reactive wind shear alarm reliability evaluation by using the trained VAE-LSTM model. According to the method, the reliability of a wind shear alarm event is evaluated and calibrated by analyzing sample data, then the wind shear recognition reliability is converted into an abnormal detection problem of a time sequence index based on a VAE-ISTM model, and the energy change rate cumulative sum is selected as a judgment index; the invention provides an approach stage reactive wind shear alarm reliability evaluation method based on energy change rate accumulation, and provides technical support for avionics equipment manufacturers to optimize reactive wind shear alarm system design and bureau to better carry out airworthiness certification on related equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of aircraft avionics airworthiness, and particularly relates to a method for evaluating the reliability of reactive wind shear warning based on flight data recorder (FDR) data. Background Art

[0002] Low-altitude wind shear is a weather climate extremely adverse to flight safety. It occurs rapidly, leaving little time for pilots to think, make decisions, and operate. If not properly and promptly handled, it will cause serious personnel and economic losses. To help pilots promptly identify wind shear, the airborne avionics equipment provides a reactive wind shear warning system, which can use aircraft inertial parameters and aerodynamic data to solve the flight dynamics equation, calculate the change rate of the horizontal wind speed component and the vertical wind speed component in the area where the aircraft is currently located, and then evaluate the degree of continuous energy loss attenuation of the aircraft. From the analysis of the crew's response and the QAR data of flights triggering the RWS warning, it can be seen that there may be some false alarms and missed alarms in the existing reactive wind shear warning, which has an adverse impact on the normal operation of flights and also increases the unnecessary burden on the crew.

[0003] In the prior art, there is an airborne wind shear detector and a wind shear detection method (201710514160.1). This invention provides an airborne wind shear detector, including a radar antenna, a signal transmission system, a wind shear data processing system, a wind shear warning output display system, etc. And it provides a wind shear detection method based on this wind shear detector. There is also an airport low-altitude wind shear detection method and system (201610005901.9). This invention conducts real-time detection of the wind field over the airport through a boundary layer wind profiler radar to obtain actual meteorological observation data. When the wind shear information exceeds the system threshold, it gives warnings for wind shear in the horizontal and vertical directions. There is also an approach phase wind shear operation handling quality evaluation method based on QAR data (202310231815.X). This invention uses QAR data to evaluate the handling quality of the crew after wind shear occurs in the approach phase, mainly used to evaluate the handling quality of the crew, and does not involve the judgment of whether the warning is effective.

[0004] Currently, the related research directions on wind shear in the prior art mainly focus on the methods and optimizations of warning recognition, and more are starting from the principles and processes of warnings on how to improve warnings. However, there is a blank in the evaluation and judgment of the accuracy and reliability of reactive wind shear warning recognition. This brings great difficulties to the airworthiness certification and improved design of airborne reactive wind shear warning systems. Currently, the research on reactive wind shear warning all focuses on the design improvement of warning detection methods, and there is no reported research on the evaluation of warning reliability.

[0005] Given that QAR data can truly reflect the pilot's control actions, aircraft status, and the operating characteristics of the external environment, a reactive wind shear warning reliability evaluation method based on flight parameter data can be established by selecting relevant parameters that can objectively reflect the aircraft's status in the wind shear area with the flight parameter data as the benchmark. Summary of the Invention

[0006] To solve the problems existing in the prior art, the present invention provides a reactive wind shear warning reliability evaluation method based on flight parameter data. First, sample data is collected, and the reliability of wind shear warning events is evaluated and calibrated by analyzing the sample data. Then, based on the VAE-ISTM model, the wind shear recognition reliability is transformed into an anomaly detection problem of time series indicators, and the cumulative sum of the rate of change is selected as the judgment index. An evaluation method for the reliability of reactive wind shear warnings in the approach stage based on the cumulative sum of the rate of change is designed and proposed, providing technical support for avionics equipment manufacturers to optimize the design of reactive wind shear warning systems and for the regulatory authorities to better conduct airworthiness certification of related equipment, and solving the problems mentioned in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: A reactive wind shear warning reliability evaluation method based on flight parameter data, comprising the following steps:

[0008] S1. Extraction of flight data of flights with wind shear

[0009] S2. Calibration of wind shear warning events

[0010] S3. Calculation of wind shear warning reliability evaluation indicators

[0011] S4. Construction and training of the VAE-LSTM model

[0012] S5. Use the trained VAE-LSTM model to evaluate the reliability of reactive wind shear warnings.

[0013] Preferably, in step S1, the flight data of flights is exported, the events of reactive wind shear warnings occurring in the approach stage are screened, and the flight data of flights with wind shear in the final approach stage that triggered the reactive wind shear warning is extracted.

[0014] Preferably, in step S2, all wind shear warning events are divided into two categories: effective wind shear warnings and ineffective wind shear warnings. The effective wind shear warning means that the wind shear is significant, specifically: the wind speed and direction change significantly, the airspeed changes by more than 15 kt in a short time or the rate of climb or descent changes by more than 500 ft / min in a short time; the ineffective wind shear warning means that there are no significant wind shear characteristics, specifically: the wind speed and direction do not change significantly or the airspeed and rate of climb or descent quickly recover after the change.

[0015] Preferably, in step S3, according to the extracted sample events that triggered the reactive wind shear warning, calculate the cumulative sum of the energy change rate in the final approach phase; specifically, it includes the following: calculate the energy change rate E of the aircraft from the calculation parameters in the flight wind shear operation data extracted in step S1 t , and the calculation formula is as follows:

[0016]

[0017] Calculate the cumulative sum of the energy change rate in the final approach phase, and the calculation formula is as follows:

[0018]

[0019] where IVV is the rate of climb and descent, TAS is the true airspeed, t start is the time point when the final approach phase starts, and t end is the time point when the final approach phase ends.

[0020] Preferably, in step S4, it specifically includes the following steps:

[0021] S41. Based on the calculated cumulative sum of the energy change rate in the final approach phase, select the normal approach data before the warning is triggered as the training set and the test set; the data format of the training set is X train =[x1, x2, …, x L , where L represents the time span of the training data; the test set data is defined as X test =[x1, x2, …, x L' , and L' is the length of the test set;

[0022] S42. Perform data preprocessing on the training set and test set data, including: perform standardization processing on the training set, calculate its mean and variance as the standardization parameters, and then use these standardization parameters to perform the same standardization processing on the test set to ensure the consistency of the data distribution between the training set and the test set. The standardization processing is represented by the following formula:

[0023]

[0024] where x is the original data, mean is the mean of the training set, and std is the variance of the training set;

[0025] S43. Train the VAE-LSTM model, use the preprocessed training set data to train the VAE-LSTM model, and then use the test set data for verification to obtain the trained VAE-LSTM model.

[0026] Preferably, in step S43, during the training phase of the VAE-LSTM model, the original data is first divided into multiple subsequences by means of a sliding window, and the window data w is defined. t , based on the given time t and window length l, the original data X = [x1,..., x L-1 , x L can be divided into multiple sliding window data W = [w1,..., w L-l , w L-l+1 , including L-l+1 window data, where each window data is represented as w t = [x t-l+1 ,..., x t-1 , x t ;

[0027] Use the divided window data W to pre-train the VAE model, then save the weights of the encoder and decoder in the VAE model, and then combine and splice the window data, and further divide it into n sliding windows Extract features by the encoder of the pre-trained VAE model, input the extracted feature vectors into the LSTM model for prediction, and then restore them to the initial data dimension through a decoder with fixed weights, and finally calculate the reconstructed data and the original input data The arithmetic square root of is used as the reconstruction error loss recon , and the formula is expressed as follows;

[0028]

[0029] Preferably, in step S5, compare the detection result of the VAE-LSTM model with the calibration level of the wind shear warning event to obtain the reliability evaluation of the wind shear warning.

[0030] Preferably, in order to detect whether the wind shear warning of the VAE-LSTM model is abnormal, a threshold is set through the F1 score. When the reconstruction error exceeds this threshold, it indicates that the reconstructed data of the current window is quite different from the real data and the warning is reasonable; the F1 score calculation formula is expressed as follows:

[0031]

[0032] In the formula: P represents the detection accuracy; R represents the recall rate; F1 represents the F1 score; n TP is the true positive, representing the number of data points that are actually abnormal when detected as abnormal; n FP is the false positive, representing the number of data points that are detected as abnormal but actually normal; n FN is the false negative, representing the number of data points that are detected as normal but actually abnormal.

[0033] The beneficial effects of the present invention are as follows: Based on QAR data, the cumulative sum of the energy change rate in the final approach phase, through the VAE-LSTM model and expert verification, a method for evaluating the reliability of wind shear warnings is designed. 1) This method can be used as a method for judging the optimization effect during the subsequent optimization of the warning system by analyzing QAR data; 2) By analyzing the cumulative sum of the energy change rate through VAE-LSTM and combining with the warnings of airborne avionics equipment, the pilot's wind shear recovery operation can be evaluated more effectively, and there is a more reasonable explanation for some operations earlier than the system warning time; 3) Through the statistical analysis of a large number of real wind shear data, the algorithm structure can be continuously optimized to obtain an index that can better reflect the wind shear intensity. Description of the Drawings

[0034] Figure 1 It is a schematic diagram of the reliability evaluation process of the wind shear warning system in the embodiment of the present invention;

[0035] Figure 2 It is a schematic diagram of the architecture of the reactive wind shear warning reliability evaluation method in the approach phase in the embodiment of the present invention;

[0036] Figure 3 It is a flow chart of the cumulative sum calculation of the energy change rate in the embodiment of the present invention;

[0037] Figure 4 It is a flow chart of the evaluation based on VAE-LSTM in the embodiment of the present invention;

[0038] Figure 5 It is an internal structure diagram of LSTM in the embodiment of the present invention;

[0039] Figure 6 It is a schematic diagram of IAS and IVV data before and after a certain RWS at Lhasa Airport in Embodiment 3 of the present invention;

[0040] Figure 7 It is a schematic diagram of IAS and IVV data before and after 2 times of RWS at Kunming Airport in Embodiment 3 of the present invention;

[0041] Figure 8 It is a schematic diagram of the reconstruction error of the wind shear section at Lhasa Airport in Embodiment 3 of the present invention;

[0042] Figure 9 It is a schematic diagram of the reconstruction error of the wind shear section at Kunming Airport in Embodiment 3 of the present invention;

[0043] Figure 10 It is a schematic diagram of wind shear detection based on VAE-LSTM at Lhasa Airport in Embodiment 3 of the present invention;

[0044] Figure 11 It is a schematic diagram of wind shear detection based on VAE-LSTM at Kunming Airport in Embodiment 3 of the present invention;

[0045] Figure 12 It is a schematic diagram of IAS and IVV data of a certain wind shear at Kunming Airport. Specific implementation manners

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

[0047] Embodiment 1

[0048] Based on the cumulative sum of the energy change rate of the aircraft in the final approach stage, the present invention designs a reactive wind shear warning reliability evaluation method based on flight parameter data, which can be used as a test standard to optimize the wind shear warning system. The specific process is as Figure 1 shown, and includes the following steps:

[0049] Step 1: Extraction of flight wind shear operation data

[0050] Through decoding software such as AirFase or AGS, export the flight data of the flight, screen the events that have reactive wind shear warnings during the approach stage, and extract the flight wind shear operation data of the final approach stage that triggered the reactive wind shear warning based on the required QAR data fields and flight phases.

[0051] Step 2: Calibration of wind shear warning events

[0052] The on-board system will record all events of wind shear warnings. However, not all warnings represent that the aircraft actually encounters wind shear. Therefore, it is necessary for flight experts to analyze the QAR data and classify all wind shear warning events into two categories: effective wind shear warnings (significant wind shear manifestations) and invalid wind shears (no significant wind shear characteristics).

[0053] Step 3: Calculation of wind shear warning reliability evaluation indicators

[0054] The energy change rate represents the aircraft's own energy level under the influence of the wind in the wind shear area, and can comprehensively reflect the state of the aircraft. Select the sample events that triggered the reactive wind shear warning extracted in Step 1 and calculate the cumulative sum of the energy change rate in the final approach stage.

[0055] Step 4: Evaluation method based on the VAE-LSTM model

[0056] The variational autoencoder (VAE) model is a generative deep learning model that can reduce the dimension of data while preserving its features. The long short-term memory network (LSTM) is a special type of recurrent neural network that can make up for the deficiencies of the VAE model in time series prediction. Based on the cumulative sum of the rate of change of energy in the final approach phase calculated in step 3, the normal approach data before the trigger warning is selected as the training set, and the data for the part of the reactive wind shear warning trigger is detected. The detection results are compared with the levels of each sample event given by the flight experts in step 2 to verify the rationality of the method, and a warning reliability evaluation method based on the cumulative sum of the rate of change of energy is given.

[0057] Example 2

[0058] A method for evaluating the reliability of reactive wind shear warnings based on flight parameter data includes the following steps:

[0059] Step 1: Extraction of flight wind shear operation data

[0060] Use decoding software such as AIRFASE or AGS to read the flight QAR data, establish a template for extracting relevant parameters of wind shear operation for different decoding libraries used by different aircraft, use this template to scan the flight QAR data, identify the flights that trigger reactive wind shear warnings, and extract the relevant data of wind shear operation for these flights. The data table is shown in Table 1.

[0061] Table 1 Flight wind shear operation data

[0062]

[0063] Step 2: Calibration of wind shear warning events

[0064] As Figure 2 shown, this method mainly learns the indicators through the model and combines them with the calibration of flight experts for comparison. Therefore, it is necessary for flight experts to analyze the wind shear operation data of the flights that trigger reactive wind shear warnings based on the judgment parameters in the flight wind shear operation data extracted in step 1, and classify the events that trigger reactive wind shear warnings according to the descriptions shown in Table 2:

[0065] Table 2 Classification and calibration of wind shear warning events

[0066]

[0067] Step 3: Calculation of wind shear warning reliability evaluation indicators

[0068] Due to the fact that the aircraft encounters strong airflow changes within a short period in the wind shear area, the impact on the aircraft's airspeed and rate of climb / descent is relatively significant. Since the unit values of speed and rate of climb / descent are not unified, the energy change rate index, which can comprehensively reflect the changes of both, is selected as the evaluation index for the reliability of wind shear warning. Figure 3 Figure Figure 3 shows the calculation process of the energy change rate index. The following details the specific calculation method of the index.

[0069] (1) Energy change rate

[0070] The energy change rate index starts from the energy change of the aircraft and characterizes the impact of wind shear on the aircraft's energy attenuation. The energy change rate E of the aircraft is obtained from the calculation parameters in the flight wind shear operation data extracted in Step 1 t , and the calculation formula is shown in Equation (1).

[0071]

[0072] (2) Cumulative sum of energy change rate

[0073] In theory, during the final approach phase, the aircraft should maintain a certain glide path for stable approach. Therefore, calculating the cumulative sum of the energy change rate and presenting the trend of the energy change rate per second can better analyze the intensity of the wind shear and evaluate the reliability of the reactive wind shear warning based on this. The calculation formula is shown in (2).

[0074]

[0075] where t start is the starting time point of the final approach phase, and t end is the ending time point of the final approach phase.

[0076] Step 4: Evaluation based on the VAE-LSTM model

[0077] This method is used to analyze the reliability of the reactive wind shear warning based on anomaly detection of univariate time series. It mainly consists of two parts: model training and anomaly detection. The data format of the training set is X train = [x1, x2,..., x L , where L represents the time span of the training data. The data of the test set is defined as X test = [x1, x2,..., x L' , and L' is the length of the test set. Since the QAR data belongs to time series data, fixed-length inputs and outputs are generated by means of a sliding window, and whether it is defined as an anomaly is measured based on the anomaly scores of the reconstructed samples and the normal samples. Figure 4 Figure Figure 4 shows the overall architecture of the model.

[0078] (1) Data preprocessing

[0079] Since the time between the triggering of the wind shear warning and the pilot's access to the control is short, in order to increase the amount of data for model training, interpolation processing is performed on the cumulative sum data of the energy change rate calculated.

[0080] Considering that QAR parameters are a type of time series data, in order to further improve the robustness of the model, standardization processing is usually carried out in the data preprocessing stage. Standardization processing aims to eliminate the dimensional differences between different features, making them have similar scales, and helps the model learn the patterns of the data more effectively. This data preprocessing method is applied to both the training set and the test set. First, standardization processing is performed on the training set to calculate its mean and variance as standardization parameters. Then, using these parameters, the same standardization processing is carried out on the test set to ensure the consistency of the data distribution between the training set and the test set. Standardization processing can be expressed by the following formula:

[0081]

[0082] where x is the original data, mean is the mean of the training set, and std is the variance of the training set.

[0083] (2) VAE model

[0084] The full name of the VAE model is Variational AutoEncoder. The purpose of the VAE model is to reconstruct the original data through a set of encoders and decoders, making the generated data x' as similar as possible to the original data x. Compared with the mapping relationship between single values found by the traditional AE model, the VAE model finds the mapping relationship between distributions, further enhancing the data generation ability. By maximizing the evidence lower bound function ELBO, it is ensured that the x' most similar to the original data x can be generated from the approximate posterior distribution p(z|x) closest to the prior distribution q(z). The evidence lower bound function is

[0085] ELBO = E q [logp(x|z)] - KL(q(z)||p(z|x)) (4)

[0086] where, E q [logp(x|z)] is the reconstruction error, which measures whether the decoder can accurately reconstruct the original data when the latent variable z is sampled from the approximate posterior distribution p(z|x). KL(q(z)||p(z|x)) is the KL divergence, which is used to measure the difference between the approximate posterior distribution p(z|x) and the prior distribution q(z) of the latent variable z.

[0087] The present invention selects a CNN convolutional neural network as the fully connected layer in the VAE model. In order to enable the model to learn the features of each part of the QAR data, the original data is divided into multiple subsequences by means of a sliding window. Define the window data w t , based on the given time t and window length l, the original data X = [x1, …, x L-1 , x L can be divided into multiple sliding window data W = [w1, …, w L-l , w L-l+1 , including L - l + 1 window data, where each window data can be expressed as w t = [x t-l+1 ,..., x t-1 , x t .

[0088] In order to enable the model to learn more features contained in the data, a convolutional layer with a stride is selected to replace the pooling layer, which enables the model to flexibly select an appropriate downsampling method according to the characteristics of the data, rather than fixedly using a certain preset pooling operation. The divided window data W is used as the input layer and input into the encoder. The input data is processed through two one-dimensional convolutions to extract the features of the time series data. Then, two fully connected layers are used to calculate the mean and logarithmic variance of the latent variable z respectively. Finally, through the reparameterization trick, the latent variable z is generated by combining the mean, logarithmic variance of the latent variable z and random noise. The decoder reconstructs and restores the latent feature z to the original input window size through convolution. The convolution operation formula is

[0089]

[0090] In the formula, is the i-th feature of the k-th layer obtained through convolution, f(·) is the ReLU activation function, is the weight coefficient for the i-th convolution kernel operation of the k-th layer, * is the convolution operation symbol, X k-1 is the feature output by the (k - 1)-th layer (the first layer is the input layer, W), is the bias of the i-th convolution kernel of the k-th layer.

[0091] (3) LSTM model

[0092] The full name of the LSTM model is Long Short Term Memory. By adding special gating structures (forget gate, input gate, and output gate), the LSTM model can choose to remember or forget specific information from before, solving problems such as gradient vanishing and gradient explosion in traditional RNN (recurrent neural network). The LSTM structure is as Figure 5 shown.

[0093] Among them, x t is the input information for this time, c t-1 and c t are the cell states retained in the previous round and the cell state of this time respectively, h t-1 and h t are the output information of the previous round and this time respectively. f t is the forget gate, which controls the information retained in the previous unit state through the activation function sigmoid. i t is the input gate, which determines the information of the input x t for this time through the activation function. By synthesizing the results of the forget gate and the input gate, the cell state c t of this time can be obtained. o t is the output gate, which determines part of the information included in the output h t by passing the updated cell state c t through the activation function. The calculation formulas for each part are as follows:

[0094]

[0095] (4) VAE-LSTM model

[0096] The present invention adopts the following process for the detection and verification of the effectiveness of wind shear warning: First, preprocess the target data, then train the VAE-LSTM model using the training data, and finally use the trained model to process the test data to verify the effect of the model.

[0097] Among them, the preprocessing stage includes the calculation of indicators, the standardization of data, and the division of windows. In the model training stage, first pre-train the VAE model using the divided window data W, then save the weights of the encoder and decoder in the VAE model, and then combine and splice the window data and further divide it into n sliding windows Extract features by the encoder of the VAE model after pre-training is completed, input the extracted feature vectors into the LSTM model for prediction, then restore them to the initial data dimension through the decoder with fixed weights, and finally calculate the reconstructed data and the original input data to calculate the arithmetic square root as the reconstruction error loss recon .

[0098]

[0099] To detect whether the wind shear warning is abnormal, a threshold also needs to be set. Once the reconstruction error exceeds this threshold, it indicates that the reconstructed data of the current window is quite different from the real data, and the warning is reasonable, and vice versa. According to the reconstruction error, 200 threshold reference values are selected, and the F1 score of each threshold reference value is calculated. Since the F1 score comprehensively considers the detection accuracy and recall rate, it can better cover all abnormalities.

[0100]

[0101] In the formula: P represents the detection accuracy, R represents the recall rate, F1 represents the F1 score, and n TP is the true positive, representing the number of data points that are actually abnormal when detected as abnormal, and n FP is the false positive, representing the number of data points that are detected as abnormal but are actually normal, and n FN is the false negative, representing the number of data points that are detected as normal but are actually abnormal. Considering that abnormal events such as wind shear warnings are usually sequence abnormalities, all windows containing abnormalities will be marked as abnormal.

[0102] Example 3

[0103] To ensure the applicability of the algorithm as much as possible, the QAR data of landing flights at two airports where wind shear often occurs are selected as the experimental objects. Although the wind shear warning is mainly triggered in the final approach phase, the duration of the final approach phase of different flights may vary. To ensure the training efficiency, 100 sets of flight data of normal approach landings are selected for each airport, including the data of the first 200 seconds before the landing phase without triggering the wind shear warning as the training data set. Considering that the triggering frequency of the wind shear warning is relatively low, 10 sets of flight data triggering the reactive wind shear warning in the final approach phase at Lhasa Airport and 12 sets at Kunming Airport are selected; at the same time, considering that after the reactive wind shear warning is triggered, the pilot will perform a go-around maneuver, and after a series of operations such as applying TOGA thrust, the aircraft will not only be affected by the external wind shear, but also change according to the pilot's maneuvers. Therefore, the data of the first 200 seconds before the reactive wind shear warning moment is used as the test data set. First, combined with the advice of flight experts, the test data set is analyzed, and it is found that although the RWS is triggered in the data of the second verification flight segment at Lhasa Airport, the changes in airspeed and rate of climb are not significant. Figure 6 are the IAS and IVV data of this flight, and the red vertical line is the warning moment shown in the QAR data. Obviously, in the second flight segment for some time before the warning, the IAS did not change by more than 15 kt, and the IVV did not change by more than 500 ft / min. In principle, it should not be determined that this flight segment encountered wind shear. Similarly, analyzing the 12 sets of verification data at Kunming Airport, it is found that two sets do not meet the loss criteria of airspeed and rate of climb, and their airspeed and rate of climb are as Figure 7As shown. After confirming that the remaining 19 sets of flight segment data are all true and valid RWSs, retain the flight segment data of this time as the invalid RWS for comparison and verification in the test set.

[0104] Use the VAE-LSTM model to detect the flight data of 10 flights at Lhasa Airport and 12 flights at Kunming Airport that triggered reactive wind shear warnings during the approach phase, and obtain the reconstruction errors as Figure 8 、 Figure 9 shown. According to the optimal F1 principle, among the preset thresholds, select the thresholds of 0.1961 and 0.2046 for Lhasa Airport and Kunming Airport respectively. The red dotted line in the figure is the selected threshold.

[0105] According to the set threshold, obtain the detection results as Figure 10 、 Figure 11 shown. The blue curve in the figure is the energy change rate of the test flight. Mark the warning moment of the airborne wind shear warning system with a red dotted line, and fill the wind shear recognition window calculated by the model with a green solid line. As shown, among the 10 reactive wind shear warnings at Lhasa Airport, 9 were successfully detected, all of which are true and valid RWSs, and the invalid RWS for comparison and verification in the second flight segment was not recognized; 10 groups of wind shear verification data at Kunming Airport were successfully detected, and the remaining 2 groups are the weaker wind shears determined above. The result accuracy is shown in Table 3.

[0106] Table 3 Wind shear effectiveness detection index parameters for different airports

[0107]

[0108] The main factors affecting indicators such as accuracy are the two groups identified multiple times. Therefore, the main reasons are analyzed below in combination with the original QAR data.

[0109] Case analysis: By combining the suggestions of flight experts and the original QAR data, it is found that the main reasons for the model to identify in advance in the example flight segments at Lhasa Airport and Kunming Airport are as follows: (1) In the window data identified by the model, there are large fluctuations in the airspeed or rate of climb and descent related to the wind shear height; (2) The radio altitude at the window position identified in advance by the model exceeds 2000 ft, while the airborne wind shear warning system usually starts only when the radio altitude is below 1500 ft. It can be considered that the recognition ability of this model can not only maintain a certain recognition rate at low altitudes, but also has certain reference value for the recognition of wind shear at high altitudes. Below, take the second group of wind shear at Kunming Airport as an example to specifically analyze whether its advance recognition is credible. As Figure 12As shown, the speed and rate of climb / descent parameters for this flight segment are presented. It is not difficult to observe that between the 80th and 150th moments, the IAS decreased significantly multiple times (by more than 15 kt), and even remained below VAPP for 8 consecutive seconds. Moreover, the IVV decreased significantly (by more than 500 ft / min) at the 130th and 183rd moments. Thus, it can be determined that the model detection results are accurate.

[0110] It should be noted that in the present invention, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also other elements not explicitly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising a..." does not exclude the existence of additional identical elements in the process, method, article or device comprising said element.

[0111] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms unless the context clearly dictates otherwise.

[0112] It should be understood that the term "and / or" used in the present invention is merely a description of the relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Additionally, the character " / " in the present invention generally represents an "or" relationship between the associated objects before and after.

[0113] Depending on the context, the word "if" as used herein can be interpreted as "when", "while" or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrase "if determined" or "if detected (stated condition or event)" can be interpreted as "when determined", "in response to determining", "when detected (stated condition or event)", or "in response to detecting (stated condition or event)".

[0114] The "first / second" mentioned in the embodiments is merely to distinguish similar objects and does not represent a specific order for the objects. It can be understood that the "first / second" can be interchanged in the allowed cases to change the specific order or sequence. It should be understood that the objects distinguished by the "first / second" can be interchanged appropriately so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0115] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A reactive wind shear warning reliability evaluation method based on flight data recorder data, characterized in that It includes the following steps: S1. Extraction of flight wind shear operation data; S2. Calibration of wind shear warning events; S3. Calculation of wind shear warning reliability evaluation indicators; S4. Construction and training of the VAE-LSTM model; S5. Use the trained VAE-LSTM model to conduct reactive wind shear warning reliability evaluation.

2. The reactive wind shear warning reliability evaluation method based on flight data recorder data according to claim 1, wherein: In step S1, export flight data, screen events that trigger reactive wind shear warnings during the approach phase, and extract the flight wind shear operation data of the final approach phase that triggered the reactive wind shear warning.

3. The method for evaluating the reliability of reactive wind shear warning based on flight data recorder data according to claim 1, wherein: In step S2, classify all wind shear warning events into two categories: effective wind shear warnings and ineffective wind shear warnings. The effective wind shear warning means that the wind shear is significant, specifically: the wind speed and direction change significantly, the airspeed changes by more than 15 kt within a short time, or the rate of climb or descent changes by more than 500 ft / min within a short time; the ineffective wind shear warning means that there are no significant wind shear characteristics, specifically: the wind speed and direction do not change significantly, or the airspeed and rate of climb or descent quickly return after the change.

4. The method for evaluating the reliability of reactive wind shear warning based on flight data recorder data according to claim 1, wherein: In step S3, based on the extracted sample events that triggered the reactive wind shear warning, calculate the cumulative sum of the energy change rate in the final approach phase; specifically, it includes the following: Calculate the energy change rate E of the aircraft from the calculation parameters in the flight wind shear operation data extracted in step S1 t , and the calculation formula is as follows: Calculate the cumulative sum of the energy change rate in the final approach phase. The calculation formula is as follows: Among them, IVV is the rate of climb and descent, TAS is the vacuum rate, t start is the time point when the final approach phase starts, t end is the time point when the final approach phase ends.

5. The method for evaluating the reliability of reactive wind shear warning based on flight data recorder data according to claim 1, wherein: In step S4, it specifically includes the following steps: S41. Select the data of normal approach before the trigger warning as the training set and the test set based on the cumulative sum of the rate of change during the final approach stage calculated; the data format of the training set is X train = [x1, x2,..., x L , where L represents the time span of the training data; the test set data is defined as X test = [x1, x2,..., x L' , and L' is the length of the test set; S42. Perform data preprocessing on the training set and test set data, including: performing standardization processing on the training set, calculating its mean and variance as standardization parameters, and then using these standardization parameters to perform the same standardization processing on the test set to ensure the consistency of the data distribution between the training set and the test set. The standardization processing is represented by the following formula: where x is the original data, mean is the mean of the training set, and std is the variance of the training set; S43. Training of the VAE-LSTM model. Use the preprocessed training set data to train the VAE-LSTM model, and then use the test set data for verification to obtain the trained VAE-LSTM model.

6. The method for evaluating the reliability of reactive wind shear warning based on flight data recorder data according to claim 5, wherein: In step S43, during the training phase of the VAE-LSTM model, the original data is first divided into multiple subsequences by means of a sliding window, and the window data w is defined t , based on the given time t and window length l, the original data X = [x1, …, x L-1 , x L can be divided into multiple sliding window data W = [w1, …, w L-l , w L-l+1 , which contains L - l + 1 window data, where each window data is represented as w t = [x t-l+1 , …, x t-1 , x t ; Pre-train the VAE model using the partitioned window data W, then save the weights of the encoder and decoder in the VAE model, and then combine and splice the window data and further divide it into n sliding windows Extract features by the encoder of the pre-trained VAE model, input the extracted feature vectors into the LSTM model for prediction, then restore them to the initial data dimension by the decoder with fixed weights, and finally calculate the reconstructed data and the original input data Take the arithmetic square root of as the reconstruction error loss recon , and the formula is expressed as follows; 7. The reactive wind shear warning reliability evaluation method based on flight data recorder (FDR) data according to claim 1, characterized in that: In step S5, compare the detection results of the VAE-LSTM model with the calibration level of the wind shear warning events to obtain the wind shear warning reliability evaluation.

8. The method for evaluating the reliability of reactive wind shear warning based on flight data recorder data according to claim 7, wherein: To detect whether the wind shear warning of the VAE-LSTM model is abnormal, set a threshold through the F1 score. When the reconstruction error exceeds this threshold, it indicates that the reconstructed data of the current window is significantly different from the real data, and the warning is reasonable. The F1 score calculation formula is expressed as follows: Where: P represents the detection accuracy; R represents the recall rate; F1 represents the F1 score; n TP is the true positive, representing the number of data points where the detected anomaly is actually an anomaly; n FP is the false positive, representing the number of data points where the detected anomaly is actually normal; n FN is the false negative, representing the number of data points where the detected normality is actually an anomaly.

Images