Proximity landing stage anomaly detection method based on flight data

Abstract

The invention relates to the technical field of flight state anomaly detection, in particular to an approach landing stage anomaly detection method based on flight data, which comprises the following steps of: extracting the flight data to determine direct and indirect influence parameters which influence the approach landing of an aircraft, and performing preprocessing; constructing a multivariate time series prediction model based on StemGNN, and carrying out prediction and over-limit monitoring on flight data; establishing a flight state transition stability analysis model according to the flight program of the approach landing stage, and performing flight state transition stability analysis on the non-overrun monitoring parameters; and performing anomaly analysis according to an output result. According to the method, by preprocessing the flight data, analyzing direct and indirect influence parameters which influence the flight safety in the approaching landing stage, performing multivariate parameter prediction analysis through the StemGNN network, and performing flight state stationary analysis through multi-feature fusion, anomaly detection and early warning are realized, so that prediction of unsafe events can be more accurately realized.

Description

Technical Field

[0001] The present invention relates to the technical field of flight status anomaly detection, and in particular to an approach and landing phase anomaly detection method based on flight data. Background Art

[0002] Improving civil aviation flight safety has always been a focus of the industry. According to the classification of aircraft flight phases, most accidents occur during the approach and landing phases. The approach phase is a phase with more operations during the flight process, and the possibility of errors is high. Flight safety during the landing phase is currently the biggest safety challenge in the aviation industry, because safety issues such as hard landing, tail strikes, and runway runoff usually occur during the landing phase. Therefore, safety analysis of the flight status during the approach and landing phase can help reduce the incidence of unsafe incidents.

[0003] Most of the current flight safety analysis methods are data-driven safety analysis based on Quick Access Record (QAR) data or Flight Data Recorder (FDR) data. QAR records the aircraft's onboard equipment parameters, pilot control parameters, aircraft flight status parameters and environmental parameters during the flight, and the data recorded by QAR has the characteristics of high frequency, multiple types and large data volume. Airline operators and aviation regulatory agencies conduct in-depth mining and analysis of QAR data to meet the needs of daily maintenance, flight inspections, system performance monitoring and flight quality monitoring. If the flight data recorded by QAR is used to achieve in-depth analysis of the aircraft's safe flight mode, the change rules of flight data under safe flight can be obtained, which can then be used for real-time flight safety analysis.

[0004] At present, civil aviation has adopted many safety warning and anti-collision systems to ensure the safety of aircraft during flight. Most of these systems predict a certain parameter through a computer network, compare the predicted value with the threshold value pre-set in the computer, and judge whether it exceeds the limit to complete the flight safety warning. They do not consider the indirect impact that other parameters may have on the predicted flight parameter changes. Therefore, the use of direct and indirect impact parameters to predict flight unsafe events can more accurately predict unsafe events.

[0005] Furthermore, the stability of flight state transition has a great impact on flight safety, and the change of flight mode is also reflected in the change law of flight parameters. Extracting effective flight state information that characterizes flight safety mode is rarely considered in current research.

[0006] Based on this, we propose a StemGNN flight approach and landing anomaly detection method based on flight data. Summary of the invention

[0007] The purpose of the present invention is to overcome the above problems or at least partially solve the above problems, and a method for detecting abnormalities in the approach and landing phase based on flight data is proposed. For the approach and landing phase, according to the monitoring items and monitoring parameters in the "Implementation and Management of Flight Quality Control (FOQA)", the main events affecting the safety of approach and landing are tail strikes and hard landings. According to the monitoring items of these two events, their monitoring parameters are determined to be pitch angles and vertical accelerations; according to flight dynamics, there are three types of parameters that affect pitch angles and vertical accelerations: control input parameters (throttle lever control amount, elevator deflection, aileron, rudder), aircraft state parameters (roll angle, roll angular velocity, pitch angular velocity, altitude, latitude, longitude, descent rate, corrected airspeed), and external environmental parameters (wind speed). A multivariate time series prediction model based on StemGNN is constructed, and direct parameters and indirect parameters are used as inputs of the model to predict pitch angles and vertical accelerations, and to make over-limit judgments. Combined with the flight program, the multivariate time series prediction model is used to extract the flight status of the control and attitude changes in the approach and landing phase, and a smooth transfer model of the flight status is established to ensure the safe approach and landing of the aircraft.

[0008] To achieve the above object, the present invention provides the following technical solution: a method for detecting abnormalities in the approach and landing phase based on flight data, comprising the following steps:

[0009] Step 1: extract flight data for preprocessing, including the following steps:

[0010] Extract flight data from historical flights from quick access recorders;

[0011] According to the Implementation and Management of Flight Quality Assessment (FOQA), the monitoring items and parameters of the approach and landing phase are determined. The monitoring parameters include direct impact parameters and indirect impact parameters.

[0012] The direct and indirect impact parameters are processed with noise and normalized;

[0013] Step 2: Build a prediction model to predict flight data and monitor over-limit, including the following steps:

[0014] Sliding window processing is performed on direct and indirect impact parameters to construct sample data as input of the prediction model;

[0015] Build a multivariate time series prediction model based on StemGNN, extract the dependencies between sequences and the temporal correlation within sequences through graph convolutional networks and frequency domain analysis;

[0016] The model is trained using prediction loss and reconstruction loss, and the classification weighting method in cost-sensitive learning is introduced to design a loss function that pays more attention to specified features;

[0017] Compare the predicted value with the monitoring parameter to determine whether it exceeds the limit. If so, issue a safety warning. If not, proceed to step three.

[0018] Step 3: According to the flight procedure of the approach and landing phase, a flight state transfer stability analysis model is established, and the flight state transfer stability analysis is performed on the monitoring parameters that are not exceeded;

[0019] Step 4: Perform anomaly analysis based on the output results.

[0020] In a preferred embodiment, in the step 1, the monitoring items analyzed during the approach and landing phase include tail strike and hard landing events, and the monitoring parameters of the tail strike and hard landing events include pitch angle and vertical acceleration, which are directly influencing parameters.

[0021] In a preferred embodiment, in step one, according to the principles of flight dynamics, the parameters affecting the pitch angle and vertical acceleration include control input parameters (throttle lever control amount, elevator deflection, aileron, rudder), aircraft state parameters (roll angle, roll angular velocity, pitch angular velocity, altitude, latitude, longitude, descent rate, corrected airspeed) and external environmental parameters (wind speed), which are indirect influencing parameters.

[0022] In a preferred embodiment, corresponding direct influencing parameters and indirect influencing parameters are determined in a quick access recorder, [CK_TLA_PCT1, FLAP_POS_FWC, RUDDER, ELEV1] are control input parameters in the aircraft motion equation, [HEIGH_DEST, LAT_AIR, LONG_AIR, ALT_RATE, ROLL_RATE1, ROLL, TAS_REC, PITCH, PITCH_RATE, VRTG] are state parameters, SIDE_WIND is an external meteorological parameter, and the control input parameters and meteorological parameters affect the state parameters of the aircraft.

[0023] In a preferred embodiment, in step 1, the Kalman filter method is used to perform noise processing on the direct and indirect influencing parameters, and the Min-Max method is used to normalize the parameters, that is,

[0024]

[0025] In the formula, x represents the original data, x min Represents the minimum value in the original data, x max Represents the maximum value in the original data, x new Represents normalized data.

[0026] In a preferred embodiment, in step 2, sliding window processing is performed on the direct and indirect influencing parameters to construct sample data as the input of the prediction model, that is, Where N represents the number of direct and indirect influencing parameters, here N = 15, S represents the sliding window step size, and T is the input sliding window length. The output is H represents the output sliding window length, that is, the prediction time step.

[0027] In a preferred embodiment, in the step 2, the dependency relationship between sequences and the temporal correlation within the sequence are extracted by graph convolutional network and frequency domain analysis. First, the model uses the potential correlation layer to extract the dependency matrix W between the input multivariate time series, and then the graph structure G = (X, W) composed of the input and the dependency matrix is ​​used as the input of the StemGNN Block. The StemGNN Block jointly extracts the temporal dependency in the multivariate time series in the spectral domain. Finally, the model uses the prediction loss and the reconstruction loss to train the model, as shown in formula (2):

[0028]

[0029] The first term represents the prediction loss, the second term represents the reconstruction loss, X t is the true value to be predicted at time t, For X t The predicted value of B represents the prediction loss. t-i (X) represents the value reconstructed by the backtracking module, where B represents the entire network that generates the backcast output. By self-encoding the output, the representation ability of the sequence is enhanced. θ Represents all parameters in the network.

[0030] In a preferred embodiment, in step 2, the loss function is as shown in formula (3), the first term is the prediction loss, the second term is the reconstruction loss, W C is the weighting matrix:

[0031]

[0032] In a preferred embodiment, in step 2, the potential correlation layer includes a gated recurrent unit (GRU) and a self-attention mechanism. First, the multivariate time series is input into the GRU, the intrinsic temporal correlation of the sequence is extracted, and the final hidden state of the GRU is obtained. Then, the self-attention mechanism is used to calculate the relationship matrix W of the multivariate time series.

[0033] In a preferred embodiment, in the step 2, StemGNN is constructed by stacking multiple StemGNN Blocks using skip connections. The StemGNN Block is designed by embedding a spectral-sequential cell (Spe-Seq) unit into a spectral graph convolution module. StemGNN converts data from the spatiotemporal domain to the frequency domain, captures the periodicity and autocorrelation within the time series through the spectral sequence convolution Spe-Seq, and extracts the spatial dependency between time series using the spectral graph convolution.

[0034] In a preferred embodiment, in step 2, two StemGNN blocks are used for residual connection, the residual between the output value of the second StemGNN block and the reconstructed value of the first StemGNN block, and finally, the outputs of the two blocks are superimposed and sent to the GLU and fully connected layers to generate a predicted value.

[0035] In a preferred embodiment, in step 3, the flight state is defined as a feature matrix extracted from the flight parameters that can characterize the aircraft motion state (control input and attitude), and the input of the last fully connected layer of the StemGNN prediction model (the output of the IDFT layer) is extracted as the expression of the flight state. Let the state matrix at time t be St, as shown in formula (4):

[0036]

[0037] Among them, f t nl is the lth value of the nth feature at time t, n is the number of features, and l is the dimension of the feature.

[0038] In a preferred embodiment, in step three, the approach and landing phase includes two phases: an approach procedure segment from 1500 feet to 800 feet and a landing segment from 800 feet to the ground.

[0039] In a preferred embodiment, in step three, the flight status transfer during the approach and landing phase of the flight process is a series of data that changes regularly according to the approach procedure. The approach procedure of a specific airport is consulted according to the China Aeronautical Information Compilation, and the changes in the aircraft status during the approach process are analyzed based on a large amount of fleet data.

[0040] In a preferred embodiment, in step 3, the distance between the flight state matrix at the current moment and the next moment is solved as the change in flight state. The change in the state matrix at time t and time t+1 is expressed as formula (5):

[0041]

[0042] Where ED is the Euclidean distance, as shown in formula (6):

[0043]

[0044] Compare the change amount from the current flight state to the next flight state. If it exceeds the state change range, it is considered that the flight state transfer is a non-smooth transfer, with a large fluctuation, that is, a state mutation occurs, which will cause an unsafe flight incident.

[0045] In a preferred embodiment, in step 4, in the event of a sudden change in state, the flight parameters at the next moment are further analyzed in detail. If it is not caused by the pilot's manipulation, then some parts of the aircraft have malfunctioned, making the aircraft unable to fly according to the flight principles. If it is caused by the pilot's manipulation, the system prompts the pilot to confirm whether the manipulation is wrong.

[0046] Compared with the prior art, the present invention has the following beneficial effects: the present invention preprocesses the fast access recorder data, combines the "Flight Quality Assurance (FOQA) Implementation and Management" and flight dynamics, analyzes the direct and indirect influencing parameters affecting flight safety during the approach and landing phases, performs multivariate parameter prediction and analysis through the StemGNN network, and performs flight status stability analysis through multi-feature fusion, realizes anomaly detection and early warning, and can more accurately predict unsafe events. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is an overall block diagram of the present invention;

[0048] Figure 2 This is a StemGNN multivariate time series data prediction model diagram of the present invention;

[0049] Figure 3 The flowchart of the stability analysis method of flight state transfer of the present invention is as follows;

[0050] Figure 4 This is a diagram of the sample data construction method based on the sliding window of the present invention;

[0051] Figure 5 This is a state change diagram of a flight approaching stage at Nanjing Lukou International Airport of the present invention;

[0052] Figure 6 This is a flight parameter variation diagram of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] Example 1

[0055] This embodiment is based on the QAR historical data of 104 Airbus A320 flights, and takes the abnormality of the approach and landing phase of Nanjing Lukou Airport as an example to illustrate its specific implementation.

[0056] The QAR data of 104 A320 flights were selected for safety warning analysis based on the division of flight altitudes in the approach and landing phases in the flight phases and flight procedures, and the data with flight altitudes of 3000ft to 0ft were selected.

[0057] Step 1: According to the monitoring parameters and flight quality monitoring parameters, 15 corresponding flight data in the QAR data are selected, and noise filtering and normalization are performed;

[0058] Step 2: Build a multivariate time series prediction model based on StemGNN to predict and exceed the limit analysis of parameters;

[0059] 1) Arrange the normalized data in time sequence to form a two-dimensional matrix, and construct the training set by sliding window interception, with sliding window step size S = 1, input sliding window length T = 20, and output sliding window length H = 10. Finally, 41,449 samples were obtained in this way, and the ratio of training set, validation set, and test set was 7:2:1.

[0060] 2) In the network structure of StemGNN, the channel size of the graph convolution layer is set to 64, and the kernel size of the 1D convolution is set to 3. The RMSprop optimizer is used, and the number of iterations is set to 30. The initial value of the learning rate is set to 0.0001, and it decays by 0.7 times after every 5 iterations. A dropout layer is added with a coefficient set to 0.5.

[0061] 3) The weighted matrix of the loss function of the network in formula (3) is set as shown in formula (7).

[0062]

[0063] 4) The root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) are used to evaluate the prediction performance of different model methods.

[0064]

[0065]

[0066]

[0067] 5) The performance of StemGNN (N=15, direct and indirect parameters, weighted loss function), StemGNN (N=15, direct and indirect parameters), StemGNN (N=2, direct parameters), LSTM and CNN-LSTM were compared and analyzed. Table 1 shows the performance indicators of each model method.

[0068] Table 1 Performance indicators of each model method

[0069]

[0070] As can be seen from Table 1, the model based on StemGNN significantly outperforms the other two models in the three indicators of vertical acceleration (VRTG). The StemGNN model based on cost-sensitive learning has the best overall performance for pitch angle (PITCH) and vertical acceleration. The model based on StemGNN also outperforms the LSTM model in predicting PITCH, which is similar to CNN-LSTM. Compared with LSTM and CNN-LSTM, StemGNN is more effective in extracting potential features of time series data, which helps to improve prediction accuracy. Compared with the StemGNN model with direct influence parameters, the StemGNN models with direct and indirect influence parameters are improved again. This shows that the indirect influence parameters affect the prediction accuracy of VRTG and PITCH. Among the StemGNN models with direct and indirect influence parameters, the StemGNN model based on cost-sensitive learning has a greater improvement in the prediction accuracy of pitch angle, indicating that adding an appropriate amount of weight to the prediction loss of the target parameter in the loss function can improve the accuracy of the prediction of the target parameter.

[0071] Step 3: According to the flight procedure of the approach and landing phase, a flight state transfer stability analysis model is established to analyze the stability of the flight state transfer process;

[0072] 1) For the case where the parameters in step 2 are within the limit, the output features of the IDFT layer in the StemGNN model are extracted as the representation matrix St of the aircraft's flight state, see formula (4), and its dimension is (15,100);

[0073] 2) According to equations (5) and (6), the changes of the two states before and after are calculated to determine whether a large fluctuation or state mutation has occurred. Assume that the difference between the minimum and maximum values ​​of the change of the flight state matrix within 4 seconds is greater than 0.02, as shown in equation (10), which means that a large fluctuation has occurred in a short period of time.

[0074]

[0075] 3) Combined with the flight approach procedure of Nanjing Lukou Airport, analyze the state changes of a certain flight during the approach phase. Figure 5 It can be seen from the figure that the aircraft will have a very obvious state change at about 1950ft between 150s and 200s. The actual flight altitude of 1950ft is close to the 1926ft of the final approach fix point of Lukou International Airport. The flight altitude change diagram is roughly consistent with the approach flight altitude profile diagram of runways 06 and 07 of Lukou International Airport. Therefore, it can be proved that the large fluctuation of the flight state occurs at the final approach fix point FAF of the flight procedure.

[0076] In step 4, anomaly analysis is performed based on the output results.

[0077] Combination Figure 6 From the analysis of the changes in flight parameters, it can be seen that the pilot performed some drastic flight operations on the aircraft at this point, that is, the drastic change in the aircraft state was caused by the pilot's operation.

[0078] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting anomalies in the approach and landing phase based on flight data, characterized in that: The following steps are involved: Step 1: extract flight data for preprocessing, including the following steps: Extract flight data from historical flights from quick access recorders; Determine the monitoring items and monitoring parameters of the approach and landing phase. The monitoring items include tail strike and hard landing events. The monitoring parameters of tail strike and hard landing events include direct influencing parameters and indirect influencing parameters. Among them, the direct influencing parameters include pitch angle and vertical acceleration. The parameters affecting pitch angle and vertical acceleration include control input parameters, aircraft state parameters and external environment parameters, which are indirect influencing parameters. The direct and indirect impact parameters are processed with noise and normalized; Step 2: Build a prediction model to predict flight data and monitor over-limit, including the following steps: Sliding window processing is performed on direct and indirect impact parameters to construct sample data as input of the prediction model; Build a multivariate time series prediction model based on StemGNN, extract the dependencies between sequences and the temporal correlation within sequences through graph convolutional networks and frequency domain analysis; The model is trained using prediction loss and reconstruction loss, and the classification weighting method in cost-sensitive learning is introduced to design a loss function that pays more attention to specified features; Compare the predicted value with the monitoring parameter to determine whether it exceeds the limit. If so, issue a safety warning. If not, proceed to step three. Step 3: According to the flight procedure of the approach and landing phase, a flight state transfer stability analysis model is established, and the flight state transfer stability analysis is performed on the monitoring parameters that are not exceeded. The approach and landing phase includes two stages: the approach procedure segment from 1500 feet to 800 feet and the landing segment from 800 feet to the ground; Step 4: Perform an abnormality analysis based on the output results. In the event of a sudden change in state, further analyze the flight state parameters at the next moment in detail. If it is not caused by the pilot's operation, then some parts of the aircraft have malfunctions, making the aircraft unable to fly according to the flight principles. If it is caused by the pilot's operation, the system prompts the pilot to confirm whether the operation is wrong.

2. The method for detecting anomalies in the approach and landing phase based on flight data according to claim 1, characterized in that: In step 1, the Kalman filter method is used to perform noise processing on the direct and indirect influencing parameters, and the Min-Max method is used to normalize the parameters, that is, In the formula, x represents the original data, x min Represents the minimum value in the original data, x max Represents the maximum value in the original data, x new Represents normalized data.

3. The method for detecting anomalies in the approach and landing phase based on flight data according to claim 2, characterized in that: In step 2, the direct and indirect influencing parameters are processed by sliding windows to construct sample data as the input of the prediction model, that is, Where N represents the number of direct and indirect influencing parameters, here N = 15, S represents the sliding window step size, and T is the input sliding window length. The output is H represents the output sliding window length, that is, the prediction time step.

4. The method for detecting anomalies in the approach and landing phase based on flight data according to claim 3, characterized in that: In the step 2, the dependency between sequences and the temporal correlation within the sequence are extracted by graph convolutional network and frequency domain analysis. First, the model uses the potential correlation layer to extract the dependency matrix W between the input multivariate time series. Then, the graph structure G = (X, W) composed of the input and the dependency matrix is ​​used as the input of the StemGNN Block. The StemGNN Block jointly extracts the temporal dependency in the multivariate time series in the spectral domain. Finally, the model is trained using the prediction loss and reconstruction loss, as shown in formula (2): The first term represents the prediction loss, the second term represents the reconstruction loss, X t is the true value to be predicted at time t, For X t The predicted value of B represents the prediction loss. t-i (X) represents the value reconstructed by the backtracking module, where B represents the entire network that generates the backcast output. By self-encoding the output, the representation ability of the sequence is enhanced. θ Represents all parameters in the network.

5. The method for detecting anomalies in the approach and landing phase based on flight data according to claim 4, characterized in that: In step 2, the loss function is as shown in formula (3), where the first term is the prediction loss and the second term is the reconstruction loss. C is the weighting matrix:

6. The method for detecting anomalies in the approach and landing phase based on flight data according to claim 5, characterized in that: In the step 2, the potential correlation layer includes a gated recurrent unit (GRU) and a self-attention mechanism. First, the multivariate time series is input into the GRU to extract the intrinsic temporal correlation of the sequence and obtain the final hidden state of the GRU. Then, the self-attention mechanism is used to calculate the relationship matrix W of the multivariate time series.

7. The method for detecting anomalies in the approach and landing phase based on flight data according to claim 5, characterized in that: In the step 2, StemGNN is constructed by stacking multiple StemGNN Blocks using skip connections. StemGNN Block is designed by embedding the spectral-sequential cell (Spe-Seq) unit into the spectral graph convolution module. StemGNN converts data from the spatiotemporal domain to the frequency domain, captures the periodicity and autocorrelation in the time series through the spectral sequence convolution Spe-Seq, and extracts the spatial dependency between time series using the spectral graph convolution.

8. The method for detecting anomalies in the approach and landing phase based on flight data according to claim 5, characterized in that: In the step 2, two StemGNN blocks are used for residual connection, the residual between the output value of the second StemGNN block and the reconstructed value of the first StemGNN block, and finally, the outputs of the two blocks are superimposed and sent to the GLU and fully connected layers to generate a predicted value.

9. The method for detecting anomalies in the approach and landing phase based on flight data according to any one of claims 1 to 8, characterized in that: In step 3, the flight state is defined as a feature matrix extracted from the flight parameters that can characterize the aircraft motion state (control input and attitude). The input of the last fully connected layer of the StemGNN prediction model (the output of the IDFT layer) is extracted as the expression of the flight state. Let the state matrix at time t be St, as shown in formula (4): Among them, f t nl is the lth value of the nth feature at time t, n is the number of features, and l is the dimension of the feature.

10. The method for detecting anomalies in the approach and landing phase based on flight data according to claim 9, characterized in that: In step 3, the distance between the flight state matrix at the current moment and the next moment is solved as the change in flight state. The change in the state matrix at time t and time t+1 is expressed as formula (5): Where ED is the Euclidean distance, as shown in formula (6): Compare the change amount from the current flight state to the next flight state. If it exceeds the state change range, it is considered that the flight state transfer is a non-smooth transfer, with a large fluctuation, that is, a state mutation occurs, which will cause an unsafe flight incident.

Images