A flight landing quality monitoring and evaluation method and system

Abstract

The present invention provides a flight landing quality monitoring and evaluation method and system, which belongs to the fields of data mining, aviation safety, and risk assessment. The method uses the original QAR data of the aircraft to obtain Measurement indicators related to hard landings, and uses the Measurement indicators to obtain hard landing risk values. The method includes: (1) collecting the original QAR data of the aircraft; (2) extracting flight parameter measurement values, i.e., Measurement data, from the original QAR data; (3) preprocessing the data; (4) constructing a hard landing risk assessment model to obtain a hard landing risk value. Using the present invention, airlines can understand the risk level of hard landings of flights and explore the causes of the risks, facilitate the management of massive flight data and automated evaluation, and can also visualize valuable information such as risk change trends.

Description

Technical Field

[0001] The present invention belongs to the fields of data mining, aviation safety and risk assessment, and specifically relates to a flight landing quality monitoring and assessment method and system. Background Art

[0002] In modern society where the aviation industry is well developed, aviation safety has always been one of the most important topics in the aviation field. According to Boeing's statistics on major flight safety accidents from 2009 to 2019, the final landing phase is a stage where flight safety accidents are very likely to occur. The main reason is that the aircraft requires a lot of pilot operations and complex environmental changes at this stage. Pilot misjudgment or improper operation may lead to serious consequences, such as triggering safety incidents. Therefore, research and analysis of the landing phase is of great significance to the field of aviation safety.

[0003] Among landing safety events, hard landing is a typical landing safety event, which refers to an event in which the impact load (generally expressed as vertical acceleration) generated by the landing gear and the ground exceeds the specified limit when the aircraft lands. Among the safety events in the landing phase, hard landing is one of the most frequent unsafe events. From 2006 to 2011, there were 125 unsafe hard landing events in my country's civil aviation, accounting for about 20% of the total number of unsafe landing events. From 1993 to 2002, 385 accidents occurred worldwide, of which hard landings caused 2 aircraft to be destroyed, 47 aircraft to be seriously damaged, and 11 aircraft to be slightly damaged. As a typical unsafe landing risk event, hard landing will not only bring a bad flight experience to passengers and damage the image of airlines, but also accelerate the fatigue damage or even fracture of wings, landing gear, engines, and the overall structure of the aircraft, increase the probability of landing safety accidents, bring huge economic losses to airlines, and in serious cases, cause catastrophic accident consequences, posing a threat to the safety of passengers' lives.

[0004] At present, the research on hard landings can be divided into two categories: the first category is based on management and psychology, and the second category is based on QAR (quick access recorder, an airborne flight data recording device with protection device) data.

[0005] The first category of researchers are generally professionals in the field of flight. They have a deep understanding of the field of civil aviation and mainly analyze the causes of hard landings and preventive measures from the perspectives of management and psychology. Shi Miao and others analyzed and thought about the problem of hard landings, and summarized the causes of hard landings into three categories: weather reasons, pilot operating technology reasons, and other reasons. Li Jiahua summarized the causes of hard landings into three categories: unstable approach, improper flight control, and misjudgment caused by environmental or special weather conditions. In response to the above-mentioned causes of hard landings, the author gave improvement suggestions from the aspects of pilot operating technology points, attention allocation, pilot form, and landing airport weather forecast.

[0006] The second category is the hard landing research based on QAR. This type of research is called data-driven research because it mainly uses machine learning and risk assessment models to analyze and mine QAR data to find the cause or probability of hard landing from the data. Compared with research based on management and psychology, this type of research speaks with data, so its results are more objective and can help flight experts find some new and useful information. Hard landing research based on QAR data has gradually become the current mainstream. Cao Haipeng and others designed a three-layer BP neural network to predict hard landings, and some scholars have studied the hard landing prediction model based on SVM. Tong Chao and others used a deep learning framework to solve the hard landing problem. According to the time series characteristics of QAR data, they proposed a hard landing prediction framework based on long short-term memory network. These methods are predictions made before the hard landing occurs, hoping to avoid the occurrence of hard landings as much as possible. In terms of risk assessment, Wang Lei proposed a quantitative assessment model for hard landing risks based on QAR. This model defines the risk of hard landing as the product of the probability of a hard landing and the severity of the hard landing event. On this basis, he further improved the risk model and took into account three types of events: hard landing, runway deviation, and tail strike.

[0007] However, the existing risk assessment methods only consider the probability and severity of occurrence to conduct risk assessment on hard landings, but do not explore the characteristic parameters of special time points or special time periods that cause the risk. Summary of the invention

[0008] The purpose of the present invention is to solve the problems existing in the above-mentioned prior art and to provide a flight landing quality monitoring and evaluation method and system, so as to facilitate airlines to understand the risk level of hard landing of flights and explore the causes of the risks, to facilitate the management and automated evaluation of massive flight data, and to visualize valuable information such as risk change trends.

[0009] The present invention is achieved through the following technical solutions:

[0010] A first aspect of the present invention provides a flight landing quality monitoring and evaluation method, wherein the method uses original QAR data of an aircraft to obtain Measurement indicators related to a hard landing, and uses the Measurement indicators to obtain a hard landing risk value.

[0011] A further improvement of the present invention is:

[0012] The method comprises:

[0013] (1) Collect the original QAR data of the aircraft;

[0014] (2) Extracting flight parameter measurements, i.e., Measurement data, from the original QAR data;

[0015] (3) Preprocess the data;

[0016] (4) Construct a hard landing risk assessment model and obtain the hard landing risk value.

[0017] A further improvement of the present invention is:

[0018] The operation of step (2) includes:

[0019] The flight parameter measurements, i.e., Measurement data, are extracted from the original QAR data using Measurement software.

[0020] A further improvement of the present invention is:

[0021] The operation of step (3) includes:

[0022] (31) Preprocessing the original QAR data;

[0023] (32) Preprocess the Measurement data.

[0024] A further improvement of the present invention is:

[0025] The operation of step (4) includes:

[0026] (41) Extracting data related to heavy landing from the preprocessed Measurement data as Measurement indicators, and drawing a box plot of each Measurement indicator;

[0027] (42) Establish reference ranges for each Measurement indicator and determine the weight of each Measurement indicator;

[0028] (43) Obtain the hard landing risk value.

[0029] A further improvement of the present invention is:

[0030] The operation of drawing a box plot of each Measurement indicator in step (41) includes:

[0031] Select a time period of 30 seconds forward and 20 seconds backward from the touchdown time, a total of 50 seconds, as the landing interval;

[0032] For each flight segment, first extract the change curve of the vertical load VRTG in the landing interval, and then obtain the peak value of the change curve. If the peak value is greater than the threshold, the flight segment is determined to be a hard landing segment. If the peak value is less than or equal to the threshold, the flight segment is determined to be a non-hard landing segment.

[0033] Draw a box plot for each Measurement indicator, including non-hard landing segments and hard landing segments.

[0034] A further improvement of the present invention is:

[0035] The operation of step (42) includes:

[0036] Select the measurement indicators that significantly distinguish between hard landing and non-hard landing segments based on the box plot;

[0037] For each Measurement indicator that can significantly distinguish between hard landing and non-hard landing segments, obtain the range of 90% of the measurement values ​​of the Measurement indicator in the landing interval, and use the range as the reference range of the Measurement indicator; according to the ability to distinguish between hard landing segments and non-hard landing segments, set the weight of the Measurement indicator, and the weight range is [0-1].

[0038] A further improvement of the present invention is:

[0039] The operation of step (43) includes:

[0040] The hard landing risk value Risk of a flight segment is calculated using the following formula:

[0041]

[0042] Where M represents the set of all Measurement indicators used to calculate the hard landing risk value, m i is the i-th Measurement indicator, I{*} is a symbolic function, which means: when the condition in the brackets is met, I{*}=1, otherwise I{*}=0, w i is the weight of the i-th Measurement indicator;

[0043] macth rule indicates that the calculation rule is satisfied, as follows: The calculation rule includes three value types: Left, Right, and Both. Left indicates that the measured value deviates to the left outside the reference range, Right indicates that the measured value deviates to the right outside the reference range, and Both indicates that the measured value deviates to the left or to the right outside the reference range.

[0044] A second aspect of the present invention provides a flight landing quality monitoring and evaluation system, the system comprising:

[0045] A collection unit, used to collect the original QAR data of the aircraft;

[0046] An extraction unit, connected to the acquisition unit, for extracting flight parameter measurement values, i.e., Measurement data, from the original QAR data;

[0047] A preprocessing unit, connected to the acquisition unit and the extraction unit respectively, for preprocessing the data;

[0048] A risk value calculation unit is connected to the preprocessing unit and is used to construct a hard landing risk assessment model to obtain a hard landing risk value;

[0049] Visualization unit: connected to the collection unit, extraction unit, preprocessing unit, and risk value calculation unit respectively, and used for visual display of data.

[0050] Compared with the prior art, the beneficial effects of the present invention are: by using the present invention, airlines can understand the risk level of hard landings of flights and explore the causes of the risks, facilitate the management of massive flight data and automated evaluation, and can also visualize valuable information such as risk change trends. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Effect diagram of different interpolation methods;

[0052] Figure 2 Wind speed and direction change diagram;

[0053] Figure 3 Example of Measurement table content:

[0054] Figure 4-1 Box plot of the aircraft's maximum descent rate from Touchdown Gate 1 (200 feet altitude) to the 50-foot stage;

[0055] Figure 4-2 Box plot of the aircraft's minimum descent rate from Touchdown Gate 1 (200 feet altitude) to the 50-foot stage;

[0056] Figure 5-1IVV curve for heavy landing;

[0057] Figure 5-2 IVV curve for non-heavy landing;

[0058] Figure 6-1 Box plot comparison results of the average descent rate from 50ft to touchdown;

[0059] Figure 6-2 Comparison results of the box plot of the maximum descent rate from 50 ft to touchdown;

[0060] Figure 6-3 Comparison results of the box plot of the minimum descent rate from 50 ft to touchdown;

[0061] Figure 7 Comparison results of box plots of 50ft to touchdown time;

[0062] Figure 8 Minimum VRTG value during landing phase;

[0063] Fig. 9 Comparison results of the box plot of the maximum pitch angle during the landing phase;

[0064] Fig.10 Comparison results of the box plot of the maximum pitch angle at the moment of touchdown;

[0065] Fig.11 Comparison results of the box plot of the maximum pitch angle from 50ft to touchdown;

[0066] Fig.12 Comparison results of the box plot of the maximum absolute roll angle at the moment of touchdown;

[0067] Fig.13 Box plot of the maximum absolute roll angle from 50ft to touchdown;

[0068] Figure 14-1 Energy box plot at the moment of touchdown;

[0069] Figure 14-2 Energy box plot at 50ft;

[0070] Figure 15-1 Aircraft ground speed box line diagram at the time of touchdown;

[0071] Figure 15-2 Aircraft ground speed box line diagram at 50ft;

[0072] Figure 16-1 Box plot of maximum ground speed during landing phase;

[0073] Figure 16-2 Box plot of maximum airspeed during landing phase;

[0074] Fig.17Box plot comparison results of maximum lateral acceleration during landing phase;

[0075] Fig.18 Comparison results of the minimum side wind box line diagram from 50ft to touchdown;

[0076] Fig.19 Box plot of hard landing risk value comparison;

[0077] Fig. 20 Scatter plot of the relationship between hard landing risk and VRTG peak value;

[0078] Fig.21 The overall framework diagram of the system of the present invention;

[0079] Fig. 22 Implementation framework diagram of the system of the present invention;

[0080] Fig.23 Data drilling analysis based on Teblueau;

[0081] Figure 24-1 Case analysis of hard landing risks;

[0082] Figure 24-2 Case analysis of hard landing risks;

[0083] Figure 24-3 Case analysis of hard landing risks;

[0084] Figure 24-4 Case analysis of hard landing risks;

[0085] Fig.25 A flowchart of the steps of the method of the present invention. DETAILED DESCRIPTION

[0086] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0087] Characteristic parameters have great practical significance for the study of the risk level and causes of hard landings. The present invention uses characteristic parameters of certain time points or time periods of the Measurement data, and considers the degree to which the characteristic deviates from the normal range to construct a hard landing risk assessment model, which is convenient for airlines to understand the risk level of hard landings of flights and explore the causes of the risks. On this basis, the present invention makes the system a tool to facilitate the management and automated evaluation of massive flight data, and can also visualize valuable information such as risk change trends.

[0088] The present invention provides a flight landing quality monitoring and evaluation method, such as Fig.25 As shown, the method includes:

[0089] (1) Collecting the original QAR data of the aircraft:

[0090] The original QAR data of the aircraft completely records various parameters of the aircraft during the entire flight phase, including the aircraft's own state parameters, the pilot's control command parameters, external environment parameters, etc. Table 1 lists some typical QAR data.

[0091]

[0092]

[0093] Table 1

[0094] (2) Extract flight parameter measurements from the original QAR data, i.e., Measurement data:

[0095] The existing tool "Measurement" can be used to extract flight parameter measurement values ​​from the original QAR data, and the extracted data is called Measurement data.

[0096] Measurement is a data measurement and collection tool developed by Teledyne for aviation applications. It can extract the flight parameter measurements that users are interested in from the original QAR data through operations such as decoding, calculation, measurement, and aggregation. The flight parameter measurements extracted by Measurement are divided into the following four categories:

[0097] Information Measurements: This type of Measurements contains basic information about a flight, such as aircraft model, departure airport ICAO code, landing airport ICAO code, flight number, etc.

[0098] Scalar Measurements: This type of Measurements information is the measurement value obtained from a single sampling time (in seconds) of a flight segment, such as the touchdown time.

[0099] Aggregate Measurements: These are measurements aggregated over a predefined time range using a predefined aggregation method. The time range may include two time points, a flight phase, or an entire flight segment. The aggregation methods include sum, max, min, average, etc.

[0100] Combined Aggregate Measurements: These measurements are obtained by aggregating the same measurement objects in multiple ranges, such as all cruise phases of a flight segment.

[0101] Measurement itself has predefined many commonly used measurement values. In addition, it also supports users to expand on existing Measurement definitions to obtain customized Measurements.

[0102] (3) Preprocess the data:

[0103] First, the original QAR data is preprocessed. For example, in order to obtain the magnitude of the aircraft's axial wind and lateral wind, the three parameters of wind speed, wind direction, and magnetic heading in the original QAR data need to be processed to obtain two new flight parameters of axial wind and lateral wind. It is also necessary to process the original QAR data of different frequencies and digitize the parameters of the state type.

[0104] After the original QAR data is preprocessed, the Measurement data is preprocessed.

[0105] The operation of step (3) is as follows:

[0106] (31) Preprocess the original QAR data:

[0107] Most of the QAR data are of numerical type, such as radio altitude, airspeed, ground speed, pitch angle, descent rate, vertical load, etc., while other parameters are discrete state variables. For example, the landing gear state mainly includes two states, AIR and GROUND, which correspond to the on and off of the landing gear air-ground switch, respectively. AIR means that the landing gear is in the air (not grounded), and GROUND means that the landing gear is grounded. In order to visualize the landing gear state of the aircraft, it is necessary to convert the AIR and GROUND states into numerical types. In order to make the result as intuitive as possible, the present invention converts the AIR state to 1 and the GROUND state to 0. Similar operations can also be taken for other discrete state parameters (AP_EGD1, AP_EGD2, ATHR_EGD, FLAP_LEVEL1, FLAP_LEVEL2).

[0108] In addition, the sampling frequencies of different QAR data are also different. In order to visualize the QAR data change curve in a fine-grained manner and to be able to compare and analyze different parameter curves in the same figure, the present invention needs to interpolate some data with lower sampling frequencies and expand their sampling frequencies to the highest sampling frequency (8Hz). Different interpolation methods are required for different types of data. For example, for numerical parameters such as radio altitude, linear interpolation can be used. Since the sampling frequency of radio altitude itself is 4Hz, to expand it to 8Hz, it is only necessary to insert their average value (x1+x2) / 2 between two adjacent data points x1 and x2. However, for discrete state parameters such as landing gear status, linear interpolation is not suitable. Assuming that the AIR state corresponds to 1 and the GROUND state corresponds to 0, the intermediate value 0.5 may be obtained after interpolation. This value has no corresponding landing gear state corresponding to it. For such parameters, the present invention adopts the "covering" interpolation method, that is, the interpolation point is covered with the previous point closest to the point (that is, the current point and the previous point have the same value). The effects of the two interpolations are as follows. Figure 1 In summary, for QAR parameters of different types and sampling frequencies, corresponding interpolation methods should be adopted according to specific circumstances.

[0109] In addition to the above data transformation and interpolation processing, some parameters also need to be transformed at an advanced level, including the two parameters of wind speed and wind direction. Studies have shown that wind shear during the landing phase is also one of the causes of unsafe landing incidents. However, wind shear is relative to the flight direction of the aircraft (headwind becomes tailwind, left wind becomes right wind, etc.). It is difficult to obtain accurate information about wind shear directly from the wind speed and wind direction parameters. Therefore, it is necessary to combine the aircraft's magnetic heading and wind speed and direction to calculate the effect of wind on the aircraft. The calculation method is: first calculate the angle θ=β-α between the wind direction and the aircraft's magnetic heading, where β is the wind direction and α is the magnetic heading. Then use the trigonometric function relationship to calculate the components of the wind speed parallel and perpendicular to the aircraft's heading, as follows:

[0110]

[0111] Among them, WIN SPD Indicates wind speed, WIN CRS Indicates the size of the crosswind. A positive value indicates a left-side wind and a negative value indicates a right-side wind. ALG Indicates the axial wind size, a positive value indicates a tailwind, and a negative value indicates a headwind. The calculation diagram is as follows Figure 2 shown.

[0112] (32) Preprocess the Measurement data:

[0113] The present invention performs data preprocessing operations based on Measurement data, including data cleaning, missing value filling, data association, etc. For AircraftType field preprocessing, the present invention first analyzes the AircraftType field of the Flight table (the Flight table is derived from the Measurement database and is a field containing various flight parameter fields (such as flight time, aircraft model, i.e., AircraftType, etc.), and finds that the same aircraft model corresponds to two names in this field, such as B737 and 737. For such fields, the present invention adopts a processing method of unifying the field values ​​into standard values, such as modifying all "737" to "B737". For the missing processing of field content, in the Measurement data, there are some empty segments in the fields related to take-off and landing airports. When performing airport-related analysis on such data, they are uniformly classified as other airport categories.

[0114] The Measurement table stores the measurement values ​​of multiple flight segments. Figure 3 As shown in the figure, it contains fields such as FlightID, MeasurementCode, and MeasurementDesc. Table 2 lists the field names and specific meanings in the Measurement table. The FlightID field specifies which flight segment the measurement value comes from, so the Flight table can be associated with the Measurement table through FlightID to obtain all the Measurement values ​​of a certain flight segment.

[0115]

[0116]

[0117] Table 2

[0118] In order to determine whether MeasurementCode can be used as a unique identifier for a certain measurement method, the present invention analyzes the correspondence between MeasurementCode and MeasurementDesc, MeasurementName, MeasurementRange, RangeType, and Aggregate, and the results are shown in Table 3. It can be seen that there is a one-to-one correspondence between MeasurementCode and MeasurementDesc, but there is no one-to-one correspondence between MeasurementCode and other fields. However, for the same MeasurementCode, its MeasurementDesc, MeasurementName, MeasurementRange, RangeType, and Aggregate are all unique, so MeasurementCode can be used as a unique identifier for a certain measurement method.

[0119]

[0120] Table 3

[0121] The present invention selects some fields related to landing safety, extracts their MeasurementValue measurements, and displays them in a box plot. It is found that some measurement values ​​have obvious anomalies. Take the descent rate (IVV) in the landing phase as an example. Figure 4-1 and Figure 4-2 As shown, Figure 4-1 This is a box plot of the maximum descent rate of the aircraft from Touchdown Gate 1 (200 feet altitude) to 50 feet. Figure 4-2 is the minimum drop rate. It can be seen that the maximum and minimum drop rates contain some values ​​that are obviously deviated from the normal value range (greater than 10000 or less than -10000).

[0122] In addition to this parameter, similar outlier values ​​were found in some other parameters. To solve the above problems, the present invention proposes a rule-based outlier filtering method, which combines expert experience to provide the normal value range of the main Measurement indicators. When cleaning the Measurement data, the Measurement values ​​that exceed this range are removed, thereby obtaining higher quality data.

[0123] (4) Constructing a hard landing risk assessment model:

[0124] (41) Extract the data related to the heavy landing from the preprocessed Measurement data as the Measurement index, and draw a box plot of each Measurement index. The box plot includes two parts: non-heavy landing and heavy landing:

[0125] First, in order to analyze the hard landing event, the present invention extracts the landing interval from the preprocessed QAR data, and selects 30 seconds forward and 20 seconds backward from the touchdown time, a total of 50 seconds as the landing interval. The present invention analyzes the change law of various aircraft parameters within this interval. For each flight segment, the change curve of the vertical load (VRTG) in the landing interval is first extracted, and then the peak value of the VRTG curve is obtained. If the peak value is greater than the threshold, it is determined that a hard landing has occurred in the flight segment, that is, a hard landing flight segment. If it is less than or equal to the threshold, it is determined that no hard landing has occurred in the flight segment, that is, a non-hard landing flight segment. The flight segment can be divided into a hard landing flight segment and a non-hard landing flight segment by the threshold. When drawing the box plot below, it is also divided into non-hard landing and hard landing parts according to the threshold.

[0126] In this embodiment, the analysis is performed based on QAR data, and the aircraft models are all A320, which includes 53 hard landing segments (VRTG peak value is greater than 1.8G) and 101 non-hard landing segments.

[0127] Since the occurrence of a hard landing is usually closely related to the aircraft descent rate, the present invention further performs a visual analysis on the descent rate (IVV) curve, such as Figure 5-1 and Figure 5-2 As shown in the figure, it can be seen that before entering the 50-foot altitude, the IVV fluctuation of the hard landing segment is larger, while the IVV of the non-hard landing segment is more stable. In addition, in the interval from 50ft to touchdown, the IVV curve characteristics of the two show significant differences, such as Figure 5-1 and Figure 5-2 As shown in the figure, the IVV curve of the heavy landing segment is "convex" in shape, while the non-heavy landing segment is "convex". This feature shows that after the aircraft reaches the 50ft altitude, the heavy landing segment does not control the aircraft's descent rate in time, because the aircraft touches the ground quickly, resulting in a large vertical load. After the non-heavy landing reaches the 50ft altitude, the descent rate control is more obvious, so the descent rate of the aircraft when touching the ground is small, avoiding the occurrence of a heavy landing. When the aircraft reaches the 50ft altitude, timely and effective control of the aircraft's descent rate is the key to avoiding a heavy landing.

[0128] Extract the data related to the heavy landing from the preprocessed Measurement data as the Measurement indicator:

[0129] Through systematic sorting, the Measurement indicators include two categories: one is Scalar Measurement, and the other is Aggregate Measurement. Table 4 gives the Scalar Measurement indicators related to hard landing, and Table 5 gives the Aggregate Measurement indicators related to hard landing (these indicators are directly extracted from QAR data, and are named different Measurement indicators according to different selected parameters and different time points, such as Airspeed at Touchdown, which means the airspeed value of the aircraft at the moment of landing.).

[0130]

[0131]

[0132] Table 4

[0133]

[0134]

[0135] Table 5

[0136] Based on the above-mentioned Measurement indicator results, 49 Measurement indicators are finally selected for extraction. The present invention uses box plots to compare the differences between hard landing segments and non-hard landing segments in different indicators, thereby assisting in discovering which indicators have a more important reference significance for calculating hard landing risks.

[0137] Figure 6-1 to Figure 6-3 The box plots of the avg_ivv_50ft_td, max_ivv_50ft_td, and min_ivv_50ft_td (average, maximum, and minimum descent rates from 50ft to touchdown) indicators are shown. It can be seen that the descent rate of the hard landing segment is significantly higher than that of the normal landing segment. Figure 6-1 to Figure 6-3 The vertical axis in the figure is the descent rate. The three figures compare the descent rate range of hard landing flights and non-hard landing flights. For example, the descent rate of hard landing flights is greater (the negative number of the vertical axis indicates the direction, and the value indicates the descent speed). Figure 6-1 to Figure 6-3 Among them, the three descent rates (average, maximum, and minimum) of heavy landing flights are all greater than the descent rate of non-heavy landings, so the descent rate parameter is more important for distinguishing between heavy landings and non-heavy landings.

[0138] Figure 7The box plot comparison results of time_50ft_td (50ft to touchdown time) are shown. Figure 7 It can be seen that the 50-foot to touchdown time for the hard landing segment is significantly lower than the 50-foot to touchdown time for the non-hard landing segment.

[0139] Figure 8 The box plot comparison results of min_vrtg_land (minimum VRTG value in the landing phase) are shown. It can be seen from the figure that in the landing phase, the minimum VRTG of the hard landing segment is significantly lower than the minimum VRTG value of the non-hard landing segment, and the fluctuation range of the minimum VRTG value of the hard landing segment is larger, indicating that the hard landing segment is usually accompanied by larger VRTG fluctuations in the landing phase.

[0140] Fig. 9 The box plot comparison results of max_pitch_land (maximum pitch angle during the landing phase) are given. It can be seen from the figure that during the landing phase, the maximum pitch angle of the heavy landing segment is significantly higher than the maximum pitch angle of the non-heavy landing segment, indicating that the heavy landing segment has a large attitude change during the landing phase and is not as stable as the non-heavy landing segment.

[0141] Fig.10 The box plot comparison results of max_pitch_td (maximum pitch angle at touchdown) are given. Similarly, the pitch angle of the hard landing segment is significantly higher.

[0142] Fig.11 The box plot comparison results of max_pitch_50ft_td (maximum pitch angle from 50ft to touchdown) are given. Unlike the previous example, the maximum pitch angle of the hard landing segment is not significantly higher than that of the non-hard landing segment. The main difference is that the maximum fluctuation range of the pitch angle of the hard landing segment is larger than that of the non-hard landing segment, indicating that the aircraft's attitude changes more dramatically during this period of time in the hard landing segment. Therefore, keeping the aircraft's attitude stable from 50ft to touchdown can also help reduce the risk of hard landing.

[0143] Fig.12 The box plot comparison results of max_abs_roll_td (maximum absolute roll angle at touchdown) are given. Fig.12 It can be seen that the absolute value of the maximum roll angle of the heavy landing segment at the touchdown moment is significantly higher than that of the non-heavy landing segment, indicating that the heavy landing segment generally has a relatively large roll attitude at the touchdown moment.

[0144] Fig.13 The box plot comparison results of max_abs_roll_50ft_td (maximum absolute roll angle from 50ft to touchdown) are given. Fig.12 similar.

[0145] Figure 14-1 and Figure 14-2 The box plots of the aircraft's energy (kinetic energy + gravitational potential energy, unit: megajoule) at touchdown and at an altitude of 50ft are given respectively. It can be seen that the average energy of the hard landing segment is significantly higher than that of the non-hard landing segment, indicating that energy has a certain influence.

[0146] Figure 15-1 and Figure 15-2 The box plot comparison results of the aircraft ground speed at touchdown and 50ft altitude are given respectively. It can be seen that the average ground speed of the hard landing segment is significantly higher than that of the non-hard landing segment, indicating that excessive ground speed will increase the risk of hard landing to a certain extent.

[0147] Figure 16-1 and Figure 16-2 The box plot comparison results of the maximum ground speed (GS) and the maximum airspeed (IAS) during the landing phase are given. It can be seen that the maximum ground speed and airspeed during the landing phase of the heavy landing segment are significantly higher than those of the non-heavy landing segment.

[0148] Fig.17 The box plot comparison results of max_abs_latg_land (maximum lateral acceleration during landing phase) are given, and it can be seen that the hard landing segment is significantly higher.

[0149] Fig.18 The box plot comparison results of min_win_lat_50ft_td (minimum crosswind from 50ft to touchdown) are given. It can be seen that the minimum crosswind in the hard landing segment is significantly larger, indicating that the hard landing segment usually suffers from a larger crosswind from 50ft to touchdown, further indicating that environmental factors (crosswind) are one of the causes of hard landing events.

[0150] Risk assessment emphasizes evaluating the risk of a hard landing for any flight data (including hard landings and non-hard landings) through comprehensive parameter information before landing. For example, if the IVV of some flights is too large at the 50ft moment, the risk of a hard landing can be evaluated through hard landing risk assessment.

[0151] According to the above statistical analysis results, the present invention designs a hard landing risk assessment model, which includes two stages, namely, the Measurement reference range establishment stage and the risk calculation stage, as follows:

[0152] (42) Establish reference ranges for each Measurement indicator:

[0153] According to the box plot, the Measurement indicators that significantly distinguish the hard landing and non-hard landing segments are selected. Then, based on these indicators, for the non-hard landing segment, the Measurement values ​​of the non-hard landing segment are calculated (the Measurement values ​​are the measurement values ​​of different parameters at different time points or time periods within the 50 seconds from 30 seconds before touchdown to 20 seconds after touchdown, such as max_ivv_50ft_td, which represents the maximum descent rate value within the time period from 50 feet before touchdown to touchdown. Other parameters are also obtained based on the Measurement data preprocessing part) 90% range (meaning that 90% of the measurement values ​​are within this range) is used as the reference range of the Measurement indicator.

[0154] For each selected Measurement indicator, a weight is given according to its ability to distinguish between hard landing and non-hard landing segments, and the weight value range is [0~1].

[0155] (43) Calculation of hard landing risk value

[0156] For a flight segment, first obtain the various Measurement values ​​of the flight segment, and then determine whether each Measurement value exceeds the reference range established in step (42). If it exceeds the reference range, add the risk value to the weight corresponding to the Measurement indicator (i.e., multiply it by the weight). Finally, normalize the risk value to obtain the hard landing risk value Risk of the flight segment.

[0157] Taking the ivv (descent rate) at touchdown as an example, the reference range of ivv for non-heavy landing is [-200 to -400] (i.e. 90% of ivvs are within this range). For a given single segment, the ivv at the touchdown time of this segment is -500, i.e. -500 exceeds the reference range by 100 (-500-(-400)). When calculating the size, the absolute value is taken, i.e. the speed of 500 exceeds the normal range (200 to 400) by 100.

[0158] The present invention provides the measurement index weights and calculation rules related to hard landing, as shown in Table 6, wherein the calculation rule Rule has three value types, namely Left, Right and Both, and Rule = Left means that when the measurement index of the flight segment deviates to the left outside the index reference range, the weight is accumulated in the hard landing risk. For example, if the reference range of a certain measurement index is [100-200], and the measurement value of the measurement index of the current flight segment is 50, then the index of this flight segment deviates to the left outside (i.e., less than the lower limit of the reference range) the index reference range.

[0159] The calculation formula (i.e., evaluation model) of the hard landing risk value of the present invention is:

[0160]

[0161] Where M represents the set of all measurement indicators used to calculate risk (i.e., all selected measurement indicators that significantly distinguish between hard landing and non-hard landing segments), m i Belongs to M, that is, the i-th Measurement indicator, I{*} is a sign function, when the condition in the brackets is met, I{*}=1, otherwise I{*}=0, the denominator is normalized by summing all weights to ensure that the calculated risk values ​​are between [0,1]. i is the weight of the i-th Measurement indicator.

[0162] For the i-th measurement indicator of a given flight segment, determine whether the measurement value of this indicator deviates to the left or to the right of the reference range. Deviating to the left means that the measurement value of this indicator is less than the lower limit of the reference range, and deviating to the right means that the measurement value of this indicator is greater than the upper limit of the reference range. For example, if the reference range is [100-200], the measurement value of the indicator is 40, which means it deviates to the left by 60 (100-40=60), and the measurement value of the indicator is 230, which means it deviates to the right by 30 (230-200=30). The "match rule" in the above formula i " is to put the m i The metric matches this rule and is calculated accordingly.

[0163] "Left" means "only the left deviation meets the rule", "right deviation" does not meet the rule, "Right" is the opposite; "both" means "either the left deviation or the right deviation" meets the rule. If the rule is "left deviation", then "matchrule" in the above formula i " means that only when it deviates to the left, I is equal to 1, and when it deviates to the right or is within the reference range, I is equal to 0; if the rule is "Both", then "match rule i " means that when it deviates to the left or right, I is equal to 1, and only when it is within the reference range, I is equal to 0.

[0164] When using the present invention, the Measurement index that significantly distinguishes the hard landing segment from the non-hard landing segment is selected each time, and then the reference range and weight of each index are established, and finally the risk value is obtained according to the calculation formula.

[0165] No Measurement Weight Rule 1 avg_ivv_50ft_td 1 Left 2 max_ivv_50ft_td 0.7 Left 3 min_ivv_50ft_td 0.5 Left 4 time_50ft_td 0.6 Left 5 min_vrtg_land 0.3 Right 6 max_pitch_land 0.4 Right 7 max_pitch_td 0.3 Right 8 max_pitch_50ft_td 0.2 Both 9 max_abs_roll_td 0.3 Right 10 max_abs_roll_50ft_td 0.2 Right 11 energy_td 0.3 Right 12 energy_50ft 0.2 Right 13 gs_td 0.3 Right 14 gs_50ft 0.2 Right 15 max_gs_land 0.1 Right 16 max_ias_land 0.2 Right 17 gw_td 0.2 Right 18 ias_td 0.1 Right 19 max_abs_latg_land 0.2 Right 20 max_win_lat_50ft_td 0.1 Right

[0166] Table 6

[0167] Based on the QAR data of 100 non-hard landing segments, the present invention counts the 90% reference range of 49 Measurement indicators, and the results are shown in Table 7, where lower_bound and upper_bound represent the left and right boundaries of the 90% reference range respectively. The 20 indicators in Table 6 are some of the related indicators listed to illustrate the weights of different indicators. The 49 indicators given in Table 7 are all indicators related to hard landing, and the boundary values ​​of the reference range are given.

[0168]

[0169]

[0170] Table 7

[0171] Based on the evaluation model, the present invention calculates the risk values ​​of 52 hard landing segments and 100 non-hard landing segments. The results are shown in the box plot comparison. Fig.19 As shown, it can be seen that the risk value of the hard landing segment (average value is about 0.4) is significantly higher than the risk value of the non-hard landing segment (average value is about 0.05).

[0172] Fig. 20 The relationship between the hard landing risk value and the VRTG peak value is further given through a scatter plot. Fig. 20 It can be seen that, overall, there is a certain positive correlation between the hard landing risk value and VRTG. For the hard landing segment, this relationship is more obvious, while for the non-hard landing segment, the correlation is not very significant.

[0173] The present invention also provides a flight landing quality monitoring and evaluation system, which is as follows:

[0174] The system comprises:

[0175] A collection unit, used to collect the original QAR data of the aircraft;

[0176] An extraction unit, connected to the acquisition unit, for extracting flight parameter measurement values, i.e., Measurement data, from the original QAR data;

[0177] A preprocessing unit, connected to the acquisition unit and the extraction unit respectively, for preprocessing the data;

[0178] A risk value calculation unit is connected to the preprocessing unit and is used to construct a hard landing risk assessment model to obtain a hard landing risk value;

[0179] Visualization unit: connected to the collection unit, extraction unit, preprocessing unit, and risk value calculation unit respectively, and used for visual display of data.

[0180] In order to facilitate the tooling of the above risk assessment model, the present invention designs the tooling system from three aspects.

[0181] First, the model evaluation results of all flights (i.e., the risk value calculated for each flight using the above method) are centrally stored and managed. Secondly, the incremental flights (newly added flights) can be automatically evaluated using the model. Finally, from an overall dimension, the landing quality of the flight is visualized, which can be used to measure the landing quality of the aircraft, i.e., the landing quality.

[0182] For the centralized management of flight key data, the flight landing quality assessment model requires some key data of the flight to complete the calculation. Faced with tens of millions of flights each year, the present invention needs to collect and manage key data in a timely manner, and promptly support efficient queries. For the automation of incremental flight evaluation, the present invention is not only used to evaluate historical flight segment data, but also needs to evaluate the flight landing quality of newly added flights. When new flight data is added, the entire model evaluation process of data collection, risk calculation, and data storage needs to be automatically implemented. Finally, for the visualization of flight landing quality, the massive flight data stored in the system of the present invention hides valuable information such as risk change trends, and such valuable information can be visualized.

[0183] The overall framework of the system of the present invention is as follows Fig.21As shown in the figure, the data source is mainly provided by AIRFASE (AIRFASE is a decoding software. The original QAR data needs to be decoded by AIRFASE and converted into a csv file for data analysis). The back-end database uses MySQL to store some intermediate information of the system. The risk calculation is implemented in Python. By calling the Python FLASK interface, the risk value of the flight segment is obtained in the form of HTTP request. The system scheduling is implemented through XXLJOB, and the front-end uses Tableau to visualize the statistical analysis of the risk on the page. Among them, XXLJOB is a lightweight distributed task scheduling platform, which is used in the system to schedule and execute scheduled tasks, complete the data acquisition of incremental flights, risk calculation and other functions; OkHttp is an open source project for processing network requests, and is a lightweight framework for the Android side; MyBatis is an excellent persistence layer framework, which supports custom SQL, stored procedures and advanced mapping, encapsulates native JDBC, and supports multiple databases. In the present invention, the data of Airface's SqlServer database is read and stored in the MySQL database; FLASK is a lightweight Web application framework written in Python, which is more flexible, lightweight, safe and easy to use than other similar frameworks. The present invention deploys the core algorithm for flight risk calculation on the WEB server built by the FLASK framework, and the system can complete the calculation of flight risk by calling the data structure. MySQL is a relational database management system that manages the database and stores data in different tables instead of putting all data in a large warehouse, which increases speed and flexibility; Tableau is a mature BI software. In the present invention, the tool visualizes the structured data such as flight risk stored in MySQL.

[0184] The system of the present invention is developed and implemented in a way of separating the front-end and the back-end. Fig. 22 shown.

[0185] A MySQL database is established at the back end of the system of the present invention for storing flight information, Measurements, References, flight risks and other data; the flight information, Measurements and other data stored in the SqlServer database of Airfase are efficiently read based on the MyBatis framework of JAVA, and after being structured according to the configuration relationship, they are written into the MySQL database of the system; the data interface is called using the OkHttp technology, and after completing the landing quality calculation of the flight, the result is stored in the MySQL database; the algorithm of the flight landing quality evaluation model is deployed on the WEB server based on the FLASK framework of Python, and the system is called to complete the calculation through the data interface.

[0186] The front end uses TABLEAU to directly connect to the MySQL database, and uses the structured data in the database to draw visual charts; line charts, box plots, etc. are used to realize trend changes, airline comparisons, and key airports. At the same time, the dashboard is used reasonably to realize data aggregation and drilling. The effect is as follows Fig. 22 shown. Fig. 22 The “screening” in the above is the screening of risk factors, that is, the screening of measurement indicators with high correlation with heavy landings. This screening is achieved by preprocessing the measurement and drawing box plots (Figure 6- Fig.18 ) is used for screening, and after screening out these measurement indicators, the risk assessment model is used to calculate the risk value.

[0187] Fig. 22 The "risk factor display" in the figure refers to the visualization of the measurement indicator through a box plot, which can intuitively show the correlation between the indicator and the hard landing. This part is the visualization part, which is the function of the tool.

[0188] Fig. 22 The term "calculate reference" refers to calculating the reference range.

[0189] Fig. 22 The "mesurement parameter configuration table" in the figure is obtained from the measurement data preprocessing part. Configuring measurement parameters is to configure all the original QAR data within 50s into different measurement indicators, such as the IVV indicator at the touchdown time and the airspeed indicator at an altitude of 50ft.

[0190] An embodiment of the present invention is as follows:

[0191] [Example]

[0192] This embodiment selects a segment with the largest risk value (risk value: 0.3594) from the non-heavy landing segments, and visualizes its vertical load (VRTG), descent rate (IVV), pitch angle (PITCH) and crosswind (CROSS_WIND). The results are as follows: Figure 24-1 to Figure 24-4As shown. It can be seen that although this segment did not trigger a hard landing over-limit event, its IVV curve has certain similarities with the IVV curve pattern of the hard landing segment analyzed previously, and the time taken for this segment to go from 50ft to touchdown is relatively short (about 5 seconds). Observing the change in its pitch angle, it can be seen that when entering the 50ft altitude, the aircraft's attitude is relatively small, and the pitch angle is less than 1 degree. Through the crosswind analysis, it can be seen that there has been a large crosswind (8 to 10 knots) during the landing process of this segment. Therefore, through preliminary analysis, it can be determined that the risk of this segment is relatively high due to the crosswind environmental factors.

[0193] The present invention first extracts the Measurement indicators related to hard landings, and then compares the distribution range of hard landings and non-hard landings with box plots of these indicators, selects indicators with obvious discrimination, and establishes a Measurement reference range based on this, and then specifies the weight of each hard landing-related indicator according to its discrimination ability, and combines the value that exceeds the indicator reference range in the previous stage to calculate the total hard landing risk value of the segment. The actual results of the experimental data show that the risk value of the hard landing segment is significantly higher than that of the non-hard landing segment, which proves the effectiveness of the hard landing risk assessment model proposed by the present invention. In addition, in order to engineer the flight landing quality model, the present invention designs a system to centrally store and manage the model evaluation results of all flights, automatically evaluate incremental flights using the model, and visualize the flight landing quality from an overall dimension.

[0194] Finally, it should be explained that the above technical solution is only one implementation method of the present invention. For those skilled in the art, it is easy to make various types of improvements or modifications based on the application methods and principles disclosed in the present invention, and it is not limited to the method described in the above specific implementation method of the present invention. Therefore, the method described above is only preferred and does not have a restrictive meaning.

Claims

1. A flight landing quality monitoring and evaluation method, characterized in that: The method uses the original QAR data of the aircraft to obtain a Measurement index related to a hard landing, and uses the Measurement index to obtain a hard landing risk value; The method comprises: (1) Collect the original QAR data of the aircraft; (2) Extracting flight parameter measurements, i.e., Measurement data, from the original QAR data; (3) Preprocess the data; (4) Construct a hard landing risk assessment model to obtain the hard landing risk value; The operation of step (4) includes: (41) Extracting data related to heavy landing from the preprocessed Measurement data as Measurement indicators, and drawing a box plot of each Measurement indicator; (42) Establish reference ranges for each Measurement indicator and determine the weight of each Measurement indicator; (43) Obtaining a hard landing risk value; The operation of step (43) includes: The hard landing risk value Risk of a flight segment is calculated using the following formula: Where M represents the set of all Measurement indicators used to calculate the hard landing risk value, m i is the i-th Measurement indicator, I{*} is a symbolic function, which means: when the condition in the brackets is met, I{*}=1, otherwise I{*}=0, w i is the weight of the i-th Measurement indicator; macth rule indicates that the calculation rule is satisfied, as follows: The calculation rule includes three value types: Left, Right, and Both. Left indicates that the measured value deviates to the left outside the reference range, Right indicates that the measured value deviates to the right outside the reference range, and Both indicates that the measured value deviates to the left or to the right outside the reference range.

2. The flight landing quality monitoring and evaluation method according to claim 1, characterized in that: The operation of step (2) includes: The flight parameter measurements, i.e., Measurement data, are extracted from the original QAR data using Measurement software.

3. The flight landing quality monitoring and evaluation method according to claim 1, characterized in that: The operation of step (3) includes: (31) Preprocessing the original QAR data; (32) Preprocess the Measurement data.

4. The flight landing quality monitoring and evaluation method according to claim 1, characterized in that: The operation of drawing a box plot of each Measurement indicator in step (41) includes: Select a time period of 30 seconds forward and 20 seconds backward from the touchdown time, a total of 50 seconds, as the landing interval; For each flight segment, first extract the change curve of the vertical load VRTG in the landing interval, and then obtain the peak value of the change curve. If the peak value is greater than the threshold, the flight segment is determined to be a hard landing segment. If the peak value is less than or equal to the threshold, the flight segment is determined to be a non-hard landing segment. Draw a box plot for each Measurement indicator, including non-hard landing segments and hard landing segments.

5. The flight landing quality monitoring and evaluation method according to claim 1, characterized in that: The operation of step (42) includes: Select the measurement indicators that significantly distinguish between hard landing and non-hard landing segments based on the box plot; For each Measurement indicator that can significantly distinguish between hard landing and non-hard landing segments, obtain the range of 90% of the measurement values ​​of the Measurement indicator in the landing interval, and use the range as the reference range of the Measurement indicator; according to the ability to distinguish between hard landing segments and non-hard landing segments, set the weight of the Measurement indicator, and the weight range is [0-1].

6. A flight landing quality monitoring and evaluation system, characterized in that: The system comprises: A collection unit, used to collect the original QAR data of the aircraft; An extraction unit, connected to the acquisition unit, for extracting flight parameter measurement values, i.e., Measurement data, from the original QAR data; A preprocessing unit, connected to the acquisition unit and the extraction unit respectively, for preprocessing the data; A risk value calculation unit is connected to the preprocessing unit and is used to construct a hard landing risk assessment model to obtain a hard landing risk value; Visualization unit: connected to the acquisition unit, extraction unit, preprocessing unit, and risk value calculation unit respectively, for visualizing the data; The risk value calculation unit performs the following operations: (41) Extracting data related to heavy landing from the preprocessed Measurement data as Measurement indicators, and drawing a box plot of each Measurement indicator; (42) Establish reference ranges for each Measurement indicator and determine the weight of each Measurement indicator; (43) Obtaining a hard landing risk value; The operation of step (43) includes: The hard landing risk value Risk of a flight segment is calculated using the following formula: Where M represents the set of all Measurement indicators used to calculate the hard landing risk value, m i is the i-th Measurement indicator, I{*} is a symbolic function, which means: when the condition in the brackets is met, I{*}=1, otherwise I{*}=0, w i is the weight of the i-th Measurement indicator; macth rule indicates that the calculation rule is satisfied, as follows: The calculation rule includes three value types: Left, Right, and Both. Left indicates that the measured value deviates to the left outside the reference range, Right indicates that the measured value deviates to the right outside the reference range, and Both indicates that the measured value deviates to the left or to the right outside the reference range.

Images