Method for evaluating flight route flight safety risk based on data driving

Abstract

The invention discloses a data driving-based route flight safety risk assessment method. The assessment method comprises the steps of obtaining historical QAR data and performing data preprocessing; constructing a flight safety risk index system which comprises a plurality of index events influencing the flight safety in the predetermined flight scene, dividing the index events into a plurality of risk levels, and setting a corresponding risk index coefficient for each risk level; carrying out standardization processing on the risk index coefficient; calculating information entropies of the plurality of index events, and determining weight coefficients of the plurality of index events according to an entropy weight method; and for the flight to be assessed, calculating a flight safety assessment risk value of the flight. According to the method, the QAR data can be comprehensively utilized to establish a safety index monitoring model of complete team operation, an airline and other related parties can be helped to more comprehensively grasp the safety risk of an aircraft in the course operation, and the safety risk in the course operation can be objectively and comparatively evaluated.

Description

Technical Field

[0001] The present disclosure relates to a method for evaluating flight safety risks in the field of aviation, and particularly to a method for evaluating route flight safety risks based on data-driven. Background Art

[0002] Domestic and foreign research on flight safety risk assessment in the field of civil aviation transportation focuses on the exploration and research of risk assessment methods, and there are many different ideas and attempts in the industry.

[0003] For example, Alireza Ahmadi mentioned using the event tree analysis method to evaluate aviation operation accidents and their resulting costs, as well as to correctly evaluate the flight safety of airlines. Daniel P. Murray proposed the application of the analytic hierarchy process, using simplified conservative assumptions instead of complex models to quantitatively analyze flight safety risks.

[0004] Some other research has adopted the fuzzy comprehensive evaluation method. For example, Wang Lei et al. proposed considering the normal data and abnormal data of QAR (i.e., Quick Access Record, which means the quick access recorder of the aircraft) as the sample space, determining the risk interval by using the principles of mathematical statistics and quantitative risk assessment, and then establishing a prediction model for hard landings, providing a new idea for the prediction research of hard landings of airlines. Tang Weizhen proposed that when analyzing the risk factors of airline flight safety, an index system of people, machine, environment, and management was constructed by the AHP method, and the flight safety was evaluated by using a multi-level matter-element model. Liu Ningmin proposed using the SHELL model to analyze the risks in the flight operation process, and combining expert fuzzy evaluation and risk matrix to obtain the weights.

[0005] With the explosive development of machine learning and artificial intelligence algorithm technologies, more and more research scholars have proposed the idea of applying algorithms such as machine learning to flight safety risk assessment driven by QAR data.

[0006] However, there is still a lack of a flight safety risk assessment method that can effectively utilize actual flight data-driven in aviation practice, especially in the field of civil aviation, and at the same time has high implementability and relatively low computational load. Therefore, there is an urgent need to provide a new flight safety risk assessment method to at least partially alleviate or solve the above problems and defects existing in the existing solutions. Summary of the Invention

[0007] An object of the present disclosure is to propose a method for evaluating route flight safety risks based on data-driven in order to overcome at least some of the above defects existing in the existing solutions.

[0008] The present disclosure provides a method for evaluating flight safety risks of a route based on data driving. The evaluation method is characterized in that the evaluation method includes:

[0009] Obtain historical QAR data associated with a target route and perform data preprocessing. Among them, the historical QAR data includes QAR data of multiple flights;

[0010] Construct a flight safety risk index system based on multiple types of flight safety influencing factors. The flight safety risk index system includes multiple index events that affect flight safety in a predetermined flight scenario. Among them, each index event is associated with a flight safety influencing factor, and the multiple index events are divided into several risk levels, and corresponding risk index coefficients are set for each risk level;

[0011] Perform standardization processing on the risk index coefficients of the multiple index events that appear in the historical QAR data to obtain standardized risk index coefficients;

[0012] Based on the historical QAR data, calculate the information entropy of the multiple index events, and determine the weight coefficients of the multiple index events according to the entropy weight method;

[0013] For a flight to be evaluated belonging to the target route, obtain the QAR data of the flight to be evaluated, and calculate the inner product of the weight coefficients and the standardized risk index coefficients of the multiple index events included therein as the flight safety evaluation risk value of the flight to be evaluated for the predetermined flight scenario.

[0014] Among them, in the step of standardizing the risk index coefficients, the standardization process is such that the maximum value of the risk index coefficients of any index event is standardized to the same value. For example, the maximum value of the risk index coefficients of the index event is standardized to 1.

[0015] And among them, calculating the inner product of the weight coefficients and the standardized risk index coefficients of the multiple index events can be understood as the inner product of the vector expressing the weight coefficients of the multiple index events and the vector expressing the standardized risk index coefficients of the multiple index events, that is, it is equivalent to calculating the sum of the products of the weight coefficients and the standardized risk index coefficients of each index event.

[0016] According to some embodiments of the present disclosure, the step of standardizing the risk index coefficients of the multiple index events that appear in the historical QAR data includes applying the maximum-minimum normalization method to the risk index coefficients of the multiple index events that appear in the historical QAR data.

[0017] According to some preferred embodiments of the present disclosure, the multiple types of flight safety influencing factors include human operation factors, mechanical factors, and environmental factors.

[0018] According to some preferred embodiments of the present disclosure, in the step of constructing the flight safety risk index system, the warning-triggering index events are set to belong to the highest risk level, and the highest risk index coefficient is set.

[0019] According to some preferred embodiments of the present disclosure, the warning-triggering index events include some or all of the following:

[0020] Wind shear warning, terrain warning, landing gear not locked down, stall warning, smoke warning, master warning.

[0021] According to some preferred embodiments of the present disclosure, in the step of constructing the flight safety risk index system, for the index events where the corresponding parameters deviate from the predetermined safety range, their risk levels are set according to the degree of deviation, and different multi-level risk index coefficients are set accordingly.

[0022] According to some preferred embodiments of the present disclosure, the index events where the corresponding parameters deviate from the predetermined safety range include some or all of the following:

[0023] The 1000 - 500 ft approach speed, 500 - 50 ft approach speed, 50 - 20 ft approach speed, 50 - 20 ft approach speed, landing gear extension speed, touchdown pitch angle, touchdown pitch rate, 1500 - 500 ft approach slope, 500 - 200 ft approach slope, 200 - 50 ft approach slope, 1200 ft glide path angle, 800 ft glide path angle, 400 ft glide path angle, 3000 - 2000 ft descent rate, 2000 - 1000 ft descent rate, ILS localizer deviation angle, ILS glide slope deviation angle, 1000 - 500 ft descent rate, 500 - 50 ft descent rate, 50 - 20 ft descent rate, approach twin-engine N1 deviation, landing configuration altitude, approach phase AP disconnect altitude.

[0024] According to some preferred embodiments of the present disclosure, the steps of determining the weight coefficients of the multiple index events according to the entropy weight method include: for a single index event sufficient to cause a flight accident, assigning its weight coefficient to a preset high weight value.

[0025] According to some embodiments of the present disclosure, the evaluation method further includes:

[0026] Calculate the flight safety assessment risk values for all flights within the predetermined time span for the predetermined flight scenario based on the historical QAR data associated with the target route within the predetermined time span;

[0027] Based on the calculated flight safety assessment risk values of all flights, use a time series prediction algorithm to predict the trend of the flight safety risk of future flights for the predetermined flight scenario.

[0028] According to some embodiments of the present disclosure, the assessment method further includes:

[0029] For the historical raw QAR data associated with the target route, decode it according to the method of converting binary to engineering values, and preprocess the decoded QAR data. The preprocessing includes linearly interpolating to complete missing values and removing outliers.

[0030] According to some embodiments of the present disclosure, the assessment method further includes:

[0031] Add the QAR data of the flight after the flight of the newly completed flight to the flight safety risk index system, obtain the flight safety assessment risk value of the newly completed flight, and then update the standardized risk index coefficient and the weight coefficient of the multiple index events.

[0032] According to some embodiments of the present disclosure, the assessment method further includes:

[0033] Analyze the flights with high-risk flight safety assessment risk values, formulate risk control measures based on the analysis results, and track and evaluate the implementation effects of the risk control measures.

[0034] On the basis of conforming to the common knowledge in the art, the above preferred conditions can be combined arbitrarily to obtain various preferred examples of the present disclosure.

[0035] The positive and progressive effects of the present disclosure are as follows:

[0036] According to the data-driven assessment method for route flight safety risks of the present disclosure, it can help relevant parties such as airlines to comprehensively utilize QAR data to better grasp the safety risks of aircraft during route operations, and provide a method that can assist in assessing flight / route operation safety risks, and its assessment method has relatively better comparability and is relatively objective. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Shows a schematic flowchart of an assessment method for data-driven route flight safety risks according to a preferred embodiment of the present disclosure.

[0038] Figure 2 Schematic diagram for extracting influencing factors of an exemplary flight scenario of an unstable approach

[0039] Figure 3 Schematic illustration of an example of setting risk levels and risk index coefficients of a flight safety risk index system constructed in an evaluation method according to a preferred embodiment of the present disclosure

[0040] Figure 4 Schematic diagram of weight coefficient examples (determined by the entropy weight method) of multiple index events in an exemplary flight scenario of an unstable approach for an evaluation method according to a preferred embodiment of the present disclosure

[0041] Figure 5 Schematic illustration of an example of a trend prediction result of flight safety risk in an exemplary flight scenario of an unstable approach for an evaluation method according to a preferred embodiment of the present disclosure Detailed implementation manners

[0042] The following further describes in detail the preferred embodiments of the present invention in conjunction with the accompanying drawings of the specification. The following description is exemplary and not a limitation to the present invention. Any other similar situations also fall within the protection scope of the present invention.

[0043] In the following detailed description, directional terms, such as "left", "right", "up", "down", "front", "rear", etc., are used with reference to the directions described in the accompanying drawings. The components of the embodiments of the present invention can be placed in a variety of different directions, and the directional terms are for illustrative purposes and not restrictive.

[0044] Refer to Figures 1-5 In particular Figure 1 As shown, an evaluation method for flight safety risk of route flight based on data driving according to a preferred embodiment of the present disclosure includes the following steps:

[0045] Obtain historical QAR data associated with a target route and perform data preprocessing, wherein the historical QAR data includes QAR data of multiple flights;

[0046] Construct a flight safety risk index system based on multiple types of flight safety influencing factors. The flight safety risk index system includes multiple index events affecting flight safety in a predetermined flight scenario. Each index event is associated with a flight safety influencing factor, and the multiple index events are divided into several risk levels, and corresponding risk index coefficients are set for each risk level;

[0047] Perform standardization processing on the risk index coefficients of the multiple index events that appear in the historical QAR data to obtain standardized risk index coefficients;

[0048] Calculate the information entropy of the multiple metric events based on the historical QAR data, and determine the weight coefficients of the multiple metric events according to the entropy weight method;

[0049] For a flight to be evaluated belonging to a target route, obtain the QAR data of the flight to be evaluated, and calculate the inner product of the weight coefficients of the multiple metric events and the standardized risk metric coefficients included therein as the flight safety evaluation risk value of the flight to be evaluated for the predetermined flight scenario.

[0050] Optionally, the evaluation method further includes the following steps that are first executed:

[0051] For the historical raw QAR data associated with a target route, perform decoding according to the method of converting binary to engineering values, and preprocess the decoded QAR data, where the preprocessing includes linearly interpolating to complete missing values and removing outliers.

[0052] For example, the original flight data file can be obtained first, and the QAR raw data can be decoded, and the decoded engineering value data can be stored in CSV format. Subsequently, data preprocessing work is carried out on situations such as missing values, noise data, incomplete data, inconsistent data, and abnormal data values that are likely to occur in the QAR data in CSV format, so as to improve the data calculation efficiency. More specifically, the QAR data conforms to the ARINC 717 or ARINC 767 specification, so data decoding is performed according to the method of converting binary to engineering values.

[0053] According to some embodiments of the present disclosure, the step of standardizing the risk metric coefficients of the multiple metric events occurring in the historical QAR data includes applying the maximum-minimum normalization method to the risk metric coefficients of the multiple metric events occurring in the historical QAR data.

[0054] According to a preferred embodiment of the present disclosure, the multiple types of flight safety influencing factors include human operation factors, mechanical factors, and environmental factors.

[0055] In the step or process of constructing a flight safety risk metric system based on multiple types of flight safety influencing factors, taking an exemplary (predetermined) flight scenario of unstable approach as an example, the safety of an aircraft's stable approach may be affected by multiple factors, including human operation factors, mechanical factors, and environmental factors. Specific factors can be for example as Figure 2 shown. These are some of the main factors affecting the stable approach of an aircraft in the figure, and the crew and the airline need to effectively manage and control these influencing factors to ensure the stability and safety of the aircraft.

[0056] According to a further preferred embodiment of the present disclosure, in the step of constructing the flight safety risk index system, the trigger warning type index event is set to belong to the highest risk level, and the highest risk index coefficient is set.

[0057] Among them, more preferably, the trigger warning type index event includes some or all of the following:

[0058] Wind shear warning, proximity warning, landing gear not locked in the down position, stall warning, smoke warning, master warning.

[0059] According to a further preferred embodiment of the present disclosure, in the step of constructing the flight safety risk index system, for the index event where the corresponding parameter deviates from the predetermined safety range, the risk level is set according to the degree of deviation, and different multi-level risk index coefficients are set accordingly.

[0060] Among them, more preferably, the index event where the corresponding parameter deviates from the predetermined safety range includes some or all of the following:

[0061] The 1000 - 500 ft approach speed, 500 - 50 ft approach speed, 50 - 20 ft approach speed, 50 - 20 ft approach speed, landing gear extension speed, touchdown pitch angle, touchdown pitch rate, 1500 - 500 ft approach slope, 500 - 200 ft approach slope, 200 - 50 ft approach slope, 1200 ft glide path angle, 800 ft glide path angle, 400 ft glide path angle, 3000 - 2000 ft descent rate, 2000 - 1000 ft descent rate, ILS localizer deviation angle, ILS glide slope deviation angle, 1000 - 500 ft descent rate, 500 - 50 ft descent rate, 50 - 20 ft descent rate, approach twin-engine N1 deviation, landing configuration altitude, approach phase AP disconnect altitude.

[0062] Taking the unstable approach scenario as an example again, the flight safety risk index of each flight segment can be calculated. Among them, the flight flight triggering a warning type event is defined as a severe index event, and this index records 3 points (i.e., the risk index coefficient is set to 3 points).

[0063] For example, for the index event where the corresponding parameter deviates from the predetermined safety range, if the deviation type safety risk index event is outside 3 sigma (3 times the standard deviation), it is defined as a moderate index event, and this index records 2 points; if the deviation type safety risk index event is outside 2 sigma, it is defined as a moderate index event, and this index records 1 point.

[0064] The risk index coefficient setting or the scoring mechanism of the index event illustrated above, for example, as Figure 3 shown. Each index event can be scored according to this rule.

[0065] According to a preferred embodiment of the present disclosure, the step of determining the weight coefficients of the multiple index events according to the entropy weight method includes: for a single index event sufficient to cause a flight accident, assigning its weight coefficient as a preset high weight value.

[0066] According to a preferred embodiment of the present disclosure, the evaluation method further includes:

[0067] Based on the historical QAR data associated with the target route within a predetermined time span, calculating the flight safety evaluation risk value of all flights within the predetermined time span for the predetermined flight scenario;

[0068] Based on the calculated flight safety evaluation risk values of all flights, using a time series prediction algorithm to predict the trend of the flight safety risk of future flights for the predetermined flight scenario.

[0069] According to a preferred embodiment of the present disclosure, the evaluation method further includes:

[0070] Adding the QAR data of the post-flight of the newly completed flight to the flight safety risk index system, obtaining the flight safety evaluation risk value of the newly completed flight, and then updating the standardized risk index coefficients and the weight coefficients of the multiple index events.

[0071] More specifically, taking the unstable approach scenario as an example again, exploring and analyzing the flight safety evaluation risk values of the historical flight unstable approach scenarios obtained by the above process, where the outlier information included is first removed. Then, using a time series prediction algorithm such as Prophet to predict the risk value trend, forming a risk trend prediction result for the unstable approach scenario (for example Figure 5 shown). The black represents the time series discrete points of the historical risk values of the unstable approach, the dark blue line represents the risk values fitted using the time series, and the light blue line represents the confidence interval of the time series, that is, the upper and lower bounds of the reasonable predicted risk values.

[0072] According to some embodiments of the present disclosure, the evaluation method further includes:

[0073] Analyzing the flights with high-risk flight safety evaluation risk values, formulating risk control measures according to the analysis results, and tracking and evaluating the implementation effects of the risk control measures.

[0074] Specifically, for example, corresponding risk control measures can be formulated for high-risk flights and risk factors, such as strengthening pilot training, optimizing flight procedures, etc. Optionally, the implementation effects of the risk control measures can be continuously monitored and evaluated, and the risk assessment methods and safety management measures can be continuously improved to improve the flight safety level.

[0075] According to the data-driven route flight safety risk assessment method of the above-mentioned embodiments of the present disclosure, it can help relevant parties such as airlines to comprehensively utilize QAR data to more comprehensively grasp the safety risks of aircraft during route operation, and provide a method that can assist in the assessment of flight / route operation safety risks, and its assessment method has relatively better comparability and relatively objectivity. Further, the present disclosure can help to quantitatively evaluate the safety of flights for various flight scenarios, and is easy to further optimize the accuracy of safety risk assessment during long-term use.

[0076] Although the specific embodiments of the present invention have been described above, those skilled in the art should understand that these are only examples, and the protection scope of the present invention is defined by the appended claims. Without departing from the principles and essence of the present invention, those skilled in the art can make various changes or modifications to these embodiments, but these changes and modifications all fall within the protection scope of the present invention.

Claims

1. A data-driven method for assessing flight safety risks of air routes, characterized in that: The evaluation methods include: Acquire historical QAR data associated with the target route and perform data preprocessing, wherein the historical QAR data includes QAR data of multiple flights; A flight safety risk indicator system is constructed based on multiple flight safety influencing factors, wherein the flight safety risk indicator system includes multiple indicator events that affect flight safety in a predetermined flight scenario, wherein each indicator event is associated with a flight safety influencing factor, and the multiple indicator events are divided into several risk levels, and a corresponding risk indicator coefficient is set for each risk level; Standardizing the risk indicator coefficients of the multiple indicator events appearing in the historical QAR data to obtain standardized risk indicator coefficients; Based on the historical QAR data, calculating the information entropy of the multiple indicator events, and determining the weight coefficients of the multiple indicator events according to the entropy weight method; For the flight to be evaluated that belongs to the target route, the QAR data of the flight to be evaluated is obtained, and the inner product of the weight coefficients of the multiple indicator events and the standardized risk indicator coefficient contained therein is calculated as the flight safety assessment risk value of the flight to be evaluated for the predetermined flight scenario.

2. The data-driven flight safety risk assessment method according to claim 1, characterized in that: The step of normalizing the risk indicator coefficients of the multiple indicator events appearing in the historical QAR data includes applying a maximum-minimum normalization method to process the risk indicator coefficients of the multiple indicator events appearing in the historical QAR data.

3. The data-driven flight safety risk assessment method of claim 2, wherein: The multiple types of flight safety influencing factors include human operation factors, mechanical factors and environmental factors.

4. The data-driven flight safety risk assessment method of claim 3, wherein: In the step of constructing the flight safety risk indicator system, the triggering alarm indicator event is set to belong to the highest risk level, and the highest risk indicator coefficient is set.

5. The data-driven flight safety risk assessment method of claim 4, characterized in that: Events that trigger alarm indicators include some or all of the following: Wind shear warning, ground proximity warning, landing gear not locked down, stall warning, smoke warning, master warning.

6. The data-driven flight safety risk assessment method of claim 5, characterized in that: In the step of constructing the flight safety risk indicator system, the risk level of the indicator event corresponding to the deviation of the corresponding parameter from the predetermined safety range is set according to the degree of deviation, and different multi-level risk indicator coefficients are set accordingly.

7. The data-driven flight safety risk assessment method of claim 6, characterized in that: Indicator events corresponding to parameters deviating from the predetermined safety range include some or all of the following: The 1000-500 feet approach speed, 500-50 feet approach speed, 50-20 feet approach speed, 50-20 feet approach speed, landing gear lowering speed, touchdown pitch angle, touchdown pitch rate, 1500-500 feet approach slope, 500-200 feet approach slope, 200-50 feet approach slope, 1200 feet glide path angle, 800 feet glide path angle, 400 feet glide path angle, 3000-2000 feet descent rate, 2000-1000 feet descent rate, ILS localizer deviation angle, ILS glide path deviation angle, 1000-500 feet descent rate, 500-50 feet descent rate, 50-20 feet descent rate, approach dual engine N1 deviation, landing configuration altitude, and AP disconnection altitude during approach.

8. The data-driven flight safety risk assessment method of claim 7, characterized in that: The step of determining the weight coefficients of the multiple index events according to the entropy weight method includes: for a single index event that is sufficient to cause an airliner accident, assigning its weight coefficient to a preset high weight value.

9. The data-driven flight safety risk assessment method according to any one of claims 1 to 8, characterized in that: The evaluation method also includes: Calculating the flight safety assessment risk value of all flights within the predetermined time span for the predetermined flight scenario based on the historical QAR data associated with the target route within the predetermined time span; Based on the calculated flight safety assessment risk values ​​of all flights, a time series prediction algorithm is used to predict the trend of flight safety risks of future flights for the predetermined flight scenarios.

10. The data-driven flight safety risk assessment method according to any one of claims 1 to 8, characterized in that: The evaluation method also includes: The historical original QAR data associated with the target route are decoded by a binary-to-engineering value conversion method, and the decoded QAR data are preprocessed, wherein the preprocessing includes linear interpolation to fill missing values ​​and removal of outliers.

11. The data-driven flight safety risk assessment method according to any one of claims 1 to 8, characterized in that: The evaluation method also includes: The post-flight QAR data of the newly completed flight is added to the flight safety risk index system, and the flight safety assessment risk value of the newly completed flight is obtained, and then the standardized risk index coefficients and the weight coefficients of the multiple index events are updated.

12. The data-driven flight safety risk assessment method according to any one of claims 1 to 8, characterized in that: The evaluation method also includes: Analyze flights with high risk values ​​based on flight safety assessment, formulate risk control measures based on the analysis results, and track and evaluate the implementation effects of the risk control measures.

Images