A flight arrival time dynamic calculation method based on multi-source data fusion

Abstract

The application discloses a kind of based on multi-source data fusion's flight arrival time dynamic calculation method, it is related to civil aviation operation management and intelligent computing technical field, including by fusing flight dynamic data and historical flight statistical data, obtain the flight execution record in the range setting of specified time window, data segmentation processing is carried out to time window division, from which extraction return flight standby identifier and state field analysis result, judge whether it is normal execution flight, if identifying return flight standby identifier, then eliminate corresponding record, obtain the flight sample set after preliminary cleaning;The based on multi-source data fusion's flight arrival time dynamic calculation method, realize more accurate, continuous dynamic calculation to flight arrival time;By the arrival time dynamic calculation result and the cargo compartment environment real-time parameter of special goods and the attribute of goods are fused, the differential arrival early warning and ground support resource pre-scheduling for cold-chain medicine, live animals, dangerous goods and other special goods are realized.

Description

Technical Field

[0001] This invention relates to the field of civil aviation operation management and intelligent computing technology, specifically to a method for dynamically calculating flight arrival times based on multi-source data fusion. Background Technology

[0002] In the field of flight operation monitoring and collaborative decision-making, flight arrival time is a crucial parameter for measuring flight operation status, ensuring airport operation organization, and supporting air-ground collaborative scheduling. To improve the accuracy of arrival time estimation, existing technologies typically utilize multiple types of information, including flight plan data, flight dynamic data, ADS-B surveillance data, and historical flight statistics, to analyze and predict flight operations. Flight plan data characterizes planned routes and schedules; flight dynamic data reflects flight execution status; ADS-B surveillance data provides real-time position and speed information during flight; and historical flight statistics can be used to extract operational deviation patterns under different routes and flight paths.

[0003] However, in practical applications, existing flight arrival time calculation methods still have many limitations due to differences in data structure, update frequency, completeness, and reliability among the aforementioned multi-source data. First, during the construction of historical samples, flight execution records often contain samples of return flights, diversions, abnormal statuses, or excessively large time deviations. Without effective data cleaning and anomaly removal mechanisms, statistical results can easily become distorted, affecting the fundamental accuracy of subsequent arrival time calculations. Especially when extracting historical samples within a specified time window, the operating environment varies significantly across different time periods. Without proper sample screening and quality control, it is difficult to obtain representative and effective flight samples. Second, existing technologies for extracting time deviation patterns from historical samples are often susceptible to extreme values, outliers, and low-frequency samples. For newly opened routes, low-frequency flights, or routes with insufficient historical records, the number of effective samples available for statistics is limited, making historical deviation characteristics unstable and difficult to form a reliable basis for correction. If the statistical results of a limited sample are directly used when samples are insufficient, or if other flight data are simply borrowed as substitutes, deviation information inconsistent with the current flight operation characteristics may be introduced, reducing the reliability of arrival time predictions. Furthermore, flight arrival times are closely related to flight routes. Different flight routes typically result in variations in flight distance, operating environment, and airspace organization, leading to different time deviation distributions. Current technologies for route-related analysis often struggle to balance fine-grained differentiation of planned route information with statistical stability after sample classification, thus affecting the relevance and applicability of deviation parameters. Inaccurate route classification or insufficient records for a particular route can easily cause the calculated arrival time to deviate from the actual flight situation. In addition, flight arrival times exhibit significant dynamic changes. The types of data available and their reliability differ at different stages, from before takeoff to after takeoff and throughout the flight. Current technologies, in some scenarios, rely solely on planned data or static historical statistics for estimation, lacking the ability to perform phased calculations incorporating the dynamic state of the flight, making it difficult to adapt to continuous changes during flight operations. Especially after the flight enters the flight phase, if the arrival time cannot be promptly corrected by incorporating real-time position and speed information, the predicted results are prone to lagging behind the actual flight status. Furthermore, when using ADS-B surveillance data to update estimated arrival times, flights may experience positional shifts, speed changes, or flight trajectory changes due to air traffic control, route adjustments, weather-related detours, or crew operations. If existing technologies lack an effective collaborative processing mechanism between surveillance data and existing estimated arrival times, it becomes difficult to promptly identify changes in flight operational status and re-estimate remaining flight time, resulting in significant fluctuations in estimated arrival times or untimely corrections.

[0004] Furthermore, existing technologies exhibit more pronounced application shortcomings in special cargo air transport scenarios. Special cargoes such as cold chain pharmaceuticals, vaccines, live animals, and dangerous goods have stringent requirements for the timeliness of ground handling upon arrival. For example, cold chain pharmaceuticals (temperature controlled at 2-8°C) risk spoilage if the temperature exceeds the limit; live animals must be unloaded and transferred as quickly as possible after landing to avoid stress reactions or death; and dangerous goods must be safely handed over within strict timeframes. However, existing flight arrival time calculation methods always use the flight as the sole unit of calculation, and their output is never linked to the cargo status and cargo hold conditions. When flight arrival times are delayed, ground support departments cannot know in advance whether the cargo hold environment has exceeded the cargo's tolerance threshold, nor can they pre-schedule scarce support resources such as cold chain containers, quarantine personnel, and isolation areas at the minute level based on the precise arrival time. This results in persistently high loss rates in special cargo transportation and low utilization rates of ground support resources. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic calculation method for flight arrival time based on multi-source data fusion, thereby solving the problems existing in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamically calculating flight arrival times based on multi-source data fusion, comprising: S1. By integrating flight dynamic data and historical flight statistics, flight execution records within a specified time window range are obtained. Data segmentation is performed based on the time window division, and return diversion flags and status field parsing results are extracted to determine whether the flight is a normal execution flight. If the return diversion flag is identified, the corresponding record is removed to obtain a pre-cleaned flight sample set. S2. Based on the pre-cleaned flight sample set, the deviation value calculation logic is used to calculate the difference between the actual and planned time for each record. Combining the deviation threshold standard and the abnormal flight judgment rule, if the difference exceeds the preset upper limit standard, it is marked as abnormal and removed to determine the effective flight sample set after cleaning. S3. For the cleaned set of valid flight samples, sort them in ascending order according to the size of the deviation based on the data sorting logic. Combine the extreme value removal ratio to remove records of the upper and lower extreme values. The data cleaning process ensures the quality of the samples and obtains the selected core flight sample set. S4. Based on the selected core flight sample set, the deviation values ​​of the retained records are averaged using the mean calculation method. Combined with the historical data integrity assessment, if the data volume is insufficient, the correction data is borrowed from similar routes by classifying the flight types to determine the final historical deviation correction value.

[0007] Preferably, S1 includes: Obtain flight dynamic data and historical flight statistics, and extract the corresponding flight execution records; Flight segment records are obtained by segmenting flight execution records. Parse the flight segment records to obtain the status field parsing results. If a return-to-origin / diversion flag is identified from the status field parsing results, the flight segment record is determined to be an abnormal flight record. Abnormal flight records are removed and normally operating flights are retained. A preliminary cleaned flight sample set is obtained based on the normally operating flights.

[0008] Preferably, S2 includes: The time deviation numerical sequence is obtained by performing interpolation calculations on the pre-cleaned flight sample set. Determine whether the time deviation numerical sequence exceeds the preset deviation threshold standard; If the time deviation numerical sequence exceeds the preset deviation threshold standard, it is determined to be an abnormal deviation record; Records that need to be removed are filtered by associating abnormal deviation records with historical weather data. Records that need to be removed are discarded, and the remaining records in the flight sample set are integrated to determine the effective flight sample set after cleaning.

[0009] Preferably, S3 includes: Obtain the cleaned set of valid flight samples, sort them in ascending order according to the deviation magnitude attribute of the valid flight sample set to determine the deviation sequence, and identify the upper and lower extreme value records in the deviation sequence to obtain the preliminary screening sequence. The boundary-adjusted sequence is obtained by removing records of extreme values ​​from the initial screening sequence. If the boundary adjustment sequence meets the preset threshold, the sample quality improvement sequence is obtained through the cleaning process. If the improved sample quality sequence meets the ascending order requirement, then the selected core flight sample set is determined.

[0010] Preferably, S4 includes: The initial average deviation value is obtained by averaging the flight deviation values. The historical data volume assessment result is obtained by evaluating historical data associated with the initial average deviation value; Determine whether the historical data volume assessment result is lower than a preset threshold; If the historical data volume assessment result is lower than the preset threshold, the borrowed correction data is obtained by matching the characteristics of the flights corresponding to the initial average deviation value. The initial average deviation value is compensated by borrowing corrected data to determine the final historical deviation correction value.

[0011] Preferably, it also includes S5, extracting route information from flight plan data, classifying the flight sample set according to route based on the sample grouping method, applying the final historical deviation correction value to each type of route, and supplementing missing fields based on record integrity verification if a certain route record is insufficient, thereby obtaining the deviation adjustment parameters corresponding to the route record, specifically including: By extracting segment node sequences from flight plan data to construct route orientation features, the core flight sample set is grouped using route orientation features to obtain the target route orientation classification set. Obtain historical deviation correction values ​​for the target route orientation classification set, and use the historical deviation correction values ​​to perform deviation mapping processing on the target route orientation classification set to obtain the initial deviation adjustment sequence; Determine whether the record integrity assessment value of the initial deviation adjustment sequence is lower than the preset integrity threshold; If the record completeness assessment value is lower than the preset completeness threshold, the missing field positions in the initial deviation adjustment sequence are extracted, and the missing field positions are interpolated and supplemented based on the historical deviation correction values ​​of adjacent flight segment nodes to determine the deviation adjustment parameters corresponding to the flight path.

[0012] Preferably, it also includes S6, which integrates flight plan data, flight dynamic data, and historical flight statistics, adjusts the deviation parameters according to the route direction, and applies the deviation adjustment parameters to different flight stages in conjunction with the takeoff time information in the flight dynamic data. If the current stage is before takeoff, the estimated takeoff time is used as the basis for time estimation to determine a preliminary estimated arrival time, specifically including: The initial dataset for flight plan data fusion is obtained, and the initial dataset is calibrated according to the deviation of the flight path to obtain the integrated sequence; If the integrated sequence is in the pre-takeoff state, the offset is calculated based on the expected takeoff base to determine the time series.

[0013] Preferably, S6 further includes: The flight phase differentiation features are extracted based on the time series data, and then fused with historical statistical data to obtain a preliminary estimated sequence. Details are extracted from the preliminary estimated sequence, and these details are processed in conjunction with the takeoff time to determine the estimated arrival time.

[0014] Preferably, it also includes S7, fusing ADS-B surveillance data with the preliminary estimated time of arrival, extracting position and speed information in real time, dynamically correcting the preliminary estimated time of arrival, adjusting parameters based on the route classification results and deviations, recalculating the remaining flight distance if the position deviation exceeds a preset range, and obtaining the final estimated time of arrival result based on the fused flight plan data, flight dynamic data, ADS-B surveillance data, and historical flight statistics. Specifically, this includes: Position and velocity information are extracted from broadcast automatic dependent surveillance data to generate a real-time flight status sequence; The position offset value is determined based on the real-time flight status sequence; If the position offset value exceeds the preset range, the remaining flight distance is calculated based on the position offset value and speed information.

[0015] Preferably, S7 further includes: The preliminary estimated arrival time is dynamically corrected by the remaining flight distance to obtain the corrected time series; The final estimated arrival time is obtained by fusing flight dynamic data with historical flight statistics using corrected time series data.

[0016] Preferably, the method further includes step S8; fusing the final estimated arrival time result with cargo attribute information and real-time cargo hold environmental parameters, wherein the cargo attribute information includes at least the cargo type and its corresponding maximum allowable waiting time after arrival, and the real-time cargo hold environmental parameters include at least temperature, humidity, and vibration data; based on the fusion result, predicting the evolution trend of the cargo hold environmental parameters in the remaining flight distance, and determining whether the state threshold corresponding to the cargo type will be exceeded; if the threshold is predicted to be exceeded, triggering differentiated early warnings for different support objects, and generating a pre-scheduling plan for ground support resources for the arriving airport.

[0017] Preferably, S8 specifically includes: S81. Establish a mapping table between cargo types and the maximum allowable waiting time after arrival. The mapping table shall include at least cold chain pharmaceuticals and their corresponding first time limits, live animals and their corresponding second time limits, and dangerous goods and their corresponding third time limits. S82. The temperature, humidity and vibration data of the cargo hold are acquired in real time through airborne sensors, and the data is spatiotemporally aligned with the final estimated arrival time to form a time series of cargo hold environmental parameters. S83. Based on the cargo hold environmental parameter time series and combined with the remaining flight time, predict the cargo hold environmental parameter value at the arrival time, and determine whether the predicted value exceeds the preset state threshold corresponding to the cargo type. S84. If the cargo hold temperature at the predicted arrival time is expected to exceed the allowable range, a Level 1 warning will be triggered, and the crew will be advised to adjust the cargo hold temperature control settings. If the situation is predicted to worsen and cannot be prevented from deteriorating by onboard control, a Level 2 warning will be triggered, and a priority unloading request containing cargo type, estimated arrival time, and urgency level will be sent to the ground support department of the destination airport. S85. Based on the type of goods and the final estimated arrival time, automatically generate a pre-schedule plan for ground support resources. The plan includes pre-positioning instructions for cold chain containers, pre-arrival instructions for live animal quarantine personnel, or pre-clearing instructions for hazardous materials isolation areas.

[0018] Preferably, the step of predicting the cargo hold environmental parameter value at the arrival time in S83 specifically involves: integrating the time series of the cargo hold environmental parameters with the statistical patterns of cargo hold environmental changes on the same route and during the same period of historical flights, and using the trend extrapolation method to calculate the slope of change of each parameter in the remaining flight distance, thereby obtaining the predicted value of the arrival time; the preset state threshold is the boundary of the temperature range of 2-8°C for cold chain medicines, and the boundary of the preset temperature and humidity comfort range for live animals.

[0019] As can be seen from the above technical solution, the present invention has the following beneficial effects: This dynamic flight arrival time calculation method based on multi-source data fusion comprehensively utilizes flight plan data, flight dynamic data, ADS-B surveillance data, and historical flight statistics to calculate flight arrival times in stages and dynamically. Compared with existing technologies, it can effectively improve the accuracy, stability, and adaptability of arrival time estimation. Specifically, by identifying return and diversion airports in historical flight execution records, removing outliers, cleaning extreme values, and screening core samples, the quality of historical samples can be improved, reducing the interference of outlier data on statistical results. By calculating historical deviation values ​​and borrowing correction data from similar flight routes when samples are insufficient, the reliability of historical deviation correction values ​​can be enhanced. This improves reliability and enhances the accuracy of estimated arrival times in scenarios with low-frequency routes or insufficient historical data. By extracting route information from flight plan data and setting deviation adjustment parameters according to the route, the specificity of estimated arrival times under different route scenarios can be improved. By combining the flight phase to estimate the initial estimated arrival time and further using real-time position and speed information from ADS-B monitoring data for dynamic correction, changes in flight operation status can be reflected in a timely manner. When position deviation, speed change, or flight trajectory adjustment occurs, the remaining flight distance is recalculated, thereby reducing the lag and volatility of the estimated arrival time results. Ultimately, this achieves more accurate, continuous, and robust dynamic calculation of flight estimated arrival times. Attached Figure Description

[0020] Figure 1 This is a signal transmission diagram of the present invention; Figure 2 This is a structural block diagram of the local terminal of an exemplary electronic device of the present invention; Figure 3 This is a structural block diagram of the network terminal of an exemplary electronic device of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] like Figure 1 As shown, the present invention provides a technical solution: a method for dynamically calculating flight arrival time based on multi-source data fusion, comprising: S1. By integrating flight dynamic data and historical flight statistics, flight execution records within a specified time window range are obtained. Data segmentation is performed based on the time window division, and return diversion flags and status field parsing results are extracted to determine whether the flight is a normal execution flight. If the return diversion flag is identified, the corresponding record is removed to obtain a pre-cleaned flight sample set. S2. Based on the pre-cleaned flight sample set, the deviation value calculation logic is used to calculate the difference between the actual and planned time for each record. Combining the deviation threshold standard and the abnormal flight judgment rule, if the difference exceeds the preset upper limit standard, it is marked as abnormal and removed to determine the effective flight sample set after cleaning. S3. For the cleaned set of valid flight samples, sort them in ascending order according to the size of the deviation based on the data sorting logic. Combine the extreme value removal ratio to remove records of the upper and lower extreme values. The data cleaning process ensures the quality of the samples and obtains the selected core flight sample set. S4. Based on the selected core flight sample set, the deviation values ​​of the retained records are averaged using the mean calculation method. Combined with the historical data integrity assessment, if the data volume is insufficient, the correction data is borrowed from similar routes by classifying the flight types to determine the final historical deviation correction value. S5. Extract route information from flight plan data, classify the flight sample set according to route by combining sample grouping method, apply the final historical deviation correction value for each type of route, and supplement the missing field based on record integrity check if a certain route record is insufficient to obtain the deviation adjustment parameter corresponding to the route. S6. Integrate flight plan data, flight dynamic data and historical flight statistics, adjust the deviation parameters according to the route direction, and combine the takeoff time information in the flight dynamic data to apply the deviation adjustment parameters for different flight stages. If the current stage is before takeoff, the estimated takeoff time is used as the basis for time estimation to determine the preliminary estimated arrival time. S7. The ADS-B monitoring data is fused with the preliminary estimated time of arrival. The position and speed information during flight are extracted in real time. The preliminary estimated time of arrival is dynamically corrected. The parameters are adjusted according to the route classification results and deviations. If the position deviation exceeds the preset range, the remaining flight distance is recalculated. Based on the fused flight plan data, flight dynamic data, ADS-B monitoring data and historical flight statistics, the final estimated time of arrival result is obtained.

[0023] S8. The final estimated arrival time result is fused with cargo attribute information and real-time cargo hold environmental parameters. The cargo attribute information includes at least the cargo type and its corresponding maximum allowable waiting time after arrival. The real-time cargo hold environmental parameters include at least temperature, humidity and vibration data. Based on the fusion result, the evolution trend of cargo hold environmental parameters is predicted in the remaining flight distance to determine whether the state threshold corresponding to the cargo type will be exceeded. If the threshold is predicted to be exceeded, differentiated early warnings for different support objects are triggered, and a pre-scheduling plan for ground support resources for the arriving airport is generated.

[0024] In this embodiment, during the data preparation stage, flight execution records within a specified time window are extracted by fusing flight dynamic data and historical flight statistics, and then segmented according to the time window division rules. Further, the return / diversion flags, status fields, and related operational status information in each flight execution record are parsed to determine whether the corresponding flight is a normally executed flight. If return, diversion, cancellation, or other obviously statistically unrepresentative abnormal operation records are identified, these records are removed, resulting in a pre-cleaned flight sample set. The method of this application allows the samples participating in subsequent statistical analysis to better reflect the normal flight operation patterns, thereby providing a reliable data foundation for subsequent deviation correction calculations.

[0025] In some embodiments, the pre-cleaned flight sample set requires further deviation calculation and anomaly removal. Specifically, the difference between the actual time and the planned time for each record is calculated, and combined with a preset deviation threshold standard and anomaly flight judgment rules, records with deviation values ​​exceeding the upper limit standard are identified. Thus, the method of this application can exclude obviously abnormal samples caused by extreme weather, air traffic control, temporary operational restrictions, or data acquisition errors, thereby determining the cleaned valid flight sample set. As a preferred embodiment, when calculating the deviation value, the time difference between the actual arrival time and the planned arrival time can be preferentially used as the core deviation indicator. If necessary, takeoff deviation, taxiing deviation, or segment operation deviation can also be combined as auxiliary judgment criteria to enhance the accuracy of anomaly identification.

[0026] In this embodiment, after obtaining the effective flight sample set, the samples are further sorted in ascending order according to the magnitude of the deviation, and records that deviate from the main distribution at both ends are removed using an extreme value removal ratio to form a refined core flight sample set. It should be noted that extreme value removal here is not simply deleting a small amount of data, but rather suppressing the tails of the sample distribution to reduce the interference of special operating days, sudden events, or extreme cases on the overall statistical results. Based on the aforementioned embodiment, it is clear that the sample set after sorting and extreme value removal has higher representativeness and statistical stability, and is therefore more suitable for generating historical deviation correction values.

[0027] In this embodiment, the deviation values ​​of the retained records are averaged based on the selected core flight sample set to determine the historical deviation correction value. As a preferred embodiment, this historical deviation correction value can characterize the overall deviation trend of a certain type of flight relative to the planned arrival time within a given time window, thus providing a statistical compensation basis for subsequent estimated arrival times. Furthermore, it should be noted that when the sample size within a certain time window is insufficient, this application does not directly abandon the correction calculation for that flight category. Instead, it borrows correction data from samples with similar operational characteristics through flight type classification and similar route matching mechanisms to improve the availability and stability of the historical deviation correction value. As can be seen from the above embodiments, this compensation strategy can still generate meaningful deviation correction results even when the target route sample is small.

[0028] In the above working principle, flight plan data is mainly used to extract route information. Specifically, flight samples can be classified according to route orientation based on waypoint sequences, airport combination relationships, planned route codes, or flight path direction characteristics. Corresponding historical deviation correction values ​​are then matched to the corresponding categories to form deviation adjustment parameters for each route orientation. The method of this application allows for differentiated corrections for flights with different route orientations and flight path characteristics, rather than uniformly using a single correction parameter. Therefore, the method of this application can enhance the adaptability of estimated arrival times to complex operational scenarios. Furthermore, if the number of records for a certain route orientation is insufficient, the missing fields can be supplemented based on the record integrity verification results, thereby ensuring the computability and usability of the deviation adjustment parameters corresponding to that route orientation.

[0029] In this embodiment, after classifying the flight path and determining the deviation adjustment parameters, flight plan data, flight dynamic data, and historical flight statistics are fused together to form a preliminary estimated arrival time. Specifically, different time estimation benchmarks are used for different flight stages: if the current stage is before takeoff, the estimated takeoff time is used as the basis, combined with the planned flight duration and the deviation adjustment parameters corresponding to the flight path for time estimation; if the current stage is after takeoff, the planned arrival time can be corrected by combining the actual takeoff time, the confirmed departure time, or the segment's progress status to form a preliminary estimated arrival time that is closer to the current operating status. As can be seen from the foregoing embodiments, this application distinguishes between flight operation stages, enabling the estimated arrival time calculation to continue from the pre-takeoff stage to the in-flight stage, thus improving the continuity of the output results.

[0030] Furthermore, it should be noted that the preliminary estimated arrival time is still mainly based on the plan execution logic and historical deviation patterns. For dynamic factors such as temporary deviations, detours, and speed changes during flight, further real-time monitoring data is needed for correction. Therefore, in this embodiment, ADS-B monitoring data is fused with the preliminary estimated arrival time to extract real-time information such as position, speed, heading, and altitude during flight, and the preliminary estimate is dynamically corrected accordingly. Specifically, the current flight position is matched with the planned flight path to determine if the flight is following the predetermined route. If a position deviation exceeds a preset range, the remaining flight distance is recalculated, and the remaining flight time is re-estimated based on real-time speed information, thereby updating the estimated arrival time. Thus, the method of this application can promptly correct the original estimation results when flight status changes, improving the real-time performance of the final estimated arrival time.

[0031] In another specific implementation of this application, when the actual flight trajectory is basically consistent with the planned path, the estimated arrival time can be slightly corrected based solely on real-time position advancement and speed changes, while retaining the original route classification results and deviation adjustment parameters. This avoids the system burden caused by frequent recalculations and maintains the timeliness and rationality of result updates. The method of this application establishes a coordinated and consistent calculation logic among three types of information: static planning, historical patterns, and dynamic monitoring, enabling the estimated arrival time to continuously approach the actual arrival time as the flight's operational status changes.

[0032] In another key embodiment of this application, a safeguard logic for special cargo transportation is also introduced. First, a mapping table is established between cargo types and the maximum allowable waiting time after arrival. This mapping table is a fixed configuration table, containing at least three records: cold chain pharmaceuticals correspond to a first time limit (e.g., 30 minutes), live animals correspond to a second time limit (e.g., 45 minutes), and dangerous goods correspond to a third time limit (e.g., 60 minutes). It should be noted that the above time limits refer to the maximum allowable time interval from when the flight lands until the cargo enters a specific safeguard stage (e.g., cold storage, unloading inspection platform, dangerous goods handover area). Through this mapping table, the abstract concept of cargo can be transformed into specific, quantifiable time constraints. Then, airborne sensors deployed in the cargo hold acquire real-time data on temperature, humidity, and vibration at a sampling frequency of no less than once per minute. This data is then transmitted in real-time to the ground processing system via an airborne communication system (such as ACARS) or a broadband satellite link. After receiving this data, the ground processing system performs spatiotemporal alignment with the final estimated arrival time calculated in the aforementioned steps to form a time series of cargo hold environmental parameters. Spatiotemporal alignment means that each frame of environmental parameters is stamped with the precise aircraft position (latitude, longitude, and altitude) and a timestamp, so that it can correspond one-to-one with the remaining flight distance in the flight process. Next, based on the time series of cargo hold environmental parameters and the remaining flight time, the cargo hold environmental parameter values ​​at the arrival time are predicted. The prediction method is not a simple linear extrapolation, but rather a fusion of the current sequence with historical data on cargo hold environmental changes during the same route and time period (e.g., the afternoon of spring). For example, historical data analysis shows that on a certain Middle East route, the cargo hold temperature rises at an average rate of 0.5°C every 10 minutes during the cruise phase. This historical pattern is weighted and fused with the currently monitored real-time temperature rise rate (e.g., 0.6°C every 10 minutes) (historical weight 0.3, real-time weight 0.7) to more accurately predict the cargo hold temperature at arrival. Then, it is determined whether the predicted value exceeds the preset state threshold corresponding to the cargo type. For cold chain pharmaceuticals, the threshold is the boundary of the temperature range of 2-8°C; for live animals, the threshold is the boundary of a preset comfortable temperature and humidity range (e.g., temperature 10-25°C, relative humidity 40%-70%). If the cargo hold temperature at the predicted arrival time is expected to exceed the permissible range, a Level 1 warning is triggered. This warning is sent to the flight crew's multifunction control display unit, which automatically suggests that the crew adjust the cargo hold temperature control setting (e.g., lowering the setting from 5°C to 3°C). If the system further predicts that even with onboard temperature control measures, the cargo hold temperature cannot be kept within the threshold for the remaining flight distance, a Level 2 warning is triggered. This warning is sent to the ground support control center at the destination airport, containing information such as cargo type (e.g., cold chain pharmaceuticals like insulin), estimated arrival time (accurate to the minute), and urgency level (e.g., urgent: predicted overheating), and automatically generates a priority unloading request instruction. This instruction can be integrated into the airport's flight support prioritization system, placing this flight at the top of the unloading priority list. Finally, based on the cargo type and the final estimated arrival time, a pre-scheduling plan for ground support resources is automatically generated. This plan is not a fixed template but a dynamically calculated result. For example, when the estimated arrival time is 30 minutes earlier than planned, the system sends a pre-positioning instruction for cold chain containers to the cold storage management system, specifying that the containers be towed to the designated aircraft position before a specific time. When the cargo is live animals and the estimated arrival time is delayed, the system sends a pre-arrival instruction for quarantine personnel to the quarantine department terminal, postponing the personnel's arrival time from the original plan to the new estimated arrival time. When the cargo is dangerous goods, the system sends a pre-clearing instruction for the dangerous goods isolation area to the airport operations command center, ensuring that the clearing is completed 15 minutes before the flight lands. In this way, granularity from flight-level forecasting to cargo hold-level support is achieved, significantly improving the safety and support efficiency of special cargo transportation.

[0033] S1 includes acquiring flight dynamic data and historical flight statistics, extracting the corresponding flight execution records; performing data segmentation on the flight execution records to obtain flight segment records; parsing the flight segment records to obtain the status field parsing results, and if a return-to-origin / alternate-landing indicator is identified from the status field parsing results, the flight segment record is determined to be an abnormal flight record; removing abnormal flight records and retaining normally executing flights, and obtaining a preliminarily cleaned flight sample set based on the normally executing flights.

[0034] In this embodiment, flight operation information within the same statistical period is first extracted from flight dynamic data and historical flight statistics. This information is then linked according to flight number, execution date, departure airport, and destination airport to form a set of original records for the same flight. As a preferred embodiment, before a record enters the processing flow, a field integrity check is performed. The check fields include flight number, execution date, departure airport, destination airport, and status time. If any of these fields is missing, the corresponding record will not proceed to the subsequent processing flow; if all fields are present, the corresponding record will proceed to the subsequent processing flow.

[0035] In this embodiment, after the original records are collected, the status change information corresponding to the same flight is arranged in ascending order according to the status time, thus forming a sequential and continuous status sequence. Based on this, the status sequence is segmented. Specifically, if the time interval between two adjacent status records does not exceed 10 minutes and the status category remains consistent, the two status records are grouped into the same segment; if the time interval between two adjacent status records exceeds 10 minutes, or the status category changes, a new segment record is generated. The aforementioned 10-minute threshold is determined according to the status data refresh rules: when the data refresh cycle of flight dynamic data is fixed at 5 minutes, the segmentation threshold is fixed at two refresh cycles, i.e., 10 minutes. This setting ensures that the segmentation boundary is consistent with the original data update rhythm, guaranteeing that each segment corresponds to a stable operating status interval.

[0036] In this embodiment, after forming segmented records, the status field corresponding to each segment is parsed. Specifically, the structured status code is directly mapped, and the textual status description is standardized before rule matching. The standardization process includes removing spaces, standardizing symbol formats, and standardizing synonymous status expressions. Subsequently, return-to-origin identification rules and diversion identification rules are established. For the return-to-origin scenario, when the flight has already taken off, the segment's ending airport is the same as the planned departure airport, and the status field contains the meaning of return-to-origin, the flight corresponding to that segment is determined to be in a return-to-origin state. For the diversion scenario, when the segment's ending airport is different from the planned destination airport, the status field contains the meaning of diversion, and this status is consistent in two consecutive segmented records, the flight corresponding to that segment is determined to be in a diversion state. Here, two consecutive segmented records are used as the confirmation condition, and the determination is based on the following: a single status record may have a momentary jump, and when two consecutive segmented records simultaneously meet the same abnormal feature, misjudgment caused by a single code error or momentary status switch can be ruled out. Since the data refresh cycle is fixed at 5 minutes, the minimum confirmation duration for two consecutive segmented records is fixed at 10 minutes. This threshold is a fixed threshold and is not subject to floating settings.

[0037] In this embodiment, in addition to the aforementioned identification rules, a validity judgment condition is set for the duration of each segment. Specifically, if the duration of a single segment is less than 5 minutes, that segment is not used as a separate basis for anomaly judgment, but is merged with the previous segment and re-judged; if the duration of a single segment is not less than 5 minutes, that segment participates in anomaly identification as an independent judgment unit. The five-minute threshold is consistent with the single state refresh cycle, the purpose of which is to eliminate the influence of instantaneous state jitter on the identification results. Through this processing method, anomaly identification can be established on stable state segments, avoiding erroneous judgments due to single short-term state changes.

[0038] In this embodiment, after completing the segmentation state parsing, an airport consistency check is performed on the parsing results. Specifically, the airport at the end of the segment is compared with the final landing airport in the historical flight statistics data; if they match, the airport information is valid and can directly proceed to anomaly detection; if they do not match, the record is marked as a field conflict record and transferred to the verification set, neither directly as a normal sample nor directly as an abnormal sample. By setting this verification step, misjudgments caused by airport code errors, field mapping errors, or data synchronization delays can be prevented.

[0039] In this embodiment, when any segment corresponding to a flight is determined to be in a return-to-origin or diversion-to-arrival state, the entire flight is marked as an abnormal flight and removed from the original sample. It should be noted that this embodiment uses the method of removing the entire flight, rather than deleting only a single abnormal segment. This is because once a return-to-origin or diversion-to-arrival occurs, the entire flight process, arrival node, and time result deviate from the original operational link, and continuing to retain the remaining segments of the flight cannot maintain sample consistency. Correspondingly, flights that do not hit the return-to-origin judgment rule, do not hit the diversion-to-arrival judgment rule, have the same departure airport as the planned departure airport, have the same destination airport as the planned destination airport, and have actual arrival times are retained as normally operating flights and added to the pre-cleaned flight sample set.

[0040] S2 includes: performing a difference operation on the pre-cleaned flight sample set to obtain a time deviation numerical sequence; determining whether the time deviation numerical sequence exceeds a preset deviation threshold standard; if the time deviation numerical sequence exceeds the preset deviation threshold standard, it is determined to be an abnormal deviation record; filtering the abnormal deviation records based on historical weather data to obtain records that need to be removed; removing the records that need to be removed and integrating the remaining records in the flight sample set to determine the cleaned valid flight sample set.

[0041] In this embodiment, the time field is first extracted and formatted uniformly on the pre-cleaned flight sample set. Specifically, the planned arrival time, actual arrival time, planned departure time, actual departure time, execution date, departure airport, destination airport, and flight unique identifier are extracted from each flight sample. All time fields are then uniformly converted to the same time system and further converted into numerical time in minutes. As a preferred embodiment, the actual arrival time minus the planned arrival time is used as the main deviation value, and this main deviation value is written into the corresponding sample record. Using the method of this application, each sample record corresponds to a specific time deviation value. Thus, after the main deviation values ​​of all sample records are arranged in order according to the flight unique identifier, a time deviation value sequence for subsequent judgment is formed.

[0042] In this embodiment, the formation process of the time deviation numerical sequence adopts a method of calculating, writing, and summarizing each value individually. Specifically, when the actual arrival time is later than the planned arrival time, the calculation result is positive; when the actual arrival time is earlier than the planned arrival time, the calculation result is negative; and when the actual arrival time is the same as the planned arrival time, the calculation result is zero. It should be noted that this embodiment uses a fixed difference calculation rule, without using a weighted conversion rule, a manual correction rule, or a range fuzzy classification rule. Therefore, each deviation value can be directly recalculated from the original time field.

[0043] In the above implementation, in this embodiment, after obtaining the time deviation numerical sequence, a preset deviation threshold standard is applied to the sequence for judgment. The threshold is determined using fixed statistical rules. Specifically, samples in the same batch are first grouped according to execution date, departure airport, and destination airport; then, the time deviation values ​​of each group of samples are summarized and calculated to obtain the average deviation value of that group of samples; then, the absolute value of the difference between each time deviation value and the average deviation value is used as the fluctuation amount, and all fluctuation amounts are averaged to obtain the fluctuation benchmark value; finally, the upper limit threshold is fixedly set as the average deviation value plus twice the fluctuation benchmark value, and the lower limit threshold is fixedly set as the average deviation value minus twice the fluctuation benchmark value. Here, the average deviation value is used to represent the overall deviation level of the current sample group, the fluctuation benchmark value is used to represent the dispersion of each sample around the overall deviation level, and twice the fluctuation benchmark value is used to form the anomaly judgment boundary.

[0044] Furthermore, in some embodiments, when the number of valid samples within a group is not less than 20, the upper and lower limit thresholds are directly used as the average deviation value plus twice the fluctuation benchmark value and the average deviation value minus twice the fluctuation benchmark value; when the number of valid samples within a group is less than 20, the fluctuation benchmark value is no longer calculated, and a fixed minute threshold is used instead. Specifically, the upper limit threshold for late arrival anomalies is fixed at 180 minutes, and the lower limit threshold for early arrival anomalies is fixed at -120 minutes. This fixed threshold is directly applicable in scenarios with insufficient sample size and does not require secondary correction. As a preferred embodiment, the statistical scope of the sample size is fixed as the number of valid flight records within the same execution date interval, from the same departure airport to the same destination airport.

[0045] In this embodiment, after determining the upper and lower thresholds, each deviation value in the time deviation numerical sequence is compared one by one. Specifically, when a deviation value is greater than the upper threshold, the sample is directly determined to be a late arrival abnormal deviation record; when a deviation value is less than the lower threshold, the sample is directly determined to be an early arrival abnormal deviation record; when a deviation value is between the upper and lower thresholds, the sample is directly determined to be a normal deviation record. Furthermore, it should be noted that this embodiment adopts a "beyond the threshold is abnormal, below the threshold is normal" judgment logic.

[0046] In this embodiment, after identifying abnormal deviation records, all abnormal deviation records are not immediately deleted. Instead, historical weather association filtering is further performed on the abnormal deviation records. Specifically, the execution date, time of deviation occurrence, departure airport, and destination airport in the abnormal deviation records are matched with weather phenomena, visibility, wind speed, precipitation status, and severe convection records in historical weather data. The matching rule here is fixed as a dual matching rule, that is, when both time matching and location matching are met, the weather data association is determined to be valid; when either condition is not met, the weather data association is determined to be invalid. Among them, the time matching rule is fixed as follows: the time of occurrence of the weather record falls within two hours before or after the scheduled flight departure time, or falls within two hours before or after the scheduled arrival time; the location matching rule is fixed as follows: the airport corresponding to the weather record is the same as the departure airport in the abnormal deviation record, or the same as the destination airport.

[0047] In this embodiment, a fixed threshold rule is used to identify weather impacts. Specifically, when historical weather data meets any of the following conditions, the corresponding abnormal deviation record is determined to be affected by weather factors: First, there is a thunderstorm record at the departure airport or destination airport during the corresponding time period; second, the runway visual range or dominant visibility at the departure airport or destination airport during the corresponding time period is less than 1000 meters; third, the average wind speed at the departure airport or destination airport during the corresponding time period reaches 15 meters per second; fourth, the gust wind speed at the departure airport or destination airport during the corresponding time period reaches 20 meters per second; fifth, there is any one of the following weather phenomena during the corresponding time period: moderate rain, heavy rain, freezing rain, snowfall, or hail; sixth, there is a recorded severe convective weather event in the area traversed by the flight route. If any one of these six conditions is met, the weather impact is determined to be valid.

[0048] In this embodiment, abnormal deviation records are classified a second time based on historical weather association screening results. Specifically, when an abnormal deviation record does not meet the weather impact identification criteria, it is directly included in the set of records to be removed; when an abnormal deviation record meets the weather impact identification criteria, it is marked as a weather-related abnormal record and retained in the pending review set. Here, the logic for generating records to be removed adopts a dual-condition rule of "abnormal deviation without weather support". As can be seen from the aforementioned embodiment, abnormal deviations not supported by weather factors are more likely to originate from time field errors, data mismatches, abnormal status acquisition, or atypical operational events. Such records do not have the representativeness to be included in subsequent correction value statistics, and therefore should be removed.

[0049] In some embodiments, if the business objective is to construct a standard operating sample set, weather-related anomaly records are also included in the set of records to be removed; if the business objective is to retain actual operating disturbance samples, weather-related anomaly records are retained in the sample set, but a weather impact identifier field is added to the records. As a preferred embodiment, the system fixes the removal mode before batch processing begins and maintains it unchanged during the current processing. When the removal mode is fixed to the standard operating mode, all abnormal deviation records enter the set of records to be removed; when the removal mode is fixed to the weather retention mode, only records with "abnormal deviations and no weather support" enter the set of records to be removed. In another specific embodiment of this application, the weather impact identifier can also be written into the sample supplementary field.

[0050] In this embodiment, after forming the set of records to be removed, the pre-cleaned flight sample set undergoes deletion and integration processing. Specifically, firstly, all records to be removed are deleted from the sample set according to the unique flight identifier; then, the remaining records are renumbered; next, they are sorted in ascending order by execution date and ascending order by planned arrival time; finally, the retained records, along with their time deviation values, anomaly detection results, and weather correlation results, are uniformly written into the cleaned valid flight sample set. It should be noted that after the integration process, each record in the sample set contains a complete flight identity field, a complete time deviation field, and a clearly defined filtering status field.

[0051] S3 includes obtaining a cleaned set of valid flight samples, sorting the valid flight samples in ascending order based on the magnitude of their deviations to determine a deviation sequence, and identifying the upper and lower extreme value records in the deviation sequence to obtain a preliminary screening sequence; removing the upper and lower extreme value records from the preliminary screening sequence to obtain a boundary adjustment sequence; if the boundary adjustment sequence meets a preset threshold, then processing it through a cleaning process to obtain a sample quality improvement sequence; if the sample quality improvement sequence meets the ascending order requirements, then determining the selected core flight sample set.

[0052] In this embodiment, a cleaned set of valid flight samples is first obtained, and the time deviation value, flight unique identifier, execution date, departure airport, and destination airport fields are extracted from each sample record. Based on this, all valid flight samples are sorted in ascending order solely by the time deviation value, forming a unique deviation sequence. Specifically, when the time deviation value of one record is less than that of another, the record is ranked first; when the time deviation value of one record is greater than that of another, the record is ranked last; when two records have the same time deviation value, they are sorted in ascending order by execution date; if the execution dates are the same, they are then sorted in ascending order by flight unique identifier. This fixed sorting rule ensures that a completely consistent deviation sequence is obtained when the same batch of samples is processed repeatedly.

[0053] In this embodiment, after the deviation sequence is formed, the extreme value records at both ends of the sequence are identified. Specifically, the total number of samples in the current deviation sequence is first counted and recorded as the sample size; then, the number of extreme values ​​to be removed at both ends is calculated. As a preferred embodiment, when the sample size is not less than 20, the number of extreme values ​​to be removed at the upper end is fixedly set to the integer obtained by multiplying the sample size by 10% and rounding down, and the number of extreme values ​​to be removed at the lower end is fixedly set to the integer obtained by multiplying the sample size by 10% and rounding down; when the sample size is less than 20, the values ​​are not proportionally selected, but the first record at the top of the sort and the last record at the bottom of the sort are fixedly identified as the upper and lower extreme value records. Here, the 10% threshold is a fixed proportion used to control the influence range of boundary samples on the main distribution; when the sample size is less than 20, one record at the top and one at the bottom are fixedly removed to ensure that the boundary shrinkage capability is still present in small sample scenarios.

[0054] In this embodiment, after identifying the upper and lower extreme value records, the lower extreme value record located at the beginning of the deviation sequence and the upper extreme value record located at the end of the deviation sequence are marked respectively, and a preliminary screening sequence is generated. It should be noted that the preliminary screening sequence in this embodiment is not the result after deleting extreme values, but rather an intermediate sequence after the extreme positions have been marked. This intermediate sequence includes at least four types of fields: original sorting position, deviation value, extreme value mark, and sample identity identifier.

[0055] In this embodiment, after forming the initial screening sequence, the marked upper and lower extreme value records are deleted to obtain the boundary adjustment sequence. Specifically, the deletion rule adopts a global deletion rule, that is: all samples marked as lower extreme value records are directly deleted from the initial screening sequence; all samples marked as upper extreme value records are directly deleted from the initial screening sequence. After deletion, consecutive serial numbers are regenerated for the remaining samples to obtain the boundary adjustment sequence.

[0056] In this embodiment, after obtaining the boundary adjustment sequence, a preset threshold judgment is performed on the sequence to determine whether to proceed to the subsequent cleaning process. The preset threshold adopts a two-condition judgment rule. The first condition is the sample retention rate threshold, and the second condition is the deviation dispersion threshold. Specifically, the sample retention rate is calculated by dividing the number of samples in the boundary adjustment sequence by the number of samples in the initial valid flight sample set; when the sample retention rate is not less than 80%, the sample quantity is deemed to meet the requirement; when the sample retention rate is less than 80%, the sample quantity is deemed not to meet the requirement. The deviation dispersion is calculated by the difference between the maximum and minimum deviation values ​​in the boundary adjustment sequence; when the difference is not greater than 120 minutes, the deviation dispersion is deemed to meet the requirement; when the difference is greater than 120 minutes, the deviation dispersion is deemed not to meet the requirement. Only when the sample retention rate is not less than 80% and the deviation dispersion is not greater than 120 minutes is the boundary adjustment sequence deemed to meet the preset threshold.

[0057] Furthermore, in some embodiments, if the boundary adjustment sequence does not meet the aforementioned dual-condition threshold, the sequence is not directly output; instead, boundary compression processing is re-executed. Specifically, the number of extreme values ​​to be removed is calculated again based on the number of samples in the current boundary adjustment sequence. The removal ratio is fixed at 5%, meaning that records at both ends are removed by rounding down to the nearest integer after multiplying the current sample size by 5%. When the current sample size is less than 20, one record is still removed from both ends. After this second removal, the sample retention rate and deviation dispersion are recalculated until a preset threshold is met, or until the number of samples drops to 10. Once the number of samples drops to 10, compression stops, and the current sequence is directly used as the boundary adjustment result. The purpose of this rule is to prevent excessive cleaning from causing sample distortion.

[0058] In this embodiment, once the boundary adjustment sequence meets a preset threshold, a cleaning process is performed on the sequence to obtain a sample quality improved sequence. The cleaning process includes duplicate value verification, field consistency verification, and sorting integrity verification. Specifically, duplicate value verification is used to delete duplicate records with identical flight unique identifiers, execution dates, and deviation values; field consistency verification is used to confirm that each sample record contains flight unique identifiers, execution dates, departure airports, destination airports, and deviation value fields; if any field is missing, the record is directly deleted; sorting integrity verification is used to confirm that the deviation values ​​in the current sequence are still arranged in ascending order; if a reversed record is found at a certain position, an ascending sort is performed again.

[0059] In this embodiment, after the cleaning process is completed, a sample quality improvement sequence is formed, and a final sorting conformity judgment is performed on the sequence. Specifically, starting from the first record of the sequence, the deviation values ​​of adjacent two records are compared one by one; if the deviation value of the previous record is not greater than the deviation value of the next record, the adjacent position is determined to meet the ascending order requirement; if the deviation value of the previous record is greater than the deviation value of the next record, the sequence is directly determined not to meet the ascending order requirement, and the sorting process is re-executed. Only when all adjacent positions satisfy the condition that the previous value is not greater than the next value is the sample quality improvement sequence determined to meet the ascending order requirement.

[0060] In this embodiment, when the improved sample quality sequence meets the ascending order requirement, the sequence is determined as the refined core flight sample set. It should be noted that the core flight sample set in this embodiment refers to the sample set after abnormal deviation removal, extreme boundary contraction, sample retention rate determination, dispersion determination, duplicate value deletion, field consistency verification, and final sorting confirmation. Based on the aforementioned embodiments, this set, compared to the cleaned effective flight sample set, has a more stable deviation distribution boundary and higher statistical consistency, thus making it more suitable as the input basis for subsequent historical deviation correction value calculations.

[0061] S4 includes averaging the flight deviation values ​​to obtain an initial average deviation value; evaluating the historical data volume based on the historical data associated with the initial average deviation value to obtain a historical data volume evaluation result; determining whether the historical data volume evaluation result is lower than a preset threshold; if the historical data volume evaluation result is lower than the preset threshold, obtaining borrowed correction data based on the feature matching of the flight corresponding to the initial average deviation value; and using the borrowed correction data to perform compensation calculations on the initial average deviation value to determine the final historical deviation correction value.

[0062] In this embodiment, the average deviation values ​​of all flights in the core flight sample set are first calculated to obtain an initial average deviation value. Specifically, all deviation values ​​involved in the calculation are summed, and the summation result is divided by the number of samples involved in the calculation to obtain the average deviation result corresponding to the current sample set. Here, all deviation values ​​use minutes as the uniform unit of measurement, with positive values ​​indicating later than the scheduled time and negative values ​​indicating earlier than the scheduled time. As a preferred embodiment, when the number of samples in the core flight sample set is 0, the averaging calculation is not performed, and the process directly proceeds to the historical data shortage handling flow; when the number of samples is not less than 1, the initial average deviation value is calculated directly according to the above rules.

[0063] In this embodiment, after obtaining the initial average deviation value, the historical data corresponding to this initial average deviation value is evaluated to obtain the historical data volume evaluation result. Specifically, the historical data volume evaluation result consists of two components: the number of valid sample entries and the number of days covered by the time period. The number of valid sample entries refers to the number of core flight samples actually used in the calculation of the initial average deviation value; the number of days covered by the time period refers to the number of dates after deduplication of the execution dates corresponding to the above samples. It should be noted that this embodiment adopts a dual-indicator evaluation rule, not a single-entry-count evaluation rule or a single-day evaluation rule. The reason is that if only the number of sample entries is considered, the samples may be concentrated on a few dates; if only the number of days covered is considered, there may be insufficient samples on a single day.

[0064] In this embodiment, after the historical data volume assessment result is generated, a threshold judgment is performed on the result to determine whether the current historical data meets the direct output requirements. The preset threshold adopts a fixed dual-threshold rule. Specifically, the effective sample number threshold is fixed at 30, and the time coverage days threshold is fixed at 7 days. Only when the effective sample number is not less than 30 and the time coverage days are not less than 7 days is the historical data volume assessment result determined to meet the requirements; if the effective sample number is less than 30, or the time coverage days are less than 7 days, the historical data volume assessment result is determined to be below the preset threshold. Here, 30 samples are used to ensure that the average statistics have a stable basis, and 7 days of coverage are used to ensure that the samples can cover the continuous operation period.

[0065] In some embodiments, if the system uses statistical windows of different lengths, the threshold can be pre-fixed to another set of corresponding thresholds. Specifically, when the statistical window is fixed to the most recent 3 days, the threshold for the number of valid samples is fixed at 15, and the threshold for the number of days covered by the time is fixed at 3; when the statistical window is fixed to the most recent 7 days, the threshold for the number of valid samples is fixed at 30, and the threshold for the number of days covered by the time is fixed at 7; when the statistical window is fixed to the most recent 14 days, the threshold for the number of valid samples is fixed at 60, and the threshold for the number of days covered by the time is fixed at 10.

[0066] In this embodiment, when the historical data volume assessment result meets the preset threshold, it indicates that the current core flight sample set is capable of independently supporting the generation of historical deviation correction values. Therefore, the initial average deviation value is directly determined as the final historical deviation correction value, and the borrowing correction process is no longer executed. It should be noted that in this case, the initial average deviation value is not amplified, reduced, or manually corrected; instead, it is directly output.

[0067] In this embodiment, when the historical data volume assessment result is lower than a preset threshold, it indicates that the current sample is insufficient to support a stable correction value output on its own, and the borrowing correction process is initiated. Specifically, the operational characteristics of the flight corresponding to the initial average deviation value are first extracted, and then matching processing is performed based on these operational characteristics to obtain the borrowed correction data. The operational characteristics include departure airport, destination airport, planned flight duration, route type, flight execution time, and flight operation type. Among them, departure airport and destination airport are used to characterize airport pair relationships, planned flight duration is used to characterize range class, route type is used to characterize flight direction, flight execution time is used to characterize the operation time interval, and flight operation type is used to characterize operational attributes. As a preferred embodiment, the matching priority is executed in a fixed order, namely, airport pair consistency, route consistency, similar planned flight duration, consistent execution time, and consistent operation type. Through the method of this application, a data set with consistent operational attributes with the target flight can be selected from the historical sample library.

[0068] In this embodiment, the acquisition of borrowed correction data is accomplished by progressively relaxing the matching conditions. Specifically, the first-level matching requirements are consistent departure airport, consistent destination airport, consistent flight route, and consistent execution time. If the number of historical samples meeting this level condition is not less than 20, the average deviation result of this group of samples is directly used as the borrowed correction data. If the first-level matching results are less than 20, the process proceeds to the second-level matching. The second-level matching requirements are consistent departure airport region, consistent destination airport region, consistent flight route, and a planned flight duration difference not exceeding 20 minutes. If the number of historical samples meeting this level condition is not less than 20, the average deviation result of this group of samples is used as the borrowed correction data. If the second level still has fewer than 20 results, the process proceeds to the third-level matching. The third-level matching requirements are consistent flight route, a planned flight duration difference not exceeding 30 minutes, and consistent execution time. If the number of historical samples meeting this level condition is not less than 20, the average deviation result of this group of samples is used as the borrowed correction data. If the third layer still has fewer than 20 data points, the overall average deviation result of similar flights will be used directly as the borrowed and corrected data. Here, the 20-sample threshold is used to ensure that the borrowed data has basic statistical stability.

[0069] In this embodiment, after obtaining the borrowed and corrected data, the initial average deviation value is compensated using the borrowed and corrected data to determine the final historical deviation correction value. Specifically, when the amount of historical data is insufficient but the number of valid samples is not less than 10, the initial average deviation value is used as the main compensation basis, and the borrowed and corrected data is used as the auxiliary compensation basis. The initial average deviation value has a fixed weight of 60%, and the borrowed and corrected data has a fixed weight of 40%. The reason for this setting is that when the number of samples is not less than 10, the current sample still has a certain representativeness, so the current statistical results should be the main basis. When the amount of historical data is insufficient and the number of valid samples is less than 10, the borrowed and corrected data is used as the main compensation basis, and the initial average deviation value is used as the auxiliary compensation basis. The borrowed and corrected data has a fixed weight of 60%, and the initial average deviation value has a fixed weight of 40%. The reason for this setting is that when the number of samples is less than 10, the current sample's ability to reflect the overall pattern is insufficient, so the influence of the borrowed and corrected data should be increased.

[0070] In this embodiment, it should also be noted that boundary constraint processing is performed on the final result after the compensation calculation is completed. Specifically, when the final historical deviation correction value is greater than 180 minutes, the result is fixed at 180 minutes; when the final historical deviation correction value is less than -120 minutes, the result is fixed at -120 minutes; when the final historical deviation correction value is between -120 minutes and 180 minutes, the calculation result remains unchanged. Here, 180 minutes is used as the upper limit for late arrival correction, and -120 minutes is used as the lower limit for early arrival correction, to prevent the compensated result from exceeding the normal operation statistical range.

[0071] S5 includes extracting segment node sequences from flight plan data to construct route orientation features; grouping the core flight sample set using route orientation features to obtain a target route orientation classification set; obtaining historical deviation correction values ​​for the target route orientation classification set; performing deviation mapping processing on the target route orientation classification set using historical deviation correction values ​​to obtain an initial deviation adjustment sequence; determining whether the record completeness assessment value of the initial deviation adjustment sequence is lower than a preset completeness threshold; if the record completeness assessment value is lower than the preset completeness threshold, extracting the missing field positions in the initial deviation adjustment sequence, interpolating and supplementing the missing field positions based on the historical deviation correction values ​​of adjacent segment nodes, and determining the deviation adjustment parameters corresponding to the route orientation.

[0072] In this embodiment, a segment node sequence is first extracted from the flight plan data, and a route orientation feature is constructed based on this sequence. Specifically, the segment node sequence includes at least the departure airport, departure route node, main route node, arrival route node, and destination airport. As a preferred embodiment, the system first sorts all nodes according to their order of appearance in the flight plan, then deletes duplicate and invalid nodes to form a standardized node sequence. Here, duplicate nodes refer to nodes with the same name that appear consecutively, and invalid nodes refer to nodes that lack a name, lack a node order, or lack a geographic location identifier.

[0073] In this embodiment, after obtaining the standardized node sequence, a route orientation feature construction process is performed on the node sequence. Specifically, the first valid node is extracted as the starting direction node, the last valid node is extracted as the ending direction node, and a main reference node located in the middle of the node sequence is extracted as the intermediate direction node. Subsequently, the route orientation feature is determined based on the combination relationship of the starting direction node, the intermediate direction node, and the ending direction node. As a preferred embodiment, when the node sequence length is not less than 5 nodes, the second node is fixed as the starting direction node, the second to last node is extracted as the ending direction node, and the node corresponding to the middle sequence number is extracted as the main reference node; when the node sequence length is less than 5 nodes, the first node is fixed as the starting direction node, the last node is extracted as the ending direction node, and the route orientation feature is obtained by combining all nodes in sequence.

[0074] In this embodiment, in addition to the node sequence characteristics, the geographical directions corresponding to the node sequence are also categorized. Specifically, the direction connecting the departure airport to the main reference node is taken as the first segment direction, and the direction connecting the main reference node to the destination airport is taken as the second segment direction. Then, the directions are classified into fixed direction categories according to the azimuth angle range. As a preferred embodiment, azimuth angles between 0 and 45 degrees and between 315 and 360 degrees are classified as north-facing; azimuth angles between 45 and 135 degrees are classified as east-facing; azimuth angles between 135 and 225 degrees are classified as south-facing; and azimuth angles between 225 and 315 degrees are classified as west-facing. Therefore, the method of this application can combine the segment node sequence information and geographical direction information to form the route orientation characteristics, so that the classification results reflect both the path structure and the operating direction.

[0075] In this embodiment, after the route orientation features are constructed, the core flight sample set is grouped using these features to obtain a target route orientation classification set. Specifically, the system reads the flight plan data corresponding to each core flight sample and assigns flight samples with the same starting direction node, the same main reference node, the same ending direction node, and the same direction category to the same classification set. It should be noted that this embodiment adopts the grouping rule of "all features being identical simultaneously" and does not adopt the fuzzy grouping rule of "partial feature similarity". That is, only when the key nodes and direction categories of the node sequence are all identical are they determined to be in the same route orientation classification. Based on the foregoing embodiments, it can be seen that the method of this application can ensure that each core flight sample belongs to only one specific route orientation classification set.

[0076] In this embodiment, after obtaining the target route orientation classification set, the system further acquires the historical deviation correction values ​​corresponding to each classification set and performs deviation mapping processing on the classification sets to obtain the initial deviation adjustment sequence. Specifically, the system first reads the historical deviation correction values ​​output in the previous step, and then writes the corresponding correction values ​​into each record in each classification set according to the route orientation classification identifier. As a preferred embodiment, when the number of samples in a certain route orientation classification set is not less than 20, the historical average deviation result corresponding to that classification set is directly used as the historical deviation correction value; when the number of samples in a certain route orientation classification set is less than 20, the historical deviation correction value under the same direction category at the next higher level is directly used as the temporary mapping value. Here, 20 samples are used as the minimum sample threshold for the classification-level correction value.

[0077] In this embodiment, the initial deviation adjustment sequence includes at least a unique flight identifier, a route classification identifier, a node sequence identifier, historical deviation correction values, a record completeness identifier, and a field missing identifier. It should be noted that each record in the initial deviation adjustment sequence retains its classification information and correction value source information during generation, so as to facilitate subsequent completeness assessment and missing data supplementation.

[0078] In this embodiment, after forming the initial deviation adjustment sequence, a record completeness assessment is performed on the sequence. Specifically, the completeness assessment adopts a fixed field counting rule. First, it is preset that each record must have six fields: flight unique identifier, route classification identifier, starting direction node, primary reference node, ending direction node, and historical deviation correction value. Then, the number of valid fields actually existing in each record is counted. Finally, the number of valid fields is divided by the required number of fields to obtain the record completeness assessment value. As a preferred embodiment, when all six fields are present, the record completeness assessment value is 100%; when one field is missing, the record completeness assessment value is 83%; when two fields are missing, the record completeness assessment value is 67%; and when three or more fields are missing, the record completeness assessment value is below 67%.

[0079] In this embodiment, when determining whether the record completeness assessment value is lower than a preset completeness threshold, the preset completeness threshold is fixed at 85%. Only when the record completeness assessment value is not lower than 85% is the record considered complete; when the record completeness assessment value is lower than 85%, the record is directly determined to be missing, and the missing field supplementation process begins. The reason for setting the completeness threshold to 85% is that when one of the six key fields is missing, the completeness assessment value is 83%, which is insufficient to directly support the output of deviation adjustment parameters, and therefore must be supplemented; when all six key fields are present, the completeness assessment value is 100%, satisfying the direct output condition.

[0080] In this embodiment, when the record completeness assessment value is below 85%, the system further extracts the positions of missing fields in the initial deviation adjustment sequence. Specifically, it first scans the six key fields in the initial deviation adjustment sequence one by one, and then marks the positions of fields that are null, blank, or invalid placeholder values ​​as missing field positions. As a preferred embodiment, if the missing field is a start direction node or an end direction node, it first checks the same category node fields in adjacent records; if the missing field is a historical deviation correction value, it directly enters the interpolation supplementation process; if two or more key fields are missing at the same time, it first supplements the node fields, and then supplements the correction value fields.

[0081] In this embodiment, after determining the location of the missing field, interpolation is performed to supplement the missing field location based on the historical deviation correction values ​​of adjacent flight segment nodes. Specifically, when the node corresponding to the missing field is located between two valid adjacent nodes, the historical deviation correction values ​​of the preceding and following adjacent nodes are averaged, and the average result is written into the missing field location; when the missing field is located at the beginning of the node sequence and only the following adjacent node exists, the historical deviation correction value of the following adjacent node is directly used as the supplementary value; when the missing field is located at the end of the node sequence and only the preceding adjacent node exists, the historical deviation correction value of the preceding adjacent node is directly used as the supplementary value.

[0082] In this embodiment, in some implementations, when there are valid adjacent nodes on both sides of the missing field location, but the difference in historical deviation correction values ​​between the two adjacent nodes exceeds 30 minutes, averaging is not directly used for supplementation. Instead, the value of the adjacent node consistent with the current flight path direction is preferentially used as the supplementary value. Here, 30 minutes is used as the threshold for determining the difference between adjacent nodes to avoid distortion of the supplementary value caused by directly averaging when the correction difference between adjacent nodes is too large. Specifically, if the current missing field is located within the preceding directional interval, the historical deviation correction value of the preceding adjacent node is used; if the current missing field is located within the following directional interval, the historical deviation correction value of the following adjacent node is used.

[0083] In this embodiment, after the missing fields are supplemented, a completeness check is performed on the supplemented results. Specifically, a completeness assessment is performed again according to the six key fields; when the completeness assessment value of the supplemented record reaches 100%, the record is retained in the deviation adjustment results; when the completeness assessment value of the supplemented record is still below 100%, the record is marked as a record to be excluded and is not included in the final output parameter set. Here, the final output parameter set only retains records with complete key fields and does not retain records that still have missing fields.

[0084] In this embodiment, after all classification sets have completed deviation mapping, completeness assessment, and missing field supplementation, the classification results are summarized to determine the deviation adjustment parameters corresponding to the route orientation. The deviation adjustment parameters include at least the route orientation classification identifier, node sequence features, direction category identifier, and corresponding historical deviation correction values. It should be noted that the final output deviation adjustment parameters no longer retain missing status identifiers, but only retain valid parameter records that have passed the completeness verification.

[0085] S6 includes acquiring an initial dataset for flight plan data fusion, calibrating the initial dataset based on the route deviation to obtain an integrated sequence; if the integrated sequence is in the pre-takeoff state, calculating the offset based on the expected takeoff base to determine the time series; extracting flight phase distinguishing features based on the time series, fusing the flight phase distinguishing features with historical statistical data to obtain a preliminary estimated sequence; extracting details from the preliminary estimated sequence, and processing the details by combining the takeoff time to determine the estimated arrival time.

[0086] In this embodiment, the initial dataset for flight plan data fusion is first obtained. Specifically, the system reads the flight plan data corresponding to the current flight and extracts the planned departure time, planned arrival time, departure airport, destination airport, planned flight duration, segment node sequence, and route classification identifier. Simultaneously, it reads the flight dynamic data corresponding to the flight and extracts the flight status, estimated departure time, actual departure time, estimated arrival time, and current status update time. Then, it reads the route deviation adjustment parameters generated in the previous steps and extracts historical deviation correction values ​​consistent with the current flight's route classification. Subsequently, the above data is associated according to the flight's unique identifier and execution date to form the initial dataset. As a preferred embodiment, the initial dataset contains at least 10 fields: flight unique identifier, execution date, departure airport, destination airport, planned departure time, planned arrival time, estimated departure time, actual departure time, route classification identifier, and historical deviation correction value. If two or more of the above 10 fields are missing, the record will not proceed to the subsequent estimation process.

[0087] In this embodiment, after obtaining the initial dataset, it is calibrated based on the route deviation to obtain an integrated sequence. Specifically, the system first reads the route classification identifier corresponding to the current flight, then retrieves the historical deviation correction value corresponding to the same classification identifier from the route deviation adjustment parameter table, and writes the historical deviation correction value into the current flight record. Subsequently, using the planned arrival time as the base time point, the historical deviation correction value is superimposed on the planned arrival time to form the calibrated arrival time; simultaneously, the planned flight duration is combined with the historical deviation correction value to form the calibrated flight duration. Here, if the historical deviation correction value is positive, it indicates that the overall operation of flights under this route is biased towards late arrival, and the calibrated arrival time is shifted backward accordingly; if the historical deviation correction value is negative, it indicates that the overall operation of flights under this route is biased towards early arrival, and the calibrated arrival time is shifted forward accordingly. Based on the aforementioned embodiment, it can be seen that the method of this application can ensure that the initial dataset has completed route differentiation correction before entering the operational stage for discrimination, thereby forming an integrated sequence.

[0088] In this embodiment, the integrated sequence undergoes further operational status identification to determine whether the current flight is in a pre-departure state. Specifically, the system sequentially checks the actual departure time field, the departure status field, and the departure confirmation flag field. Only when the actual departure time is empty, the departure status field does not display a departed status, and the departure confirmation flag field is in an unconfirmed state, is the current integrated sequence determined to be in a pre-departure state. As a preferred embodiment, if the time difference between the estimated departure time and the current status update time is greater than or equal to 0 minutes and less than 240 minutes, the flight is determined to be in an estimable pre-departure state; if the time difference is less than 0 minutes, it indicates that the estimated departure time has expired, and the actual departure time field needs to be checked first; if the time difference is greater than 240 minutes, the flight is marked as a long-term plan state and will not enter the immediate estimation process in this step. Here, 240 minutes is used as the maximum time window threshold for immediate pre-departure estimation.

[0089] In this embodiment, if the integrated sequence is in the pre-takeoff state, the offset is calculated based on the expected takeoff baseline to determine the time series. Specifically, the system uses the expected takeoff time as the starting point and adds the planned flight duration to the historical deviation correction values ​​corresponding to the route direction to form the expected total time after takeoff. Then, the expected total time after takeoff is extended backward by the expected takeoff time to form candidate arrival times. At the same time, the difference between the planned takeoff time and the expected takeoff time is determined as the takeoff offset, and this takeoff offset is written into the current flight record. Subsequently, a time series is generated using the current status update time, expected takeoff time, candidate arrival time, and takeoff offset as sequential nodes. It should be noted that the time series in this embodiment includes at least four time nodes: the current status update time, the expected takeoff time, the end time after calibrated takeoff, and the candidate arrival time. Through this processing method, the current flight can be mapped from the "not yet taken off" state to a set of estimated time nodes with sequential relationships, providing a basis for distinguishing subsequent flight phases.

[0090] In this embodiment, flight phase differentiation features are extracted based on time series data. Specifically, the system extracts four features from the time series: takeoff waiting time, estimated flight time, takeoff offset, and planned arrival offset. Takeoff waiting time is determined by the time difference between the estimated takeoff time and the current status update time; estimated flight time is determined by the time difference between the candidate arrival time and the estimated takeoff time; takeoff offset is determined by the time difference between the estimated takeoff time and the planned takeoff time; and planned arrival offset is determined by the time difference between the candidate arrival time and the planned arrival time. As a preferred embodiment, when the takeoff waiting time is greater than 0 minutes but not more than 180 minutes, the flight is classified into the near-term waiting phase; when the takeoff waiting time is greater than 180 minutes but not more than 240 minutes, the flight is classified into the far-term waiting phase; and when the takeoff waiting time is equal to 0 minutes, the flight is classified into the critical takeoff phase. Here, 180 minutes and 240 minutes are fixed phase division thresholds.

[0091] In this embodiment, after extracting the flight phase distinguishing features, the flight phase distinguishing features are fused with historical statistical data to obtain a preliminary estimation sequence. Specifically, the system reads historical statistical data consistent with the airport pair, route classification, and execution time of the current flight, and extracts three statistical values: historical average ground waiting time, historical average departure offset time, and historical average air time correction. These three statistical values ​​are then fused with the corresponding takeoff waiting time, takeoff offset, and estimated flight time. As a preferred embodiment, if the current flight is in the near-term takeoff phase, the fusion is completed primarily with the current takeoff waiting time and secondarily with the historical average ground waiting time; if the current flight is in the far-term takeoff phase, the fusion is completed primarily with the historical average ground waiting time and secondarily with the current takeoff waiting time; if the current flight is in the critical takeoff phase, the current takeoff waiting time is directly used as the ground phase estimation result. Further, for the air time portion, the calibrated flight time combined with the historical average air time correction is used to form the air phase estimation result. Subsequently, the ground phase estimation result and the air phase estimation result are written into the same record in chronological order to form a preliminary estimation sequence. Therefore, the method of this application can be used to estimate the pre-flight status by combining the current time series with historical statistical patterns, so that the preliminary estimation results reflect both the current dynamic status and the historical operating patterns of similar flights.

[0092] In this embodiment, details are extracted from the preliminary estimation sequence, and these details are processed in conjunction with the takeoff time to determine the estimated arrival time. Specifically, the system extracts five detailed information items from the preliminary estimation sequence: estimated ground waiting time, estimated flight time, historical departure offset time, planned arrival time, and calibrated arrival time. Then, processing is performed according to the takeoff time combination rules. When the actual takeoff time is empty, the estimated takeoff time is used as the takeoff reference time; when the actual takeoff time has been generated but state synchronization has not yet been completed, the actual takeoff time is used as the takeoff reference time. Subsequently, the estimated flight time is superimposed on the takeoff reference time, and a consistency comparison is performed in conjunction with the historical departure offset time and the calibrated arrival time to determine the estimated arrival time. As a preferred embodiment, when the difference between the "takeoff reference time plus estimated flight time" and the calibrated arrival time does not exceed 20 minutes, the calibrated arrival time is directly used as the estimated arrival time; when the difference exceeds 20 minutes, the result of the takeoff reference time plus the estimated flight time is used as the estimated arrival time. Here, 20 minutes is used as the consistency verification threshold for the estimation result to eliminate conflicts when the differences between the two path estimation results are too large.

[0093] In this embodiment, it should also be noted that after the initial estimated sequence is generated, a boundary rationality check can be performed on the estimation result. Specifically, when the estimated arrival time is earlier than the current state update time, the result is directly deemed invalid, and a backup estimation result is generated by adding the historical deviation correction value to the planned arrival time; when the estimated arrival time is later than the planned arrival time by more than 720 minutes, the result is directly limited to 720 minutes after the planned arrival time. Here, 720 minutes is used as the upper limit threshold of the estimated arrival time to prevent the output of significantly distorted estimation results due to field abnormalities or data mismatches.

[0094] S7 includes extracting position and speed information from automatic dependent surveillance (ADS) data to generate a real-time flight status sequence; determining the position offset value based on the real-time flight status sequence; calculating the remaining flight distance based on the position offset value and speed information if the position offset value exceeds a preset range; dynamically correcting the preliminary estimated arrival time using the remaining flight distance to obtain a corrected time sequence; and fusing the corrected time sequence with flight dynamic data and historical flight statistics to obtain the final estimated arrival time result.

[0095] In this embodiment, the current position, historical continuous positions, ground speed, heading, and timestamp are first extracted from the broadcast automatic correlation surveillance data, and a real-time flight status sequence is generated accordingly. Specifically, the system associates the received surveillance data according to the flight's unique identifier and the surveillance timestamp, arranges the surveillance records reported by the same flight within a continuous time period in chronological order, and then writes the longitude, latitude, altitude, ground speed, heading, and recording time of each surveillance record into the same sequence. As a preferred embodiment, the sampling interval of the real-time flight status sequence is fixed at 5 seconds. When the reporting interval of the original surveillance data is less than 5 seconds, the record closest to the 5-second node is retained; when the reporting interval of the original surveillance data is greater than 5 seconds but does not exceed 15 seconds, the original record is directly retained; when the time interval between two adjacent surveillance records exceeds 15 seconds, this time period is marked as a status discontinuity interval.

[0096] In this embodiment, after forming a real-time flight state sequence, a validity check is performed on the sequence. Specifically, the system sequentially checks whether the position field, velocity field, and time field are complete. If the current position, ground speed, or timestamp is missing, the corresponding record is not included in the subsequent calculation process; if the number of consecutive valid records is less than 3, dynamic correction is not performed, and the preliminary estimated arrival time is directly used; if the number of consecutive valid records is not less than 3, the position offset determination process is initiated. It should be noted that this embodiment uses 3 consecutive valid records as the minimum data threshold for entering dynamic correction because a single record can only reflect the instantaneous position, 2 records can only reflect the direction of a single displacement, while 3 consecutive records are needed to form a stable flight trend judgment. As can be seen from the above, the method of this application can ensure that subsequent offset identification and distance recalculation are based on continuous and valid flight states.

[0097] In this embodiment, the position offset value is determined based on the real-time flight status sequence. Specifically, the system first extracts the planned route node sequence corresponding to the current flight from the flight plan data, and then projects the latest valid position in the real-time flight status sequence onto the reference track formed by connecting the planned route nodes to determine the nearest track projection point corresponding to the latest valid position. Subsequently, the lateral distance between the latest valid position and the nearest track projection point is calculated, and this lateral distance is determined as the position offset value. As a preferred embodiment, when the latest valid position is between two adjacent planned route nodes, the nearest track projection point is determined only within the current planned segment formed by these two nodes; when the current planned segment cannot be determined, the segment closest to the latest valid position is selected from all planned segments as the projection basis. Through this processing method, it can be ensured that only one determined position offset value is formed at each moment.

[0098] In this embodiment, to improve the stability of position offset judgment, a continuous consistency check is performed on the latest three valid position records in the real-time flight status sequence. Specifically, when the position offset values ​​corresponding to the latest three valid position records all change in the same offset direction, and the difference between the maximum and minimum position offset values ​​among the three records does not exceed 8 kilometers, the current offset state is determined to be stable; when the above conditions are not met, the current offset state is determined to be unstable, and the system continues to wait for the next valid monitoring record. Here, 8 kilometers is used as the offset fluctuation consistency threshold to prevent misjudgment caused by instantaneous trajectory jumps or monitoring data jitter. Based on the foregoing embodiments, it can be seen that the method of this application can make position offset judgment consider not only the latest single-point result, but also the continuous trajectory change trend.

[0099] In this embodiment, after obtaining the position offset value, it is determined whether the value exceeds a preset range. The preset range adopts a fixed threshold rule. Specifically, when the position offset value does not exceed 15 kilometers, it is determined that the current flight is still within the allowable offset range of the planned route, and the remaining flight distance recalculation process is not initiated; when the position offset value exceeds 15 kilometers, it is determined that the current flight has deviated from the allowable range of the planned route, and the remaining flight distance recalculation process is initiated. Here, 15 kilometers is used as the position offset determination threshold.

[0100] In this embodiment, when the position offset exceeds 15 kilometers, the remaining flight distance is calculated based on the position offset and speed information. Specifically, the system first determines the nearest planned route node corresponding to the current flight position, and then determines the remaining planned flight segment from that nearest planned route node to the destination airport. Subsequently, the actual distance connecting the current flight position to the nearest planned route node is added to the cumulative distance of the remaining planned flight segment to obtain the basic remaining flight distance. Further, when the position offset exceeds 15 kilometers but does not exceed 40 kilometers, the basic remaining flight distance is increased by 1 times the position offset as a detour compensation distance; when the position offset exceeds 40 kilometers, the basic remaining flight distance is increased by 1.5 times the position offset as a detour compensation distance. Here, 40 kilometers is used as a threshold to distinguish between severe and significant offsets. This processing method allows the remaining flight distance to simultaneously reflect the planned flight distance and the additional flight length caused by the current offset.

[0101] In this embodiment, before calculating the remaining flight time using speed information, the speed information is first smoothed. Specifically, the system extracts the ground speeds corresponding to the latest three valid records in the real-time flight status sequence, and averages these three ground speeds, determining the average result as the current valid speed. If the difference between the maximum and minimum values ​​in the latest three ground speed records exceeds 120 kilometers per hour, the speed fluctuation is deemed too large. Instead of directly using the average result of the current three ground speeds, the average speed during the mid-flight phase of the historical flight along the same route is used as the current valid speed. Here, 120 kilometers per hour is used as the speed fluctuation judgment threshold.

[0102] In this embodiment, after obtaining the current effective speed, the remaining flight time is determined using the remaining flight distance and the current effective speed. Specifically, when the current altitude is not lower than 3000 meters, it indicates that the flight is in a stable climb, cruise, or the early stage of descent, and the remaining flight time is directly calculated based on the remaining flight distance and the current effective speed. When the current altitude is lower than 3000 meters, it indicates that the flight has entered the approach or pre-landing stage, and in this case, it no longer relies entirely on the current effective speed, but instead adds the historical average approach time to the airport to the remaining flight time. As a preferred embodiment, the historical average approach time to the airport is fixed at 18 minutes. Through this phased processing method, the remaining flight time can be calculated using different logics in the cruise and approach phases, thereby improving the adaptability of the dynamic correction results to the actual flight phase.

[0103] In this embodiment, the preliminary estimated arrival time is dynamically corrected using the remaining flight distance to obtain a corrected time series. Specifically, the system uses the timestamp of the latest valid monitoring record as the current correction start time, combines the current correction start time with the remaining flight time in sequence to obtain the current corrected arrival time; then, multiple consecutive corrected arrival times are arranged in monitoring time order to form a corrected time series. As a preferred embodiment, when the difference between two consecutive corrected arrival times does not exceed 3 minutes, the newer corrected arrival time is retained; when the difference between two consecutive corrected arrival times exceeds 3 minutes, a smooth update is performed on the newer corrected arrival time, that is, it is only allowed to be adjusted forward or backward by 3 minutes relative to the previous corrected arrival time, and the excess is retained for correction in the next correction cycle. Here, 3 minutes is used as a correction change smoothing threshold to prevent drastic jumps in the result during continuous updates. Based on the foregoing embodiments, it can be seen that the method of this application can enable the dynamic correction process to respond to real-time trajectory changes while maintaining continuous and stable result output.

[0104] In this embodiment, when the position offset does not exceed 15 kilometers, the remaining flight distance is not recalculated. Instead, the current position, current effective speed, and previous correction arrival time are used to perform a fine-tuning update. Specifically, if the difference between the current effective speed and the effective speed used in the previous calculation does not exceed 60 kilometers per hour, the previous correction arrival time remains unchanged. If the difference between the current effective speed and the effective speed used in the previous calculation exceeds 60 kilometers per hour, the previous correction arrival time is adjusted forward or backward by 2 minutes according to the new speed difference. Here, 60 kilometers per hour is used as the speed fine-tuning threshold, and 2 minutes is used as the single update time amplitude threshold in the non-offset state.

[0105] In this embodiment, after forming the corrected time series, the system merges flight dynamic data with historical flight statistics to obtain the final estimated arrival time. Specifically, the system first extracts the current flight status, whether descent has begun, whether it has entered the terminal area, whether it has landed, and the latest estimated arrival time from the flight dynamic data; then, it extracts the average approach time, average terminal area delay compensation value, and average taxiing time before landing under the same airport, same time period, and same route conditions from the historical flight statistics; subsequently, it merges the latest corrected arrival time in the corrected time series with the aforementioned dynamic and historical statistical fields. As a preferred embodiment, when the flight dynamic data indicates that the flight has entered the descent phase but has not yet entered the terminal area, the average terminal area delay compensation value is added to the latest corrected arrival time; when the flight dynamic data indicates that the flight has entered the terminal area, the average approach time is fixedly added to the latest corrected arrival time; when the flight dynamic data indicates that the flight has landed, the actual arrival time is directly used as the final estimated arrival time.

[0106] In this embodiment, it should also be noted that a reasonableness check is performed on the result before obtaining the final estimated arrival time. Specifically, if the final estimated arrival time is earlier than the current monitoring time, the result is directly deemed invalid, and the previous valid time in the corrected time series is used instead; if the final estimated arrival time is more than 360 minutes later than the planned arrival time, the result is directly limited to 360 minutes after the planned arrival time; if the difference between the final estimated arrival time and the latest estimated arrival time in the flight dynamics data exceeds 45 minutes, the latest corrected arrival time in the corrected time series is retained first, and the difference result is written to the anomaly flag field. Here, 360 minutes is used as the upper limit threshold of the final result, and 45 minutes is used as the result conflict determination threshold.

[0107] S8. The final estimated arrival time result is fused with cargo attribute information and real-time cargo hold environment parameters. First, a mapping table is established between cargo type and maximum allowable waiting time after arrival. This mapping table is stored in a fixed configuration and contains at least three records: cold chain pharmaceuticals correspond to the first time limit (e.g., 30 minutes), live animals correspond to the second time limit (e.g., 45 minutes), and dangerous goods correspond to the third time limit (e.g., 60 minutes). This mapping table can be updated quarterly or annually according to the airport's support capacity. Then, airborne sensors deployed in the cargo hold acquire real-time data on temperature, humidity, and vibration at a sampling frequency of no less than once per minute. This data is then transmitted in real-time to the ground processing system via an airborne communication system (such as ACARS) or a broadband satellite link. Upon receiving this data, the ground processing system performs spatiotemporal alignment with the final estimated arrival time calculated in the aforementioned steps to form a time series of cargo hold environmental parameters. Spatiotemporal alignment refers to binding each frame of environmental parameters with the aircraft's current position (latitude, longitude, and altitude) and timestamp. Next, based on the cargo hold environmental parameter time series and the remaining flight time, the cargo hold environmental parameter values ​​at the arrival time are predicted. The prediction method is as follows: the cargo hold environmental parameter time series is fused with the statistical patterns of cargo hold environmental changes on the same route and during the same period of historical flights. The trend extrapolation method is used to calculate the slope of change of each parameter in the remaining flight distance, thereby obtaining the predicted value at the arrival time. For example, by analyzing historical data to obtain the average temperature rise rate of the route in similar time periods, and performing weighted fusion with the currently monitored temperature rise rate (historical weight 0.3, real-time weight 0.7), the temperature at arrival is predicted. Then, it is determined whether the predicted value exceeds the preset state threshold corresponding to the cargo type. For cold chain pharmaceuticals, the threshold is the temperature range boundary of 2-8°C; for live animals, the threshold is the preset temperature and humidity comfort range boundary (e.g., temperature 10-25°C, relative humidity 40%-70%). If the cargo hold temperature at the predicted arrival time is expected to exceed the permissible range, a Level 1 warning is triggered. This warning is sent to the flight crew's multifunction control display unit, which automatically suggests that the crew adjust the cargo hold temperature control setpoint (e.g., lowering the setpoint from 5°C to 3°C). If the system further predicts that even with onboard temperature control measures, the cargo hold temperature cannot be kept within the threshold for the remaining flight distance (e.g., based on historical data of the maximum adjustment rate of the aircraft's cargo hold temperature control system), a Level 2 warning is triggered. This warning is sent to the ground support control center at the destination airport, containing the cargo type, estimated arrival time, and urgency level, and automatically generates a priority unloading request instruction. Finally, based on the cargo type and the final estimated arrival time, a ground support resource pre-schedule plan is automatically generated. This plan is a dynamic instruction set: when the cargo is cold chain medicine, a cold chain container pre-positioning instruction is sent to the cold storage management system, specifying that the container be towed to the designated aircraft position 15 minutes before the estimated arrival time; when the cargo is live animals, a quarantine personnel pre-arrival instruction is sent to the quarantine department terminal, setting the personnel's arrival time to 10 minutes before the estimated arrival time; when the cargo is dangerous goods, a dangerous goods isolation area pre-clearing instruction is sent to the airport operations command center, ensuring that the clearing is completed 15 minutes before the estimated arrival time. In this way, granularity from flight-level forecasting to cargo hold-level support is achieved.

[0108] The present invention also discloses a machine-executable program that can be automatically executed by a machine to realize the dynamic calculation method for flight arrival time based on multi-source data fusion as described above.

[0109] The machine (electronic device) mentioned above is, for example, a microcontroller, a single-board computer, a desktop computer, a laptop computer, a server, a programmable controller, or a field-programmable gate array.

[0110] Figure 2 This is a structural block diagram of the local terminal of an exemplary electronic device (machine) of the present invention; as shown... Figure 2 As shown, the electronic device of the present invention includes a processor 31, a memory 32 and a storage space 33 for storing a machine-executable program 34, the machine-executable program 34 being used to execute the above-described control logic.

[0111] Figure 3 This is a structural block diagram of the network end of an exemplary electronic device of the present invention; as shown below. Figure 3 As shown, the present invention also provides an electronic device (machine), which may include at least one processor 410, at least one memory 430 communicatively connected to the processor, and a communication bus 440 and a communication interface 220 connecting different system components (including the memory 430 and the processor 410); the processor 410, the memory 430, and the communication interface 420 are connected through the communication bus 440 and communicate with each other; the communication interface 420 is used for data interaction with external devices. The memory 430 stores a machine-executable program that can be executed by the processor, and the processor 410 can execute the aforementioned control logic by calling the machine-executable program.

[0112] Communication bus 440 represents one or more of several bus architectures, including a memory bus or memory controller, peripheral bus, graphics acceleration port, processor, or local bus using any of the various bus architectures. Examples of these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnection (PCI) bus.

[0113] Electronic devices typically include a variety of computer system readable media, which can be any available media that can be accessed by the electronic device, including volatile and non-volatile media, and removable and non-removable media.

[0114] Memory 430 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The electronic device may further include other removable / non-removable, volatile / non-volatile computer system storage media. Memory 430 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the control logic described above.

[0115] A program / utility having a set (at least one) of program modules can be stored in memory 430. Such program modules include, but are not limited to, an operating system, one or more applications, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment.

[0116] Machine-executable programs for performing this invention can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages ​​such as Java, C++, and Python, and may also include specialized engineering languages ​​such as R. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0117] The present invention also discloses a storage medium on which a machine-executable program as described above is stored.

[0118] The aforementioned storage medium may be any combination of one or more computer-readable media. Computer-readable media may be, for example, computer-readable signal media or computer-readable storage media. Computer-readable storage media include, but are not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or flash memory, optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium may be, for example, any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0119] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0120] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0121] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for dynamically calculating flight arrival times based on multi-source data fusion, characterized in that, include: S1. By integrating flight dynamic data and historical flight statistics, flight execution records within a specified time window range are obtained. Data segmentation is performed based on the time window division, and return diversion flags and status field parsing results are extracted to determine whether the flight is a normal execution flight. If the return diversion flag is identified, the corresponding record is removed to obtain a pre-cleaned flight sample set. S2. Based on the pre-cleaned flight sample set, the deviation value calculation logic is used to calculate the difference between the actual and planned time for each record. Combining the deviation threshold standard and the abnormal flight judgment rule, if the difference exceeds the preset upper limit standard, it is marked as abnormal and removed to determine the effective flight sample set after cleaning. S3. For the cleaned set of valid flight samples, sort them in ascending order according to the size of the deviation based on the data sorting logic. Combine the extreme value removal ratio to remove records of the upper and lower extreme values. The data cleaning process ensures the quality of the samples and obtains the selected core flight sample set. S4. Based on the selected core flight sample set, the deviation values ​​of the retained records are averaged using the mean calculation method. Combined with the historical data integrity assessment, if the data volume is insufficient, the correction data is borrowed from similar routes by classifying the flight types to determine the final historical deviation correction value.

2. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 1, characterized in that: S1 includes: Obtain flight dynamic data and historical flight statistics, and extract the corresponding flight execution records; Flight segment records are obtained by segmenting flight execution records. Parse the flight segment records to obtain the status field parsing results. If a return-to-origin / diversion flag is identified from the status field parsing results, the flight segment record is determined to be an abnormal flight record. Abnormal flight records are removed and normally operating flights are retained. A preliminary cleaned flight sample set is obtained based on the normally operating flights.

3. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 1, characterized in that: S2 includes: The time deviation numerical sequence is obtained by performing interpolation calculations on the pre-cleaned flight sample set. Determine whether the time deviation numerical sequence exceeds the preset deviation threshold standard; If the time deviation numerical sequence exceeds the preset deviation threshold standard, it is determined to be an abnormal deviation record; Records that need to be removed are filtered by associating abnormal deviation records with historical weather data. Records that need to be removed are discarded, and the remaining records in the flight sample set are integrated to determine the effective flight sample set after cleaning.

4. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 1, characterized in that: S3 includes: Obtain the cleaned set of valid flight samples, sort them in ascending order according to the deviation magnitude attribute of the valid flight sample set to determine the deviation sequence, and identify the upper and lower extreme value records in the deviation sequence to obtain the preliminary screening sequence. The boundary-adjusted sequence is obtained by removing records of extreme values ​​from the initial screening sequence. If the boundary adjustment sequence meets the preset threshold, the sample quality improvement sequence is obtained through the cleaning process. If the improved sample quality sequence meets the ascending order requirement, then the selected core flight sample set is determined.

5. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 1, characterized in that: S4 includes: The initial average deviation value is obtained by averaging the flight deviation values. The historical data volume assessment result is obtained by evaluating historical data associated with the initial average deviation value; Determine whether the historical data volume assessment result is lower than a preset threshold; If the historical data volume assessment result is lower than the preset threshold, the borrowed correction data is obtained by matching the characteristics of the flights corresponding to the initial average deviation value. The initial average deviation value is compensated by borrowing corrected data to determine the final historical deviation correction value.

6. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 1, characterized in that, This also includes S5, which extracts route information from flight plan data, classifies the flight sample set according to route based on the sample grouping method, applies the final historical deviation correction value to each route, and supplements missing fields based on record integrity checks if a certain route record is insufficient, thereby obtaining the deviation adjustment parameters corresponding to the route. Specifically, this includes: By extracting segment node sequences from flight plan data to construct route orientation features, the core flight sample set is grouped using route orientation features to obtain the target route orientation classification set. Obtain historical deviation correction values ​​for the target route orientation classification set, and use the historical deviation correction values ​​to perform deviation mapping processing on the target route orientation classification set to obtain the initial deviation adjustment sequence; Determine whether the record integrity assessment value of the initial deviation adjustment sequence is lower than the preset integrity threshold; If the record completeness assessment value is lower than the preset completeness threshold, the missing field positions in the initial deviation adjustment sequence are extracted, and the missing field positions are interpolated and supplemented based on the historical deviation correction values ​​of adjacent flight segment nodes to determine the deviation adjustment parameters corresponding to the flight path.

7. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 6, characterized in that, This also includes S6, which integrates flight plan data, flight dynamics data, and historical flight statistics, adjusts parameters based on the corresponding deviations of the flight path, and applies these adjustment parameters to different flight phases in conjunction with takeoff time information from the flight dynamics data. If the current phase is before takeoff, the estimated takeoff time is used as the basis for time estimation to determine a preliminary estimated arrival time. Specifically, this includes: The initial dataset for flight plan data fusion is obtained, and the initial dataset is calibrated according to the deviation of the flight path to obtain the integrated sequence; If the integrated sequence is in the pre-takeoff state, the offset is calculated based on the expected takeoff base to determine the time series.

8. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 7, characterized in that: S6 further includes: The flight phase differentiation features are extracted based on the time series data, and then fused with historical statistical data to obtain a preliminary estimated sequence. Details are extracted from the preliminary estimated sequence, and these details are processed in conjunction with the takeoff time to determine the estimated arrival time.

9. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 7, characterized in that, It also includes S7, which fuses ADS-B surveillance data with preliminary estimated arrival time, extracts real-time position and speed information during flight, dynamically corrects the preliminary estimated arrival time, adjusts parameters based on route classification results and deviations, recalculates the remaining flight distance if the position deviation exceeds a preset range, and obtains the final estimated arrival time based on the fused flight plan data, flight dynamic data, ADS-B surveillance data, and historical flight statistics. Specifically, it includes: Position and velocity information are extracted from broadcast automatic dependent surveillance data to generate a real-time flight status sequence; The position offset value is determined based on the real-time flight status sequence; If the position offset value exceeds the preset range, the remaining flight distance is calculated based on the position offset value and speed information.

10. The method for dynamically calculating flight arrival time based on multi-source data fusion according to claim 9, characterized in that: The S7 also includes: The preliminary estimated arrival time is dynamically corrected by the remaining flight distance to obtain the corrected time series; The final estimated arrival time is obtained by fusing flight dynamic data with historical flight statistics using corrected time series data.

Images