A data-driven ebt training scenario generation method and system

Abstract

The present application relates to the technical field of data processing, and provides a data-driven EBT training scene generation method and system, wherein flight control data flow, physiological feedback data flow and voice interaction data flow are obtained in real time, the physiological feedback data flow contains pupil diameter change rate; based on the physiological feedback data flow, a quantitative score of system management ability dimension is generated; when the score is lower than a preset ability threshold, the complexity level of a concurrent system failure event is determined; according to the voice interaction data flow, a collaboration defect feature set is generated, the complexity level is combined, a candidate failure event set is retrieved from a pre-constructed aviation threat knowledge graph, a global visual dynamic parameter instruction set is generated, and an EBT training scene is generated based on the global visual dynamic parameter instruction set; the present application solves the problems of insufficient authenticity and effectiveness of training, and improves the individualization degree and real combat fidelity of the EBT training scene.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a data-driven EBT training scenario generation method and system. Background Art

[0002] In modern flight training, developing pilots' decision-making skills for complex and volatile mid-air emergencies, particularly those facing multiple concurrent system failures, has become a crucial issue in enhancing aviation safety. In particular, in high-stress, multi-task environments, dynamically constructing realistic training scenarios to assess and enhance pilots' systems management and collaborative response capabilities has become a key technical requirement within current EBT (Evidence-Based Training) systems.

[0003] To address these technical needs, current research has attempted to combine multi-source sensor data with knowledge graph technology to construct a training scenario generation mechanism based on behavioral pattern recognition. This mechanism collects operational logs, voice communication records, and some physiological signals from flight simulators, uses deep learning models to extract pilot behavioral characteristics in different scenarios, and then performs matching reasoning based on a predefined aviation fault event library to automatically generate corresponding fault injection sequences. However, existing solutions still have significant limitations in practical applications. For example, it is difficult to accurately reflect the load changes of pilots in high-pressure environments, resulting in a lack of effective matching between the complexity of the generated fault scenarios and the individual's tolerance. There are also deficiencies in multimodal data fusion and intent understanding, resulting in unstable logical coherence and threat priority of the generated concurrent fault event set, which affects the authenticity and effectiveness of training. Summary of the Invention

[0004] The present invention provides a data-driven EBT training scenario generation method and system to address the problems in the prior art such as the lack of effective matching between the complexity of fault scenarios and individual tolerance, as well as the insufficient authenticity and effectiveness of training, thereby improving the personalization and actual combat fidelity of EBT training scenarios.

[0005] In a first aspect, the present invention provides a data-driven EBT training scenario generation method, comprising:

[0006] acquiring, in real time, a flight control data stream, a physiological feedback data stream, and a voice interaction data stream, wherein the flight control data stream includes a flight control integrated deviation parameter, the flight control integrated deviation parameter including an aircraft attitude expected deviation parameter, a track-to-expected deviation parameter, a standard operating procedure execution deviation parameter, and a control voice instruction deviation parameter, and the physiological feedback data stream includes a pupil diameter change rate;

[0007] quantifying a first cognitive load state based on the physiological feedback data stream;

[0008] generating a quantitative score of a system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream;

[0009] When the quantitative score of the system management capability dimension is lower than a preset capability threshold, determining a complexity level of the concurrent system failure event according to the first cognitive load state;

[0010] generating a collaborative defect feature set based on the voice interaction data stream, and retrieving a candidate fault event set from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set;

[0011] The concurrent system fault event set is converted into a global vision dynamic parameter instruction set matching the complexity level, so as to generate an EBT training scenario based on the global vision dynamic parameter instruction set.

[0012] Optionally, generating a quantitative score of a system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream includes:

[0013] Correcting the flight control comprehensive deviation parameter according to the first cognitive load state to obtain a corrected comprehensive deviation parameter;

[0014] parsing the voice interaction data stream to obtain voice response interval and semantic integrity information, and generating a crew collaboration behavior feature value based on the voice response interval and semantic integrity information;

[0015] Calculating a system operation stability index based on the fluctuation range and fluctuation duration of the corrected comprehensive deviation parameter;

[0016] The system operation stability index, the first cognitive load state, and the crew collaborative behavior characteristic value are integrated to generate a quantitative score of the system management capability dimension.

[0017] Optionally, when the quantitative score of the system management capability dimension is lower than a preset capability threshold, determining the complexity level of the concurrent system failure event according to the first cognitive load state includes:

[0018] Dividing the second cognitive load state of the historical training scenario into a plurality of load intensity intervals, and matching the first cognitive load state with the load intensity intervals to generate a basic load level;

[0019] Querying a predefined fault event complexity benchmark library to obtain the maximum number of concurrent faults and the upper limit of the physical disturbance amplitude allowed by the basic load level, so as to generate an initial fault event complexity calibration value;

[0020] Calculating the instantaneous load change rate and the change direction consistency coefficient based on the change trajectory of the first cognitive load state;

[0021] According to the instantaneous load change rate, the quantity permission weight of the initial fault event complexity calibration value is adjusted, and at the same time, according to the change direction consistency coefficient, the physical disturbance amplitude permission weight is adjusted to generate the complexity level of the concurrent system fault event.

[0022] Optionally, adjusting the quantity allowable weight of the initial fault event complexity calibration value according to the instantaneous load change rate, and adjusting the physical disturbance amplitude allowable weight according to the change direction consistency coefficient, so as to generate the complexity level of the concurrent system fault event, including:

[0023] When the instantaneous load change rate exceeds a high change rate critical value, the quantity permission weight of the initial fault event complexity calibration value is reduced according to a first reduction ratio; when the instantaneous load change rate is lower than a low change rate critical value, the quantity permission weight is increased according to a first gain ratio to obtain an adjusted quantity permission weight;

[0024] Retrieving a physical disturbance amplitude scaling factor corresponding to the change direction consistency coefficient from a pre-built directional stability rule library, and updating the physical disturbance amplitude permissible weight of the initial fault event complexity calibration value according to the physical disturbance amplitude scaling factor to obtain an adjusted physical disturbance amplitude permissible weight;

[0025] Performing a multiplication operation on the adjusted quantity permission weight and the physical disturbance amplitude permission weight to generate a comprehensive complexity scalar;

[0026] The complexity level of the concurrent system fault event corresponding to the comprehensive complexity scalar is matched in a preset complexity level conversion rule library.

[0027] Optionally, a collaborative defect feature set is generated based on the voice interaction data stream, and a candidate fault event set is retrieved from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set, including:

[0028] identifying, based on the voice interaction data stream, command response delay frequencies and key semantic missing markers to generate a collaboration defect feature set;

[0029] According to the collaborative defect feature set, a predefined defect pattern library is searched to determine the defect type code;

[0030] Calculating the physical disturbance constraint boundary based on the physical disturbance amplitude allowable weight corresponding to the complexity level and the dynamic parameter safety threshold of the flight phase;

[0031] Retrieving a set of candidate fault events that are causally related to the defect type code from a pre-built aviation threat knowledge graph;

[0032] Removing candidate fault events that exceed the physical disturbance constraint boundary from the candidate fault event set to obtain multiple target fault events;

[0033] The functional coupling strength values ​​between different target fault events are calculated, and combined with the repair priority weight distribution table corresponding to the defect type code, a concurrent system fault event set is generated.

[0034] In a second aspect, the present invention provides a data-driven EBT training scenario generation system, comprising:

[0035] an acquisition module, configured to acquire, in real time, a flight control data stream, a physiological feedback data stream, and a voice interaction data stream, wherein the flight control data stream includes a flight control integrated deviation parameter, the flight control integrated deviation parameter including an aircraft attitude expected deviation parameter, a track-to-expected deviation parameter, a standard operating procedure execution deviation parameter, and a control voice command deviation parameter, and the physiological feedback data stream includes a pupil diameter change rate;

[0036] a quantification module, configured to quantify a first cognitive load state based on the physiological feedback data stream;

[0037] a generating module, configured to generate a quantitative score of a system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream;

[0038] a determination module, configured to determine a complexity level of a concurrent system failure event according to the first cognitive load state when the quantitative score of the system management capability dimension is lower than a preset capability threshold;

[0039] a retrieval module, configured to generate a collaborative defect feature set based on the voice interaction data stream, and retrieve a candidate fault event set from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set;

[0040] A conversion module is used to convert the concurrent system fault event set into a global vision dynamic parameter instruction set matching the complexity level, so as to generate an EBT training scenario based on the global vision dynamic parameter instruction set.

[0041] In a third aspect, the present invention provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a data-driven EBT training scenario generation method as described in the first aspect above.

[0042] In a fourth aspect, the present invention provides a computer storage medium storing a computer program, which, when executed by a computer, implements a data-driven EBT training scenario generation method as described in the first aspect.

[0043] In the present invention, a flight control data stream, a physiological feedback data stream, and a voice interaction data stream are acquired in real time. The flight control data stream includes flight control comprehensive deviation parameters, including aircraft attitude expected deviation parameters, track-to-expected deviation parameters, standard operating procedure execution deviation parameters, and control voice command deviation parameters. The physiological feedback data stream includes pupil diameter change rate. A first cognitive load state is quantified based on the physiological feedback data stream. A quantitative score of a system management capability dimension is generated based on the first cognitive load state, the flight control comprehensive deviation parameters, and the voice interaction data stream. When the quantitative score of the system management capability dimension is lower than a preset capability threshold, a complexity level of concurrent system fault events is determined based on the first cognitive load state. A collaborative defect feature set is generated based on the voice interaction data stream. In combination with the complexity level, a candidate fault event set is retrieved from a pre-constructed aviation threat knowledge graph to generate a concurrent system fault event set. The concurrent system fault event set is converted into a global visual dynamic parameter instruction set matching the complexity level, so as to generate an EBT training scenario based on the global visual dynamic parameter instruction set. The technical solution provided by the present invention overcomes the limitation of the existing technology of missing biometric data dimensions. It quantifies cognitive load in real time through the rate of change of pupil diameter, and realizes in-depth analysis of voice interaction by combining voice interaction data streams, thus building a full-dimensional data foundation covering operation, physiology, and collaboration. It overcomes the shortcomings of the rough cognitive state assessment of traditional methods, converts physiological feedback data into an actionable cognitive load indicator, and provides a precise physiological basis for dynamic difficulty control. It solves the problem of fuzzy identification of capability gaps in existing technologies. By integrating cognitive load, flight errors and voice behavior, it constructs a multi-dimensional capability profile and accurately identifies weak points in system management capabilities. It overcomes the shortcomings of static scene adaptation and triggers complexity generation only when the capability score falls below a threshold, ensuring that the complexity of fault events strictly matches the pilot's real-time cognitive state and avoiding mechanical difficulty injection. It overcomes the pain point of logical breakage in threat injection. Based on the characteristics of collaborative defects and complexity levels, it retrieves a set of candidate fault events with causal associations from the aviation knowledge graph, ensuring the physical rationality of the event chain and the targeted training. It also solves the problem of insufficient scene realism. By driving digital twin rendering through a global visual parameter instruction set, it achieves accurate mapping of fault events to three-dimensional scenes, enhancing training immersion. Furthermore, based on the real-time cognitive load status, the basic load level is generated by matching the historical load intensity interval; the complexity benchmark library is queried to obtain the initial fault constraints; the instantaneous change rate and directional consistency are calculated based on the cognitive load change trajectory; and the number of faults and the weights of the physical disturbance amplitude are dynamically adjusted accordingly to generate a complexity level that strictly adapts to the pilot's cognitive fluctuations.The complexity of faults is controlled in real time through the trajectory of cognitive load changes to prevent overload or undertraining in high-pressure environments; the quantity permission weight constrains the upper limit of concurrent faults to avoid multi-task decision-making collapse; the disturbance amplitude weight limits the physical parameters from exceeding the limit to ensure the safety of special situation simulation.

[0044] These and other aspects of the present invention will become more readily apparent from the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 A flow chart of a data-driven EBT training scenario generation method provided by the present invention is shown;

[0047] Figure 2 A schematic diagram of the structure of a data-driven EBT training scenario generation system provided by the present invention is shown;

[0048] Figure 3 A schematic structural diagram of a computing device provided by the present invention is shown. DETAILED DESCRIPTION

[0049] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0050] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence, nor do they limit "first" and "second" to be different types.

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0052] While existing technologies have introduced automated scenario generation mechanisms driven by multi-source data fusion and knowledge graphs to address the need to cultivate multi-task decision-making capabilities in emergency situations within general flight decision-making training scenarios, they still have significant shortcomings in dynamically matching pilots' real-time states with the complexity of faults. Traditional methods often rely on coarse-grained behavioral pattern recognition and pre-set rule reasoning, making it difficult to accurately quantify the changing trends in pilots' cognitive load and the characteristics of their intentional expression in collaborative communication. This results in a disconnect between the generated concurrent system fault event sets and the individual's actual response capabilities, compromising the authenticity and challenge of training outcomes. To address these issues, the present invention proposes a data-driven EBT training scenario generation method. The core of the method lies in constructing a dynamic evaluation model for pilots' system management capabilities by collecting multiple data streams such as flight control, physiological feedback, and voice interaction in a timely manner. On this basis, the cognitive load status and the semantic analysis of the crew's collaborative intention are combined to intelligently match the complexity level of concurrent faults that adapt to their capability level. Relying on the aviation threat knowledge graph, a logically coherent and prioritized set of fault events is generated, which are finally converted into dynamic parameter instructions executable by the visual simulation system. This realizes the construction of intelligent training scenarios based on individual performance, real-time feedback, semantic understanding, and knowledge guidance in a true sense, thereby significantly improving pilots' decision-making and collaborative combat response capabilities under multiple pressure environments. Figure 1 A flowchart of a data-driven EBT training scenario generation method is provided for an embodiment of the present invention, such as Figure 1 As shown, the method includes:

[0053] Step 101: Acquire flight control data streams, physiological feedback data streams, and voice interaction data streams in real time, wherein the flight control data streams include flight control integrated deviation parameters, including aircraft attitude expected deviation parameters, track-to-expected deviation parameters, standard operating procedure execution deviation parameters, and control voice command deviation parameters, and the physiological feedback data streams include pupil diameter change rate;

[0054] In this step, the flight control data stream refers to the time-series data stream collected in real time by flight control sensors, including the flight control integrated deviation parameter, which reflects the pilot's control accuracy of flight attitude. The physiological feedback data stream refers to the time-series data of physiological indicators collected by biosensors, including the pupil diameter change rate (the change in pupil diameter per unit time divided by the baseline diameter), which is used to quantify the pilot's neurocognitive load. The voice interaction data stream refers to the real-time audio stream of cabin voice communication, which is used to identify crew collaborative behavior patterns. The flight control integrated deviation parameter is a composite parameter that integrates four types of deviations: aircraft attitude, track, standard operating procedure execution, and control command consistency. It reflects the degree of deviation of all dimensions of flight control from the expected target. The aircraft attitude expected deviation parameter refers to the absolute angular deviation between the actual flight attitude and the target attitude, including pitch angle deviation and roll angle deviation, reflecting flight attitude control accuracy. The track-to-expected deviation parameter refers to the spatial deviation between the actual flight trajectory and the planned trajectory, including altitude deviation and heading deviation, reflecting navigation accuracy. The SOP execution deviation parameter refers to the degree of alignment between pilot operating steps and SOP specifications, reflecting the completeness of procedural execution. The SOP execution deviation parameter is calculated as the number of omitted steps divided by the total number of SOP steps. The control voice command deviation parameter refers to the temporal consistency between control actions and voice commands, reflecting the efficiency of crew resource management collaboration. This parameter is generated through timestamp matching. For example, a deviation of 0 indicates that the action was executed within 3 seconds of the command, while a deviation of 1 indicates a timeout or non-execution. The pupil diameter change rate refers to the maximum change in pupil diameter per unit time (1 second) as a percentage of the initial diameter, serving as a physiological marker of cognitive load.

[0055] In this embodiment of the present invention, the flight control data stream collects control surface sensor data (including control surface deflection angle and throttle position), inertial navigation system attitude data, global positioning system track data, and operational event logs in real time through the flight control computer, and generates comprehensive flight control deviation parameters. Specifically, these include calculating the expected aircraft attitude deviation parameter, which calculates the absolute difference between the actual pitch / roll angle and the target angle; calculating the difference (in meters) between the actual altitude and the target altitude, and the angular deviation (in degrees) between the actual heading and the planned heading; comparing operational event timestamps with the standard operating procedure timing to detect missed steps (e.g., if the speed knob is not pulled out within 5 seconds of a fault, this is marked as 1); and using time window matching, generating a voice command deviation parameter, which is marked as 1 if a control action (e.g., flap retraction) is not executed within 3 seconds of a voice command. The physiological feedback data stream captures raw pupil diameter data at a 100Hz sampling rate using a head-mounted eye tracker, which is filtered through a sliding window and outputs the pupil diameter change rate (%). The voice interaction data stream is collected by the cabin microphone array. The above three types of data streams are synchronized and aligned through hardware timestamps, and the alignment accuracy is controlled within The data is then encapsulated into a data packet in a unified format and sent to the distributed message middleware for transmission, providing a data basis for subsequent processing.

[0056] Step 102: quantifying a first cognitive load state based on the physiological feedback data stream;

[0057] In this step, the first cognitive load state refers to a continuous value of 0-1.0 calculated based on the pupil diameter change rate, which is obtained by the formula, first cognitive load state = (current change rate - resting change rate) / (stress change rate threshold - resting change rate), which is used to dynamically characterize the working memory load level.

[0058] In this embodiment of the present invention, a cognitive load calculation model based on pupil diameter change rate is used to perform sliding window peak detection on the physiological feedback data stream. The pupil diameter change rate is converted into a continuous load value ranging from 0 to 1.0 to generate the first cognitive load state. The specific process is: the maximum pupil diameter change rate within each 5-second window is divided by the historical baseline change rate, and then multiplied by the load factor of 0.8.

[0059] Step 103: Generate a quantitative score of the system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream;

[0060] In this embodiment of the present invention, the first cognitive load state is used to correct the flight control comprehensive deviation parameter to generate a corrected comprehensive deviation parameter. Secondly, the voice interaction data stream is parsed to obtain the crew collaborative behavior characteristic value. Then, based on the fluctuation range and fluctuation duration of the corrected comprehensive deviation parameter, the system operation stability index is calculated. Finally, the system operation stability index (weight 0.6), the first cognitive load state (weight 0.2), and the crew collaborative behavior characteristic value (weight 0.2) are input into the fully connected layer, and a quantitative score of 0-100 is output.

[0061] Step 104: When the quantitative score of the system management capability dimension is lower than a preset capability threshold, determining the complexity level of the concurrent system failure event according to the first cognitive load state;

[0062] In an embodiment of the present invention, the second cognitive load state of the historical training scenario is divided into multiple load intensity intervals and matched with the first cognitive load state to generate a basic load level; a predefined fault event complexity benchmark library is queried to obtain the maximum number of parallel faults and the upper limit of the physical disturbance amplitude allowed by the basic load level to generate an initial fault event complexity calibration value; based on the change trajectory of the first cognitive load state, the instantaneous load change rate and the change direction consistency coefficient are calculated to adjust the quantity permission weight and the physical disturbance amplitude permission weight of the initial fault event complexity calibration value accordingly to generate the complexity level of the concurrent system fault event.

[0063] Step 105: Generate a collaborative defect feature set based on the voice interaction data stream, and retrieve a candidate fault event set from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set;

[0064] In an embodiment of the present invention, the command response delay frequency and key semantic missing markers are identified based on the voice interaction data stream, a collaborative defect feature set is generated, and a predefined defect pattern library is queried to determine the defect type code; the physical disturbance constraint boundary is calculated based on the physical disturbance amplitude permission weight of the complexity level and the dynamic parameter safety threshold of the flight phase; a set of candidate fault events that are causally related to the defect type code is retrieved from a pre-constructed aviation threat knowledge graph, and the candidate fault events that exceed the physical disturbance constraint boundary are removed to obtain multiple target fault events; the functional coupling strength values ​​between different target fault events are calculated, and a concurrent system fault event set is generated based on a repair priority weight distribution table corresponding to the defect type code.

[0065] Step 106: converting the concurrent system fault event set into a global visual dynamic parameter instruction set matching the complexity level, so as to generate an EBT training scenario based on the global visual dynamic parameter instruction set;

[0066] In an embodiment of the present invention, an event type identifier of each concurrent system fault event in a concurrent system fault event set is parsed, and a predefined three-dimensional model rule library is retrieved to obtain a basic model identifier and a standard action script; the standard action script is adjusted according to the permitted weight of the physical disturbance amplitude of the complexity level to generate a disturbance enhancement action script; based on the permitted weight of the quantity of the complexity level, the trigger timing delay time between different concurrent system fault events is calculated; the basic model identifier, the disturbance enhancement action script and the trigger timing delay time are combined to generate an event-level three-dimensional dynamic parameter unit corresponding to the concurrent system fault event, and all parameter units are aggregated to generate a global visual dynamic parameter instruction set, based on which an EBT training scene is generated using a real-time rendering pipeline driver of a flight simulation device.

[0067] The embodiments of the present invention solve the three major defects of traditional EBT training, namely, static scenarios, single data dimension, and mechanical threat injection, and realize the precise, personalized, and adaptive generation of aviation special situation scenarios.

[0068] The present invention provides a specific embodiment, in which step 103 generates a quantitative score of the system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream, and specifically includes the following steps:

[0069] Step 301: Correcting the flight control comprehensive deviation parameter according to the first cognitive load state to obtain a corrected comprehensive deviation parameter;

[0070] In this step, the corrected error rate refers to the pitch angle control error rate after dynamic adjustment based on the first cognitive load state, and is used to eliminate the manipulation error amplification effect caused by high cognitive load.

[0071] In an embodiment of the present invention, a dynamic correction coefficient for cognitive load is used to correct the comprehensive flight control deviation parameter. The corrected comprehensive deviation parameter = dynamic correction coefficient for cognitive load = 1 + (first cognitive load state × 0.2). For example, when the first cognitive load state is 0.8, the dynamic correction coefficient for cognitive load = 1.16, and the original comprehensive flight control deviation parameter of 37.24 is corrected to 43.20 (i.e., 37.24 × 1.16).

[0072] Step 302: Parse the voice interaction data stream to obtain voice response interval and semantic integrity information, and generate a crew collaboration behavior feature value based on the voice response interval and semantic integrity information;

[0073] In this step, the voice response interval refers to the millisecond time difference between the end of the flight instruction announcement and the start of the pilot's voice response. This information is calculated using an audio endpoint detection algorithm and reflects the level of crew decision-making delay. Semantic integrity information refers to the binary mark (0 / 1) generated by keyword matching of the pilot's response statement. This information is used to detect the omission of necessary operational items based on a predefined vocabulary of key instructions (e.g., oil pressure check, flap setting). The crew collaborative behavior characteristic value is a weighted value that combines the standardized delay index of the voice response interval with the number of semantic integrity omissions. This value is used to quantify the degree of customer relationship management collaboration defects.

[0074] In this embodiment of the present invention, a voice endpoint detection algorithm is used to segment command segments within a voice interaction data stream. The time difference between the end of the command and the start of the response is calculated to obtain the voice response interval. Simultaneously, a semantic parsing model is used to detect whether key command words (such as flaps stuck) appear in the response. If missing, the semantic integrity information is marked as 0; otherwise, it is marked as 1. The average voice response interval of five consecutive commands is calculated and divided by the historical baseline response time to generate a normalized delay index. The number of commands with a semantic integrity of 0 is accumulated to generate a semantic missing count. The normalized delay index (weighted 0.6) is linearly combined with the semantic missing count (weighted 0.4) to generate a crew collaborative behavior characteristic value ranging from 0 to 10.

[0075] Step 303: Calculating a system operation stability index based on the fluctuation range and fluctuation duration of the corrected comprehensive deviation parameter;

[0076] In this step, the system operation stability index refers to the integral value of the fluctuation intensity of the corrected error rate within the time window, reflecting the stability of flight attitude control.

[0077] In the embodiment of the present invention, within the 30-second time window, the standard deviation of the corrected comprehensive deviation parameter is calculated as the fluctuation range, and the statistical correction of the comprehensive deviation parameter exceeds The duration of the 1σ threshold is used as the fluctuation duration. According to the following formula: System Operation Stability Index = (Fluctuation Range × Exceeding Threshold Fluctuation Duration) ÷ Total Window Duration, a system operation stability index ranging from 0 to 1.0 is generated.

[0078] Step 304: Fusing the system operation stability index, the first cognitive load state, and the crew collaborative behavior characteristic value to generate a quantitative score of the system management capability dimension;

[0079] In this step, the quantitative score of the system management capability dimension refers to a percentage score that is a weighted fusion of stability indicators, cognitive load state inverse values, and collaborative behavior characteristics, which is used to comprehensively evaluate the shortcomings of system management capabilities.

[0080] In the embodiment of the present invention, a weighted fusion formula is adopted: quantitative score = (system operation stability index × 0.5) + (1-first cognitive load state) × 0.3 + (crew collaborative behavior characteristic value / 10) × 0.2, and the calculation result is multiplied by 100 to convert it into a score of 0-100.

[0081] The embodiment of the present invention compensates for cognitive load interference through the corrected error rate, thereby improving the authenticity of the manipulation accuracy assessment; the dual factors of voice response interval and semantic integrity are used to generate collaborative behavior characteristics to accurately identify customer relationship management defects; the fluctuation range and duration are integrated to calculate the stability index to capture the risk of transient loss of control.

[0082] For example, when the pilot encountered a flap failure during a simulated flight, the system detected that the first cognitive load state was 0.9 (high load), and the original aircraft attitude expected deviation parameter was 0.8 (standardized value). The system made corrections based on the dynamic correction coefficient of cognitive load. The corrected comprehensive deviation parameter = 0.8×(1+0.9×0.2)=0.944; then the system parsed the voice interaction data stream and identified that the voice response interval for the command to check the flaps was 4.2 seconds (the baseline response interval was 2 seconds), and calculated the standardized delay index to be 4.2 / 2=2.1 seconds. At the same time, the semantic integrity information was marked as 0 (key terms were not repeated, flaps were stuck), that is, the number of instructions with semantic integrity information of 0 was accumulated to obtain a semantic missing count of 1. Based on this, the computer group collaborative behavior characteristics The eigenvalue is 2.1 × 0.6 + 1 × 0.4 = 1.66 points (out of 10 points). Within the 30-second time window, the standard deviation of the corrected comprehensive deviation parameter is 0.6, and the duration of the above-threshold fluctuation is 18 seconds. Based on this, the system operational stability index is calculated as (0.6 × 18) / 30 = 0.36. Finally, the system integrates the above three indicators according to the weights to generate a quantitative score for the system management capability dimension: (0.36 × 0.5) + (1-0.9) × 0.3 + (1.66 / 10 × 0.2) = 0.18 + 0.03 + 0.033 = 0.243, corresponding to a score of 24.3 points. Because this score is lower than the set threshold of 60 points, the system determines that the current unit's response capability is insufficient, triggering the subsequent fault scenario generation process for further evaluation or intervention.

[0083] The present invention provides a specific embodiment, in step 104, when the quantitative score of the system management capability dimension is lower than a preset capability threshold, determining the complexity level of the concurrent system failure event based on the first cognitive load state specifically includes the following steps:

[0084] Step 401: Divide the second cognitive load state of the historical training scenario into a plurality of load intensity intervals, and match the first cognitive load state with the load intensity intervals to generate a basic load level;

[0085] In this step, the second cognitive load state refers to a dataset of cognitive load states for a group of pilots in historical training scenarios. This dataset is divided into discrete intervals through cluster analysis and used to calibrate the real-time load grading benchmark. The load intensity interval refers to a continuous value range (e.g., [0, 0.3]) based on the distribution of the second cognitive load state. This interval is used to discretize the real-time continuous load values ​​into three levels: low, medium, and high. The base load level refers to the discrete label (L / M / H) output after the first cognitive load state matches the load intensity interval. This determines the initial fault constraint.

[0086] In this embodiment of the present invention, the second cognitive load status data (continuous values ​​between 0 and 1.0) of 100 pilots during historical training were collected and divided into three load intensity intervals using the K-means clustering algorithm: low load (0-0.3), medium load (0.3-0.7), and high load (0.7-1.0). The current pilot's first cognitive load status value (e.g., 0.75) was matched to the intervals, and if it fell into the high load interval, the basic load level H was generated.

[0087] Step 402: querying a predefined fault event complexity benchmark library to obtain the maximum number of concurrent faults and the upper limit of the physical disturbance amplitude allowed by the basic load level, so as to generate an initial fault event complexity calibration value;

[0088] In this step, the fault event complexity benchmark library refers to a database that stores fault constraint rules for different load levels. Its structure is {H:{maximum number of concurrent faults: 2, upper perturbation limit: 0.8}}. The maximum number of concurrent faults refers to the maximum number of system fault events that can be triggered simultaneously in a single scenario (e.g., engine failure + hydraulic leak = 2). The upper limit of the physical perturbation amplitude refers to the maximum physical parameter deviation ratio that a fault event can cause (e.g., rudder deflection angle ≤ 80% of the standard value). The initial fault event complexity calibration value is a two-tuple consisting of the maximum number of concurrent faults and the upper limit of the physical perturbation amplitude, such as (2, 0.8), which serves as the complexity adjustment benchmark.

[0089] In this embodiment, a fault event complexity benchmark database is retrieved based on the basic load level H. For example, the maximum number of concurrent faults allowed for the high load level is 2, and the upper limit of the physical disturbance amplitude is 80% of the normal value. These two parameters are combined to form the initial fault event complexity calibration value, in the format of (maximum number of faults: 2, upper limit of disturbance ratio: 0.8).

[0090] Step 403: Calculate the instantaneous load change rate and the change direction consistency coefficient based on the change trajectory of the first cognitive load state;

[0091] In this step, the instantaneous load change rate refers to the absolute change in the first cognitive load state per unit time (0.1 second), reflecting the intensity of cognitive fluctuations. The change direction consistency coefficient refers to the proportion of load changes in the same direction within consecutive time periods (monotonic interval duration / total duration), quantifying trend stability.

[0092] In this embodiment of the present invention, the time series data of the first cognitive load state for the last 30 seconds is extracted, and the instantaneous load change rate is calculated using first-order differences: instantaneous load change rate = current first cognitive load state - previous first cognitive load state / sampling interval. The proportion of segments in the change trajectory that show continuous monotonically increasing or decreasing behavior is also counted to calculate the change direction consistency coefficient. For example, if 15 seconds out of 20 seconds show a continuous increase, the change direction consistency coefficient is 15 / 20 = 0.75.

[0093] Step 404: adjusting the quantity allowable weight of the initial fault event complexity calibration value according to the instantaneous load change rate, and adjusting the physical disturbance amplitude allowable weight according to the change direction consistency coefficient, so as to generate a complexity level of the concurrent system fault event;

[0094] In this step, the permitted quantity weight refers to the actual permitted number of faults after adjustment for the instantaneous load change rate. The permitted physical disturbance amplitude weight refers to the upper limit of the actual disturbance amplitude after adjustment for the consistency coefficient. The complexity level of the concurrent system fault event refers to the final discrete level label output, which is used to control the scenario generation module.

[0095] In an embodiment of the present invention, when the instantaneous load change rate exceeds a high change rate critical value, the quantity permission weight of the initial fault event complexity calibration value is reduced according to a first reduction ratio; when the instantaneous load change rate is lower than a low change rate critical value, the quantity permission weight is increased according to a first gain ratio to obtain an adjusted quantity permission weight; a physical disturbance amplitude scaling factor corresponding to the change direction consistency coefficient is retrieved from a pre-constructed directional stability rule library to update the physical disturbance amplitude permission weight of the initial fault event complexity calibration value to obtain an adjusted physical disturbance amplitude permission weight; the two adjusted values ​​are multiplied to generate a comprehensive complexity scalar; and the complexity level of the corresponding concurrent system fault event is matched in a preset complexity level conversion rule library.

[0096] The embodiments of the present invention avoid subjective threshold setting deviations by calibrating real-time load levels based on historical experience; suppress overload risks under high fluctuation conditions through instantaneous rate of change; use directional consistency coefficients to prevent loss of control under oscillating trends; and accurately match the pilot's cognitive state with a complexity level exclusive to aviation special situation training.

[0097] The present invention provides a specific embodiment, step 404, adjusting the quantity allowable weight of the initial fault event complexity calibration value according to the instantaneous load change rate, and adjusting the physical disturbance amplitude allowable weight according to the change direction consistency coefficient to generate the complexity level of the concurrent system fault event, specifically including the following steps:

[0098] Step 411: When the instantaneous load change rate exceeds a high change rate threshold, the quantity allowance weight of the initial fault event complexity calibration value is reduced according to a first reduction ratio; when the instantaneous load change rate is lower than a low change rate threshold, the quantity allowance weight is increased according to a first gain ratio to obtain an adjusted quantity allowance weight;

[0099] In this step, the high change rate critical value refers to the cognitive load change rate threshold that triggers the compression of the number of faults. It is determined based on aviation human factors experiments and reflects the critical point of the pilot's stress state. The first reduction ratio refers to the weight scaling coefficient at high change rates, which is obtained through regression analysis of the correlation between load mutations and operational errors in historical accident statistics. The low change rate critical value refers to the lower limit of the change rate allowed to increase the number of faults, which is set according to the normal flight load fluctuation baseline. The first gain ratio refers to the weight amplification coefficient at low change rates, which is used to strengthen the training intensity under low load conditions. The adjusted quantity permission weight refers to the upper limit of the actual number of fault events after adjustment by the instantaneous load change rate.

[0100] In an embodiment of the present invention, for example, a high change rate critical value is set to 10% / second and a low change rate critical value is set to 3% / second; if the instantaneous load change rate (such as 15% / second) is greater than 10% / second, the quantity permission weight of the initial fault event complexity calibration value is multiplied by a first reduction ratio of 0.7; if the instantaneous load change rate (such as 1% / second) is less than 3% / second, it is multiplied by a first gain ratio of 1.3; for example, if the initial quantity permission weight is 2 and an instantaneous load change rate of 18% / second is detected, the adjusted quantity permission weight = 2×0.7=1.4.

[0101] Step 412: Retrieving a physical disturbance amplitude scaling factor corresponding to the change direction consistency coefficient from a pre-built directional stability rule library, and updating the physical disturbance amplitude permissible weight of the initial fault event complexity calibration value according to the physical disturbance amplitude scaling factor to obtain an adjusted physical disturbance amplitude permissible weight;

[0102] In this step, the directional stability rule base refers to a database that stores the mapping between the change direction consistency coefficient and the physical disturbance scaling factor. Its structure is {coefficient interval: scaling factor}. The physical disturbance amplitude scaling factor is a disturbance amplitude adjustment coefficient (0.6-1.2) determined based on the consistency coefficient, used to mitigate the risk of state oscillation. The adjusted physical disturbance amplitude allowable weight refers to the maximum physical disturbance amplitude after the scaling factor is updated.

[0103] In this embodiment of the present invention, based on the mapping relationship stored in the pre-built directional stability rule base: a change direction consistency coefficient <0.4 corresponds to a scaling factor of 0.6, a change direction consistency coefficient between 0.4 and 0.8 corresponds to 1.0, and a change direction consistency coefficient >0.8 corresponds to 1.2. The scaling factor 1.2 corresponding to the current change direction consistency coefficient of 0.9 is retrieved, and the initial physical disturbance amplitude allowable weight (e.g., 0.8) is multiplied by this factor to obtain the adjusted physical disturbance amplitude allowable weight, e.g., 0.8 × 1.2 = 0.96.

[0104] Step 413: performing a multiplication operation on the adjusted quantity permission weight and the physical disturbance amplitude permission weight to generate a comprehensive complexity scalar;

[0105] In this step, the comprehensive complexity scalar is the product of the quantity permission weight and the amplitude permission weight, which represents the overall strength of the scene.

[0106] In the embodiment of the present invention, the adjusted quantity permission weight (0.96) is directly multiplied by the adjusted physical disturbance amplitude permission weight (1.4×0.96=1.344), such as 1.4×0.96=1.344, to generate a comprehensive complexity scalar value of 1.344.

[0107] Step 414: matching the complexity level of the concurrent system fault event corresponding to the comprehensive complexity scalar in a preset complexity level conversion rule base;

[0108] In this step, the complexity level conversion rule base refers to a set of conversion rules that discretizes continuous scalars into aviation levels (including l1-4), which is constructed based on the airline special situation classification standards.

[0109] In this embodiment of the present invention, a preset complexity level conversion rule base is queried: a comprehensive complexity scalar of 0-0.5 corresponds to complexity level 1, a comprehensive complexity scalar of 0.5-1.0 corresponds to complexity level 2, a comprehensive complexity scalar of 1.0-1.5 corresponds to complexity level 3, and a comprehensive complexity scalar of 1.5-2.0 corresponds to complexity level 4. If the comprehensive complexity scalar is 1.344, the complexity level of the concurrent system fault event is matched to complexity level 3.

[0110] The embodiments of the present invention automatically reduce the number of faults when high load changes suddenly, preventing cognitive overload; increase the disturbance amplitude under a stable trend, strengthening the training of high-capacity personnel; reflect the interactive effect of the number of faults and the disturbance amplitude through multiplicative coupling; output discrete levels that meet aviation standards, and accurately control the scenario generation module.

[0111] The present invention provides a specific embodiment, step 105, generating a collaborative defect feature set based on the voice interaction data stream, and retrieving a candidate fault event set from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set, specifically comprising the following steps:

[0112] Step 501: Identifying the command response delay frequency and key semantic missing markers based on the voice interaction data stream to generate a collaboration defect feature set;

[0113] In this step, command response delay frequency refers to the number of times a pilot's response to a command timeout occurs per unit time. This is measured by the number of times (response start time - command end time) exceeds 2 seconds, reflecting the severity of the decision delay. The key semantic missing flag is a binary flag (0 / 1) set based on whether a predefined key operational command word is omitted, used to detect collaboration vulnerabilities. The collaboration defect feature set is a time series dataset consisting of delay frequency and semantic missing flags in the format of [timestamp, frequency, flag], used to quantify crew resource management deficiencies.

[0114] In this embodiment of the present invention, a time window segmentation algorithm is used to process the voice interaction data stream. The frequency of command response delays per minute is calculated, with the number of times the time between the end of a command and the start of a response exceeding two seconds. A keyword matching engine is also used to detect whether a predefined set of key commands (such as "check oil pressure" and "confirm flaps") is missing from the response text. If missing, the key semantic missing flag is set to 1; otherwise, it is set to 0. The per-minute delay frequency and the semantic missing flag are combined into a two-tuple [command response delay frequency, key semantic missing flag] to generate a collaborative defect feature set.

[0115] Step 502: According to the collaborative defect feature set, query the predefined defect pattern library to determine the defect type code;

[0116] In this step, command response delay frequency refers to the number of times a pilot failed to respond to a command within a unit of time. This is measured by the number of times (response start time - command end time) exceeds 2 seconds, reflecting the severity of the decision delay. The predefined defect pattern library is a rule library that stores the mapping between defect characteristics and type codes. Its structure is {command response delay frequency > 3, key semantic missing flag = 1:101} and is used for automated defect classification. The defect type code is a three-digit code (e.g., 101) representing a specific collaborative defect type (101 = communication failure) and driving knowledge graph retrieval.

[0117] In this embodiment of the present invention, the predefined defect pattern library stores the following rules: If the command response delay frequency is >3 times / minute and the key semantic missing flag is 1, the code is 101 (CRM severe defect); if the command response delay frequency is 1-3 times / minute, the code is 102 (CRM moderate defect). The current collaborative defect feature set [command response delay frequency = 4, key semantic missing flag = 1] is input into the pattern library for matching, and the output defect type code is 101.

[0118] Step 503: Calculating the physical disturbance constraint boundary based on the physical disturbance amplitude allowable weight corresponding to the complexity level and the dynamic parameter safety threshold of the flight phase;

[0119] In this step, the dynamic parameter safety threshold refers to the maximum physical perturbation allowed during the current flight phase (e.g., a 30° roll angle threshold during climb), derived from the aircraft manufacturer's operating manual. The physical perturbation constraint boundary is a safety threshold adjusted for complexity and used to filter out high-risk failure events.

[0120] In an embodiment of the present invention, a dynamic parameter safety threshold (e.g., a pitch angle safety threshold = 25°) of the current flight phase (e.g., the approach phase) is obtained, the physical disturbance amplitude permissible weight (e.g., 0.8) in the complexity level is extracted, and the physical disturbance constraint boundary is calculated as follows: dynamic parameter safety threshold × physical disturbance amplitude permissible weight = 25° × 0.8 = 20°.

[0121] Step 504: Retrieve a set of candidate fault events that are causally associated with the defect type code from the pre-built aviation threat knowledge graph;

[0122] In this step, the aviation threat knowledge graph refers to a graph database that stores the causal relationships between fault events. Nodes represent fault events (e.g., engine failure) and edges represent causal strengths (0-1.0). The candidate fault event set refers to the initial set of events retrieved from the knowledge graph that are causally associated with the defect code (e.g., {radio failure, oil pressure alarm}).

[0123] In this embodiment of the present invention, the pre-built aviation threat knowledge graph stores event causal chains, such as code 101, which links radio failure, navigation error, and oil pressure alarm. Using defect type code 101 as the search key, all directly causal events are returned, forming the candidate fault event set {radio failure, navigation error, oil pressure alarm}.

[0124] Step 505: removing candidate fault events that exceed the physical disturbance constraint boundary from the candidate fault event set to obtain multiple target fault events;

[0125] In this step, the target fault event refers to the set of events retained after filtering by the physical constraint boundary (e.g., oil pressure alarm disturbance 22° > boundary 20° is eliminated).

[0126] In this embodiment of the present invention, a set of candidate fault events is traversed, such as radio failure event A (disturbance amplitude 15°), navigation error event B (18°), and oil pressure warning event C (22°). By comparing the physical disturbance constraint boundary of 20°, oil pressure warning event C is removed (22° > 20°), resulting in two target fault events: radio failure event and navigation error event.

[0127] Step 506: Calculate the functional coupling strength values ​​between different target fault events, and generate a concurrent system fault event set based on the repair priority weight distribution table corresponding to the defect type code;

[0128] In this step, the functional coupling strength value refers to the weighted sum of the logical dependency and physical connectivity between event pairs, ranging from 0 to 1.0. The repair priority rule refers to the weight coefficient assigned by defect type code (e.g., code 101 = 1.2), which is used to increase the priority of selecting events associated with specific defects. The concurrent system fault event set refers to the final generated fault event combination (e.g., {radio failure event A, navigation error event B}), which is used to construct training scenarios.

[0129] In an embodiment of the present invention, the functional impact factors and physical association tags between each fault event pair are retrieved from a pre-constructed aviation system functional dependency graph and are superimposed to generate a functional coupling strength value for each fault event pair. Based on the defect type code, the repair weight coefficient of each fault event pair is retrieved from a preset repair priority weight allocation table. Based on all fault event pairs, different fault event combinations are generated, and the functional coupling strength values ​​of all fault event pairs in each fault event combination are accumulated to generate a corresponding raw coupling total score. Based on the repair weight coefficients of all fault event pairs in each fault event combination, a corresponding combined repair priority weighted value is calculated. The raw coupling total score of each fault event combination is multiplied by the repair priority weighted value to obtain a comprehensive score for the event combination, and the fault event combination with the highest comprehensive score is selected as the concurrent system fault event set.

[0130] The embodiment of the present invention accurately identifies customer relationship management collaboration defects through delay frequency and semantic loss markers; defect type coding enables targeted retrieval of knowledge graphs to ensure that events are relevant to training objectives; physical constraint boundaries are dynamically adjusted with complexity levels to ensure scenario safety; functional coupling strength values ​​optimize event combinations to strengthen the logical rationality of fault chains (for example, radio failure will inevitably lead to navigation errors).

[0131] The present invention provides a specific embodiment, step 506, calculating the functional coupling strength values ​​between different target fault events, combining the repair priority weight distribution table corresponding to the defect type code, and generating a concurrent system fault event set, specifically including the following steps:

[0132] Step 511: Retrieve the functional impact factor and physical association mark between each fault event pair from the pre-built aviation system functional dependency graph, wherein the fault event pair consists of two different target fault events;

[0133] In this step, the functional impact factor refers to the causal influence strength value (0-1.0) between failure events defined in the aviation system functional dependency graph. It is calculated based on the frequency statistics of event chains in historical accident reports (e.g., the probability that A causes B). The physical association flag is a binary identifier (1 / 0) that determines whether there is a direct mechanical or electrical connection between events based on the aircraft system physical topology (e.g., hydraulic line connection = 1).

[0134] In this embodiment of the present invention, for any two different target fault events (e.g., radio failure event A and navigation error event B), the aviation system functional dependency graph (Neo4j graph database) is retrieved to obtain the functional impact factor (the causal strength value from radio failure event A to navigation error event B is 0-1.0) and the physical association flag (e.g., 1 if a mechanical / electrical connection exists, 0 otherwise). For example, the functional impact factor from radio failure event A to navigation error event B is 0.9, and the physical association flag is 1.

[0135] Step 512: superimpose the functional impact factor and the physical association mark to generate a functional coupling strength value for each fault event pair;

[0136] In this embodiment of the present invention, a weighted superposition formula is used: Functional Coupling Strength Value = Functional Impact Factor × 0.7 + Physical Association Flag × 0.3. For example, the functional coupling strength value of the fault event pair (radio failure event A, navigation error event B) = 0.9 × 0.7 + 1 × 0.3 = 0.93.

[0137] Step 513: According to the defect type code, query the repair weight coefficient of each fault event pair from the preset repair priority weight allocation table;

[0138] In this step, the repair priority weight allocation table refers to a database that stores the mapping between defect type codes and weight coefficients. Its structure is {101:1.2,102:1.0} and is used to increase the priority of specific defect-related events. The repair weight coefficient is the weight value (≥1.0) assigned to each fault event pair. It is obtained by querying the defect type code and reflects the targeted training objectives.

[0139] In the embodiment of the present invention, a storage rule of the weight allocation table is preset, such as defect type code 101 corresponds to repair weight coefficient 1.2. The table is searched using current code 101 as a key, and a repair weight coefficient of 1.2 is uniformly assigned to each fault event pair.

[0140] Step 514: Based on all fault event pairs, generate different fault event combinations, accumulate the functional coupling strength values ​​of all fault event pairs in each fault event combination, and generate a corresponding raw coupling total score, wherein each fault event combination contains at least two different target fault events;

[0141] In this step, the original total coupling score refers to the arithmetic sum of the functional coupling strength values ​​of all events in the fault event combination, reflecting the overall correlation strength between events in the combination.

[0142] In this embodiment of the present invention, for the target fault event {radio failure event A, navigation error event B, oil pressure alarm event C}, all possible combinations are enumerated: the functional coupling strength value of fault event combination 1 {radio failure event A, navigation error event B} is 0.93; the functional coupling strength value of fault event combination 2 {radio failure event A, oil pressure alarm event C} is 0.8; the functional coupling strength value of fault event combination 3 {navigation error event B, oil pressure alarm event C} is 0.75; fault event combination 4 {radio failure event A, navigation error event B, oil pressure alarm event C}: contains the fault event pairs AB (0.93) + AC (0.8) + BC (0.75), that is, the original total coupling score corresponding to fault event combination 4 = 0.93 + 0.8 + 0.75 = 2.48.

[0143] Step 515: Calculate the corresponding combined repair priority weighted value based on the repair weight coefficients of all fault event pairs in each fault event combination;

[0144] In this step, the repair priority weighted value refers to the arithmetic mean of the repair weight coefficients of all fault events in the combination, which represents the matching degree between the combination and the training target.

[0145] In this embodiment of the present invention, the arithmetic mean of the repair weight coefficients of all fault event pairs in the combination is taken: fault event combination 4 {radio failure event A, navigation error event B, oil pressure alarm event C} contains 3 fault event pairs, their repair weight coefficients are all 1.2, and the corresponding arithmetic mean = 1.2; fault event combination 1 {radio failure event A, navigation error event B} contains 1 fault event pair, and the corresponding arithmetic mean = 1.2.

[0146] Step 516: Multiply the original coupling total score of each fault event combination by the repair priority weighted value to obtain a comprehensive score of the event combination, and select the fault event combination with the highest comprehensive score as the concurrent system fault event set;

[0147] In this step, the comprehensive score of the event combination refers to the product of the original coupling total score and the weighted value of the repair priority, which is used to quantify the technical rationality and training value of the combination.

[0148] In an embodiment of the present invention, for example, the event combination comprehensive score of fault event combination 4 = 2.48 × repair priority weighted value 1.2 = 2.976, and the event combination comprehensive score of fault event combination 1 = 0.93 × repair priority weighted value 1.2 = 1.116. Fault event combination 4 corresponding to the highest event combination comprehensive score of 2.976 is selected as the concurrent system fault event set.

[0149] The embodiment of the present invention ensures the engineering feasibility of fault combinations through physical association marking (for example, hydraulic failure will not cause radio failure); functional impact factors ensure that the event chain conforms to the evolution law of aviation accidents; repair weight coefficients enhance the targeted training of key defects; and a comprehensive scoring mechanism balances technical rationality and training goal priority to output the optimal fault event set.

[0150] The present invention provides a specific embodiment, step 106, converting the concurrent system fault event set into a global visual dynamic parameter instruction set matching the complexity level, and generating an EBT training scenario based on the global visual dynamic parameter instruction set, specifically comprising the following steps:

[0151] Step 601: parsing the event type identifier of each concurrent system fault event in the concurrent system fault event set, and searching a predefined three-dimensional model rule library to obtain a basic model identifier and a standard action script corresponding to the event type identifier;

[0152] In this step, the event type identifier is a string code that uniquely identifies the fault event and is used to match it to the 3D model library. The 3D model rule library is a database that stores the mapping between fault events and 3D assets. The base model identifier is the resource path of the fault object in the 3D engine. The standard action script is a predefined set of event dynamic parameters (such as a vibration amplitude of 5mm) that contains the initial values ​​for the physical simulation.

[0153] In an embodiment of the present invention, the system reads each fault event (e.g., an engine failure event) in a concurrent system fault event set, extracts its event type identifier, and then queries a predefined 3D model rule library (stored in a relational database) to obtain the basic model identifier bound to the identifier (e.g., the 3D model file path and the standard action script (e.g., a text instruction containing vibration amplitude and frequency parameters).

[0154] Step 602: dynamically scaling the motion amplitude parameters in the standard action script according to the permitted weight of the physical disturbance amplitude of the complexity level to generate a disturbance-enhanced action script;

[0155] In this step, the motion amplitude parameters refer to the values ​​in the action script that control the displacement / rotation of the model, which are used for physical simulation. The perturbation-enhanced action script refers to the dynamic parameters scaled by complexity weights to adapt to the pilot's state.

[0156] In an embodiment of the present invention, the system extracts the permitted weight value of the physical disturbance amplitude corresponding to the complexity level (such as 0.85), multiplies the motion amplitude parameter in the standard action script (such as a vibration amplitude of 5mm) by the weight value (5×0.85=4.25mm), and keeps other parameters (such as frequency) unchanged to generate a disturbance enhancement action script to achieve dynamic adaptation of the physical disturbance intensity.

[0157] Step 603: Calculate the triggering timing delay time between different concurrent system failure events based on the quantity permission weight corresponding to the complexity level;

[0158] In this step, the trigger timing delay time refers to the minimum trigger interval (in seconds) between multiple fault events.

[0159] In an embodiment of the present invention, based on the quantity permission weight value corresponding to the complexity level (such as 1.6), the delay calculation formula is: trigger timing delay time (seconds) = reference interval time / quantity permission weight, where the preset reference interval is 2 seconds, and the trigger timing delay time is calculated (2 / 1.6=1.25 seconds). This value controls the trigger time interval of multiple fault events in the scenario.

[0160] Step 604: combining the basic model identifier, the disturbance enhancement action script, and the trigger timing delay time to generate an event-level three-dimensional dynamic parameter unit corresponding to the concurrent system fault event;

[0161] In this step, the event-level 3D dynamic parameter unit refers to the rendering instruction unit of a single fault event (model ID + action script + trigger delay).

[0162] In this embodiment of the present invention, the basic model identifier, disturbance enhancement action script, and trigger timing delay parameters are encapsulated as structured data units. For example, the parameter unit for an engine failure event includes: model path, action script: vibration amplitude = 4.25mm, frequency = 10Hz, trigger timing delay time 1.25 seconds.

[0163] Step 605: Aggregate event-level 3D dynamic parameter units of all concurrent system failure events to generate a global view dynamic parameter instruction set;

[0164] In an embodiment of the present invention, all concurrent system fault events (such as engine failure events and oil pressure leakage events) are traversed, and after generating event-level three-dimensional dynamic parameter units for each fault event, these units are arranged in the order of event triggering, and global control tags are added (such as whether the event triggering mode is parallel or serial). Finally, a global view dynamic parameter instruction set is packaged and generated. The instruction set uses a key-value pair structure to organize data and can be directly parsed by the rendering engine.

[0165] Step 606: Based on the global visual dynamic parameter instruction set, generate an EBT training scene using a real-time rendering pipeline driver of a flight simulation device;

[0166] In this step, the global visual dynamic parameter instruction set refers to the aggregated instructions for all event units, controlling the overall behavior of the scene. The real-time rendering pipeline refers to the graphics processing pipeline of the flight simulator, which converts the instructions into real-time images. The EBT training scenario refers to the evidence-based cockpit visual environment (such as shaking instruments and warning sound effects) used for pilot training.

[0167] In this embodiment of the present invention, a global visual dynamic parameter instruction set is input into the flight simulator's rendering pipeline (such as the Unity engine's programmable rendering pipeline). The rendering pipeline then loads a 3D model file according to the instruction set, applies physical parameters in an action script to drive the model's dynamic behavior (e.g., vibrating the engine model with a 4.25mm amplitude). Delay parameters are used to control the timing of multiple event triggers (e.g., triggering an oil pressure leak event 1.25 seconds after an engine failure). This creates a cockpit visual environment with multiple failure effects in real time, completing the EBT training scenario.

[0168] The embodiment of the present invention automatically binds fault events to three-dimensional models to eliminate manual configuration errors; the physical disturbance amplitude is dynamically adjusted with the complexity level; the multi-event triggering interval is inversely proportional to the weight of the fault number to prevent event stacking; the global view dynamic parameter instruction set ensures multi-platform rendering consistency and improves scene generation efficiency.

[0169] Figure 2 The present invention provides a structural diagram of a data-driven EBT training scenario generation system, as shown in FIG. Figure 2 As shown, the system includes:

[0170] an acquisition module 21 for acquiring, in real time, a flight control data stream, a physiological feedback data stream, and a voice interaction data stream, wherein the flight control data stream includes flight control integrated deviation parameters, including aircraft attitude expected deviation parameters, track-to-expected deviation parameters, standard operating procedure execution deviation parameters, and control voice command deviation parameters; and the physiological feedback data stream includes pupil diameter change rate;

[0171] a quantification module 22, configured to quantify a first cognitive load state based on the physiological feedback data stream;

[0172] A generating module 23 is configured to generate a quantitative score of a system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream;

[0173] a determination module 24 configured to determine a complexity level of a concurrent system failure event based on the first cognitive load state when the quantitative score of the system management capability dimension is lower than a preset capability threshold;

[0174] A retrieval module 25 is configured to generate a collaborative defect feature set based on the voice interaction data stream, and retrieve a candidate fault event set from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set;

[0175] The conversion module 26 is configured to convert the concurrent system fault event set into a global vision dynamic parameter instruction set matching the complexity level, so as to generate an EBT training scenario based on the global vision dynamic parameter instruction set.

[0176] Figure 2 The data-driven EBT training scenario generation system can be executed Figure 1 The implementation principles and technical effects of the data-driven EBT training scenario generation method described in the illustrated embodiment are not further elaborated. The specific manner in which each module and unit performs operations in the data-driven EBT training scenario generation system in the above embodiment has been described in detail in the relevant embodiments of the method and will not be further elaborated here.

[0177] In one possible design, Figure 2 A data-driven EBT training scenario generation system of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0178] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0179] The processing component 32 is used for the above Figure 1 The embodiment provides a data-driven EBT training scenario generation method.

[0180] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0181] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0182] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0183] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0184] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0185] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0186] The embodiment of the present invention further provides a computer storage medium storing a computer program, which can achieve the above-mentioned Figure 1 A data-driven EBT training scenario generation method of the illustrated embodiment.

[0187] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0188] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0189] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0190] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A data-driven EBT training scenario generation method, characterized in that: include: acquiring, in real time, a flight control data stream, a physiological feedback data stream, and a voice interaction data stream, wherein the flight control data stream includes a flight control integrated deviation parameter, the flight control integrated deviation parameter including an aircraft attitude expected deviation parameter, a track-to-expected deviation parameter, a standard operating procedure execution deviation parameter, and a control voice instruction deviation parameter, and the physiological feedback data stream includes a pupil diameter change rate; quantifying a first cognitive load state based on the physiological feedback data stream; generating a quantitative score of a system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream; When the quantitative score of the system management capability dimension is lower than a preset capability threshold, determining a complexity level of the concurrent system failure event according to the first cognitive load state; generating a collaborative defect feature set based on the voice interaction data stream, and retrieving a candidate fault event set from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set; converting the concurrent system fault event set into a global vision dynamic parameter instruction set matching the complexity level, so as to generate an EBT training scenario based on the global vision dynamic parameter instruction set; Generating a quantitative score of a system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream includes: Correcting the flight control comprehensive deviation parameter according to the first cognitive load state to obtain a corrected comprehensive deviation parameter; parsing the voice interaction data stream to obtain voice response interval and semantic integrity information, and generating a crew collaboration behavior feature value based on the voice response interval and semantic integrity information; Calculating a system operation stability index based on the fluctuation range and fluctuation duration of the corrected comprehensive deviation parameter; fusing the system operation stability index, the first cognitive load state, and the crew collaborative behavior characteristic value to generate a quantitative score of the system management capability dimension; Converting the concurrent system fault event set into a global visual dynamic parameter instruction set matching the complexity level, and generating an EBT training scenario based on the global visual dynamic parameter instruction set, comprising: Parsing the event type identifier of each concurrent system fault event in the concurrent system fault event set, and searching a predefined three-dimensional model rule library to obtain a basic model identifier and a standard action script corresponding to the event type identifier; Dynamically scaling the motion amplitude parameters in the standard action script according to the permitted weight of the physical disturbance amplitude of the complexity level to generate a disturbance-enhanced action script; Calculate the triggering timing delay time between different concurrent system failure events based on the quantity permission weight corresponding to the complexity level; Combining the basic model identifier, the disturbance enhancement action script, and the trigger timing delay time to generate an event-level three-dimensional dynamic parameter unit corresponding to the concurrent system failure event; Aggregate the event-level 3D dynamic parameter units of all concurrent system failure events to generate a global view dynamic parameter instruction set; Based on the global visual dynamic parameter instruction set, an EBT training scene is generated by using a real-time rendering pipeline driver of a flight simulation device.

2. The method according to claim 1, characterized in that When the quantitative score of the system management capability dimension is lower than a preset capability threshold, determining the complexity level of the concurrent system failure event according to the first cognitive load state includes: Dividing the second cognitive load state of the historical training scenario into a plurality of load intensity intervals, and matching the first cognitive load state with the load intensity intervals to generate a basic load level; Querying a predefined fault event complexity benchmark library to obtain the maximum number of concurrent faults and the upper limit of the physical disturbance amplitude allowed by the basic load level, so as to generate an initial fault event complexity calibration value; Calculating the instantaneous load change rate and the change direction consistency coefficient based on the change trajectory of the first cognitive load state; According to the instantaneous load change rate, the quantity permission weight of the initial fault event complexity calibration value is adjusted, and at the same time, according to the change direction consistency coefficient, the physical disturbance amplitude permission weight is adjusted to generate the complexity level of the concurrent system fault event.

3. The method according to claim 2, characterized in that According to the instantaneous load change rate, the quantity allowable weight of the initial fault event complexity calibration value is adjusted, and according to the change direction consistency coefficient, the physical disturbance amplitude allowable weight is adjusted to generate the complexity level of the concurrent system fault event, including: When the instantaneous load change rate exceeds a high change rate critical value, the quantity permission weight of the initial fault event complexity calibration value is reduced according to a first reduction ratio; when the instantaneous load change rate is lower than a low change rate critical value, the quantity permission weight is increased according to a first gain ratio to obtain an adjusted quantity permission weight; Retrieving a physical disturbance amplitude scaling factor corresponding to the change direction consistency coefficient from a pre-built directional stability rule library, and updating the physical disturbance amplitude permissible weight of the initial fault event complexity calibration value according to the physical disturbance amplitude scaling factor to obtain an adjusted physical disturbance amplitude permissible weight; Performing a multiplication operation on the adjusted quantity permission weight and the physical disturbance amplitude permission weight to generate a comprehensive complexity scalar; The complexity level of the concurrent system fault event corresponding to the comprehensive complexity scalar is matched in a preset complexity level conversion rule library.

4. The method according to claim 1, wherein Generate a collaborative defect feature set based on the voice interaction data stream, and retrieve a candidate fault event set from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set, including: identifying, based on the voice interaction data stream, command response delay frequencies and key semantic missing markers to generate a collaboration defect feature set; According to the collaborative defect feature set, a predefined defect pattern library is searched to determine the defect type code; Calculating the physical disturbance constraint boundary based on the physical disturbance amplitude allowable weight corresponding to the complexity level and the dynamic parameter safety threshold of the flight phase; Retrieving a set of candidate fault events that are causally related to the defect type code from a pre-built aviation threat knowledge graph; Remove candidate fault events that exceed the physical disturbance constraint boundary from the candidate fault event set to obtain multiple target fault events; The functional coupling strength values ​​between different target fault events are calculated, and combined with the repair priority weight distribution table corresponding to the defect type code, a concurrent system fault event set is generated.

5. The method according to claim 4, characterized in that Calculate the functional coupling strength values ​​between different target fault events, combine them with the repair priority weight distribution table corresponding to the defect type code, and generate a concurrent system fault event set, including: Retrieving a functional impact factor and a physical correlation marker between each fault event pair from a pre-built aviation system functional dependency graph, wherein the fault event pair consists of two different target fault events; Superimposing the functional impact factor and the physical association mark to generate a functional coupling strength value for each fault event pair; According to the defect type code, query the repair weight coefficient of each fault event pair from a preset repair priority weight allocation table; Based on all fault event pairs, different fault event combinations are generated, and the functional coupling strength values ​​of all fault event pairs in each fault event combination are accumulated to generate the corresponding raw coupling total score, where each fault event combination contains at least two different target fault events; Based on the repair weight coefficients of all fault event pairs in each fault event combination, the corresponding combined repair priority weighted value is calculated; The original coupling total score of each fault event combination is multiplied by the repair priority weighted value to obtain the comprehensive score of the event combination. The fault event combination with the highest comprehensive score is selected as the concurrent system fault event set.

6. A data-driven EBT training scenario generation system, characterized by: include: an acquisition module, configured to acquire, in real time, a flight control data stream, a physiological feedback data stream, and a voice interaction data stream, wherein the flight control data stream includes a flight control integrated deviation parameter, the flight control integrated deviation parameter including an aircraft attitude expected deviation parameter, a track-to-expected deviation parameter, a standard operating procedure execution deviation parameter, and a control voice command deviation parameter, and the physiological feedback data stream includes a pupil diameter change rate; a quantification module, configured to quantify a first cognitive load state based on the physiological feedback data stream; a generating module, configured to generate a quantitative score of a system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream; a determination module, configured to determine, when the quantitative score of the system management capability dimension is lower than a preset capability threshold, a complexity level of the concurrent system failure event according to the first cognitive load state; a retrieval module, configured to generate a collaborative defect feature set based on the voice interaction data stream, and retrieve a candidate fault event set from a pre-built aviation threat knowledge graph in combination with the complexity level to generate a concurrent system fault event set; a conversion module, configured to convert the concurrent system fault event set into a global visual dynamic parameter instruction set matching the complexity level, so as to generate an EBT training scenario based on the global visual dynamic parameter instruction set; Generating a quantitative score of a system management capability dimension based on the first cognitive load state, the flight control comprehensive deviation parameter, and the voice interaction data stream includes: Correcting the flight control comprehensive deviation parameter according to the first cognitive load state to obtain a corrected comprehensive deviation parameter; parsing the voice interaction data stream to obtain voice response interval and semantic integrity information, and generating a crew collaboration behavior feature value based on the voice response interval and semantic integrity information; Calculating a system operation stability index based on the fluctuation range and fluctuation duration of the corrected comprehensive deviation parameter; fusing the system operation stability index, the first cognitive load state, and the crew collaborative behavior characteristic value to generate a quantitative score of the system management capability dimension; Converting the concurrent system fault event set into a global visual dynamic parameter instruction set matching the complexity level, and generating an EBT training scenario based on the global visual dynamic parameter instruction set, comprising: Parsing the event type identifier of each concurrent system fault event in the concurrent system fault event set, and searching a predefined three-dimensional model rule library to obtain a basic model identifier and a standard action script corresponding to the event type identifier; Dynamically scaling the motion amplitude parameters in the standard action script according to the permitted weight of the physical disturbance amplitude of the complexity level to generate a disturbance-enhanced action script; Calculate the triggering timing delay time between different concurrent system failure events based on the quantity permission weight corresponding to the complexity level; Combining the basic model identifier, the disturbance enhancement action script, and the trigger timing delay time to generate an event-level three-dimensional dynamic parameter unit corresponding to the concurrent system failure event; Aggregate the event-level 3D dynamic parameter units of all concurrent system failure events to generate a global view dynamic parameter instruction set; Based on the global visual dynamic parameter instruction set, an EBT training scene is generated by using a real-time rendering pipeline driver of a flight simulation device.

7. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a data-driven EBT training scenario generation method as described in any one of claims 1 to 5.

8. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a data-driven EBT training scenario generation method as described in any one of claims 1 to 5 is implemented.

Images