A graph theory-based dynamic assessment method for flight number confusion risk

Abstract

This invention relates to a graph theory-based dynamic assessment method for flight number confusion risk. The steps are as follows: Based on flight dynamics and plans, extract sets of flight numbers to construct flight number node sets; calculate the auditory and visual similarity of each randomly combined flight number pair in the flight number node set, and obtain a confusion risk coefficient after fusion processing; set a high-risk threshold, retain high-risk edges with a confusion risk coefficient greater than or equal to the high-risk threshold, resulting in a weighted undirected graph, and construct a weighted confusion network; calculate the node weighted degree centrality, the node-level chain confusion coefficient for one-step propagation, and the set confusion vulnerability index, and rank and issue early warnings for flights involved in the flight number node set within a set time window. This invention can adapt to dynamic air traffic control environments, provide early warnings, and adapt to mixed communication environments; by visualizing the assessment results, controllers can intuitively perceive the similarity structure within the flight number node set, thereby reducing their perceptual and decision-making load.

Description

Technical Field

[0001] This invention relates to the fields of air traffic management and aviation safety technology, specifically to a graph theory-based method for dynamic assessment of flight number confusion risk. Background Technology

[0002] During civil aviation air traffic control, controllers use aircraft call signs (i.e., flight numbers) via radio communication to identify and direct specific aircraft. When two or more flight numbers that are highly similar in character composition or pronunciation appear simultaneously on the same control channel, it can easily induce visual identification errors and auditory confusion for controllers and flight crews, resulting in the "similar flight number" problem. Such confusion is a significant cause of mishearing, mis-calling instructions, and even flight conflicts.

[0003] Currently, air traffic control automation systems generally incorporate similar flight number identification and alarm functions. The basic implementation involves setting discrete logical judgment rules. For example, if two flight numbers share the same three-letter airline code and at least two numbers in their numerical components are identical in both value and position, they are considered similar and an alarm is triggered. This approach has significant shortcomings: First, it can only provide a qualitative "yes" or "no" conclusion, failing to continuously quantify the degree of similarity and associated risks. This leads to a large number of low-risk combinations frequently triggering "ceiling" alarms, consuming controller attention, while truly high-risk critical combinations may be overlooked because they haven't reached the hard threshold. Second, the judgment rules are based solely on character form, failing to incorporate the core perception channel of pronunciation confusion in radio communications into the quantification system. Third, existing methods compare isolated "flight number pairs," unable to handle the cascading similarity effects between multiple aircraft. For example, flight numbers A and B are highly similar, and B and C are highly similar, but the direct similarity between A and C may not be the highest. In this case, the controller is very likely to miscall C or misidentify A due to the strong activation state of B in the working memory of frequently switching call signs. This risk of transmission confusion is completely outside the current assessment system.

[0004] Therefore, there is an urgent need for a method that can continuously assign confusion probabilities to any pair of flight numbers based solely on the attributes of the flight number string itself, and can quantitatively evaluate the chain similarity network of multiple call signs within the flight number set, providing a refined algorithmic kernel for source desimilarity of flight schedules and automated early warning. Summary of the Invention

[0005] To address the technical challenges of automated air traffic control systems, such as their inability to continuously quantify the similarity and associated risks of "similar flight numbers" and their inability to handle the chain-like similarity effects between multiple aircraft, this invention provides a graph-based dynamic assessment method for flight number confusion risk. By constructing a confusion risk network graph, the method calculates node confusion probability, node weighted degree centrality, chain confusion coefficient, and set confusion vulnerability index, achieving accurate risk ranking and early warning. Combined with measures such as manual signage language coverage, temporary changes to flight numbers, and pre-coordination to connect flights to the next sector, the method effectively reduces the risk of flight number confusion.

[0006] The technical solution adopted by this invention to achieve the above objectives is: a graph theory-based dynamic assessment method for flight number confusion risk, comprising the following steps: S1: Based on the current flight dynamics and plans within the controlled sector, extract the set of flight numbers to construct the flight number node set V; S2: Calculate the auditory and visual similarity of each pair of randomly combined flight numbers in the flight number node set, and obtain the confusion risk coefficient after fusion processing; S3: Set a high-risk threshold Retain those with a confusion risk coefficient greater than or equal to the high-risk threshold. High-risk edges yield a weighted undirected graph. Where V is the set of flight number nodes and E is the set of edges, a weighted confusion network is constructed; S4: In a weighted confusion network, calculate the node weighted degree centrality, the node-level chain confusion coefficient for one-step propagation, and the set confusion fragility index; S5: Utilizing the confusion risk coefficient of flight numbers, node weighted degree centrality, node-level chain confusion coefficient, and set confusion vulnerability index, the concentration of flight number nodes within a set time window is considered. The relevant flights are subject to risk ranking and early warning output.

[0007] The set of flight numbers in step S1 includes: Airline code + call sign in numeric form, language identifier, company code, company code voice, numeric, numeric voice, planned sector entry and exit, and actual entry and exit times.

[0008] In step S2, let the confusion risk coefficient for each pair of flight numbers be... Auditory similarity is Visual similarity is i and j represent different flight numbers, and the specific calculation is as follows: S201. Calculate and obtain auditory similarity. ; ; In the formula, Auditory weighting for airline codes , The auditory similarity of the airline code portions of the i-th and j-th flight numbers. The auditory similarity between the numerical components of the i-th flight number and the j-th flight number is calculated as follows: ; ; In the formula, The Levenstein distance function is used to calculate the degree of difference between two strings. This indicates that the string length is being retrieved. To find the maximum value function, The audio text is the airline code portion of the i-th flight number. The audio text is the airline code portion of the j-th flight number. The audio text is the numerical part of the i-th flight number. The audio text of the numerical part of the j-th flight number; S202. Calculate and obtain visual similarity. ; Visual similarity is calculated using string-weighted edit distance. Break it down into visual similarity of the airline code portion and visual similarity of the numerical portion: ; In the formula, For the visual weight of airline codes, , Visual similarity of airline code portions. The visual similarity of the digits is calculated using the following formula: ; ; In the formula, The text represents the airline code portion of the i-th flight number. For the airline code portion of the j-th flight number, The text is the numerical part of the i-th flight number. The text is the numerical part of the j-th flight number; S203. Confusion Risk Factor Based on auditory similarity Visual similarity Weighted fusion, as follows: ; In the formula For auditory weight, , This is the time overlap factor, which is 1 if the two flights have overlapping times, and 0 otherwise.

[0009] Step S4 involves calculating the node weighted degree centrality, node-level chain confusion coefficient, and set confusion vulnerability index in the weighted confusion network, including the following steps: S401. Calculate the weighted degree centrality of nodes. as follows: ; In the formula, For nodes The set of neighboring nodes, For nodes The weight of the edge between u and u; S402. Calculate the node-level chain confusion coefficient: Assuming nodes If called, the final set of nodes that may be affected includes... and the set of all its neighboring nodes Then the node-level chain confusion coefficient The calculation is as follows: ; In the formula, For nodes In a weighted undirected graph, the number of neighbors represents the percentage of direct confusion that a single air-to-ground call to a flight number could potentially cause. For calculating the flight number node set of the function The number of flight nodes; S403. Calculate the set confusion vulnerability index : ; ; ; In the formula, To pass the high-risk threshold The number of high-risk edges selected For edge density weights, The weight that determines the maximum impact of a node's ability in a single step. and ; For edge density, The maximum impact capability of a node in one step. To calculate the total number of edges in a network using permutations and combinations.

[0010] Step S4 also includes: S404. For situations requiring calls to multiple flight numbers, what is the probability of confusion regarding flight numbers from other adjacent nodes in the network? for: ; In the formula, For a single call, For nodes and The weight of the edges between them; Chaining confusion coefficient of a single multi-flight call for: ; ; In the formula, yes The union of all high-risk neighbors of all nodes. For nodes Adjacent and edge weight greater than The set of all neighboring nodes.

[0011] Step S5 includes static evaluation and dynamic evaluation, specifically: S501. During static evaluation, assessments are conducted at both the individual network node level and the overall network level to improve the targeting of intervention strategies and overall situational awareness; a time window is set. During the time window The confusion risk coefficient for each pair of flight numbers entering the sector will be calculated separately. Node weighted degree centrality coefficient Node-level chain confusion coefficient Vulnerability index and set confusion ; S502. When calling a flight number, enter dynamic evaluation, capture the flight number according to the voice command, and only evaluate the confusion probability of adjacent network nodes that are affected in real time, so as to improve the targeting of the controller's real-time monitoring, including calling one flight number at a time and calling multiple flight numbers at a time.

[0012] Step S501 involves evaluating each network node level and the network as a whole, specifically as follows: Evaluation at each network node level: For the overall encoding of the confusing network, the node size is adjusted by sorting the nodes according to their weighted degree centrality coefficients, and the edges are adjusted according to their confusion risk coefficients and whether they exceed a high-risk threshold. Adjust the edge thickness and the contrast between solid and void; For local coding of scrambled networks, nodes and edges are colored. All nodes in the same scrambled network have the same color. The influence of nodes is distinguished by the intensity of their colors. Edges are colored according to a color threshold range. Conduct an overall assessment of the network: Assess the overall network obfuscation risk based on the size of the set obfuscation vulnerability index; After manual verification by controllers, the system autonomously sorts flight numbers by node weighted degree centrality from high to low, outputting the set of flight numbers with predicted confusion risk and the corresponding chain confusion coefficients. Set a high-risk warning threshold and issue warnings for flight number sets that exceed the high-risk warning threshold; at the same time, issue simultaneous warnings for the confusion of medium and high-risk periods in the overall network.

[0013] In step S502, the affected obfuscation network is first activated. For a single call to a flight number, the obfuscation probability of adjacent nodes is the edge weight. The called flight number is captured in real time according to the voice command, and the set of affected real-time obfuscation risk flights is output. The nodes are sorted according to their obfuscation probabilities, and the node-level chain obfuscation coefficient of the calling node is output. Furthermore, the activated obfuscated network is visually re-encoded; For scenarios involving multiple flights in a single call, the confusion probability of the affected nodes has been revised. and in accordance with Sort the data from highest to lowest size, and output the real-time set of flight numbers at risk of obfuscation and the node-level chain obfuscation coefficients of the obfuscated node set. Furthermore, the activated obfuscated network is visually re-encoded; The system can automatically sort nodes by their confusion probability from high to low, set a high-risk threshold, and issue warnings for flight numbers that exceed the high-risk threshold.

[0014] The activated obfuscated network is visually re-encoded, specifically as follows: Activated obfuscated network nodes and lines are displayed with highlights and a flowing visual effect, respectively. The node size is based on the obfuscation probability of the node's flight number. Size adjustment, color intensity based on the confusion probability of the node flight number. Size adjustment, with a flowing display that moves from the calling node to the affected node, while highlighting the calling node for differentiation.

[0015] The present invention has the following beneficial effects and advantages: 1. This invention provides a graph theory-based dynamic assessment method for flight number confusion risk. It divides flight numbers into company codes and numbers, and obtains accurate results through weighted calculation of character and phonetic similarity. It has strong universality, low safety risk, can adapt to dynamic air traffic control environments, provides early warning, supports differences between Chinese and English voices and human language coverage, and is suitable for mixed communication environments. The human language coverage mechanism respects the professional judgment of controllers, reduces false alarms, and improves safety.

[0016] 2. By visualizing the evaluation results, this invention enables controllers to intuitively perceive similar structures within flight numbers, thereby reducing their perception and decision-making load. At the same time, it can display active network nodes for real-time command, thereby improving the targeting of command monitoring.

[0017] 3. This invention can transform flight number confusion risk into graph theory quantitative indicators, enabling multi-level automatic assessment from nodes to the whole; the optional single-call-multiple-flight call confusion assessment function provides safety assurance for rare multi-call-signal instruction scenarios, ensuring low-risk steady-state operation of each flight, reducing the workload of controllers, balancing safety assurance and operational efficiency, and effectively improving the economic benefits of airlines while improving the quality of air traffic control services. Attached Figure Description

[0018] Figure 1 This is a flowchart of a graph theory-based dynamic assessment method for flight number confusion risk according to the present invention. Figure 2 This is a static evaluation network diagram generated by the method of the present invention in the embodiment; Figure 3 This is a dynamic evaluation network diagram generated by the method of the present invention in the embodiment; Figure 4 The revised obfuscated network graph generated by activating the obfuscated network in this embodiment; Figure 5 This is the obfuscated network diagram after being modified by artificial language overlay in the embodiment. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0020] like Figure 1 As shown, the present invention provides a graph theory-based dynamic assessment method for flight number confusion risk, comprising the following steps: S1: Based on the current flight dynamics and plans within the controlled sector, extract the set of flight numbers to construct the flight number node set V; S2: Calculate the auditory and visual similarity of each pair of randomly combined flight numbers in the flight number node set, and obtain the confusion risk coefficient after fusion processing; S3: Set a high-risk threshold Retain those with a confusion risk coefficient greater than or equal to the high-risk threshold. High-risk edges yield a weighted undirected graph. Where V is the set of flight number nodes and E is the set of edges, a weighted confusion network is constructed; S4: In a weighted confusion network, calculate the node weighted degree centrality, the node-level chain confusion coefficient for one-step propagation, and the set confusion fragility index; S5: Utilizing the confusion risk coefficient of flight numbers, node weighted degree centrality, node-level chain confusion coefficient, and set confusion vulnerability index, the concentration of flight number nodes within a set time window is considered. The relevant flights are subject to risk ranking and early warning output.

[0021] In step S1, each flight number set includes: airline code + call sign in numeric form, language identifier, company code, company code voice, numeric, numeric voice, planned sector entry and exit, and actual entry and exit times.

[0022] In step S2, let the confusion risk coefficient for each pair of flight numbers be... Auditory similarity is Visual similarity is i and j represent different flight numbers, and the specific calculation is as follows: S201. Calculate and obtain auditory similarity. ; ; In the formula, Auditory weighting for airline codes , The auditory similarity of the airline code portions of the i-th and j-th flight numbers. The auditory similarity between the numerical components of the i-th flight number and the j-th flight number is calculated as follows: ; ; In the formula, The Levenstein distance function is used to calculate the degree of difference between two strings. This indicates that the string length is being retrieved. To find the maximum value function, The audio text is the airline code portion of the i-th flight number. The audio text is the airline code portion of the j-th flight number. The audio text is the numerical part of the i-th flight number. The audio text of the numerical part of the j-th flight number; S202. Calculate and obtain visual similarity. ; Visual similarity is calculated using string-weighted edit distance. Break it down into visual similarity of the airline code portion and visual similarity of the numerical portion: ; In the formula, For the visual weight of airline codes, , Visual similarity of airline code portions. The visual similarity of the digits is calculated using the following formula: ; ; In the formula, The text represents the airline code portion of the i-th flight number. For the airline code portion of the j-th flight number, The text is the numerical part of the i-th flight number. The text is the numerical part of the j-th flight number; S203. Confusion Risk Factor Based on auditory similarity Visual similarity Weighted fusion, as follows: ; In the formula For auditory weight, , This is the time overlap factor, which is 1 if the two flights have overlapping times, and 0 otherwise.

[0023] In step S4, the weighted degree centrality of nodes, the node-level chain confusion coefficient, and the set confusion vulnerability index are calculated in the weighted confusion network. This includes the following steps: S401. Calculate the weighted degree centrality of nodes. as follows: ; In the formula, For nodes The set of neighboring nodes, For nodes The weight of the edge between u and u; S402. Calculate the node-level chain confusion coefficient: Assuming nodes If called, the final set of nodes that may be affected includes... and the set of all its neighboring nodes Then the node-level chain confusion coefficient The calculation is as follows: ; In the formula, For nodes In a weighted undirected graph, the number of neighbors represents the percentage of direct confusion that a single air-to-ground call to a flight number could potentially cause. For calculating the flight number node set of the function The number of flight nodes; S403. Calculate the set confusion vulnerability index : ; ; ; In the formula, To pass the high-risk threshold The number of high-risk edges selected For edge density weights, The weight that determines the maximum impact of a node's ability in a single step. and ; For edge density, The maximum impact capability of a node in one step. To calculate the total number of edges in a network using permutations and combinations; S404. For situations requiring calls to multiple flight numbers, what is the probability of confusion regarding flight numbers from other adjacent nodes in the network? for: ; In the formula, For a single call, For nodes and The weight of the edges between them; Chaining confusion coefficient of a single multi-flight call for: ; ; In the formula, yes The union of all high-risk neighbors of all nodes. For nodes Adjacent and edge weight greater than The set of all neighboring nodes.

[0024] Step S5 performs risk ranking and early warning output for both real-time and future time windows, including static assessment and dynamic assessment, specifically as follows: S501. During static evaluation, assessments are conducted at both the individual network node level and the overall network level to improve the targeting of intervention strategies and overall situational awareness; a time window is set. During the time window The confusion risk coefficient for each pair of flight numbers entering the sector will be calculated separately. Node weighted degree centrality coefficient Node-level chain confusion coefficient Vulnerability index and set confusion ; S502. When calling a flight number, enter dynamic evaluation, capture the flight number according to the voice command, and only evaluate the confusion probability of adjacent network nodes that are affected in real time, so as to improve the targeting of the controller's real-time monitoring, including calling one flight number at a time and calling multiple flight numbers at a time.

[0025] Step S501 involves evaluating each network node level and the network as a whole, specifically as follows: Evaluation at each network node level: For the overall encoding of the confusing network, the node size is adjusted by sorting the nodes according to their weighted degree centrality coefficients, and the edges are adjusted according to their confusion risk coefficients and whether they exceed a high-risk threshold. Adjust the edge thickness and the contrast between solid and void; For local coding of scrambled networks, nodes and edges are colored. All nodes in the same scrambled network have the same color. The influence of nodes is distinguished by the intensity of their colors. Edges are colored according to a color threshold range. Conduct an overall assessment of the network: Assess the overall network obfuscation risk based on the size of the set obfuscation vulnerability index; After manual verification by controllers, the system autonomously sorts flight numbers by node weighted degree centrality from high to low, outputting the set of flight numbers with predicted confusion risk and the corresponding chain confusion coefficients. Set a high-risk warning threshold and issue warnings for flight number sets that exceed the high-risk warning threshold; at the same time, issue simultaneous warnings for the confusion of medium and high-risk periods in the overall network.

[0026] In step S502, the affected obfuscation network is first activated. For a single call to a flight number, the obfuscation probability of adjacent nodes is the edge weight. The called flight number is captured in real time according to the voice command, and the set of affected real-time obfuscation risk flights is output. The nodes are sorted according to their obfuscation probabilities, and the node-level chain obfuscation coefficient of the calling node is output. Furthermore, the activated obfuscated network is visually re-encoded; For scenarios involving multiple flights in a single call, the confusion probability of the affected nodes has been revised. and in accordance with Sort the data from highest to lowest size, and output the real-time set of flight numbers at risk of obfuscation and the node-level chain obfuscation coefficients of the obfuscated node set. The activated obfuscated network is then visually re-encoded; The system can automatically sort nodes by their confusion probability from high to low, set a high-risk threshold, and issue warnings for flight numbers that exceed the high-risk threshold.

[0027] In steps S501-S502, the activated obfuscation network is visually re-encoded, specifically as follows: Activated obfuscated network nodes and lines are displayed with highlights and a flowing visual effect, respectively. The node size is based on the obfuscation probability of the node's flight number. Size adjustment, color intensity based on the confusion probability of the node flight number. Size adjustment, with a flowing display that moves from the calling node to the affected node, while highlighting the calling node for differentiation.

[0028] The method of this invention constructs a confusion risk network graph, calculates the node confusion probability, node weighted degree centrality, chain confusion coefficient, and set confusion vulnerability index, and combines it with manual signage language coverage to achieve accurate risk ranking and early warning. It balances safety assurance and operational efficiency, and effectively improves the economic benefits of airlines while enhancing the quality and safety level of air traffic control services, thus achieving the purpose of the invention.

[0029] The following describes the implementation process of the present invention in detail with a specific example (12 flights).

[0030] Step S1: Based on the flight dynamics and plans within the current controlled sector, extract the set of flight numbers to construct the flight number node set V. The basic data can be found in the attribute table of the flight number set within the current sector. Current sector flight number set attribute table: .

[0031] Step S2: Calculate the auditory similarity and visual similarity of each pair of randomly combined flight numbers in the flight number node set, and obtain the confusion risk coefficient after fusion processing. Parameter settings: Auditory weighting Visual weight Airline code auditory weighting Digital auditory weighting is Airline code visual weight Digital visual weights are High-risk threshold Time overlap factor (real time).

[0032] Taking a random pair of flight numbers, i = CCA8465 (Air China) and j = CXA8465 (Egret 8465), as an example, the auditory similarity... Based on the pronunciation standards of ICAO Doc 9432 (Manual of Radiotelephony) and the Civil Aviation Administration of China's Air Traffic Management Bureau's regulations on air traffic radio communication, and given that "Air China 8465" and "Egret 8465" have different airline codes, the auditory similarity of the airline code portions of the i-th and j-th flight numbers is as follows: If the value is 0, and the digits are the same, then the auditory similarity of the digits of the i-th flight number and the j-th flight number is... The overall auditory similarity is 1, therefore... .

[0033] Visual similarity of airline codes Visual similarity with the digital part calculate: ; ; Visual similarity ; Confusion risk factor ; Similarly, the confusion risk coefficients for other flight number combinations are calculated.

[0034] Step S3: Set a high-risk threshold A value of 0.5 is used to retain the high-risk value (confusion risk coefficient is greater than or equal to the high-risk value). The high-risk edges are used to obtain a weighted undirected graph, where V is the set of flight number nodes and E is the set of edges, and a weighted confusion network is constructed.

[0035] Step S4: In the weighted confusion network, calculate the weighted degree centrality of nodes, the node-level chain confusion coefficient, the set confusion vulnerability index, and the weighted degree centrality of each node. The sorting table is as follows: Calculate the weighted degree centrality of each node. Sorting table: .

[0036] Calculate the chain confusion coefficient at the node level The results are shown in the node-level chain confusion coefficient table: Node-level chain confusion coefficient table: .

[0037] Calculate the set's confusion vulnerability index: Passing high risk threshold =0.5, the number of high-risk edges selected. =11, see the list of 11 high-risk edges: List of 11 high-risk edges: .

[0038] Edge density ; Maximum impact capability of a node in one step ; Set Confusion Vulnerability Index .

[0039] Step S5: Concentrate flight number nodes within a set time window. The relevant flights are subject to risk ranking and early warning output, including static assessment and dynamic assessment. Static evaluation: A time window ΔT is defined, where ΔT = 0 represents real-time evaluation and greater than 0 represents future evaluation. Flights that will enter the sector within the predicted time window are analyzed. It is then determined whether the planned time intervals of each pair of randomly combined flight numbers overlap; if so, a time overlap factor is applied. =1, otherwise 0, recalculate the confusion risk coefficient. Node weighted degree centrality Node-level chain confusion coefficient Set Confusion Vulnerability Index Based on the calculation results, a confused network graph is generated, with nodes of the same type having the same color. The size and color intensity of nodes are determined according to the weighted degree centrality of the nodes. Arranged from largest to smallest, nodes and Weight of the edges between Values ​​greater than or equal to 0.7 are represented by a solid red line, values ​​between 0.5 and 0.7 are represented by a solid orange line, and values ​​less than or equal to 0.5 are represented by a dashed gray line. The size is adjusted by edge thickness, and finally a network graph for evaluation at each network node level is generated, as shown below. Figure 2 As shown, the weighted degree centrality of nodes is output while sorting them by their influence. The node-level chain confusion coefficient corresponding to the node and the aggregate mixed fragility coefficient .

[0040] Conduct an overall assessment of the network: Based on the set hybrid fragility coefficient The size is used to assess the overall network confusion risk, with low risk (0, 0.4], medium risk (0.4, 0.7], and high risk (0.7, 1]. This example uses the set's vulnerability coefficient for confusion. The value is 0.25, therefore the overall network confusion risk is low.

[0041] Dynamic assessment: If only one flight number is called at a time and misheard, the confusion probability of adjacent nodes is equal to the edge weight and does not need to be adjusted. For example, if CCA8465 is called and misheard, its confusion probability P is... conf If the value is 1, then there is a risk of confusion due to infection of neighboring nodes CXA8465 and CXA8467, with confusion probabilities of 0.8533 and 0.6533 respectively. (CCA8465, node-level chain confusion coefficient) The value remains unchanged at 3 / 12. Finally, output the confusion probabilities of the activated nodes (including source nodes and non-source nodes; source nodes are the nodes that misheard, and non-source nodes are the nodes adjacent to the source node with an edge weight greater than 0.5) from largest to smallest. and the source node At the same time, the activation of the confusion network was corrected, and the node size was adjusted according to... Size correction, among which A value greater than or equal to 0.7 is red, and a value less than 0.7 is yellow, with the color intensity depending on... Size and edges are displayed with a sense of flow, moving from the source node to non-source nodes, specifically as follows: Figure 3 As shown, the node colors are arranged according to... Correction: The source node, i.e. the calling node, is now highlighted, and you can see that the aperture is larger.

[0042] If a call to two flight numbers is misheard simultaneously, such as calling both CCA8465 and CXA8467, then the confusion probabilities of adjacent nodes need to be revised. The confusion probability of CXA8465... The confusion probabilities of other nodes are determined by the edge weights. Finally, the confusion probabilities of CCA8465, CXA8467, CXA8465, and CXA8489 are 1, 1, 0.9491, and 0.6000, respectively. The chain confusion coefficient for a single multi-flight call is... The value is 4 / 12. In this case, the output should also be arranged in descending order of the confusion probability of the active nodes. And the chain confusion coefficient of a single multi-flight call Simultaneously, the obfuscation network is revised and activated, using the same method as above. The revised obfuscation network is as follows: Figure 4 As shown, the node colors are arranged according to... Correction: The source node, i.e. the calling node, is now highlighted, and you can see that the aperture is larger.

[0043] Artificial language coverage effect: Taking the network containing CCA101, CCA107, and CCA121 as an example, CCA101 is a central node, and its weighted degree centrality is... The value is at its maximum. Assuming CCA101 has not entered the controlled sector, the controller determines through navigation reports that CCA101 is speaking in Chinese. Simultaneously, CCA107 and CCA121, both within the controller's jurisdiction, are also speaking in Chinese. To better cut off the propagation path, the coordinating controller needs to coordinate with adjacent sectors to change their communications to English, thereby reducing similarity and cutting off the propagation path. After the controller's intervention, the auditory similarity between CCA101 and CCA121 significantly decreased (from 0.7333 to 0), reducing the confusion risk factor. The auditory similarity between CCA101 and CCA107 decreased significantly (from 0.7333 to 0), indicating a decrease in the confusion risk coefficient. The value decreased from 0.7333 to 0.2933. Both pairs of nodes were below the threshold, and no more alarms were triggered, effectively eliminating false alarms. The modified obfuscated network is as follows: Figure 5 As shown, similar adjustments can be made to the communication language of the network nodes where CSN and CXA are located in the future. For example, changing the communication language of CXA8465 to English and CES3557 to Chinese can effectively reduce the similarity with neighboring nodes, cut off the propagation path, thereby reducing the network chain coefficient and edge density, and improving the overall stability of the network.

[0044] In addition to the methods mentioned above, controllers can also take other strategies depending on the situation, such as coordinating to contact the next sector in advance or temporarily changing the flight number, which can also achieve the effect of avoiding confusion.

Claims

1. A dynamic assessment method for flight number confusion risk based on graph theory, characterized in that, Includes the following steps: S1: Based on the current flight dynamics and plans within the controlled sector, extract the set of flight numbers to construct the flight number node set V; S2: Calculate the auditory and visual similarity of each pair of randomly combined flight numbers in the flight number node set, and obtain the confusion risk coefficient after fusion processing; S3: Set a high-risk threshold Retain those with a confusion risk coefficient greater than or equal to the high-risk threshold. High-risk edges yield a weighted undirected graph. Where V is the set of flight number nodes and E is the set of edges, a weighted confusion network is constructed; S4: In a weighted confusion network, calculate the node weighted degree centrality, the node-level chain confusion coefficient for one-step propagation, and the set confusion fragility index; S5: Utilizing the confusion risk coefficient of flight numbers, node weighted degree centrality, node-level chain confusion coefficient, and set confusion vulnerability index, the concentration of flight number nodes within a set time window is considered. The relevant flights are subject to risk ranking and early warning output.

2. The graph theory-based dynamic assessment method for flight number confusion risk according to claim 1, characterized in that... The set of flight numbers in step S1 includes: Airline code + call sign in numeric form, language identifier, company code, company code voice, numeric, numeric voice, planned sector entry and exit, and actual entry and exit times.

3. The graph theory-based dynamic assessment method for flight number confusion risk according to claim 1, characterized in that, In step S2, let the confusion risk coefficient for each pair of flight numbers be... Auditory similarity is Visual similarity is i and j represent different flight numbers, and the specific calculation is as follows: S201. Calculate and obtain auditory similarity. : ； In the formula, Auditory weighting for airline codes , The auditory similarity of the airline code portions of the i-th and j-th flight numbers. The auditory similarity between the numerical components of the i-th flight number and the j-th flight number is calculated as follows: ； ； In the formula, The Levenstein distance function is used to calculate the degree of difference between two strings. This indicates that the string length is being retrieved. To find the maximum value function, The audio text is the airline code portion of the i-th flight number. The audio text is the airline code portion of the j-th flight number. The audio text is the numerical part of the i-th flight number. The audio text of the numerical part of the j-th flight number; S202. Calculate and obtain visual similarity. : Visual similarity is calculated using string-weighted edit distance. Break it down into visual similarity of the airline code portion and visual similarity of the numerical portion: ： In the formula, For the visual weight of airline codes, , Visual similarity of airline code portions. The visual similarity of the digits is calculated using the following formula: ； ； In the formula, The text represents the airline code portion of the i-th flight number. For the airline code portion of the j-th flight number, The text is the numerical part of the i-th flight number. The text is the numerical part of the j-th flight number; S203. Confusion Risk Factor Based on auditory similarity Visual similarity Weighted fusion, as follows: ； In the formula For auditory weight, , This is the time overlap factor, which is 1 if the two flights have overlapping times, and 0 otherwise.

4. The graph theory-based dynamic assessment method for flight number confusion risk according to claim 1, characterized in that, Step S4 involves calculating the node weighted degree centrality, node-level chain confusion coefficient, and set confusion vulnerability index in the weighted confusion network, including the following steps: S401. Calculate the weighted degree centrality of nodes. as follows: ； In the formula, For nodes The set of neighboring nodes, For nodes The weight of the edge between u and u; S402. Calculate the node-level chain confusion coefficient: Assuming nodes If called, the final set of nodes that may be affected includes... and the set of all its neighboring nodes Then the node-level chain confusion coefficient The calculation is as follows: ； In the formula, For nodes In a weighted undirected graph, the number of neighbors represents the percentage of direct confusion that a single air-to-ground call to a flight number could potentially cause. For calculating the flight number node set of the function The number of flight nodes; S403. Calculate the set confusion vulnerability index : ； ； ； In the formula, To pass the high-risk threshold The number of high-risk edges selected For edge density weights, The weight that determines the maximum impact of a node's ability in a single step. and ; For edge density, The maximum impact capability of a node in one step. To calculate the total number of edges in a network using permutations and combinations.

5. The graph theory-based dynamic assessment method for flight number confusion risk according to claim 4, characterized in that, Step S4 also includes: S404. For situations requiring calls to multiple flight numbers, what is the probability of confusion regarding flight numbers from other adjacent nodes in the network? for: ； In the formula, For a single call, For nodes and The weight of the edges between them; Chaining confusion coefficient of a single multi-flight call for: ； ； In the formula, yes The union of all high-risk neighbors of all nodes. For nodes Adjacent and edge weight greater than The set of all neighboring nodes.

6. The graph theory-based dynamic assessment method for flight number confusion risk according to claim 1, characterized in that, Step S5 includes static evaluation and dynamic evaluation, specifically: S501. During static evaluation, assessments are conducted at both the individual network node level and the overall network level to improve the targeting of intervention strategies and overall situational awareness; a time window is set. During the time window The confusion risk coefficient for each pair of flight numbers entering the sector will be calculated separately. Node weighted degree centrality coefficient Node-level chain confusion coefficient Vulnerability index and set confusion ; S502. When calling a flight number, enter dynamic evaluation, capture the flight number according to the voice command, and only evaluate the confusion probability of adjacent network nodes that are affected in real time, so as to improve the targeting of the controller's real-time monitoring, including calling one flight number at a time and calling multiple flight numbers at a time.

7. The graph theory-based dynamic assessment method for flight number confusion risk according to claim 6, characterized in that, Step S501 involves evaluating each network node level and the network as a whole, specifically as follows: Evaluation at each network node level: For the overall encoding of the confusing network, the node size is adjusted by sorting the nodes according to their weighted degree centrality coefficients, and the edges are adjusted according to their confusion risk coefficients and whether they exceed a high-risk threshold. Adjust the edge thickness and the contrast between solid and void; For local coding of scrambled networks, nodes and edges are colored. All nodes in the same scrambled network have the same color. The influence of nodes is distinguished by the intensity of their colors. Edges are colored according to a color threshold range. Conduct an overall assessment of the network: Assess the overall network obfuscation risk based on the size of the set obfuscation vulnerability index; After manual verification by controllers, the system autonomously sorts flight numbers by node weighted degree centrality from high to low, outputting the set of flight numbers with predicted confusion risk and the corresponding chain confusion coefficients. Set a high-risk warning threshold and issue warnings for flight numbers that exceed the high-risk warning threshold; At the same time, early warnings are issued synchronously for periods of medium to high risk of confusion in the overall network.

8. The graph theory-based dynamic assessment method for flight number confusion risk according to claim 6, characterized in that: In step S502, the affected obfuscation network is first activated. For a single call to a flight number, the obfuscation probability of adjacent nodes is the edge weight. The called flight number is captured in real time according to the voice command, and the set of affected real-time obfuscation risk flights is output. The nodes are sorted according to their obfuscation probabilities, and the node-level chain obfuscation coefficient of the calling node is output. Furthermore, the activated obfuscated network is visually re-encoded; For scenarios involving multiple flights in a single call, the confusion probability of the affected nodes has been revised. and in accordance with Sort the data from highest to lowest size, and output the real-time set of flight numbers at risk of obfuscation and the node-level chain obfuscation coefficients of the obfuscated node set. Furthermore, the activated obfuscated network is visually re-encoded; The system can automatically sort nodes by their confusion probability from high to low, set a high-risk threshold, and issue warnings for flight numbers that exceed the high-risk threshold.

9. The graph theory-based dynamic assessment method for flight number confusion risk according to claim 8, characterized in that, The activated obfuscated network is visually re-encoded, specifically as follows: Activated obfuscated network nodes and lines are displayed with highlights and a flowing visual effect, respectively. The node size is based on the obfuscation probability of the node's flight number. Size adjustment, color intensity based on the confusion probability of the node flight number. Size adjustment, with a flowing display that moves from the calling node to the affected node, while highlighting the calling node for differentiation.

Images