INNOVATSION TEXNOLOGIYALAR Maxsus son, 2023-yil, dekabr
ISSN 2181-4732
31
the network had different levels of vulnerability depending on the centrality measure used to select
the nodes for removal. For instance, the network was more vulnerable to attacks based on
betweenness centrality than on degree centrality, because betweenness centrality captured the
importance of nodes as bridges or bottlenecks in the network.
The evaluation of the network resilience to disruptions and failures can provide guidance for
designing and implementing contingency plans and recovery strategies for the ATM system. For
instance, random failures can be mitigated by increasing the redundancy and diversity of the network,
while targeted attacks can be prevented by enhancing the security and protection of the network.
Moreover, different types of disruptions and failures can require different types of responses
depending on their impact on different parts of the network [4, 5].
Network optimization for performance improvement.
The ATM system aims to achieve high
levels of performance in terms of safety, capacity, efficiency, predictability, environment, and cost-
efficiency. These performance areas are interrelated and often conflicting, meaning that improving
one area can compromise another area. Therefore, there is a need to find optimal solutions that can
balance and harmonize these performance areas according to the objectives and preferences of the
stakeholders involved in the ATM system.
One example of applying CNT to optimize the network performance is the work by Zhang et
al. (2019), who studied the US ATM network using data from FAA. They constructed a network
where nodes were airports and links were flights between them. They proposed a multi-objective
optimization model that considered three performance indicators: flight delay (the difference between
actual and scheduled arrival times), flight distance (the length of flight routes), and flight fuel
consumption (the amount of fuel consumed by flights). They used a genetic algorithm to find Pareto-
optimal solutions that minimized these indicators simultaneously. They compared these solutions
with the actual flight data and evaluated their potential benefits [5, 6].
They found that their optimization model could reduce flight delay by 19.8%, flight distance
by 2.4%, and flight fuel consumption by 2.6%, compared with the actual flight data. They also found
that their optimization model could generate different trade-offs among these indicators depending
on the weights assigned to them. For instance, they found that reducing flight delay had a positive
effect on reducing flight fuel consumption, but a negative effect on increasing flight distance.