曝光台 注意防骗
网曝天猫店富美金盛家居专营店坑蒙拐骗欺诈消费者
aircraft at same or lower flight levels)
SEPARATION SAFETY MODELING
7-10
[This page intentionally blank.]
DATA NEEDS AND MODEL CALIBRATION
8-1
8.0 DATA NEEDS AND MODEL CALIBRATION
8.1 THE REQUIREMENT FOR DATA
In order to provide a reasonably accurate output, any model used to estimate collision risk
for an air traffic management (ATM) system must be based on reliable system data.
Where data are unavailable, it may be necessary to estimate values for input parameters,
which can lead to a very high level of uncertainty in results. The lack of data can,
therefore, restrict the applicability and reliability of a collision risk model.
8.2 SOURCES OF DATA
The availability and source of data for risk modeling depends on the type of situation
being modeled. When modeling an existing system, data can be obtained by direct
observation, although this may be expensive and time consuming. An example of this is
the collection of data on traffic levels and lateral track-keeping errors that is undertaken
for the annual risk assessment for the North Atlantic region. Another example is the
measurement of aircraft height-keeping performance at high altitudes, done as part of a
study of the feasibility of reducing vertical separation minima above FL 290.
Obtaining reliable data for hypothetical systems can be much more difficult. For such
systems, it may be necessary to estimate many of the important performance parameters.
This results in high levels of uncertainty in the risk estimates. For this reason, it is much
easier to assess risk for existing systems (or existing systems with small modifications)
than for completely new concepts of air traffic management.
The best source of data for collision risk modeling is information taken directly from
current systems. Such data fully reflect the actual performances of all parts of the systems.
It is usually possible to obtain reliable estimates for the error in data values obtained from
observations. This allows the error in the final risk estimates to be assessed.
Failing “real” data, system specifications can be used as a source of model input
parameters. The greatest problem with specification data, particularly for engineered
systems, is that the real performance is often much different (better or worse) than the
designed performance. This may result in overly pessimistic or optimistic risk estimates -
sometimes by orders of magnitude.
A final source of information which is commonly used is simulation, real- or fast-time.
However, information from simulations can be very misleading: Simulations can contain
many hidden assumptions about how the system will behave and real-time (i.e., human-inthe-
loop) simulations can only model a limited number of potential situations. Simulations
should, therefore, be considered part of the modeling process rather than a source of (realworld)
input data.
SEPARATION SAFETY MODELING
8-2
8.3 TYPES OF DATA
For the purpose of collection, data for risk modeling can be divided into three general
categories as follows:
· Physical data on the system,
· Equipment performance data, and
· Human performance data.
Physical data on an ATM system include such items as aircraft sizes and speeds, flow
rates, route structures, separation standards, etc. Equipment performance data include
both normal and abnormal performance such as the normal variability in aircraft navigation
and radar failure rates. Human performance data include the capability and reliability of
the human elements of the system. These three components are not entirely independent
and, in practice, it is not always possible to completely distinguish system performance
from human performance.
In general, the collection of physical data on the system to be modeled is the most
straightforward, both for existing systems and for future systems. The most difficult
physical data to obtain for future systems is the traffic demand pattern. Predicting traffic
demand for future scenarios can require highly complex models.
Equipment performance data can be obtained either from direct observations of operating
systems or from performance specifications. The preferred method is to use observational
data. Performance specifications may be overly pessimistic or optimistic, as mentioned
above and are also often not sufficiently detailed for risk modeling. For instance, Required
Navigation Performance (RNP) specifications require aircraft navigation systems to ensure
that aircraft remain within some containment range of their nominal position for 95% of
the time. Unfortunately, it is the 5% of time that the aircraft may spend outside of the
中国航空网 www.aero.cn
航空翻译 www.aviation.cn
本文链接地址：
a concept paper for separation safety modeling(54)
