Page 133 - 360.revista de Alta Velocidad - Nº 5

Probabilistic Safety Analysis of High Speed and Conventional Railway Lines

The methodology of the model can be summarized in different stages described below:

The first step is to identify and reproduce the most relevant variables that play an important
role in the safety of a railway line to model its multidimensional random behavior.

The second stage consists of reproducing the elements that the driver and the train observe
when they travel along the line. Particularly, we reproduce the evolution of the level of attention
of the driver, the speed and occurrence of incidents.

2.1 Proposed Bayesian network model

2.1.1 Introduction

As event trees and fault trees, although very powerful present some important limitations,
Bayesian networks have been selected, given the previous experience of the authors in this
very important tool of representation of the probabilistic structure of multidimensional random
variables.
A Bayesian network consists of two elements: A directed acyclic graph and a set of conditional
probability tables. In this way, any joint multidimensional probability can be reproduced with
no restriction, which implies being able to treat any multidimensional random variable. To
properly perform a Probabilistic Safety Analysis (PSA) it is convenient to use the following sta-
ges: The first requires to identify and reproduce all the items or elements which are relevant
to the safety of a railway line. The second stage must identify the variables that influence the
elements previously described and how they influence them. In particular, the evolution of
the driver’s level of attention (variables M), speed (V variables), the presence of signals of all
kinds, tunnels, viaducts, occurrence of accidents (variables A), etc. are required. Finally, the
third step must be devoted to define the structure of the model and quantify the conditional
probabilities.

With this aim, a video from the train’s cabin recording the railway line becomes an essential
tool to identify all these possible hazardous items that the engineer encounters when tra-
velling along the railway line. Some elements that can be considered are crossing switches,
tunnel crossings, cutting and embankments, viaducts, infrastructure crossings (overpass and
underpass), level crossings, landslides, etc. not forgetting the light signals and signs of speed
limitations, since a large distance is required to be able to reduce the speed and errors are not
always protected by the automatic protection systems (ATP).

2.1.2 Variables used in the model

From the previous discussion, we can identify the following list of variables that are very im-
portant to the line safety:

1. Tr: Driver’s tiredness. Since the driver is subject to an increase of tiredness with driving
time a variable is needed to analyze how it changes along the line when travelling. Since ti-
redness is known to be one important contribution to human error, it must not be forgotten.
2. D: Driver’s attention. It refers to the driver’s attention level that in our model is simplified
to three states: distracted, attentive and alert. We assume that an alert situation always
leads to a correct decision and that a distracted situation leads to a no action at all. Con-
trary, an attentive situation is subject to both correct and incorrect actions with given
probabilities, such that the probability of the first is much bigger that the probability of the
second.

International Congress on High-speed Rail: Technologies and Long Term Impacts - Ciudad Real (Spain) - 25th anniversary Madrid-Sevilla corridor 131