Abstract

Tilting is a worldwide accepted technology concept in railway transportation. The
particular benefit from tilting trains use is reduction in journey times due to speed
increase on track corners (while maintaining acceptable passenger comfort), a point
that facilitates improved customer service. An additional benefit is cost effectiveness
due to the train running on existing rail tracks. Many countries opted to using tilting
trains as means of fast public transportation (UK, USA, Canada, Sweden, Norway,
Switzerland, Germany, Japan).
The industrial norm of tilting high speed trains is that of precedence tilt whereby
preview tilt enabling signals are used to provide the required information to the
vehicles (it can also use a combination of track database information or GPS but the
concept is the same). Precedence tilt tends to be complex (mainly due to the signal
interconnections between vehicles and the advanced signal processing required for
monitoring). Research studies of earlier than precedence schemes,i.e. the so-called
nulling-type schemes whereby local-per-vehicle signals are used to provide tilt (a
disturbance rejection-scheme although tends to suffer from inherent delays in the
control feedback), are still an important research aim due to the simple nature
and most importantly due to the more straightforward fault detection compared to
precedence. Use of nulling-type tilt has been supported by recent studies in this
context.
The research presented in this thesis highly contributes to simplified single-inputsingle-output robust tilt control using the simplest rail vehicle tilt structure, i.e. an
Active Anti-Roll Bar. Proposed are both robust conventional (integer-type) control
approaches and non-conventional (non-integer) schemes with a rigorous investigation of the difficult to achieve deterministic/stochastic tilt trade-off. Optimization
has been used extensively for the designs. A by-product of the work is the insight
provided into the relevant tilting train model Non Minimum Phase characteristics
and its link to uncertainty for control design. Work has been undertaken using
Matlab, including proper assessment of tilt ride quality considerations.