1. THE CHARACTERISATION OF RELATIONSHIPS
BETWEEN DYNAMIC IMPACT AND THE RESPONSE
OF RAILWAY TRACK-LIKE STRUCTURES
By
Min Leong Lee
BCOM., BE (HONS)
THIS THESIS IS SUBMITTED IN FULFILMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
Department of Mechanical Engineering
Monash University (Clayton),
Victoria, Australia
August 2005

2. I
SUMMARY
The aim of this research was to characterise the relationship between dynamic impacts
resultant from wheel defects and the response of the railway track. Specifically, this
study assessed the viability of two proposed techniques based on using the track
acceleration response to reconstruct or predict a force magnitude resultant from a
wheel-flat impact. One proposed method known as the inverse analysis method (IAM)
involves using a classical signal deconvolution method whereby a calibrated transfer
function can be used to reconstruct impact waveforms. The second method involves
using the acceleration response only to form a correlation between impact force
magnitude and the root mean square value (RMS) of the track response. This study
contributed towards the utilisation of acceleration signals in wheel impact force
prediction and towards improvement in the overall monitoring of wheel impact forces.
In totality, this work can be split into three major components. The first two having
been performed in parallel comprised of experimental impact testing on a laboratory
based railway test track and finite element (FE) simulations of impact events on a
modelled 3 crib track structure. Both the experimental and FE techniques were used to
identify parameters important to impact force reconstruction by the proposed
methods.
It was initially found through experimental testing that both proposed methods
showed promise in terms of correctly determining the impact force magnitude. Firstly
the linear behaviour of the test structure enabled the IAM to successfully reconstruct
impact waveforms of varying magnitudes. Good correlation between the impact
magnitude and the track response RMS indicated potential for successful impact
magnitude determination by interpolation along a calibrated impact force vs. RMS
curve. It was then shown experimentally that the severity of errors for both proposed
methods increased as the attempted predictions occurred increasingly remote from the
original calibration location along the rail. Experimental and simulation work then
revealed that impact duration is a parameter which can affect the transmittance of
impact energy into the track structure. It was shown that impacts of a given magnitude
and with varying durations excite different levels of response from the track structure.
Whilst this affects the RMS method, it was shown that the IAM contained a measure

3. II
of immunity because the method relies on structural linearity for accurate impact
force reconstruction.
The third major component of this study was the full scale field testing. A piece of in-
service track was instrumented with shear-bridges and accelerometers. Shear-bridges
are widely accepted as an accurate means of determining impact magnitudes. They
however provide only limited coverage about the mid-span within a rail crib. In total
the instrumentation was active for 15 days, and 36 freight trains were logged. The
proposed signal deconvolution method (IAM) proved to be a useful tool with 80% of
impacts detected within the useful shear-bridge region being within 30% error in
magnitude. Whilst not recommended as a direct replacement for the traditional shear-
bridge system, the high numbers of accurate reconstructions give confidence that the
proposed methodology can be used as a complement with the biggest benefit in aiding
detection of impacts occurring outside the range of the shear-bridges. Furthermore,
the proposed inverse method was very successful in detecting more complex style
impact events such as when multiple strikes occur within a single crib of rail.
With regards to using pure acceleration responses (RMS) of a railway track to predict
impact magnitude, it was found that whilst a fair correlation existed between
acceleration response of the railway track and the impact force, there was a significant
spread in the relationship such that a given impact magnitude can be represented by a
significant range in RMS values and vice versa. Therefore the predictive value of
using the pure response only to surmise impact magnitude must be considered along
with the inherent errors associated with the method. Finally, the method has no
mechanism for dealing with complexities such as multiple impacts occurring within a
single crib of rail.