Understanding and modelling low adhesion risk in the wheel-rail interface

As the global drive for more sustainable and environmentally friendly travel continues, rail companies and operators are being put under more pressure to improve safety and reliability, as well as encourage commuters to switch to rail for their commute. Leaves are known to contribute to low adhesion in the wheel-rail interface, which is a massive problem for train operators. These problems include, but are not limited to, wheel sliding, Signals Passed At Danger (SPADs), station overruns and collisions. This project aimed to investigate the current understanding of how leaves get to the railhead and how they bond to the rail while causing low adhesion, including the differences between leaf species. The outcomes of these investigations were then used alongside KPI data from industry to develop a low adhesion risk assessment model.
A literature review and paper grading were conducted to find gaps in the current knowledge and steer the direction of this work. These gaps included; mechanisms of leaf fall, chemical composition of certain deciduous tree species native to the UK and their effects on friction. Specific bonding mechanisms and bond strength also remain unclear. Hypotheses regarding the specific leaf layer formation and low adhesion mechanisms were identified during the literature review. Throughout the autumn period data is gathered, including monitoring leaf levels on trees across the UK using photographs taken by leaf fall observers. These form part of the input to the adhesion prediction(s), which are used by the Train Operating Companies (TOC's), to plan their journeys and timetables. A more detailed, location specific model that takes other physical factors into account is needed, hence the development of the low adhesion risk assessment model in this work.
Through Autumn of 2018 and 2019 ambient humidity, pressure and temperature as well as railhead temperature were recorded at various known low adhesion points across the Supertram network in Sheffield, UK and at several heritage rail locations. This was achieved using a sensor box to record the ambient parameters (air temperature, pressure and humidity) and an infrared thermometer to record railhead temperature. The purpose was to determine which environmental conditions correlate with leaf fall times and low adhesion incidents. These were then fed into the adhesion risk assessment prediction model.
A study was conducted to assess the leaf fall behaviours of three tree species under still and artificially windy conditions. Leaf retention on ballast was tested with dry, slightly wet and fully saturated leaves, where the wind speed at which leaves were removed was tested. This information was fed into the adhesion risk assessment prediction model.
The bonding and low adhesion hypotheses identified from the literature review were detailed along with their own specific testing plans. The hypotheses were tested using a mixture of mechanical testing, chemical testing and analysis. The findings of these tests contribute to the wider understanding of leaf layer formation.
A large part of this project involved the development of an improved adhesion prediction model. Historical Key Performance Indicator (KPI) data concerning Wheel Slide Prevention (WSP) was provided by Chiltern Railways Company Limited (CRCL) and formed the basis of a case study and initial formation of a leaf layer induced low adhesion risk assessment model. Locations on the CRCL network were organised by the frequency of WSP activation, then split into high, medium and low groups. A combination of Google maps and physical site visits were used to assess the vegetation levels and physical track parameters. The scores for half of the locations were mathematically analysed in order to rank the impact of the parameters. The model was then applied to the second half of the locations to validate the parameter rankings. Outcomes of the other parts of work looking at leaf fall classification, times and associated weather conditions as well as friction and bonding hypotheses also fed into the scoring method where applicable.