Management and understanding of rolling contact fatigue

Transcription

1 Research Brief Management and understanding of rolling contact fatigue Background Rolling contact fatigue (RCF) has been a safety concern to the railway industry for many years. The cost of managing RCF on Network Rail infrastructure, from the premature replacement of cracked rail and maintenance activities such as rail grinding, is estimated to be in the order of 250 million per year. An earlier RSSB research project (T115) developed a Whole Life Rail Model (WLRM) to help understand and manage the causes of RCF. That research, following earlier work on understanding RCF, was completed in Key stakeholders considered the outcomes of that research and in January 2004 identified a series of questions to be answered through further research. The questions relating to crack initiation and crack propagation formed the basis of this research project. Aims and Objectives The programme of work was split into two work packages: the first aimed to understand the mechanisms of crack initiation; the second aimed to understand crack propagation. The objectives of the project were delivered by implementing a methodology that relied on numerical modelling of RCF. More specifically, the objective of the research in WP1 was to increase the understanding of the mechanisms of crack initiation and to underpin and improve the energy-based model of RCF which had been developed under research project T115. Work was also undertaken to analyse the effect of material properties on the RCF resistance so that a similar function could be developed for other rail materials. The objective of research in WP2 was to develop a mathematical model of Phase II RCF crack growth, and to use this model to investigate the key variables influencing RCF crack growth so that RCF cracks could be better managed and their growth rates reduced. Full scale experimental work was also conducted to determine to what extent fluid can penetrate surface breaking RCF cracks and promote crack growth. Numerical Modelling of RCF A cross section through a section of rail reveals intense plastic deformation in a thin layer of material at the running surface. This is caused by the vertical and shear forces generated within the contact patch. Each wheel pass over the material generates a small increment of deformation which can accumulate to very large values. This process is known as ratchetting. The material cannot accumulate this deformation indefinitely and when the accumulated deformation T355 July 2008 RSSB Research and Development Programme Floor 6, Central House Upper Woburn Place London WC1H 0HY

2 exceeds the ductility of the material, that piece of material fails producing either wear debris or crack-like flaws (Figure 1). Figure 1: Development of wear and cracklike flaws by ratchetting. Left: Computer simulation. Right: Twin-disc rail steel sample. The so-called 'Brick Model' has been developed by Newcastle University to simulate ratchetting and the associated failure of material through wear or cracking. A sample of material is modelled as a matrix of small elements (bricks), each of which can accumulate deformation as a result of vertical and shear forces applied to the material surface, and fail when their ductility is exhausted. For the current research work a hexagonal microstructure was developed to simulate a pearlitic rail steel (Figure 2) with different material properties specified for the pearlitic grains and the ferrite grain boundaries. The brick model is able to predict material wear rates, the time taken for the development of any crack-like flaws and the lengths of such flaws under different loading conditions. Very small crack-like flaws generated through ratchetting may not propagate if the stress intensity at the crack tip does not exceed the fatigue threshold stress intensity for that material. However, if flaws are long enough to develop a sufficiently large stress intensity they will grow. If their growth rate is faster than the wear rate then their length will increase, otherwise they may grow at the crack tip but will remain of constant length due to wear led shortening at the crack mouth on the surface. Crack growth rates were predicted using a linear-elastic fracture mechanics crack growth model. A 'Rotated Crack Model' was developed to correctly account for the lateral traction within the contact. The effects of material properties on the rate of crack growth rate were included through changes to the threshold stress intensity. The model allows for the effect of contact pressure, shear tractions, surface friction, crack face friction, residual stresses and continuously welded rail stress on the crack growth rate to be investigated. Figure 2: Simulation of pearlitic rail steels. Methodology The objectives of the research were achieved through utilisation of the brick model and crack growth model to predict ratchetting and crack growth behaviour under a variety of loading conditions. The contact patch data and forces which formed the input to the ratchetting and fracture mechanics models were taken from vehicle dynamics simulations of vehicles at a number of sites in Britain, and had been used in earlier studies to develop the WLRM.

3 These simulations utilised measured track geometry and models of a range of vehicles, with typical wheel profiles, seen at two sites on the East Coast Main Line (Harringay and Sandy), and one site on c2c route (Leigh-on-Sea). For the ECML sites the following vehicles were simulated: Class 43 and Class 91 locomotives; Mk 3 and Mk 4 coaches; and Class 365 electric multiple units (EMUs). These were deemed to represent the most numerous vehicles at each site. For each case simulations were performed for three wheel profiles, representing new, moderate and heavily worn, as in previous studies. For the site at Leigh-on-Sea results were used from simulations of Class 312, Class 317 and Class 357 EMUs. From the results for each site a number of locations were chosen at which the contact patch forces for each vehicle type were sampled and these were used as inputs to the brick model and crack growth model. The locations were chosen such that they represented a range of observed RCF states, from locations where no cracks were found to locations where large RCF cracks were observed. Findings and recommendations: WP1 The results from the brick model showed a good correlation with the wear number, Tγ, which forms the basis of the existing WLRM (Figure 3). This was particularly true for values of Tγ below about 65J/m, where the model identified a threshold for the onset of ratchetting damage and a linear increase in ratchetting damage with energy input above the threshold. Figure 3: The Tγ model relates energy expended at the contact patch (the product of shear force, T, and creepage, y), shown on the x-axis to damage to the rail, shown on the y-axis. At higher Tγ values the expected increase in wear rate, which would result in a decrease in RCF initiation risk (in agreement with site observations), was not predicted by the brick model. Although it was not possible to explain this discrepancy, one explanation may be that the increased microslip in the contact patch at higher values of contact patch energy may lead to an increase in temperature and a subsequent increase in the wear rate. Such thermal effects are not currently included in the brick model, but are an area which needs further investigation. The models also identified that the wear number per unit contact patch area (( Tγ) A ) may be a better parameter to quantify RCF than the Tγ parameter, particularly when analysing conditions where flange contact occurs. Through analysis of existing site and modelling data, a new damage function based on the ( Tγ) A parameter was proposed and tested which has a similar shape to that of the earlier Tγ parameter. There was little difference between the Tγ and ( Tγ) A parameters when contact patch positions towards the crown of the rail were considered. The ratchetting and crack growth models were also used to analyse the effect of

4 material prehardening on RCF initiation, and to develop a model for RCF initiation in different grades of rail material. These showed that material hardening by, for example, a fleet of vehicles which did not generate enough ratchetting to initiate RCF defects, had little effect on the resistance of the material to initiating RCF under the subsequent passage of a fleet of vehicles which generated higher ratchetting damage. This was because, under high compressive hydrostatic stress, the time for the material to reach its maximum hardness is a small proportion of the time to exhaustion of ductility and subsequent material failure. To investigate the effects of material properties on resistance to RCF, three different materials were simulated through changes to the microstructure in the brick model and the fatigue threshold in the fracture mechanics model. Although these analyses did not provide conclusive results on the effects of material properties on RCF initiation, they did show that the simulated harder grade materials had enhanced resistance to ratchetting failure, but the reduction in wear rate resulted in an increase in crack growth rate, should a defect be initiated. It was not possible to substantiate these predictions with experimental results or field observations. Findings & recommendations: WP2 The full-scale track test conducted during this project has confirmed that it is possible for water or water-based fluids to enter RCF cracks, with or without wheel loading cycles (Figure 4). This evidence of crack penetration by fluids supports the use of crack growth mechanisms which depend on the presence of fluids inside cracks. However, the tests could not, and were not, intended to identify which mechanisms were present, and the study recommended undertaking laboratory studies using instrumented cracks to further understand them. Crack growth predictions for the two ECML sites were found to correlate well with the Tγ parameter generated in the vehicle dynamics simulations. The investigation of key parameters on Phase II crack growth predictions concluded that when wheel-rail interface friction conditions are constant, increasing the crack face friction coefficient (i.e. a 'dry' crack) results in crack faces locking and Phase II crack growth reducing, or even stopping altogether. Phase II crack growth was possible at all the surface friction levels considered ( μ ), although it begins at the lowest number of contact cycles for a high surface friction coefficient. As the surface friction coefficient was reduced the number of contact cycles which would occur before rapid crack growth began increased, showing that surface friction is a driver of crack growth. A further trend with surface friction is that cracks stabilise at shorter lengths under high friction conditions. Imposing a compressive residual stress pattern into the rail head results in lower crack growth rates after early crack growth (> 500,000 load cycles). It was found that, even a compressive longitudinal residual stress results in higher crack growth rates in early life of the crack. Applying small tensile vertical residual stresses does increase crack propagation rates. For Phase II RCF crack growth, vertical rail bending does not contribute significantly to the rate of propagation until the cracks are extremely long. However, RCF cracks are observed to grow in three dimensions and there is a need to identify whether lateral rail bending is a driver for the growth of RCF cracks in the field direction.

5 Figure 4: A green florescence indicates the penetration of the crack by UV dye during testing. The presence of the dye is also indicated along the other crack mouths at the rail surface Tensile or compressive longitudinal thermal stresses were found to increase or decrease crack growth (compared to the base case of a stress-free rail) depending on the crack length considered. A tensile thermal stress does result in consistently higher crack growth rates than a compressive thermal stress. Next Steps The research project increased the understanding of the mechanisms of crack initiation and crack propagation, and validated the WLRM for crack initiation for Tγ values up to 65J/m. Further work is required to understand material behaviour and the reasons for increasing wear rates at higher values of contact patch energy. The influence of various factors on crack growth rate has been estimated, and a comprehensive review of the current state of the art in this area of RCF has been produced. These findings have been shared with the technical advisory group of V/T SIC, which is coordinating further RCF research, including further development of the WLRM (RSSB project T775). Contact Head of Engineering Research R&D programme RSSB research@rssb.co.uk