Project A2 - Predicting the Life of Various Grades of Steel
INVESTIGATORS: Dr Claire Davis (University of Birmingham); Prof. Ajay Kapoor (University of Newcastle upon Tyne)
OBJECTIVES AND RESEARCH SUMMARY:
RESEARCHER(S): Dr John E. Garnham (University of Birmingham); Dr Francis J. Franklin, Dr David I. Fletcher (University of Newcastle upon Tyne)
INDUSTRIAL COLLABORATORS: Bombardier Transportation; Carillion Rail; Corus Rail Technologies; Metronet
objectives as presented in the grant proposal were as follows:
- To provide greater fundamental understanding of the metallurgical and mechanical changes in rail steels during the initial stages of service loading (1.1)
- To produce an accurate mechanics model for the wear-fatigue interaction in rails during service (1.2)
- To produce an accurate model for predicting rail life for different steel grades and in-service conditions (1.3)
Project summary at time of proposal: Wear-fatigue interaction, especially crack truncation, plays an important role in determining the life of rails. When the cracks are long, crack propagation is driven by the distribution of stress within the railhead whereas crack truncation, caused by wear, is dominated by the stress at the railhead surface. For short cracks, however, these mechanisms are not distinct. An existing model shows how failure at the surface leads to wear, as well as crack-like features. Further research is required using this model to include detailed microstructural and mechanical property data for the steels of interest at the micron (x 10-6m) and nanometre (x10-9m) scales. Detailed micro- / nano-hardness and ductility values from the surface are required along with microstructural characterisation of the steel both before service and during service. This information, along with the links to the micro-mechanisms of fatigue crack initiation,
will form an input into the model of wear-fatigue interaction during crack initiation and early growth. Important parallel research work on collecting field data and developing empirical models is on-going in the rail research community. This field data will be used for validation and for developing better and accurate life prediction models for the industry.
Summary of outcomes:
The main outcomes from the work are:
- Robust data, from worn rail examinations and twin-disc tests, linking pro-eutectoid ferrite fraction in ferrite-pearlite rail steels to rolling contact fatigue (RCF) life (i.e. crack initiation and early propagation).
- A fundamental understanding of the role of pro-eutectoid ferrite on crack initiation.• Mechanical property data for pro-eutectoid ferrite and pearlite during deformation for input into mechanics modelling of crack initiation and wear.
- A mechanics model for the wear-fatigue interaction in ferrite-pearlite rail steels taking into account microstructural data including failure mechanisms.
Summary of work done: The ‘brick’ model, a computer model of plastic strain accumulation, wear and rolling contact fatigue crack initiation in rail steel, has been developed to simulate a microstructure of multi-phase materials (e.g., pro-eutectoid ferrite and pearlite in pearlitic rail steel). To provide the required material property and microstructural data for the model, sections of worn and fatigued rail, removed from service, have been metallurgically analysed. To obtain further data on rail-steel deformation and rolling contact fatigue crack initiation mechanisms, twin-disc tests have been performed using discs cut from across a railhead and wheel rim. Two heat treatments were applied to some rail discs to investigate the effect on fatigue life of pro-eutectoid ferrite phase distributions and volume fractions. Tests were run to failure (defined by an eddy current crack-detection system) and to percentages of fatigue lives. Micro- and nano-hardness tests, and microstructural observations, have suggested a micro-mechanism of fatigue crack initiation for the model, which has led to a proposed new method of differential plastic strain accumulation.The interface between short plastic ratchetting driven crack growth and longer crack growth described by fracture mechanics has been investigated to understand how these mechanisms interact, and how transition between mechanisms takes place. Correlation with wear and crack growth field data has been sought.
- Metallurgical analysis (optical and scanning electron metallography; macro-, micro- and nano-hardness
testing; chemical analysis)
- Heat treatment
- Twin-disc testing
- Computer modelling
- Correlation and validation with field data
Results and Analysis: Rails taken from service with RCF cracking were sectioned axially, transversely and normal to primary
crack alignments, this latter, angled direction being determined by the combination of longitudinal,
transverse and spin creepage at that part of the rail surface. These angled sections show the true regularity of crack formation. The microstructural observations from the rails removed from service indicate that the pro-eutectoid (prior austenite grain boundary) ferrite can influence the fatigue crack initiation and early propagation stages, in agreement with our previous work and recent literature. Consequently the effect of pro-eutectoid ferrite volume fraction on fatigue crack initiation life and crack initiation mechanisms was investigated using twin-disc tests. Discs were machined from a railhead
and tested in 3 conditions: untreated (identified as ‘RN’); heat-treated to maximise the pro-eutectoid
ferrite content (identified as ‘R84’); and heat-treated to minimise pro-eutectoid ferrite content (identified as ‘R115’). For all three rail microstructures, pro-eutectoid ferrite does not completely surround
the pearlitic areas; it is non-continuous (in two-dimensional views), but there can be a considerable
extent of ferrite along the boundary.
The results from twin-disc testing showed that the R115 discs had a longer mean fatigue life than the RN discs, which in turn had longer fatigue lives than the R84 discs; however, there was also greater variability between the fatigue lives for the R115 discs. A possible cause can be found in the traction coefficient for the low-life tests, which was uncharacteristically high compared to the other tests. Also, singular defects, rather than the multiple flakes seen for the other discs, were produced in the short life tests. The shorter fatigue lives for the R84 discs were found to be associated with faster crack growth rates, confirming that the different microstructural conditions have a significant effect on fatigue life. Surface RCF crack initiation features were similar to those seen on rail removed from service with early RCF damage. Observations from circumferential microsections indicated that numerous fatigue cracks in the detected ‘defect zone’ had initiated along highly strained, pro-eutectoid ferrite band boundaries. The crack propagation direction appears primarily determined by the strain field. However, there was microstructural modification, with many crack path diversions along the edges of strained pro-eutectoid ferrite bands being observed, with occasional branching to strain-flattened MnS-based inclusions.
The strain hardening of the pearlite and pro-eutectoid ferrite phases, as reflected by percentage nano-hardness increase over core hardness, indicated that the percentage mean pro-eutectoid ferrite hardness increase, in the zone of maximum strain-hardening at around 200µm sub-surface and near-surface, is higher than that of the pearlite phase, as observed with worn rail examinations. The fatigue life results
from the twin-disc tests (shorter life for the microstructures containing the higher fractions of pro-eutectoid ferrite), coupled with the findings in the literature and detailed metallographic analysis, indicated that the thicker the initial areas of pro-eutectoid ferrite the higher the degree of straining in these regions (due to the lower constraint factor of the more rigid surrounding pearlite) hence the earlier strain exhaustion and crack initiation.Hence, Objective 1.1 of the work was achieved for ferrite-pearlite rail steels through the identification of the role of pro-eutectoid ferrite on rolling contact fatigue crack initiation and early stage growth mechanisms. Additionally nano-hardness results allowed the strain accumulation in the separate microstructural phases to be quantified.
The ‘brick’ model, developed during previous work on wear-fatigue interaction and used in this project, simulates accumulation of plastic shear strain in the rail or disc. A cross-section, parallel to the resultant
direction of traction, is represented as a mesh of rectangular elements (or ‘bricks’). In order to model the rail steel microstructure, each element is assigned a material type and related properties, in particular the initial hardness of the material and how much plastic shear strain can be accumulated before the element’s ductility is exhausted and the material ‘fails’. Twin-disc testing or wheel-rail contact
can be simulated in this way. With each load pass, each element in the model is subject to a stress cycle, and if the maximum orthogonal shear stress (in the plane of the mesh) exceeds the element’s shear yield stress (which increases as the material hardens) then there is an increment of plastic shear strain proportional to the difference.
This process of strain accumulation is plastic ratcheting. Under certain conditions, elements that are exposed to the surface can be removed as wear debris, and thereby the wear rate can be estimated. Aligned sub-surface failed bricks can be taken to represent cracks, as discussed below.
A simple representation of a pearlitic rail steel microstructure was developed for the model based on the microstructures analysed and the nano-hardness data generated in the project. A number of simulations representing the twin-disc experiments were performed using the ‘brick’ model. The predicted
shear strain and hardness are a close match to those observed in the twin-disc tests. However, in contrast to the experimental micro-hardness data, which predicts the peak hardness at depths in the range 150-180µm, the model predicts peak hardness at about 120µm; this needs further investigation. Further analysis of the twin-disc results will be used to refine the parameters and further calibrate the model.
In the ratchetting simulation the elements are shaded according to the amount of strain they have accumulated,
so the pro-eutectoid ferrite areas at the prior austenite grain boundaries appear relatively dark. Since the pro-eutectoid ferrite elements accumulate strain faster, and since the critical strain at which failure occurs is assumed to be the same for pro-eutectoid ferrite and pearlite (this assumption requires further study), if the stresses are high enough for failure to occur then the pro-eutectoid ferrite fails sooner. A failed element is material in which crack initiation is likely, and therefore the pro-eutectoid
ferrite zones at prior austenite grain boundaries provide clear paths for cracks to initiate along. (In fact, this is an overly simple interpretation. In practice, cracks are more likely to initiate along the interface between a pro-eutectoid ferrite region and a pearlite region where there is a sharp difference in strain. This aspect of the model needs further work.) The results of the twin-disc tests suggest that crack initiation along strained pro-eutectoid ferrite bands occurs sooner as the percentage volume of ferrite in the rail steel increases. As the thickness of the pro-eutectoid ferrite bands decrease, the pearlite
grains (i.e., the prior-austenite grains) are increasingly connected and may provide a shielding effect which constrains plastic deformation of the relatively soft pro-eutectoid ferrite. Based on this hypothesis,
a new shielding mechanism has been developed for the ‘brick’ model.
Hence, Objective 1.2 was achieved by further development of the ‘brick’ model to account for actual rail microstructures and relating the model predictions to twin-disc test results.
Cracks initially produced by accumulation of strain and ductility exhaustion can grow to larger sizes at which long crack growth mechanisms described by fracture mechanics begin to drive growth. The transition between these mechanisms required further study to understand how they act in combination,
and at what crack size the transition is complete. The influence of microstructure on crack growth rate is expected to be high during this transition.
Objective 1.3, that the model developed during the project should accurately predict rail life for different
steel grades, has been the most challenging of the modelling tasks. Data is available from the twin-disc tests for three different steel microstructures (i.e. three differently heat treated steels). These can be taken to represent grades such as 220 grade and 260 grade or wear resistant “head-hardened” steel, used in rails which offer enhanced wear resistance, but at the expense of reduced ductility. This data has formed an input to the model to understand the wear and crack growth performance of steels with a pearlitic microstructure that has undergone heat treatment to alter its properties. However, some properties
of such steels remain to be quantified, for example the stress-strain curve of head hardened steels under the very highly compressive stress regime present at the rail-wheel contact. This data is to be generated in a parallel project (INNOTRACK) using a specially built high pressure test cell. Currently, modelling output has been generated based on estimates of this stress-strain curve, but when better data is available this will be applied using the microstructural model developed within RRUK.
To assess the outputs of the microstructural model (based on the input data available) extensive comparisons
have been made between its predictions and data collected by industrial partners (AEAT Rail and Corus) at sites on the UK rail network. The line at Leigh-on-Sea (London to Southend route) was considered first, using vehicle dynamics data to supply rail surface loads for several traffic types operating
on this route. Good correlation was found between predictions of plastic deformation, wear and crack initiation and the sites on the line at which these did and did not occur. In addition, the model was able to distinguish between older vehicles which had been less damaging to the track, and newer vehicles in which higher track forces lead to increased rail damage. Further work correctly predicted differences in behaviour between sites on the East Coast Mainline at which cracking did and did not occur. The interaction between microstructurally controlled crack initiation which can be predicted with the model described here, and crack propagation was also examined at these sites. This showed that increased microstructural damage can dramatically reduce subsequent rail life through elimination of the small crack growth stage, which is usually slow, and responsible for a large proportion of crack life.
Hence, Objective 1.3, has been completed through prediction of wear and crack initiation behaviour, based on several assumptions, for pearlitic rail steel microstructures. This has been demonstrated for twin disc tests in which all variables were carefully controlled, and by comparison with field data on crack growth for several sites on the UK rail network. Further effort in this area is being concentrated on obtaining better data to further calibrate the model.
Conclusions: Detailed examination of worn rail taken from service has indicated that RCF crack initiation and propagation is facilitated by strain partitioning between the pro-eutectoid ferrite and pearlite, with higher percentage hardening and earlier ductility exhaustion of the pro-eutectoid ferrite phase. Many surface micro-cracks initiated down the border between the strained pro-eutectoid ferrite phase and strained pearlite. Early RCF crack propagation was facilitated running along these pearlite and highly strained, pro-eutectoid ferrite border zones, within the constraint of the overall strain field. Propagation
was also occasionally facilitated by highly strain-flattened, MnS-based inclusions.
The twin-disc tests, micro- and nano-hardness tests and metallurgical examinations indicate that the higher the amount of pro-eutectoid ferrite in the rail microstructure, the lower the mean RCF crack initiation life, due to more rapid straining of the pro-eutectoid ferrite when relatively unconstrained by neighbouring pearlite nodules. However, low pro-eutectoid ferrite samples gave both the highest mean life and the widest spread of RCF lives. Whether this outcome is characteristic of the higher hardness steel with minimised pro-eutectoid ferrite content remains to be investigated.
The twin-disc test programme was designed to provide a picture of the evolution of shear strain accumulation
and hardening with depth and number of load cycles. The detailed nano-hardness data distinguishing pro-eutectoid ferrite and pearlite has made it possible to define a two-material hexagonal
microstructure for use in a ‘brick’ model. Using this model, the predicted shear strain and hardness are a close match to those observed in the twin-disc tests.
Work with rail industry partners has given access to field data on crack initiation and wear for several UK locations. Comparison of modelling predictions with these data has given confidence in the predictions
of the microstructural model, and highlighted how predicted changes to the rail steel microstructure
can influence the early stages of long crack growth, dramatically changing overall crack and rail life.
Accurate material parameters and microstructural modelling are vital for predicting crack initiation and wear. This work represents an important step, but the current model uses a two-dimensional representation of the microstucture, making it difficult to completely represent different rail microstructures.
To compare crack initiation behaviour in different microstructures the model must therefore be enhanced to a three-dimensional representation.
Future work (see RRUK2 Project A6):
Future model development will address two key issues highlighted in the work to date:
- The generation and deformation of realistic three-dimensional microstructures will allow prior austenite grain boundaries to be mapped and probable paths of crack initiation to be identified.
- The interaction of pearlite zones and pro-eutectoid ferrite will be studied and modelled so that time to crack initiation can be predicted for different grades of pearlitic rail steel.
Future microstructural analysis will concentrate on the following areas:
- The 3D nature of initial crack growth and how it interacts with the 3D microstructure.
- The role of inclusions (e.g., MnS-based) on crack initiation and early stage growth, particularly in microstructures with little pro-eutectoid ferrite (e.g., grade 260 rail).
- Fletcher, D.I., Hyde, P., Kapoor, A. (2007) Investigating fluid penetration of rolling contact fatigue cracks in rails using a newly developed full-scale test facility. Proc. IMechE Vol. 221-1 Part F: J. of Rail & Rapid Transit. pp. 35-44.
- Franklin, F.J., Kapoor, A. (2007). Modelling wear and crack initiation in rails. Proc. IMechE Vol. 221-1 Part F: J. of Rail & Rapid Transit. pp. 23-34.
- Garnham, J.E., Franklin, F.J., Fletcher, D.I., Kapoor, A., Davis, C.L. (2007). Predicting the life of steel rails. Proc. IMechE Vol. 221-1 Part F: J. of Rail & Rapid Transit. pp.45-58.
- Fletcher, D.I., Hyde, P., Kapoor, A. (2006). Modelling and full scale trials to investigate fluid pressurisation of rolling contact fatigue cracks. Contact Mechanics of Rail Wheel Systems (CM2006), Brisbane, Australia, Sept 2006.
- Fletcher, D.I., Kapoor, A. (2006). A rapid method of stress intensity factor calculation for semi-elliptical surface breaking crack under elliptical contact loading. Proceedings of the IMechE Part F Journal of Rail and Rapid Transit, Volume 220, Number 3, 2006, pp. 219-234.
- Franklin, F.J., Garnham, J.E., Fletcher, D.I., Kapoor, A., Davis, C.L. (2006). Modelling Rail Steel Microstructure and its Effect on Wear and Crack Initiation. Contact Mechanics of Rail Wheel Systems (CM2006), Brisbane, Australia, Sept 2006.
- Garnham, J.E., Davis, C.L. (2006). Rolling contact fatigue initiation in pearlitic rail steels. Contact Mechanics of Rail Wheel Systems (CM2006), Brisbane, Australia, Sept 2006.
- Widiyarta, I.M., Franklin, F.J., Kapoor, A. Modelling thermal effects in ratcheting-led wear and rolling contact fatigue. Contact Mechanics of Rail Wheel Systems (CM2006), Brisbane, Australia, Sept 2006.
- Kapoor, A., Fletcher, D.I., Franklin, F.J., Vasic, G., Smith, L. (2005). New Developments In Rail-Wheel Contact Research At The University Of Newcastle. Proc 11th International Conference on Fracture (ICF11), Turin, Italy, March 20-25, 2005.
More detailed information will be available to members of the Engineering Interfaces Theme Network