Low-adhesion Modelling


    Some aspects of railways are still not completely understood, nearly 200 years after the railways were first developed. Indeed, your writer generally sees many research papers every year seeking to further advance knowledge about what is exactly happening in the tiny, one square centimetre contact patch between the wheel and rail.

    The contact patch is a key reason why rail is so successful, in the form of low rolling resistance, and is also a cause of many of its challenges, such as rolling contact fatigue and, the subject of this article, adhesion issues. If there is not a complete understanding of the steel to steel contact patch, imagine the extra complication when contamination is added to the interface, be it to the rail head or the wheel or both.

    Great Britain has led the way in understanding how the contact patch works in terms of vehicle dynamics and for traction and braking. Understanding how friction works is vital. RSSB hosts a wide range of industry stakeholder groups including the Adhesion Research Group, which has sponsored excellent work to develop strategies, tactics, and techniques to improve the certainty that a train can stop on demand.

    Rail Engineer has reported on many of the ARG/RSSB activities, particularly the development and testing of double variable-rate sanders. Other work has included the development of models to predict behaviour in low adhesion conditions at both wheel-rail contact and at train level. Low adhesion and how it affects the railway are, together, a complex problem, with simulation models and laboratory testing providing scientific understanding of complex issues confronting the operational railway.

    Given the complexity of these issues, different models are needed to fully explore and understand the factors involved, how they contribute to low adhesion and, ultimately, how we can develop better treatment and management techniques.

    A recent Adhere webinar focussed on three simulation and modelling tools for braking under low adhesion conditions. This article covers the University of Huddersfield’s work on LABRADOR and Sheffield University and Virtual Vehicle on LILAC. A presentation about DB ESG’s Wheel Slide Protection Evaluation Rig (WSPER) was also included in the Webinar, but WSPER has been covered in Rail Engineer before (issue 176, July 2019) and is not repeated here.


    No model is any use unless it replicates what happens in real life. Gaining the knowledge to create the model and then to validate it is often the hardest part.

    Roger Lewis from the University of Sheffield described the research carried out for RSSB (Project T1149) to understand in detail the impact of the leaf layer and to provide input to the Huddersfield’s LABRADOR low-adhesion braking model as a module called LILAC (Leaf Induced Low Adhesion Creep force model), which can also be used as a standalone tool or in a multi-body dynamics simulation.

    This work built on the successful project to understand and model the impact of small amounts of water with iron oxide contaminants that had already been incorporated in LABRADOR as WILAC (Water Induced Low Adhesion Creep force model – RSSB project T1077).

    The project started with extensive data collection, including from past low adhesion incidents. This added to the researchers’ body of knowledge about how vehicles have performed and helped form ideas for laboratory trials using Huddersfield University’s full-scale bogie test rig HAROLD (issue 145, November 2016).

    In brief, HAROLD consists of a bogie with one wheelset running on a two-metre-diameter rail roller which is driven by an electric motor. Vertical load is applied by hydraulic actuators through the secondary suspension and the bogie can be moved laterally and in yaw with respect to the rail rollers. A range of sensors and instrumentation is available to collect results, which include measurement of creepage and creep forces between the wheel and rail.

    A Y25 freight bogie was used for these tests, with one wheel jacked clear of the rail. This wheel was used for braking whilst the contamination (leaf layer/paper tape etc.) was applied to the other wheel. This was necessary simply because the bogie used was tread braked, and this approach avoided results being adulterated by the effect of tread braking on the contaminant.

    Roger described what seemed to be a really tedious process of assembling lengths of leaves, stuck together with adhesive tape, to artificially contaminate the active wheel/rail interface, adding that the set up took a lot longer than running the tests! These had to be carefully fed into the wheel-rail “nip” and rolled into the wheel and rail of the rig at low speed.

    The team later found that feeding broken up leaves into the nip using a scoop was quicker and just as effective.

    The surfaces only resembled that black surface seen on the real railway after a sliding event; rolling alone was insufficient.

    The standard test procedure was to run the test rig up to load and velocity, apply and increase the brake force until wheel slide or maximum brake force application, and then decelerate the roller to zero velocity. A simple real-time controller was used to release the brakes when a preset level of creepage was reached, in order to avoid wheel flats forming.

    Tests were carried out in dry and wet conditions and with leaves, paper tape and soap. Paper tape and soap are frequently used for practical tests on full scale vehicles. The results below compare dry conditions with leaf film.

    For dry conditions, Roger illustrated the results from the HAROLD rig. The panel on the next page describes what actually happens if the retarding force on the wheelset exceeds the available adhesion. Suffice to say, the dry results delivered characteristic creep curves with the coefficient of friction peaking at between 0.4 and 0.5.

    For the leaf film tests – where the leaf film was wetted before each test, the same characteristic curves were obtained, but with an exceptionally low peak coefficient of friction of <0.02. Roger put context to this figure by explaining that best quality oil in a car’s engine does not produce a friction level this low, but he advised that leaves should not be used as engine lubricant!

    Gerald Trummer of Austria’s Virtual Vehicle Research Institute described how the outputs of these tests and other information were used to create the LILAC module in a form ready for integration into LABRADOR. He demonstrated how the user interface works and the results obtained by varying input parameters, comparing the results with real world results.

    For future research, the developers want to investigate the role of transients – how contamination levels, environmental conditions, wheel/rail interface conditions and surface roughness, which vary along the railway, impact on the use of the model.


    LABRADOR – Low Adhesion BRAking Dynamic Optimisation for Rolling stock – is a modelling tool developed by the University of Huddersfield as part of its strategic partnership with RSSB. It was developed to simulate modern multiple-unit train braking in normal and low adhesion conditions and incorporates several different models for low adhesion in the wheel-rail interface.

    Huddersfield’s Julian Stow described the LABRADOR model, paying tribute to his colleagues Dr Hamid Alturbeh, who developed the model, and Jose Santos, who carried out its validation.

    Julian described the many challenges that modelling braking has to overcome:

    Braking performance depends on the complex interaction of a large number of components and sub systems on a train, with each vehicle delivering a different performance from its neighbours and, to an extent, each vehicle’s performance is influenced by the behaviour of its neighbours.

    It is not easy to quantify the low adhesion creep force relationship, as illustrated by the WILAC/LILAC research.

    Creating consistent low adhesion for laboratory or field tests is costly, time consuming and hard to arrange.

    Modelling real-life adhesion variation spatially and temporally is very difficult and there is little real-world data to feed into the models.

    There is limited validation of the wheel/rail contact temperature model.

    Characterising modern ‘black box’ WSP is a challenge as suppliers, understandably, may not wish to disclose detailed algorithms.

    Speed, load and contact temperature affect braking performance and have to be taken into account.

    All this means that the model has to cover variations in both the train and the infrastructure.

    For the train, it needs to provide for variable train length and to model the configuration of the braking system on each vehicle, how they are controlled and the interactions between vehicles. LABRADOR can model a four-car multiple unit (extendable to any length) with user defined parameters, and user configurable braking arrangement, controller architecture and configurable WSP. It is not just braking systems that need to be considered, but also masses, suspension configuration (for weight transfer effects) and knowledge about rotating masses.

    The infrastructure must also be modelled, including track gradients as well as adhesion profiles.

    The next step is to model how the train model interacts with the infrastructure with constant and variable brake demand, allowing for dynamic braking with blending and over-braking, the impact of sanding and conditioning effects, and weight transfer due to pitch effects (body and bogie).

    All this allows the stopping distance to be predicted, together with any tread damage due to wheel slide and the volume of compressed air used by sanding and braking (WSP). LABRADOR also generates a large amount of detailed information on what is happening to all the elements of the system throughout the simulated brake application.

    The modelling software is configured in a modular structure and can ‘plug & play’ new modules. Indeed, WILAC and LILAC are modules in LABRADOR.

    Julian then gave a professional demonstration of the software. The range of features and variables demonstrated very clearly his comment that, to get credible results, experience of braking systems and of using engineering modelling software is necessary.

    He concluded that low adhesion is a complex and persistent problem, that simulating low adhesion braking presents considerable challenges, and that the LABRADOR model allows the study of specific brake-control features, including WSP strategies, sanding effectiveness, dynamic brake utilisation and traction performance. He added that good simulation tools can support industry efforts to provide reliable (“seasonally agnostic”) braking in these conditions, and that further development of the model is underway.

    This work is at the forefront of the industry’s efforts to combat low adhesion problems. A point that was stressed by all the speakers is that there is now a full suite of tools for investigating and improving low-adhesion braking performance. These range from simulation of the complex behaviour of the contact conditions and train braking system, through small and full-scale lab testing, to WSP and sander optimisation and accreditation using DB ESG’s WSPER rig. These hold out real hope for future improvements to the autumn performance problems which are the cause of so much cost to the industry and pain to its customers.

    One of the key pieces of learning your writer took away from this webinar is that really low adhesion cannot be overcome by wheel/rail braking alone. It was frequently stated that, on a fairly long train, wheelside protection activity on the front cars ‘cleans’ the rail, allowing for better adhesion further back on the train, as shown in the illustration below taken from a LABRADOR simulation. The illustration (above) shows that this effect does not occur in the case of exceptionally low adhesion. Therefore, both railhead cleaning and the use of an adhesion enhancer such as sanding remain absolutely essential to manage the risk of poor braking in the autumn.


    Creep and creep curves

    Creep is a term used to describe when the train wheels are not simply rolling but are slipping (traction) or sliding. Clearly, a wheel that is spinning out of control or is locked is the most extreme manifestation of this, but small amounts of creep can be beneficial.

    If the applied tractive or braking effort is greater than the coefficient of friction multiplied by the vertical load, the wheel will start to slip or spin. But, if that happens, there is a curious effect – the coefficient of friction starts to increase and continues to do so until the wheel is rotating at about 10% slower (using a braking example) than the rolling speed. At this point, known as the saturation point, if the speed of rotation continues to fall, the friction falls away quite rapidly and the wheel will lock – effectively 100% creep.

    In the case of traction, creep values can be in excess of 100%, which is the effect of wheelspin.

    This phenomenon is often used in locomotives to deliver more tractive effort than the nominal adhesion level would permit. With modern control systems, creep control is comparatively straightforward and, in principle, could be applied to dynamic braking. However, with pneumatic friction braking, the response times are too slow to permit creep control; the accepted method of controlling brakes in poor adhesion is with wheelslide protection.